Population‐based interventions for preventing falls and fall‐related injuries in older people

Sharon R Lewis; Lisa McGarrigle; Michael W Pritchard; Alessandro Bosco; Yang Yang; Ashley Gluchowski; Jana Sremanakova; Elisabeth R Boulton; Matthew Gittins; Anneliese Spinks; Kilian Rapp; Daniel E MacIntyre; Roderick J McClure; Chris Todd

doi:10.1002/14651858.CD013789.pub2

Intervenciones poblacionales para la prevención de las caídas y las lesiones derivadas en personas mayores

Declaraciones de intereses de los autores

Versión publicada: 05 enero 2024 Historial de versiones

https://doi.org/10.1002/14651858.CD013789.pub2

Contraer todo Desplegar todo

Resumen

disponible en

Antecedentes

Alrededor de un tercio de los adultos mayores de 65 años que viven en la comunidad sufren caídas cada año. Las intervenciones para prevenir las caídas se pueden dirigir a toda la comunidad, en lugar de a individuos seleccionados. Estas intervenciones a nivel poblacional pueden facilitarlas distintos agentes sanitarios, sociales y comunitarios. Su objetivo es abordar los factores determinantes del riesgo de caídas en las personas mayores, e incluyen componentes como políticas comunitarias de administración de suplementos de vitamina D a las personas mayores, la reducción de los riesgos de caídas en la comunidad o en los hogares o el suministro de información de salud pública o la aplicación de programas de salud pública que reduzcan el riesgo de caídas (p. ej., la inscripción gratuita o a bajo coste en gimnasios para las personas mayores con el fin de fomentar una mayor actividad física).

Objetivos

Revisar y resumir la evidencia actual sobre los efectos de las intervenciones poblacionales para la prevención de las caídas y las lesiones relacionadas con las caídas en personas mayores. Las intervenciones poblacionales se definieron como iniciativas comunitarias para cambiar las condiciones sociales, culturales o ambientales subyacentes que aumentan el riesgo de caídas.

Métodos de búsqueda

Se realizaron búsquedas en CENTRAL, MEDLINE, Embase, otras tres bases de datos y dos registros de ensayos en diciembre de 2020, y se realizó una búsqueda complementaria en CENTRAL, MEDLINE y Embase en enero de 2023.

Criterios de selección

Se incluyeron los ensayos controlados aleatorizados (ECA), los ECA por conglomerados, los ensayos con diseños escalonados y los estudios controlados no aleatorizados que evaluaran intervenciones poblacionales para la prevención de las caídas y las lesiones derivadas en adultos ≥ 60 años de edad. Las intervenciones poblacionales se dirigen a comunidades enteras. Se excluyeron los estudios dirigidos solo a personas con alto riesgo de caídas o con comorbilidades específicas, o a residentes en centros asistidos.

Obtención y análisis de los datos

Se utilizaron los procedimientos metodológicos estándar previstos por Cochrane y se utilizó GRADE para evaluar la certeza de la evidencia. Se priorizaron siete desenlaces: tasa de caídas, número de personas que sufrieron caídas, número de personas que sufrieron una o más lesiones relacionadas con caídas, número de personas que sufrieron una o más fracturas relacionadas con caídas, número de personas que requirieron ingreso hospitalario por una o más caídas, eventos adversos y análisis económico de las intervenciones. Otros desenlaces de interés fueron: número de personas que sufrieron una o más caídas que requirieron atención médica, calidad de vida relacionada con la salud, mortalidad relacionada con las caídas y preocupaciones por las caídas.

Resultados principales

Se incluyeron nueve estudios: dos ECA por conglomerados y siete ensayos no aleatorizados (de los cuales cinco eran estudios controlados tipo antes y después [ECAD] y dos eran series temporales interrumpidas controladas [STIC]). El número de adultos mayores en las regiones de intervención y control varió entre 1200 y 137 000 residentes mayores en siete estudios. Los otros dos estudios solo proporcionaron el tamaño total de la población en lugar del número de adultos mayores (67 300 y 172 500 residentes). La mayoría de los estudios utilizaron sistemas de registros hospitalarios para recoger los datos de los desenlaces, pero tres solo utilizaron datos de cuestionarios de una muestra al azar de los residentes; un estudio utilizó ambos métodos de recogida de datos. Los estudios duraron entre 14 meses y 8 años.

Se utilizó la taxonomía Prevention of Falls Network Europe (ProFaNE) para clasificar los tipos de intervenciones. Todos los estudios evaluaron intervenciones multicomponentes de prevención de caídas. Un estudio (n = 4542) también incluyó una intervención de medicación y nutrición. No se agruparon los datos debido a la falta de homogeneidad en los diseños de los estudios.

Medicación o nutrición

A las personas mayores del área de intervención se les ofrecieron gratuitamente suplementos diarios de carbonato cálcico y vitamina D ₃ . Aunque las residentes expuestas a este programa de prevención de caídas tuvieron menos ingresos hospitalarios relacionados con las caídas (sin evidencia de una diferencia con los residentes varones) en comparación con un área control, no se tiene confianza en este hallazgo porque la certeza de la evidencia fue muy baja. Este ECA por conglomerados incluyó riesgos de sesgo altos e inciertos en varios dominios, y no fue posible determinar los niveles de imprecisión en la estimación del efecto proporcionada por los autores del estudio. Dado que esta evidencia es de certeza muy baja, no se han incluido aquí los resultados cuantitativos. Este estudio no informó sobre ninguno de los otros desenlaces de la revisión.

Intervenciones multicomponentes

Los tipos de intervenciones incluyeron componentes de ejercicios, modificación del entorno (domicilio; comunidad; espacios públicos), formación del personal, y conocimientos y educación. Los estudios incluyeron algunos o todos estos componentes en el diseño de sus programas.

No está clara la efectividad de las intervenciones multicomponentes para la prevención de las caídas en todos los desenlaces presentados. Los dos ECA por conglomerados incluyeron un riesgo de sesgo alto o incierto, y no hubo razones para aumentar la certeza de la evidencia de los diseños de ensayos no aleatorizados (que comenzó como evidencia de certeza baja). También se observaron posibles imprecisiones en algunas estimaciones de los efectos y resultados incoherentes entre los estudios. Dada la evidencia de certeza muy baja para todos los desenlaces, no se han proporcionado en la presente resultados cuantitativos.

Un ECA por conglomerados informó de tasas más bajas de caídas en el área de intervención que en el área control, y hubo menos personas en el área de intervención que sufrieron una o más caídas y lesiones relacionadas con las caídas, pero poca o ninguna diferencia en el número de personas que sufrieron una o más fracturas relacionadas con las caídas. En otro ECA por conglomerados (un estudio de varios grupos), los autores del estudio señalaron que no hubo evidencia de diferencias en el número de residentes mujeres u hombres que sufrieron caídas que provocaran ingresos hospitalarios tras una intervención multicomponente ("programa ambiental y de salud") o una combinación de este programa y el programa de calcio y vitamina D ₃ (mencionado anteriormente).

Un ECAD no informó de diferencias en la tasa de caídas entre las áreas de intervención y de control, y otro ECAD no informó de diferencias en la tasa de caídas dentro o fuera del domicilio. Dos ECAD no encontraron evidencia de diferencias en el número de caídas, y otro ECAD no encontró evidencia de diferencias en las lesiones relacionadas con las caídas. Una STIC no encontró evidencia de diferencias en el número de personas con una o más fracturas relacionadas con las caídas.

Ningún estudio informó sobre eventos adversos.

Conclusiones de los autores

Debido a la evidencia de certeza muy baja, no está claro si las intervenciones poblacionales multicomponentes o de nutrición y medicación son efectivas para reducir las caídas y las lesiones relacionadas con las caídas en los adultos mayores. Se necesitan ECA por conglomerados metodológicamente sólidos con comunidades y números de conglomerados suficientemente grandes. Establecer un índice de muestreo para los estudios poblacionales ayudaría a determinar el tamaño de las comunidades que se deben incluir. Las intervenciones se deben describir detalladamente para permitir la investigación de la eficacia de los componentes individuales de las intervenciones multicomponentes; el uso de la taxonomía ProFaNE para ello mejoraría la consistencia entre los estudios.

PICO

Population

Intervention

Comparison

Outcome

El uso y la enseñanza del modelo PICO están muy extendidos en el ámbito de la atención sanitaria basada en la evidencia para formular preguntas y estrategias de búsqueda y para caracterizar estudios o metanálisis clínicos. PICO son las siglas en inglés de cuatro posibles componentes de una pregunta de investigación: paciente, población o problema; intervención; comparación; desenlace (outcome).

Para saber más sobre el uso del modelo PICO, puede consultar el Manual Cochrane.

Resumen en términos sencillos

disponible en

¿Son útiles las intervenciones poblacionales (dirigidas a comunidades enteras en lugar de a individuos) para prevenir las caídas y las lesiones derivadas en personas mayores?

Mensajes clave :

• No está claro si los enfoques de prevención de las caídas dirigidos a toda la comunidad reducen las caídas y las lesiones relacionadas con ellas.

• Los estudios futuros deben estar bien diseñados y utilizar descripciones actualizadas de sus intervenciones. Lo ideal sería que los estudios se realizaran en varias comunidades (en lugar de solo dos comunidades de estudio), cada una de ellas con grandes poblaciones, y que los tipos de personas mayores que vivieran en las comunidades de estudio fueran representativos del país en el que se realizara el estudio.

¿Por qué es importante prevenir las caídas?

Las caídas en las personas mayores son muy frecuentes. Aproximadamente un tercio de las personas de 65 años o más se cae cada año, y algunas personas mayores pueden sufrir varias caídas al año. Las caídas en personas mayores pueden ser muy graves y provocar fracturas óseas y tratamiento hospitalario. Una mala caída puede afectar gravemente la calidad de vida de la persona y posiblemente conlleve una larga recuperación. Como las caídas de las personas mayores pueden requerir tratamiento hospitalario, incluida la cirugía para las fracturas de huesos, también son costosas en términos económicos para los servicios sanitarios. Encontrar formas de prevenir las caídas beneficiará a las personas mayores y reducirá la carga que estas suponen para los servicios sanitarios.

¿Qué son los enfoques poblacionales para la prevención de las caídas?

Las estrategias de prevención de las caídas en la tercera edad se suelen dirigir a las personas con mayor riesgo de sufrirlas. Las personas con mayor riesgo pueden haber sufrido ya al menos una caída o tener otras afecciones que aumenten su riesgo de caídas (como problemas para caminar o desplazarse o de equilibrio). Los enfoques poblacionales son diferentes porque se dirigen a comunidades enteras y no a individuos. Entre los ejemplos de enfoques poblacionales de prevención de caídas se incluyen las iniciativas de salud pública dirigidas a informar al público sobre los beneficios de las actividades físicas (p. ej., ejercicios de fuerza y equilibrio); las visitas a todas las personas mayores en sus domicilios para ayudarles a identificar y reducir los riesgos de caídas; o los ayuntamientos que mejoran los paseos públicos y la iluminación en pueblos o ciudades.

¿Qué se quería averiguar?

Se quería averiguar la eficacia de los enfoques poblacionales para prevenir las caídas o las lesiones relacionadas con ellas en los adultos mayores.

¿Qué se hizo?

Se buscaron estudios que compararan las caídas y las lesiones derivadas en comunidades que utilizaron estrategias de prevención de las caídas en toda la comunidad (es decir, enfoques basados en la población) con comunidades que no recibieron ninguna intervención. Se compararon y resumieron los resultados de estos estudios y la confianza en la evidencia se evaluó sobre la base de factores como la metodología y el tamaño de los estudios.

¿Qué se encontró?

Se encontraron nueve estudios dirigidos a participantes de al menos 60 años de edad de comunidades de ocho países diferentes. Las comunidades estudiadas variaban en tamaño. La mayoría de los estudios informaron el número de residentes de edad avanzada, que oscilaba entre 1200 y casi 137 000 residentes de edad avanzada. Otros estudios solo informaron el tamaño de la población total de las comunidades estudiadas, que fue de entre 67 300 y 172 500 residentes. Los estudios duraron entre 14 meses y 8 años. Por lo general, los enfoques incluían varios componentes, como el ejercicio, la educación o la reducción de los riesgos de caída en el hogar (como el uso de barandillas o alfombras antideslizantes) o la reducción de los riesgos de caída en la comunidad (mejora de las aceras y del alumbrado público). Un estudio también analizó el beneficio de un suplemento diario gratuito de calcio y vitamina D.

Resultados principales

No está claro si ofrecer suplementos de calcio o vitamina D a todas las personas mayores de la comunidad reduce el número de personas que necesitan tratamiento hospitalario por caídas.

Tampoco está claro si los enfoques poblacionales que de varios componentes reducen el número de caídas o el número de personas que sufren una o más caídas. Tampoco está claro si estos enfoques suponen alguna diferencia en el número de personas con fracturas óseas relacionadas con las caídas, o si reducen el número de personas con lesiones relacionadas con las caídas o los ingresos hospitalarios relacionados con las caídas. Además, no está claro que estos enfoques hayan supuesto un ahorro para los servicios sanitarios.

¿Cuáles son las limitaciones de la evidencia?

No se tiene confianza en la evidencia porque en algunos de los estudios incluidos las comunidades no fueron elegidas al azar para recibir las estrategias de prevención de caídas. Se trata de un diseño habitual en los estudios poblacionales, pero podría entrañar diferencias entre las comunidades que afecten los resultados. Los estudios no proporcionaron suficiente información para evaluar si estaban bien realizados. Además, los resultados a menudo diferían entre los estudios, y no fue posible identificar el motivo.

¿Cuál es el grado de actualización de esta evidencia?

La evidencia está actualizada hasta enero de 2023.

Authors' conclusions

Implications for practice

Owing to the very low‐certainty evidence in this review, we are uncertain whether nutrition or medication interventions or multicomponent interventions reduce rate of falls, number of fallers, number of people with fall‐related injuries, fractures or hospital admissions, adverse events, or cost savings. We found no data for other specific falls prevention intervention types (exercise and physical activity; environmental interventions; educational interventions).

Implications for research

There is often a false dichotomy between all older adults and those who started at risk of falls. As the World Guidelines point out, being at low risk is not at no risk; everyone is at risk of falling (Montero‐Odasso 2022). Population‐based interventions aim at the maintenance and improvement of health, and are directed at a population level, rather than having the individual as the main focus. In this review, interventions for falls prevention were aimed at all members of a population, regardless of risk. Introducing population‐based interventions likely needs to commence at an early age in order to prevent or at least delay the development of risk.

Establishing a unit of analysis or ‘population unit’ is complex for this type of review. It encompasses a variety of settings (e.g. medical practices, hospitals, or entire communities). In our review, we found it difficult to compare studies. Often this was because studies lacked sufficient information about population characteristics, as well as detail about the falls prevention interventions used in the intervention areas. More studies would increase the certainty of the evidence for population‐based falls prevention interventions. We encourage future studies to use the Prevention of Falls Network Europe (ProFaNE) taxonomy when describing study interventions (Lamb 2011); to follow the ProFaNE core outcome measure recommendations (Lamb 2005); and to use clear guidance on the definitions of injurious falls (Schwenk 2012). More intervention detail allows for the investigation of the effectiveness of individual components, particularly for study designs that use multicomponent interventions. In addition, intervention details will allow for better replicability of the interventions for future studies or for other stakeholders and policymakers. Including a large and populated area is likely to better support generalisability of the findings to other populations or communities. For studies that include sampling of residents for the collection of outcome data, large sample sizes would also provide more reliable data. Other review designs may enable exploration of the mechanisms by which interventions work (such as in the work by Boulton 2020), with the potential for primary research to feed into these reviews, and such review designs would also require sufficient intervention details.

We recognise that population‐based studies often lend themselves to using non‐randomised study designs. Studies using these designs should still include efforts to ensure that confounders are accounted for and that designs are methodologically robust. However, we believe that cluster randomised controlled trials are the best study design to evaluate population‐based interventions, and, if done well, are likely to provide more certain evidence than non‐randomised designs. Reporting guidelines should be followed to ensure clarity of design and conduct of studies (CONSORT 2010), including the statement extension for cluster randomised trials (Campbell 2012), and stepped‐wedge designs (Hemming 2018).

We were unable to establish whether the sample of studies was representative of older populations residing in the area being assessed. Establishing a rate of sampling for population‐based studies would support this review. This information could ensure that recruited participants are representative in numbers and socio‐demographics of the entire population for the community/area of interest.

Summary of findings

Open in table viewer

Summary of findings 1. Medication or nutrition fall prevention interventions versus control: evidence from RCTs

Outcomes	Impact of the intervention	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Population: community‐dwelling older adults ≥ 65 years of age Settings: communities Intervention: free‐of‐charge daily supplement of calcium carbonate and vitamin D₃ Comparison: no falls prevention intervention
Rate of falls	Not estimable	‐	‐	No studies reported this outcome.
Number of fallers	Not estimable	‐	‐	No studies reported this outcome.
Number of people experiencing 1 or more fall‐related injuries	Not estimable	‐	‐	No studies reported this outcome.
Number of people experiencing 1 or more fall‐related fractures	Not estimable	‐	‐	No studies reported this outcome.
Number of people experiencing 1 or more falls resulting in hospital admission Measured using Danish Hospital Registration Database Follow‐up: fall data collected during 42‐month study period	Female residents exposed to a "Calcium and Vitamin D" falls prevention programme had fewer fall‐related hospital admissions than female residents in the control area (RR 0.89; P < 0.10). For male residents, there was no evidence of a difference between the intervention and control areas (RR 1.08).^a	4542 (1 cluster RCT)	Very low^b
Number of people who experienced 1 or more adverse events	Not estimable	‐	‐	No studies reported this outcome.
Economic analysis	Not estimable	‐	‐	No studies reported this outcome.
Abbreviations: RCT: randomised controlled trial; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aData as reported in the study report. These data were reported separately for female and male residents, and were reported without confidence intervals. In addition, no P value was reported with the effect estimate for male residents. ^bWe downgraded by two levels owing to very serious risk of bias (high and unclear risk of bias). We also downgraded by one level for imprecision because the effect estimates were reported without confidence intervals, and we were unable to determine the degree of imprecision in the data (particularly for male residents).

Open in table viewer

Summary of findings 2. Multicomponent fall prevention interventions versus control: evidence from RCTs

Outcome	Impact of the intervention (findings as reported by study authors, unless specified otherwise)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Population: community‐dwelling older adults at least 60 years of age Settings: communities Intervention: multicomponent falls prevention interventions (details of components in each study are described in footnotes) Comparison: no falls prevention intervention
Rate of falls Measured using questionnaire data (self‐reported) Follow‐up: end of study follow‐up at 18 months	In a cluster RCT,^a the rate of falls was lower in the intervention area than in the control area (RaR 0.356, 95% CI 0.253 to 0.501).	1422 (1 cluster RCT)	⊕⊝⊝⊝ Very low^b
Number of fallers Measured using questionnaire data (self‐reported) Follow‐up: end of study follow‐up at 18 months	In a cluster RCT,^a fewer people had falls in the intervention area than in the control area (RR 0.34, 95% CI 0.19 to 0.62).^c	1422^d (1 cluster RCT)	⊕⊝⊝⊝ Very low^e
Number of people experiencing 1 or more fall‐related injuries Measured using questionnaire data (self‐reported) Follow‐up: end of study follow‐up at 18 months	In a cluster RCT,^a fewer people had injurious falls in the intervention area than in the control area (RR 0.39, 95% CI 0.20 to 0.77).^c	1422^d (1 cluster RCT)	⊕⊝⊝⊝ Very low^e
Number of people experiencing 1 or more fall‐related fractures Measured using questionnaire data (self‐reported) Follow‐up: end of study follow‐up at 18 months	In a cluster RCT,^a there was no evidence of a difference between the intervention and control group areas in fall‐related fractures (RR 0.55, 95% CI 0.17 to 1.85).^c	1422^d (1 cluster RCT)	⊕⊕⊝⊝ Very low^e
Number of people experiencing 1 or more falls resulting in hospital admission Measured using hospital records Follow‐up: falls data collected during 42‐month study period	In a cluster RCT^f evaluating an "Environment and Health" programme, there was no evidence of a difference between the intervention and control areas in number of females (RR 0.96) or males (RR 1.07) with falls leading to hospital admission. In the same cluster RCT, evaluating this intervention in combination with a nutrition and medication intervention ("Calcium and Vitamin D" programme), there was also no evidence of a difference between the intervention and control areas for females (RR 0.90) and males (RR 1.14).	7179 (1 cluster RCT)	⊕⊝⊝⊝ Very low^g
Adverse events	‐	‐	‐	No studies reported this outcome.
Cost‐effectiveness (economic analysis)	‐	‐	‐	No studies reported this outcome.
Abbreviations: CI: confidence interval; RaR: rate ratio; RCT: randomised controlled trial; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aInterventions in this study included: education programme (to reduce risk of falls, and including information on diet and exercise); home hazard assessment; modification of community settings (removing obstacles on pavements, roads, lawns; installing handrails). ^bDowngraded two levels for very serious risk of bias, and one level for imprecision because we could not be certain whether the effect estimate included an adjustment for clustering (and therefore may represent an overestimation of the true effect). ^cCalculated using Review Manager 2020 from data in the study report (Review Manager 2020). ^dIncluded data from a sample of the whole target population. To account for clustering, we calculated effect sample sizes to use in our analysis. ^eDowngraded two levels for very serious risk of bias, and one level for imprecision because the sample size for this population‐level study was small. ^fInterventions included: an "Environmental and Health Program" with home safety inspection, ways to avoid falls, health and dietary correction; a "Calcium and Vitamin D Program" in which residents were offered free‐of‐charge daily supplements; or a combination of both programmes. ^gWe downgraded by two levels for very serious risk of bias because this cluster RCT had high and unclear risk of bias. We also downgraded by one level for imprecision because we could not determine the level of imprecision in these data, which were reported without CIs and may not have been adjusted for the clustering effect.

Open in table viewer

Summary of findings 3. Multicomponent fall prevention interventions versus control: evidence from non‐randomised trials

Outcome	Impact of the intervention (findings as reported by study authors, unless specified otherwise)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Population: community‐dwelling older adults at least 60 years of age Settings: communities Intervention: multicomponent falls prevention interventions (details of components in each study are described in footnotes) Comparison: no falls prevention intervention
Rate of falls Measured using questionnaire data (self‐reported) or hospital records Follow‐up: falls data collected during study periods ranging from 14 months to 4 years	In a CBA,^a the reduction in rate of falls in the intervention group was not statistically significant (0.066 falls/person/year; P = 0.14). In another CBA,^b there was no evidence of a difference in rate of falls inside the home (RaR 1.07, 95% CI 0.39 to 2.99) or outside the home (RaR 0.91, 95% CI 0.61 to 1.37).^c	4197^d (2 non‐randomised trials)	⊕⊝⊝⊝ Very low^e
Number of fallers Measured using questionnaire data (self‐reported) Follow‐up: falls data collected during study periods ranging from 2 to 4 years	In a CBA,^a there was no evidence of a difference between the intervention and control areas in the number of fallers (OR 0.95, 95% CI 0.79 to 1.15). In another CBA,^f there was no evidence of a difference between the intervention and control areas in the number of fallers (RR 1.03, 95% CI 0.81 to 1.31).^c	3047^d (2 non‐randomised trials)	⊕⊝⊝⊝ Very low^e
Number of people experiencing 1 or more fall‐related injuries Measured using healthcare centre records Follow‐up: falls data collected during study period (5 years)	In a CBA,^g there was no evidence of a difference between the intervention and control areas in the number of people having injurious falls (OR 0.89, 95% CI 0.77 to 1.03).	67,300^h (1 non‐randomised trial)	⊕⊝⊝⊝ Very low^e
Number of people experiencing 1 or more fall‐related fractures Measured using hospital injury record system Follow‐up: falls data collected during 18‐month study period	In a CITS,ⁱ there was no evidence of a difference in the number of fractures that were prevented as a result of the intervention (14% prevented fractures in intervention group, 95% CI 9% more fractures to 37% fewer fractures).	24,365 (1 non‐randomised trial)	⊕⊝⊝⊝ Very low^j
Number of people experiencing 1 or more falls resulting in hospital admission Measured using hospital records Follow‐up: falls data collected during 42‐month study period	‐	‐	‐	No studies reported direct evidence for this outcome.
Adverse events	‐	‐	‐	No studies reported this outcome.
Cost‐effectiveness (economic analysis) Measured using healthcare records Follow‐up: falls data collected during study periods ranging from 4 to 8 years	A CBA^a reported a cost‐benefit in favour of the intervention with savings for avoided hospital admissions and indirect/direct costs (SCR 87.18, 95% CI 84.6 to 89.8). Another CBA^k reported cost reductions in favour of the intervention for hospital admissions (16.1%), hospital bed‐days (16.7%), and operations related to falls (35.1%).	163,683 (2 non‐randomised trials)	⊕⊝⊝⊝ Very low^l
CBA: controlled before‐and‐after study (a non‐randomised trial design); CI: confidence interval; CITS: controlled interrupted time‐series (a non‐randomised trial design); OR: odds ratio; RaR: rate ratio; RR: risk ratio; SCR: standardised cost ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aInterventions included: footwear, vision, physical activity, balance and gait, medication use, chronic conditions, plus home and public environmental hazards modification. ^bInterventions included: traffic safety, balance training, physical activity, home safety hazard modification, home modification, safe pavement (removing obstacles from pavements in the community), and medication use. ^cCalculated using Review Manager 2020 from data in the study report (Review Manager 2020). ^dIncluded data from a sample of the whole target population. ^eThere were no reasons to upgrade the certainty of the evidence from the non‐randomised trials. In addition, we downgraded the certainty of the evidence for imprecision. We also noted that the findings were inconsistent with the findings from the randomised controlled trial‐derived evidence. ^fInterventions included: an educational programme, government involvement with architectural consultations, exercise programmes, risk assessment, dietary and medicine guidance, and prevention of falls risk at home. ^gInterventions included: elimination of environment hazards (e.g. improvements to roads, pavements, street lighting); behavioural safety education and information programmes; injury prevention features in local media; availability of safety products; home modifications; and exercise support. ^hWhole population rather than target population of people aged ≥ 65 years. ⁱInterventions included: information on fall risk factors, and identifying and modifying hazards in the home and surrounding areas. Interventions aimed to reduce physical hazards, age‐debilitating illnesses, psychiatric illnesses, improper use of medication, diet insufficiency, and physical inactivity. ^jThere were no reasons to upgrade the certainty of the evidence from the non‐randomised trials. In addition, we downgraded the certainty of the evidence for imprecision. ^kInterventions included: identifying and remedy of home hazards, promoting environmental safety, health, diet, and lifestyle, and reduction in isolation and inactivity; pensioners providing skilled low‐cost services to improve physical environments in other older people's homes; availability of safety items including spiking of boots for icy pavements. ^lThere were no reasons to upgrade the certainty of the evidence from these non‐randomised trials. In addition, we downgraded by one level for indirectness. Although the population and interventions were eligible for this review, we believe the time at which these studies were conducted meant that economic analyses are less reliable because of other changes in the healthcare settings, and this may impact the directness of these results.

Background

Description of the condition

Falls represent a global public health concern, with nearly 700,000 accidental deaths globally every year, and are the second‐leading cause of mortality caused by unintentional injury after road traffic injuries (World Health Organization 2021). People aged 60 years and older represent the majority of fall‐related mortality cases (World Health Organization 2021). There is no agreed age cut‐off to determine older people; usually a cut‐off of 60 years and older or 65 years and older is used.

A fall can be defined as “an unexpected event in which the participant comes to rest on the ground, floor or lower level” (Lamb 2005). Physiological, cognitive, and environmental risk factors are responsible for an increased risk of falls and fall‐related injuries in older people (Becker 2017). Risk factors include socio‐demographic factors (such as age or sex), psychological factors (cognitive impairment, depression, concerns about falling, etc.), medical conditions (number of comorbidities, Parkinson’s, stroke, etc.), medication use (total number of medications, and types of medication, e.g. antiepileptics, sedatives, antihypertensive), and mobility and sensory issues (history of falls, walking aid use, gait problems, disability, vision or hearing impairment, physical inactivity, etc.) (Kinney 2004; Uusi‐Rasi 2017). Cognitive risk factors are related to deficits in executive function (e.g. slower processing speed and reaction times) (Welmer 2017). Environmental factors include poor footwear, inappropriate lighting, and uneven surfaces (Clemson 2019).

Nearly one‐third of community‐dwelling people aged 65 years and over experience a fall every year (World Health Organization 2007), with an estimated 30% to 50% annual risk in older people living in long‐term care institutions. Fall severity can be minor with short‐term consequences such as bruising and abrasions, or have long‐term consequences for the health of the person; more severe cases may lead to death (Gillespie 2012). In about a quarter of cases, a fall results in the older person either seeking medical help or restricting their activities for at least a day, or both (Bergen 2016). About 10% of falls lead to fractures, dislocations, or concussions (Kelsey 2012). In 2017, nearly 17 million years of life were lost from falls (James 2017). Falls account for two‐thirds of deaths associated with unintentional injuries in older people (Rubenstein 2006).

Furthermore, falls are associated with reduced physical functioning, loss of independence, and concerns about future falls, which can lead to reduced physical activity and social engagement (Frieson 2018; Gillespie 2012). Reduced levels of physical activity can negatively affect an individual’s strength and balance, increasing fall risk (Deandrea 2010). Reduced social engagement can lead to depression and poor quality of life (Delbaere 2010).

Serious injury in older people is a major risk factor for hospitalisation and long‐term care (World Health Organization 2021), representing an economic burden for national healthcare systems. For example, falls cost the UK National Health Service approximately GBP 2.3 billion per year (NICE 2013), whilst medical costs attributable to falls are approximately USD 50 billion in the USA (Florence 2018). Overall, it is estimated that high‐income countries spend about 1% of their healthcare budgets on falls (Montero‐Odasso 2022). Consequently, falls represent a serious public health problem, particularly in the context of an ageing population.

Description of the intervention

Interventions targeting falls prevention and fall‐related injuries usually target individuals with known susceptibility to modifiable risk factors or having a history of falls (Gillespie 2012), that is people identified as being at medium or high risk of falling. These fall prevention interventions usually include one or more of the following components: exercise (e.g. strength, balance, general physical activity); medication (e.g. vitamin D supplementation, medication review); medical intervention (e.g. cataract surgery, management of urinary incontinence, fluid or nutrition therapy); environmental intervention (e.g. home adaptation, mobility aids); psycho‐social intervention (e.g. cognitive behavioural therapy, home care services); and educational intervention (e.g. written material, videos, lectures) (Hopewell 2018). Falls prevention interventions that target individuals can be more readily evaluated through randomised controlled trials, and existing evidence supports their effectiveness in at‐risk groups (Gillespie 2012; Hopewell 2018; Sherrington 2019).

Population‐based interventions differ from approaches to falls prevention for individuals at medium or higher risk of falling. We conceptualise population‐based interventions based on both the Prevention of Falls Network Europe (ProFaNE) taxonomy, and the work of Geoffrey Rose (Rose 1985). ProFaNE defines population‐based interventions as "approaches in which the entire population of older people are targeted” (Lamb 2005). In this review, we have expanded on this definition by drawing on Rose 1985, who defined population‐based approaches in public health as those prioritising the change of determinants leading to the distribution of risk in specific populations, a distribution influenced by contextual conditions.

Falls prevention strategies in population‐based interventions may include:

government policies targeting vitamin D supplementation that might apply to entire states, regions, or municipalities;
local councils or local government providing general recommendations or maintenance programmes (at the population level) for hazard reduction in homes (e.g. good lighting; non‐slip surfaces) or in public places (e.g. care and maintenance of public walkways; railings on steps) for villages, towns, and cities;
public health initiatives providing communities with information or access to interventions (e.g. strength and balance exercise), irrespective of risk status and without assessment of individual risk (e.g. leaflet campaign targeting an entire city that provides general information on the importance of strength and balance training and details of accredited local training programmes); or
implementation of public health programmes enabling fall prevention behaviours, such as engaging in physical activity at the UK Chief Medical Officers’ recommended levels (e.g. all gyms in a town providing free membership for people over 60) (Department of Health and Social Care 2019; McClure 2010; Skelton 2005).

An effective intervention should focus on changing these conditions instead of targeting the risk profiles of individuals, as in the latter case, preventative measures for falls may reduce the impact on already vulnerable individuals. There is often a misconception that population‐based means interventions aiming to impact on the individual factors of a large number of people (Frohlich 2014). However, population‐based interventions should attempt, through programmes and policies, to change the underlying social, cultural, or environmental conditions of risk for the whole population (e.g. smoking bans in public places or promoting exercise). Population‐based interventions can therefore be seen as ecological interventions, rather than interventions delivered at the individual level to a large number of people. The size or scope of the population, or community (these terms are used interchangeably for the purposes of this review), depends on the type of intervention, and could aim at a large catchment population within a geographic location, or entire villages, towns, cities, regions, or, indeed, countries. To understand how interventions are implemented, it is useful to follow a standard reporting procedure (Campbell 2018).

A population‐based intervention can include shared ownership of the problem (falls and fall‐related injury) and its solution (preventing falls and fall‐related injuries). Interventions may therefore include experts and community members in determining the priorities and appropriate interventions. It should acknowledge the causal link between social and organisational structures, and any multicomponent strategy should optimise community involvement (McClure 2005; Moller 1991).

A central requirement for population‐based interventions is that the focus of the intervention is on the community rather than the individual. However, this presents some challenges, as there are examples of nationwide intervention studies using the individual as the unit of randomisation, yet still aiming to cover an entire cohort (e.g. aged 50 years and over) across several states (Le Boff 2020). The use of randomised controlled trial (RCT) designs in programme evaluation is often precluded due to difficulties with blinding and ensuring that members of a control group are not exposed to intervention material (Kempton 2000). Instead, separate communities, towns, cities, or regions with comparable demographic attributes, can be used as intervention and control areas in assessing programme effectiveness. Cluster RCTs – ideally, using multiple separate communities in each cluster – can be used to test the effectiveness of population‐based interventions (Hussey 2007), as well as stepped‐wedge designs, where all identified clusters begin as control areas, before one by one at random each cluster is switched to be an intervention area at set time points, until all clusters are intervention areas.

How the intervention might work

Through programmes and policies, population‐based interventions attempt to change the social and environmental contexts influencing health (Fuller 2012). To be included in a population‐based programme, an intervention should have demonstrated effectiveness, and have been tested by means of an RCT as a single measure, and should address a key risk factor for falls (Campbell 2010). For example, evidence suggests that vitamin D supplementation effectively reduces the rate of falls in people with insufficient levels of vitamin D (Gillespie 2012). There is further evidence that when vitamin D is associated with calcium, it helps reduce the risk of fracture in older people (Avenell 2014). Government provision of vitamin D and calcium intake for whole regions or communities may therefore have an impact on the rates of falls and fracture risk.

Population‐based interventions utilise an 'upstream' approach to reduce risk factors for falls across the whole population, before they manifest as proximal risk factors requiring clinical interventions (McClure 2010). Population‐based interventions thus work by reducing risk exposure in the cohorts of people within the setting being investigated. It is an approach that differs notably from interventions targeting specific individuals identified as being at risk of falling when the intervention is delivered to one person at a time (Hawe 2012). As selective approaches target mainly high‐risk individuals, a complementary approach which includes a non‐selective population‐based intervention is advisable, as it supports a tailored and appropriate implementation of a recommended intervention (Skelton 2005), through involving a wider range of individuals at the societal level, including those at low (but not at no) risk of falling (McClure 2010). A complementary approach may thus help reach the whole community, or a large proportion of it.

Why it is important to do this review

The current state of evidence on the effectiveness of population‐based approaches for preventing falls and fall‐related injuries is scant. To our knowledge, no systematic review to determine whether population‐based approaches are effective for falls prevention has been conducted to date. A Cochrane Review of population‐based interventions for the prevention of fall‐related injuries was published in 2005 (McClure 2005); though the authors reported consistency in the reduction of fall‐related injuries across six prospective controlled community studies, they concluded that no relevant RCTs had been carried out at the time.

For this reason, and because of the development of Cochrane methodology since 2005, updated work using new methods to assess the effectiveness of population‐based approaches for the prevention of fall‐related injury (and with the addition to include fall incidence) is needed. Since the 2005 review, study designs, such as stepped‐wedge designs (a cluster‐cross‐over randomised trial where clusters transition between control and intervention conditions at different time points, the order of which is determined using a random process), have been increasingly employed as they provide evidence comparable with other randomised designs (Haines 2018). We therefore conducted a review of population‐based controlled studies to update and extend the McClure 2005 review using more recent studies and report on the effectiveness of population‐based strategies in falls prevention.

Objectives

To review and synthesise the current evidence on the effects of population‐based interventions for preventing falls and fall‐related injuries in older people. We defined population‐based interventions as community‐wide initiatives to change the underlying societal, cultural, or environmental conditions increasing the risk of falling.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled trials (RCTs), cluster RCTs, stepped‐wedge designs, and non‐RCTs that evaluated the effects of population‐based falls prevention interventions in entire communities or large parts of communities.

For non‐randomised controlled trials, we used Reeves 2017 to describe the studies against their design features. We excluded before‐after studies and interrupted time series studies without a control group, and any study using historic controls.

We excluded studies using the individual as a unit of randomisation rather than communities.

We initially planned to include only studies with at least two intervention sites and two control sites, to improve population diversity (such as social, economic, and other demographic factors) amongst the study populations. However, we identified some non‐randomised study designs with single intervention and control sites but with large geographical areas that we expected to be diverse. In addition, we could not always determine the diversity of the populations in the included studies, therefore we included all studies, regardless of the number of intervention sites, that otherwise met the review inclusion criteria.

Types of participants

We included community‐dwelling older adults at least 60 years of age. We defined the term 'community' as any type of geographic location (e.g. villages, towns, cities, regions) or large catchment population within a geographic location. Given that this was a population‐level review, we expected that participants may be living independently in the community or residing in institutions (e.g. residential care homes, assisted care facilities, sheltered housing, retirement communities, or hospitals). We included studies in which the target population lived both independently or in institutions, but we excluded studies that only targeted older people living in institutions, as these participants would not be representative of the whole community.

We also excluded studies of participants selected according to a specific disease, condition, or risk status. We therefore excluded studies that only targeted individuals with a history of falling, who were at risk of falling due to the presence of intrinsic risk factors other than age.

Types of interventions

We included population‐based interventions targeting entire communities (or a large part of a community) that aimed to reduce the incidence of falls, fall‐related injuries, or both, in older people compared with no intervention or usual care (control). We defined population‐based interventions as community‐wide initiatives to change the underlying societal, cultural, or environmental conditions increasing the risk of falling. The control included communities, towns, cities, or regions with comparable demographic attributes to the intervention that received no intervention (or usual care), or a delayed intervention providing a comparison group for a fixed period of time. We excluded studies that included an active falls prevention component in the control group.

We categorised the components of interventions in the individual studies into comparison groups (intervention versus control). Although we based these groupings on those in the Cochrane Review by Hopewell 2018 and the ProFaNE taxonomy (Lamb 2011), we modified and condensed the components into fewer categories to fit with anticipated population‐level designs. The six broad groupings in this review were:

exercise and physical activity: interventions based on evidence‐based falls prevention exercises, such as strength and balance or Tai chi (Sherrington 2019), or physical activity generally (e.g. community‐based falls prevention exercise classes or free gym membership for older people);
medication or nutrition: interventions providing a medical or nutritional intervention to the entire community (e.g. vitamin D and calcium supplementation);
environmental: interventions involving local councils or governments providing general recommendations or maintenance programmes to entire communities for falls hazard reduction in homes or public places (e.g. targeting the care and maintenance of public walkways);
educational: interventions informing the community about risk factors and consequences of falls, or ways to prevent falls. These could be communicated through a mix of strategies (e.g. television, radio, social media, poster or leaflet campaigns with general information on falls prevention);
other initiatives: not already included in the above groupings and that align with our inclusion criteria; and
multicomponent interventions including more than one of the previous intervention types.

We also planned to further subcategorise these intervention groupings where we noted sufficient differences between the interventions in the included studies. For example, in the environmental grouping, we would have treated home hazard reduction as distinct from public health hazard reduction.

Types of outcome measures

Primary outcomes

Our primary outcomes were similar to Sherrington 2019.

Rate of falls (number of falls; falls per person‐year).
Number of fallers (number of people experiencing one or more falls).
Number of people experiencing one or more fall‐related injuries. Given the considerable heterogeneity in the definitions of injurious falls in the literature, we reported study authors’ definition of injurious falls alongside outcome data (Schwenk 2012).

Secondary outcomes

Number of people experiencing one or more fall‐related fractures.
Number of people experiencing one or more falls resulting in hospital admission.
Number of people experiencing one or more falls requiring medical attention.
Health‐related quality of life (HRQoL), measured using a validated scale.
Fall‐related mortality.
Concerns about falling, measured using a validated scale such as the Falls Efficacy Scale International (FES‐I), Yardley 2005, or short FES‐I, Kempen 2008.
Number of people experiencing one or more adverse events (e.g. increased falls or fall‐related injuries, heart attack, or death). We expected these data to vary according to the type of intervention.

Other outcomes:

For the economic analysis, we also extracted health economic data on cost utility and cost‐effectiveness in any economic analyses reported by study authors.

Search methods for identification of studies

Electronic searches

We searched the following electronic databases: the Cochrane Bone, Joint and Muscle Trauma Group Specialised Register (10 December 2020), the Cochrane Central Register of Controlled Trials (CENTRAL) via the Cochrane Register of Studies (CRS‐Web 10 December 2020 Issue 12), MEDLINE (Ovid MEDLINE(R) and Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, Daily and Versions(R) 1946 to 9 December 2020), Embase (Ovid 1980 to 10 December 2020), Cumulative Index to Nursing and Allied Health Literature (CINAHL) (EBSCO 1982 to 10 December 2020), and PsycINFO (Ovid 1967 to 9 December 2020). We also conducted a top‐up search of CENTRAL, MEDLINE, and Embase on 20 January 2023. There were no limitations based on language or publication status.

In MEDLINE, we combined the subject‐specific terms with the sensitivity‐maximising version of the Cochrane Highly Sensitive Search Strategy for identifying randomised trials (Lefebvre 2019). See Appendix 1 for search strategies.

We also searched the following trial registries for ongoing and recently completed studies: the World Health Organization International Clinical Trials Registry Platform (WHO ICTRP) (trialsearch.who.int/Default.aspx) (15 December 2020) and ClinicalTrials.gov (clinicaltrials.gov/) (16 December 2020).

Searching other resources

We checked the reference lists of included studies and of other systematic reviews identified from the database searches, including McClure 2005, and contacted researchers in the field to identify any ongoing and unpublished studies.

Data collection and analysis

Selection of studies

Due to a large volume of search results in the first database search (December 2020), we adopted a modified screening process as agreed upon with the Cochrane editorial team (Cochrane Bone, Joint and Muscle Trauma). Each of three review authors (LM, JS, and AG), working independently, was provided with a random set of records and engaged in a preliminary screening to eliminate results based on the record title. A senior review author (CT) checked the first 250 screening decisions to ensure agreement. After this preliminary screening, two review authors independently screened the selected titles and abstracts for relevance. Two review authors (AB, YY or CT) independently assessed the full‐text records for eligibility. We resolved any disagreements through initial discussion and then with a third review author if disagreement persisted.

In a top‐up search in January 2023, two review authors (SL and MP) used a standard procedure for screening results. For this search, two review authors (SL and MP) independently screened all titles and abstracts, reaching consensus through discussion. The two review authors then independently assessed full‐text records for eligibility and reached consensus through discussion with other review authors (LM, AB, CT).

We contacted study authors for more information when necessary. We recorded the reasons for excluding studies, and illustrated the selection process using a PRISMA flowchart.

Data extraction and management

Two review authors independently extracted data from the included studies using a predefined data extraction form. We were guided by previous reviews and protocols on falls interventions for data extraction (Clemson 2019; Sherrington 2019), and made adjustments for population‐based interventions. We extracted the following data.

General information: study authors; year of publication; date of data extraction; study objectives.
Study details: design; location; setting; population size; inclusion and exclusion criteria; comparability of control and intervention groups or sites; study dates, duration and length of follow‐up; funding source.
Characteristics of population: population composition by age, sex, ethnicity, residential status (e.g. living independently in the community, residential care homes, assisted care facilities, sheltered housing, elder or retirement communities), and socio‐economic status.
Interventions: experimental and control interventions; timing of intervention; mode of delivery; and information on uptake and adherence, when available.
Outcomes: review outcome descriptions, quantitative data including methods of analysis and adjustment for clustering or confounders.
Other details: relevant additional information specific to the study.

We resolved any disagreements through discussion between review authors. We contacted the authors of the included studies to request additional information when necessary.

Assessment of risk of bias in included studies

Two review authors independently conducted risk of bias assessment using tools appropriate to the study design. We resolved any disagreements through discussion.

For RCTs, we planned to use the Cochrane RoB 1 tool (Higgins 2011), which incorporates the following domains.

Sequence generation (selection bias)
Allocation concealment (selection bias)
Blinding of participants and personnel (performance bias)
Blinding of outcome assessors (detection bias)
Incomplete outcome data (attrition bias)
Selective reporting (reporting bias)
Other risk of bias

For consistency with other falls prevention reviews in this Cochrane Library series, we also assessed the risk of recall bias (in which we assessed any biases related to the ascertaining of falls).

For cluster RCTs, we assessed the following domains, as described in Chapter 16 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), in addition to those described above.

Risk of additional bias relating to recruitment
Baseline imbalance
Loss of clusters
Incorrect analysis
Comparability with individually randomised trials

For non‐randomised trials, we used the Effective Public Health and Practice Project (EPHPP) tool (EPHPP 2010), as recommended by Cochrane Public Health, and adapted it for specific confounders as appropriate. We assessed risk of bias based on the following domains.

Selection bias
Study design
Confounders (group differences prior to intervention: age; sex; ethnicity; residential status; socioeconomic status; and health status)
Blinding
Data collection methods
Withdrawals and dropouts
Intervention integrity
Analyses

We created the overall (or global) risk of bias assessment following EPHPP guidance and rated studies as strong, moderate, or weak (EPHPP Quality Assessment Dictionary). We note that this global rating does not include evaluation of the following domains: intervention integrity or analyses. We conducted the assessment at the study level, and based outcome‐specific questions in the tool on the rate of falls; if a study did not report rate of falls, we based outcome‐specific questions on the next outcome in the order as listed in Types of outcome measures.

For stepped‐wedge designs (for which we found no studies), we planned to adapt the EPHPP tool to assess the following (Eldridge 2016): bias arising from the randomisation process; bias arising from identification or recruitment of individual participants within clusters; bias due to deviations from intended interventions; analytical biases; and chance imbalance. We also planned to extend this tool to assess the additional risk of contamination across treatment conditions in stepped‐wedge designs (e.g. the intervention condition may take longer to embed in practice than planned, or there may be a delayed assessment of outcome in a sample that had long exposure to the intervention condition).

Measures of treatment effect

For the rate of falls, we reported rate ratios (RaR) with 95% confidence intervals (CIs). The rate of falls measures falls per person‐year (the total number of falls per unit of person‐time that falls were monitored). For number of fallers, people with injurious falls or fall‐related fractures, fall‐related hospital admission, or fall‐related mortality, we aimed to report risk ratios (RR) and 95% CIs. If these were not available (or no data were available to allow us to calculate RRs), we reported effect sizes as described by the study authors (e.g. using odds ratios (OR) or P values). If study authors reported rate data for other outcomes (such as for hospital admission), rather than for number of people who had at least one fall leading to hospital admission, we included these data in the review but noted that this was not our intended outcome measure. If both adjusted and unadjusted RaRs were reported, we used the unadjusted estimate unless the adjustment was for clustering. We used the calculator in Review Manager 5 to calculate effect estimates from individual studies (Review Manager 2020), guided by the Cochrane Handbook when calculating standard errors for rate ratios and when calculating effective sample sizes in cluster randomised trials (Higgins 2021).

For continuous outcomes (HRQoL and concerns about falling), we planned to report the mean difference (MD) with 95% CIs, or the standardised mean difference (SMD) in the event that we pooled data for outcomes measured with different tools.

Unit of analysis issues

Unless cluster RCTs reported effect estimates that had been adjusted for clustering by study authors, we adjusted for clustering in any analyses that we performed as guided by Chapter 23 of the Cochrane Handbook (Higgins 2021). We used an intracluster correlation coefficient (ICC) of 0.01 for this calculation (Smeeth 2002). We did not conduct meta‐analysis, and therefore did not need to make any adjustments to the control group data in multi‐arm studies. In addition, there were no unit of analysis issues related to the reporting of outcomes at different follow‐up times in the included studies.

Dealing with missing data

We contacted study authors for additional information when required, including to support decision‐making about inclusion criteria. Whilst we planned to explore missing data in sensitivity analysis, this was no longer relevant as we did not conduct meta‐analysis. In addition, no studies included continuous data, and plans to calculate missing standard deviations were not relevant in this review.

Assessment of heterogeneity

We assessed clinical and methodological heterogeneity to determine the feasibility of combining study results. Had pooling of data been appropriate, we would have assessed the statistical heterogeneity between studies using the Chi² test and the I² statistic as described in the Cochrane Handbook (Higgins 2021).

Assessment of reporting biases

Because we did not combine studies in analyses and there were too few studies, we did not investigate the possibility of reporting biases through funnel plots. Instead, we assessed the risk of selective reporting bias as part of the risk of bias assessment. We planned to compare pre‐published protocols or clinical trials registration documents with published reports of the completed studies with regard to the reporting of outcome data.

Data synthesis

We tabulated a detailed description of each study alongside their key characteristics. We based our decisions on whether to pool data on the comparability of interventions and settings described in the included studies, as well as the type of available quantitative data within the study reports. As the studies were heterogeneous, we did not pool any data in the review.

We were guided by Campbell 2019 when narratively reporting the review findings. In the presentation of data, we stratified the results according to the categories of interventions in Types of interventions. We then presented outcome data according to study design, always firstly reporting studies at lower risk of bias (i.e. cluster RCTs) before reporting studies at higher risk of bias (i.e. non‐randomised trials).

Subgroup analysis and investigation of heterogeneity

Although we planned to formally explore the effect of participant characteristics (age, gender, ethnicity, residential status, and socioeconomic status) on the review findings, we were unable to do this because we did not pool data.

Sensitivity analysis

We planned to explore the impact of study design, missing data, the inclusion of unpublished data, the inclusion of studies at high risk of bias, the choice of statistical model for pooling, and the effect of different ICCs for adjustment of sample sizes in cluster RCTs. We did not pool any data in the review, which precluded sensitivity analysis.

Summary of findings and assessment of the certainty of the evidence

We prepared summary of findings tables for each category of interventions, and included data from RCTs and non‐randomised trials in separate tables. We therefore presented summary of findings tables for:

medication or nutrition interventions (evidence derived from RCTs);
multicomponent interventions (evidence derived from RCTs); and
multicomponent interventions (evidence derived from non‐randomised trials).

In the event that robust data were available from more than one study for these categories, we described the intervention components in the footnotes of the summary of findings table. Some studies reported outcomes using measures that did not precisely meet our outcome criteria (e.g. rate of events rather than number of people experiencing one or more events). Whilst we reported these data in the Effects of interventions section for completeness, we did not include these data in the summary of findings tables. Similarly, if studies reported a breakdown of data according to subgroups (such as participant age) in addition to the overall group, we included these data in the Effects of interventions section but not the summary of findings tables.

We used the GRADE approach to assess the certainty of the evidence as it relates to the primary and secondary outcomes listed in Types of outcome measures (Schünemann 2020). The rating of high certainty is reserved for a body of evidence based on RCTs. We downgraded the certainty of the evidence to moderate, low, or very low depending on the presence and extent of five factors: study limitations, inconsistency of effect, imprecision, indirectness, and publication bias. All non‐randomised trials start as low‐certainty evidence. There were rare circumstances in which we considered upgrading the evidence from non‐randomised trials to moderate certainty. These included: a large, estimated effect (e.g. RR > 2 or RR < 0.5) in the absence of plausible confounders, or a very large effect (e.g. RR > 5 or RR < 0.2) in studies with no major threats to validity; the presence of a dose‐response gradient; or the presence of plausible biases that may lead to an underestimation of an apparent effect.

We reported the certainty of the evidence in the summary of findings tables for the following outcomes.

Rate of falls.
Number of fallers.
Number of people experiencing one or more fall‐related injuries.
Number of people experiencing one or more fall‐related fractures.
Number of people experiencing one or more falls resulting in hospital admission.
Adverse events.
Economic analysis.

Results

Description of studies

Results of the search

We screened a total of 44,707 records from the following databases: Cochrane Bone, Joint and Muscle Trauma Group Specialised Register (146), CENTRAL (4321), MEDLINE (9767), Embase (16,379), CINAHL (5793), PsycINFO (2361), the WHO ICTRP (2830), and ClinicalTrials.gov (2910). We also identified four reports from handsearches and searches of reference lists. After removal of duplicates, we screened 30,206 records. We excluded 30,030 records based on title and abstract and obtained the full texts of the remaining 176 records. We excluded 155 full‐text articles, and described 13 of these studies (with 18 records) in detail in the review. We categorised one study (with one record) as awaiting classification (Characteristics of studies awaiting classification) and found one ongoing study (Characteristics of ongoing studies). We included nine studies (with 19 records). For a detailed description of our screening process, see Figure 1.

Figure 1

Study flow diagram.

We contacted the contact authors of 10 trials either to find out additional information to support decision‐making or to support data extraction (Barker 2016; Clegg 2018; Guse 2015; Ivers 2020; Kempton 2000; Lindqvist 2001; Paul 2021; Poulstrup 2000; Rapp 2022; Robson 2003); see below.

Included studies

For details of the nine included studies, see Characteristics of included studies.

We contacted the authors of three included studies to request additional information regarding study participants' adherence to interventions, dropout rates, and numbers lost at follow‐up (Kempton 2000; Lindqvist 2001; Poulstrup 2000). Only Poulstrup 2000 provided additional information; most information was either not available or was insufficient to use in our analysis.

Study design

Two studies were cluster RCTs (Larsen 2005; Xia 2009). The remaining studies were non‐randomised (or quasi‐experimental) studies. Using characteristics from the study reports and guidance from Reeves 2017, we categorised these non‐randomised studies as controlled before‐and‐after (CBA) studies, Kempton 2000; Lindqvist 2001; Pujiula Blanch 2010; Wijlhuizen 2007; Ytterstad 1996, and controlled interrupted time‐series (CITS) studies, Poulstrup 2000; Svanström 1996.

Allocation to the intervention or control group was at a population level in all studies. Study investigators selected geographical regions in which people were exposed to the intervention, and this was compared with control group regions in which participants had no exposure to the intervention.

In the cluster RCTs, geographical regions were allocated randomly to the intervention or control groups:

Larsen 2005: a municipality was divided into four blocks, with blocks randomly allocated to one of three intervention groups or a control group;
Xia 2009: four communities were randomly allocated to an intervention or control group, with two communities in each group.

In the non‐randomised studies, control groups were mostly selected because population characteristics were comparable to the intervention group.

Kempton 2000: the control group was matched based on geography, demography, and climatic factors, and was remote enough not to be influenced by the study intervention in the single study region.
Pujiula Blanch 2010: the intervention and the control group areas were in the same area of the city. In this study, the reason for selecting the control group was not specified, although the decision to allocate the intervention to one area was made after collection of baseline data from a random sample of the population in each community.
Lindqvist 2001: the control group was a neighbouring municipality in the same county as the municipality that was selected as the intervention region. In this study, the reason for selecting the control group was not specified, but the study authors note baseline characteristics for the number of people who were urban‐living residents, gainfully employed, and their average income as a percentage of the national mean; these characteristics were comparable between the control and intervention regions.
Poulstrup 2000: five municipalities were selected as the intervention group, and four municipalities acted as control. Demographic and social characteristics were accounted for when selecting control group regions, as well as other potential confounders (such as fluoride water content, general practitioner prescribing habits, and distances to casualty wards or hospitals). Regions were geographically separate to avoid any overspill effect.
Svanström 1996: an area was selected for the intervention, and baseline and outcome data were compared with the county in which the area belonged as well as with the whole country. We included this study in the review but note that the larger populations in the control groups (county and country) inevitably also included participants from the smaller intervention area.
Wijlhuizen 2007: a community was selected for the intervention and was compared with a control group of two other communities. The control communities were suggested by the Area Health Authority based on their knowledge of the general population characteristics and were geographically separate from the intervention community.
Ytterstad 1996: a municipality was selected for the intervention and was compared with six neighbouring municipalities as well as a larger city. The study authors acknowledge that a spill‐over effect of the intervention was possible when comparing with the neighbouring municipalities, and note that the choice of city was also not ideal because it was larger and had some baseline differences in fracture rates. The study authors note that the mean age variations of residents in the selected regions were similar.

The included studies were carried out in seven countries: Australia (Kempton 2000), China (Xia 2009), Denmark (Larsen 2005; Poulstrup 2000), the Netherlands (Wijlhuizen 2007), Norway (Ytterstad 1996), Spain (Pujiula Blanch 2010), and Sweden (Lindqvist 2001; Svanström 1996).

Participants

Target populations and their matched controls were limited to residents who were > 60 years old (Kempton 2000; Xia 2009), ≥ 66 years (Larsen 2005), ≥ 70 years (Pujiula Blanch 2010), or ≥ 65 years in the remaining studies.

It was not feasible to provide a total target population size for all included studies because two studies only reported the size of the whole population rather than the target population (Lindqvist 2001; Ytterstad 1996). However, the total target population in the other seven studies (i.e. the number of intervention and control group residents that met the target age requirement) was approximately 254,004. We have reported an approximate total target population size because exact population sizes were not described in two studies (Wijlhuizen 2007 reported population sizes for the control groups to the nearest 1000, and Xia 2009 reported approximate population sizes for both intervention and control groups). Svanström 1996 included both county and country as the control; in this total population, we used population numbers at county level. In these seven studies, the target population size exposed to an intervention for preventing falls was approximately 115,320 older adult residents. The smallest intervention region of the target population had approximately 1800 older adult residents (Xia 2009), and the largest had 79,425 older adult residents (Kempton 2000). The control groups in these seven studies included approximately 136,978 older adult residents. We included county‐level population numbers for Svanström 1996. The smallest and largest regions for the control group region was in Pujiula Blanch 2010 (1212 older residents), and the largest control group region was in Kempton 2000 (61,758 older adult residents).

Total population sizes (i.e. the number of residents of any age) in Lindqvist 2001 were 41,000 residents in the intervention group and 25,900 residents in the control group. In Ytterstad 1996, these numbers were 22,500 residents and 135,000 residents, respectively.

A summary of the study designs, locations, and participant information is provided in Table 1.

Open in table viewer

Table 1. Included studies: study design, location, and trial size

Study ID	Study design^a	Country	Communities	Target population	Total population size or total size of target group (intervention + control)	Sample size^b
Kempton 2000	CBA	Australia	Intervention: North Coast of New South Wales Control: Queensland Sunshine Coast	Community‐dwelling, ≥ 60 years of age	Total target group: 141,183 Intervention: 79,425 Control: 61,758	2445
Larsen 2005	Cluster RCT	Denmark	3 intervention groups: each in 1 block within Randers municipality Control: 1 block within Randers municipality	Community‐dwelling, ≥ 66 years of age	Total target group: 9605 Intervention Block 1 ("Environmental and Health Program": 2532) Intervention Block 2 ("Calcium and Vitamin D Program"): 2426 Intervention Block 3 (both programmes): 2531 Control Block 4: 2116	N/A
Lindqvist 2001	CBA	Sweden	Intervention: Motala (municipality in Östergötland County) Control: Mjölby (municipality in Östergötland County)	≥ 65 years of age	Total whole population (before intervention): 67,300 Intervention whole population (before intervention): 41,000 Control whole population (before intervention): 25,900	N/A
Poulstrup 2000	CITS	Denmark	Intervention: 5 municipalities of County Council of Vijle Control: 4 municipalities of County Council of Vijle	Community‐dwelling, ≥ 65 years of age	Total target group: 26,221 Intervention: 12,905 Control: 11,460	N/A
Pujiula Blanch 2010	CBA	Spain	Intervention: Salt (health district in Girona) Control: Girona‐4 (health district in Girona)	≥ 70 years of age	Total target group: 3727 Intervention: 2515 Control: 1212	602
Svanström 1996	CITS	Sweden	Intervention: Lidköping (municipality in Skaraborg County) Control: Skaraborg County, and whole of Sweden	≥ 65 years of age	Total target group (first study year): 1,173,698 Intervention (first study year): 6817 Control (first study year): 50,052 (Skaraborg County); 1,116,829 (Sweden)	N/A
Wijlhuizen 2007	CBA	Netherlands	Intervention: Sneek (in Area Health Authority of Fryslân) Control: Harlingen and Heerenveen (2 communities within Area Health Authority of Fryslân)	≥ 65 years of age, living independently	Total target group: 12,769 Intervention: 4369 Control: 8400^c	1752
Xia 2009	Cluster RCT	China	Intervention: 2 residential communities, Shanghai Control: 2 residential communities, Shanghai	≥ 60 years of age	Total target group: 3600^c Intervention: 1800^c Control: 1800^c	1422
Ytterstad 1996	CBA	Norway	Intervention: Harstad Control: 6 municipalities ("close to Harstad"), and Trondheim	≥ 65 years of age	Total whole population: 172,500 Intervention whole population: 22,500 Control whole population: 15,000 (6 municipalities); 135,000 (Trondheim)	N/A

CBA: controlled before‐and‐after study (non‐randomised study design); CITS: controlled interrupted time‐series study (non‐randomised study design); N/A: not applicable; RCT: randomised controlled trial

^aClassification using Reeves 2017 as a guide.
^bSample of target population who were contacted in telephone or questionnaire surveys for some or all outcome data. In some studies, data were collected for all residents, and there were no sample data in these studies.
^cApproximate values, as reported in the study report.

Interventions

Larsen 2005 was a four‐arm study with three intervention arms (two distinct interventions, and one that was a combination of these two interventions) and one control arm. The remaining studies had one intervention group and either one control group, Kempton 2000; Lindqvist 2001; Poulstrup 2000; Pujiula Blanch 2010; Wijlhuizen 2007, or two control groups, Svanström 1996; Ytterstad 1996.

The interventions in most studies included multiple components. We categorised these interventions according to the primary intervention modality. One multi‐arm study included interventions that fit into more than one of these categories (Larsen 2005).

Exercise and physical activity only: we included no studies with this primary intervention modality.

Medication or nutrition: one study evaluated the effectiveness of free‐of‐charge daily supplements in the "Calcium and Vitamin D Program" (Larsen 2005).

Environmental only: we included no studies with this primary intervention modality.

Educational: we included no studies with this primary intervention modality.

Other initiatives: we included no studies with other initiatives.

Multicomponent population‐based interventions: all nine studies included multicomponent interventions. According to the ProFaNE taxonomy (Lamb 2011), components broadly included exercise, medication, environment (home and community‐level), social environment (staff training), and knowledge/education.

Kempton 2000: interventions included an information campaign and exercise classes. The intervention addressed footwear, vision, physical activity, balance and gait, medication use, chronic conditions, plus home and public environmental hazards. This intervention is also known as "Stay on Your Feet (SOYF)". It was delivered in a community setting including participants' homes and other locations within the community. General practitioners and healthcare workers were trained to deliver the intervention. Other components of the intervention were delivered through local media, including TV, radio, and newspapers.
Larsen 2005: interventions included home safety inspection, ways to avoid falls, health and dietary correction in the "Environment and Health Program"; a free‐of‐charge daily supplement of 1000 mg calcium carbonate and 400 international units vitamin D₃ in the "Calcium and Vitamin D Program"; or a combination of the "Environment and Health Program" and the "Calcium and Vitamin D Program". The safety inspections were delivered by trained community nurses.
Lindqvist 2001: interventions included elimination of hazards in the environment (e.g. improvements to roads, pavements, street lighting), behavioural safety education and information programmes, injury prevention features in local media, availability of safety products, home modifications, and exercise support. This intervention is also known as "WHO Safe Community" programme. Some components of the intervention were delivered by staff at social care facilities who were given additional training. Other components of the intervention were delivered through injury prevention features in local media, and safety products displayed in public places.
Poulstrup 2000: interventions included information on risk factors, and identifying and correcting hazards in the home and surrounding areas. Interventions aimed to reduce physical hazards, age‐debilitating illnesses, psychiatric illnesses, improper use of medication, diet insufficiency, and physical inactivity. Interventions were delivered through mailed leaflets, talks in clubs for older people and at welfare centres, and home visits.
Pujiula Blanch 2010: interventions included an educational programme, government involvement with architectural consultations, exercise programmes, risk assessment, dietary and medicine guidance, and risk prevention at home. Interventions were delivered through pamphlets, media, and conferences as well as home visits and community settings.
Svanström 1996: interventions included an educational programme targeting health hazards and how to reduce risk, changes regarding traffic environment (street lights control, new cycle paths), two newspaper articles that advertised updates on preventive work and increased awareness on how to target risk factors. This intervention is also known as "Lidkōping Accident Prevention Programme". Interventions were delivered to groups and individuals.
Wijlhuizen 2007: interventions included information and education (home safety, physical activity, safe medication use, traffic safety), training and exercise (home safety training for professionals and volunteers working in home care, balance training course, traffic safety when riding bikes), and modifications to the environment (modifications in the home, removing obstacles from pavements). Interventions were delivered through media sources (leaflets, posters, newspaper articles), presentations, training courses, home visits, and technical assistance.
Xia 2009: interventions included an education programme (to reduce risk of falls, and including information on diet and exercise), home hazard assessment, modification of community settings (removing obstacles on pavements, roads, lawns; installing handrails). Interventions were delivered by a multidisciplinary group including individuals from local government, the community health centre, and other members of the community. Education programmes were delivered to groups and individuals.
Ytterstad 1996: interventions included home visits to identify and remedy home hazards, to promote safety in the environment, a healthy diet and lifestyle and reduction in isolation and inactivity; introduction of a pensioners' services in which pensioners could provide skilled low‐cost services to improve physical environments in others' homes; availability of safety items including spiking of boots for icy pavements. Interventions were delivered through community meetings, promotion in the media, home visits, and involvement of voluntary organisations working with older people.

The duration of the intervention ranged from 14 months, in Wijlhuizen 2007, to eight years, in Ytterstad 1996. A summary of the interventions, the site of intervention delivery, and the people who delivered the interventions is provided in Table 2.

Open in table viewer

Table 2. Included studies: intervention characteristics

Study ID	Type of intervention	Duration of intervention (study dates)	Main sites of delivery	Interventions delivered by
Kempton 2000	Multicomponent interventions Footwear, vision, physical activity, balance and gait, medication use, chronic conditions, plus home and public environmental hazards	4 years (1992 to 1995) Baseline data collection in 1991	Community setting (in participants' homes, organisations and other locations in the community)	General practitioners and healthcare workers
Larsen 2005	Medication or nutrition Block 1: Free‐of‐charge daily supplement 1000 mg calcium carbonate or 400 IU (10 μg) vitamin D₃ Multicomponent interventions Block 2: home safety inspection, ways to avoid falls, health and dietary correction Block 3: combined interventions from Block 1 and Block 2	42 months (1995 to 1998) Baseline data collection from 1990 to 1994	Community setting (in participants' homes, organisations and other locations in the community, including the pharmacy)	Trained community nurses
Lindqvist 2001	Multicomponent interventions Elimination of environmental hazards (e.g. improvements to roads, pavements, street lighting), behavioural safety education and information programmes, injury prevention features in local media, availability of safety products, home modifications, and exercise support	5 years (1984 to 1989) Pre‐implementation phase 1983 to 1984	Community setting (in participants' homes, organisations and other locations in the community)	Staff in social care facilities were given additional training.
Poulstrup 2000	Multicomponent interventions Information on fall risk factors, and identifying and correcting hazards in the home and surrounding areas. Interventions aimed to reduce physical hazards, age‐debilitating illnesses, psychiatric illnesses, improper use of medication, diet insufficiency, and physical inactivity	18 months (1985 to 1988)	Community setting (in participants' homes, organisations and other locations in the community)	General practitioners, district nurses, and home helpers
Pujiula Blanch 2010	Multicomponent interventions Educational programme through information pamphlets, media, and conferences; government involvement with architectural consultations; exercise programmes, risk assessment, dietary and medicine guidance, and prevention of fall risks at home	2 years (2001 to 2003)	Community setting (in participants' homes, organisations and other locations in the community)	Primary care team, including professionals and "individuals"
Svanström 1996	Multicomponent interventions Educational programme targeting health hazards and how to reduce risk, environmental changes regarding traffic environment (streetlights control, new cycle paths), 2 newspaper articles which advertised updates on preventive work and increased awareness on how to target risk factors	7 years (1985 to 1992; gradual implementation over each year) Baseline data collection in 1987	Community setting (in participants' homes, organisations and other locations in the community)	District nurse, policymakers, home visitors, service apartment personnel
Wijlhuizen 2007	Multicomponent interventions Traffic safety, balance training, physical activity, home safety hazard, home modification, safe pavement (removing obstacles from pavements in the community), and medication use	14 months (1999 to 2002) Baseline data during 10 months of 1999. Final data in 10 months after 14‐month intervention period	Community setting (in participants' homes, organisations and other locations in the community)	Professionals (types not specified) Trained volunteer peer consultants
Xia 2009	Multicomponent interventions Education programme (to reduce risk of falls, and including information on diet and exercise), home hazard assessment, modification of community settings (removing obstacles on pavements, roads, lawns; handrails)	18 months (2006 to 2007)	Community setting (in participants' homes, organisations and other locations in the community)	A multidisciplinary group including individuals from local government, the community health centre, and other members of the community
Ytterstad 1996	Multicomponent interventions Identifying and remedy of home hazards, promote environmental safety, health, diet, and lifestyle and reduction in isolation and inactivity; pensioners providing skilled low‐cost services to improve physical environments in others' homes; availability of safety items including spiking of boots for icy pavements	8 years (1985 to 1993; intervention was mostly delivered in the latter 5 years of study period)	Community setting (in participants' homes, organisations and other locations in the community)	An injury prevention group comprising representatives from a hospital and several public and private organisations, including "health personnel" Nurses, nurse‐aides, and other home helpers. Physiotherapists. Peer‐support (older people offering skilled work to other older people or delivery of spiked boots)

IU: international unit

Outcomes

No studies reported data for the number of people experiencing one or more falls requiring medical attention, HRQoL, fall‐related mortality, concerns about falling, or adverse events. However, data for all other outcomes were available from at least one included study.

Most studies used hospital or healthcare record systems in the selected regions to collect data, therefore the outcome data from these studies included the whole target population.

Three studies only used the results from questionnaires or telephone surveys to collect data (Pujiula Blanch 2010; Wijlhuizen 2007; Xia 2009), and one study reported findings from both hospital records and telephone surveys (Kempton 2000). The outcome data from these four studies therefore included a subsection of the whole population; data from these sources were available for 3451 participants who were exposed to the intervention and 2770 participants in the control group regions.

Funding sources

Six studies received funding from national or regional sources (e.g. funding from healthcare or government organisations), and we judged that these funding sources were likely to be independent of the study investigation (Kempton 2000; Larsen 2005; Lindqvist 2001; Wijlhuizen 2007; Xia 2009; Ytterstad 1996). Three studies did not report the funding source (Poulstrup 2000; Pujiula Blanch 2010; Svanström 1996).

Excluded studies

We contacted the authors of five studies for additional information to support our decision‐making process (Barker 2016; Clegg 2018; Paul 2021; Rapp 2022; Robson 2003); we did not receive a reply from the study authors of Robson 2003, but believed we had sufficient information to exclude this study.

We excluded 154 articles at full‐text review. We have reported details on the exclusion of 13 key studies (see Characteristics of excluded studies). We excluded four studies because the interventions targeted individually allocated participants rather than whole communities (Barker 2016; Le Boff 2020; Robson 2003; Scronce 2021), and an additional study that targeted a subset of the whole community from general healthcare practices (Bruce 2016). We excluded Mazza 2021 because it did not include a comparative (control) community as a reference, and Lin 2006 because the comparison groups all had active components. We excluded five studies because the interventions were designed to target only older people who were at higher risk of falling (Guse 2015; Iliffe 2014; Paul 2021); had a fragility fracture (Rapp 2022); or were described as home bound (Clegg 2018). We excluded Johnston 2019 because it was the wrong type of intervention (screening people for risk of falling), which was not population‐wide and included no control group.

Ongoing studies

We identified one ongoing trial (Ivers 2020). We contacted the study authors to ask whether the study has been completed. The trial has experienced delays due to the COVID‐19 pandemic, and the study authors advised that findings will not be available for a number of years. This trial aims to assess the effectiveness of a weekly exercise and discussion programme aimed at preventing falls (Ironbark: Standing Strong and Tall programme) in older Australian Aboriginal people. See Characteristics of ongoing studies.

Studies awaiting classification

We could not source the full text for Bos 2021, and did not have sufficient information to judge its eligibility.

Risk of bias in included studies

Risk of bias in cluster RCTs

See Figure 2.

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included cluster RCT.

Both cluster RCTs reported insufficient methods used to randomise areas to the intervention and control groups, and we judged the studies to have an unclear risk of selection bias for sequence generation. However, because the number of clusters relative to the number of intervention groups was small in both studies, we judged risk of bias for allocation concealment to be high.

We judged performance bias to be high in both studies because it is not possible to blind any healthcare professionals involved in the delivery of the interventions. Outcome data were collected from hospital record systems in Larsen 2005 and from participant questionnaires in Xia 2009, therefore we judged both studies to have a low risk of detection bias.

We judged risk of attrition bias to be low in Larsen 2005. However, risk of attrition bias was unclear in Xia 2009 because only a sample of the participants were used to collect outcome data. Neither study reported a protocol or clinical trials registration, and we could not adequately determine risk of selective reporting bias.

We thought that there was a high risk of other bias in Xia 2009 because the study communities were geographically close to one another such that a spill‐over effect was possible. We also judged Xia 2009 to have a high risk of recall bias, as study investigators used a self‐report method for ascertaining falls.

For the risk of bias assessment specific to the cluster RCT design, we judged Larsen 2005 to have a high risk of baseline imbalance, as study authors reported that there were statistically significant baseline differences between study group participants regarding age and marital status. In addition, we could not be certain whether the environmental and health interventions programme was comparable to individually randomised trials because the study authors did not report this information. Owing to unclear reporting in Xia 2009, we could not determine whether correct analysis methods had been used. However, we judged risk of additional bias relating to recruitment and loss of clusters for the two cluster RCTs to be low.

Risk of bias in non‐randomised studies

Using the EPHPP tool, we judged the overall methodological quality of the non‐randomised studies in this review to be weak in six studies, Kempton 2000; Poulstrup 2000; Pujiula Blanch 2010; Svanström 1996; Wijlhuizen 2007; Ytterstad 1996, and moderate in one study, Lindqvist 2001; see Table 3. A detailed assessment is provided in Appendix 2.

Open in table viewer

Table 3. Summary and global rating of quality assessment for non‐randomised studies

Study ID	Selection bias	Study design	Confounders	Blinding	Data collection method	Withdrawals and dropouts	Overall
Kempton 2000	MODERATE	MODERATE	MODERATE	WEAK	WEAK	MODERATE	WEAK
Lindqvist 2001	MODERATE	MODERATE	STRONG	WEAK	MODERATE	MODERATE	MODERATE
Poulstrup 2000	MODERATE	MODERATE	MODERATE	WEAK	WEAK	MODERATE	WEAK
Pujiula Blanch 2010	MODERATE	MODERATE	WEAK	WEAK	WEAK	MODERATE	WEAK
Svanström 1996	MODERATE	MODERATE	WEAK	WEAK	STRONG	MODERATE	WEAK
Wijlhuizen 2007	MODERATE	MODERATE	MODERATE	WEAK	WEAK	WEAK	WEAK
Ytterstad 1996	MODERATE	MODERATE	WEAK	WEAK	STRONG	MODERATE	WEAK

We judged it very likely that all seven non‐randomised studies selected participants that were likely to be representative of the target population.

In three studies, we could not tell if there were important differences because the study authors did not report sufficient information about population characteristics (Pujiula Blanch 2010; Svanström 1996; Ytterstad 1996). One study reported that groups were similar in sex, employment status, income, and urban residency (Lindqvist 2001). Three studies noted that there were important differences between groups prior to the intervention (Kempton 2000; Poulstrup 2000; Wijlhuizen 2007).

No studies explained whether participants were aware that their community was involved in a falls prevention intervention trial, and similarly it was often unknown if outcome assessors were aware that they were collecting data from an intervention or control group area. Three studies used self‐reported methods to collect outcome data, and we judged that these methods were weak. Other studies used hospital record systems, which we expected to be reliable (Svanström 1996; Ytterstad 1996), or a separate injury record system (Lindqvist 2001; Poulstrup 2000), and Lindqvist 2001 stated that this system had been previously used and was expected to be reliable.

For the three studies that collected data from a sample of residents in the whole intervention areas, withdrawals and dropouts were adequately reported in Kempton 2000 and Wijlhuizen 2007, but not in Pujiula Blanch 2010. The sample of residents in Wijlhuizen 2007 was approximately 30% of the whole intervention area population, with a drop of less than 20% for data collection. In Kempton 2000, the sample of residents was between 60% and 79%, with very little participant dropout. However, the information in Pujiula Blanch 2010 was unclear, and we could not determine the total population sizes. Nevertheless, we noted that the sample sizes were unequal, with 11% of a large sample (or the whole population) providing outcome data in the intervention area and 26% of a large sample in the control area.

Regarding intervention integrity, although some studies identified the reach of certain intervention components, we could not tell how many people received all the interventions to which their community had been exposed. It was also unclear whether any attempts were made to measure the consistency of intervention delivery. We expected that a spill‐over effect was likely in two studies (Kempton 2000; Ytterstad 1996), and possible in another two studies despite lack of detail in the study reports (Lindqvist 2001; Pujiula Blanch 2010); spill‐over was likely or possible because the intervention and control group areas were neighbouring or geographically close to one another.

In all seven studies, the community was the unit of allocation, and the individual was the unit of analysis. We judged that the statistical methods in each study were appropriate for their study design and that analysis was conducted according to the intervention allocation status rather than the actual intervention received.

Effects of interventions

See: Summary of findings 1 Medication or nutrition fall prevention interventions versus control: evidence from RCTs; Summary of findings 2 Multicomponent fall prevention interventions versus control: evidence from RCTs; Summary of findings 3 Multicomponent fall prevention interventions versus control: evidence from non‐randomised trials

We have reported here the effects of interventions according to the types of interventions in the included studies.

Medication or nutrition interventions

Only one study (a cluster RCT) included an intervention (a "Calcium and Vitamin D" falls prevention programme) that fit within the ProFaNE classification (Larsen 2005). See summary of findings Table 1. Data were available for only one outcome from this study. No outcome data were available for: rate of falls, number of fallers, number of people who experienced one or more fall‐related injuries, number of people experiencing one or more fall‐related fractures, number of people experiencing one or more falls requiring medical attention, HRQoL, fall‐related mortality, concerns about falling, adverse events, or an economic analysis of the intervention.

Number of people experiencing one or more falls resulting in hospital admission

Whilst Larsen 2005 reported falls leading to hospital admission, we could not be certain whether the data accounted for people having more than one fall. As this study report also included risk ratios that were calculated to account for age and marital status, we selected these data for the review. We note, however, that study authors did not indicate whether these data were adjusted for the clustering effect in this study design, therefore we advise caution in the interpretation of these data. Data were reported separately according to sex using the intention‐to‐treat principle. In this study, female residents who were exposed to a "Calcium and Vitamin D" falls prevention programme had fewer fall‐related hospital admissions than female residents in the control area (RR 0.89; P < 0.10). The study authors reported that there was no evidence of a difference when male residents were exposed to this intervention (RR 1.08). Data were reported without confidence intervals, and no P value was reported with the effect estimate for male residents. These data were collected from the Danish Hospital Registration Database over a 42‐month period; the population size exposed only to the calcium and vitamin D₃ programme or in the control group was 4542 residents.

We judged the certainty of this evidence to be very low. We downgraded by two levels owing to very serious risk of bias (high and unclear risk of bias in several domains). We also downgraded by one level for imprecision because the effect estimates were reported without CIs, and we were unable to determine the level of imprecision in the data (particularly for male residents).

Multicomponent interventions

All nine studies included fall prevention interventions that have multiple components. For evidence from RCTs, see summary of findings Table 2. For evidence from non‐randomised trials, see summary of findings Table 3. No studies in this comparison group measured or reported outcome data for: number of people experiencing one or more falls requiring medical attention, HRQoL, fall‐related mortality, concerns about falling, or adverse events.

Rate of falls

One cluster RCT, Xia 2009, and three non‐randomised trials, Kempton 2000; Pujiula Blanch 2010; Wijlhuizen 2007, reported rate of falls; we did not include data for Pujiula Blanch 2010 in the summary of findings table.

Xia 2009 reported a lower incidence rate ratio of falls for people exposed to a multicomponent falls prevention intervention compared with a control group (rate ratio (RaR) 0.356, 95% CI 0.253 to 0.501). The intervention exposure time was 18 months, and data were collected from a sample of 723 people in the intervention area and 699 people in the control area. We judged the certainty of the evidence to be very low. We downgraded by two levels for very serious risk of bias because the study included high and unclear risk of bias. We also downgraded by one level for imprecision because we could not be certain whether the rate ratio reported in this study included an adjustment for clustering, we therefore we advise caution in the interpretation of these data. See summary of findings Table 2.

Kempton 2000 reported that the reduction in rate of falls for people exposed to the "Stay on Your Feet" (SOYF) falls prevention programme was not statistically significant when compared to rate of falls in the control area (reduction of 0.066 falls/person/year; P = 0.14). This analysis was adjusted for age and sex, and data were collected at the end of a four‐year study period from survey responses from 2445 residents who were a representative sample of residents in the intervention and control group areas.

For Wijlhuizen 2007, we used P values and other data reported in the study report to calculate rate of falls. These data were reported separately for falls inside and outside the home. We found no evidence of a difference between residents exposed to a multicomponent falls prevention intervention and residents in the control areas in rate of falls inside the home (RaR 1.07, 95% CI 0.39 to 2.99; Analysis 1.1) and rate of falls outside the home (RaR 0.91, 95% CI 0.61 to 1.37; Analysis 1.1). We used postintervention data per 1000 persons per year in this analysis, collected over a 10‐month period after the end of the 14 month‐long study, and note that these effect estimates do not account for possible confounders.

The certainty of the evidence for rate of falls derived from non‐randomised trials was very low. There were no reasons to upgrade the certainty of the evidence. In addition, we downgraded the certainty of the evidence for imprecision in the effect estimate. We also observed inconsistency with the findings from the RCT‐derived evidence, above. See summary of findings Table 3.

Pujiula Blanch 2010 reported that people had 1.56 falls per year in the intervention group and 1.65 falls per year in the control group; these data were collected at the end of a two‐year study period. We did not know the number of falls in each group and could not calculate a standard error for these data in order to report a RaR of falls. These data were incomplete, and were not included in our summary of findings table. In addition, we could not account for the possible differences between participants at baseline. This study used a non‐randomised trial design (CBA), and there were no reasons to upgrade the certainty of the evidence, which we judged to be very low. As well as downgrading for study design, we also downgraded by one level for imprecision because these data were collected from a small sample size of only 402 residents; we also noted that the sample size in the control group represented a larger percentage of the whole population than the sample size in the intervention area.

Number of fallers

One cluster RCT, Xia 2009, and three non‐randomised trials, Kempton 2000; Pujiula Blanch 2010; Wijlhuizen 2007, reported data about fallers; we did not include data for Wijlhuizen 2007 in the summary of findings table.

Xia 2009 reported the number of people who had one fall and the number of people who had at least two falls during a 12‐month period; we used these data to calculate the number of people who had at least one fall. This cluster RCT used data from a sample of residents in the intervention and control group areas. To account for the clustering effect, we adjusted the sample size from 723 to 159 participants in the intervention group and from 699 to 154 participants in the control group; we used the same formula to adjust the number of events in each group. We found that there were fewer fallers amongst residents exposed to a multicomponent falls prevention intervention (RR 0.34, 95% CI 0.19 to 0.62; very low‐certainty evidence; Analysis 1.2). We downgraded the evidence by two levels for serious risk of bias because the study included high and unclear risk of bias, and one level for imprecision because the sample size for this population‐level study was small.

Kempton 2000 reported no evidence of a difference in the number of people who fell at least once between residents exposed to the SOYF falls prevention programme and residents in the control area (odds ratio (OR) 0.95, 95% CI 0.79 to 1.15). This logistic regression analysis accounted for age and sex of residents and included data from 2445 residents who were a representative sample of residents in the intervention and control group areas.

In Pujiula Blanch 2010, there was no evidence of a difference in the number of people who fell (RR 1.03, 95% CI 0.81 to 1.31; Analysis 1.2). This study collected information from a sample of 602 people living in the intervention and control group areas that was collected at the end of a two‐year study period. In our analysis of the data reported in Pujiula Blanch 2010, we did not account for possible differences between participants at baseline, and we advise caution in the interpretation of these data.

We downgraded the certainty of the evidence derived from these non‐randomised trials to very low. There were no reasons to upgrade the certainty of the evidence for these non‐randomised trial designs (CBAs), and we downgraded one level for imprecision because the effect estimates included the possibility of benefit and no benefit. In addition, we noted inconsistency with the findings from the RCT‐derived evidence that we could not explain. See summary of findings Table 3.

For Wijlhuizen 2007, we used data reported in the study report to calculate the rate of fallers rather than the number of fallers. We found no evidence of a difference between residents exposed to a multicomponent falls prevention intervention and residents in the control areas in rate of people having more than one fall inside the home (RaR 1.21, 95% CI 0.75 to 1.96; Analysis 1.3) and outside the home (RaR 0.79, 95% CI 0.51 to 1.21; Analysis 1.3). Again, we used the postintervention data per 1000 persons per year having at least one fall, collected over a 10‐month period after the end of the 14 month‐long study, and our effect estimates do not account for possible confounders. This study used a non‐randomised trial design (CBA), and there were no reasons to upgrade the certainty of the evidence, which we judged to be very low. We also downgraded the certainty of the evidence by one level for indirectness because these data provided indirect evidence for this outcome; for this reason we did not include this information in the summary of findings table.

Number of people experiencing one or more fall‐related injuries

One cluster RCT, Xia 2009, and one non‐randomised trial, Lindqvist 2001, reported data on the number of people experiencing one or more fall‐related injuries.

Xia 2009 reported fall‐related injury data, although we note that this study presented no definition of injurious falls. Based on the information in the study report, we could not be certain if these data were the number of people who had one fall‐related injury or the number of people who had at least one fall‐related injury. We used the adjusted sample sizes as described above, as well as adjusting the number of events in each group for this outcome, in order to account for the clustering effect in this study design. We found that people had fewer fall‐related injuries when exposed to a multicomponent falls prevention intervention (RR 0.39, 95% CI 0.20 to 0.77; very low‐certainty evidence; Analysis 1.4). We downgraded the certainty of the evidence by two levels for serious risk of bias because the study included high and unclear risk of bias, and one level for imprecision because the sample size for this population‐level study was small. See summary of findings Table 2.

Lindqvist 2001 reported no evidence of a difference in the number of people who experienced a fall‐related injury after being exposed to the "Safe Community" falls prevention intervention (OR 0.89, 95% CI 0.77 to 1.03); these data were collected at the end of a five‐year study period. The study authors also report data for different age groups; we have included these data in Appendix 3. We noted a reduction in fall‐related injuries for people aged 75 to 79 years of age (OR 0.71, 95% CI 0.52 to 0.99), but little or no difference in effect for other age groups. Fall‐related injury data for the control group were not specifically reported, but the study authors state that there was no evidence of a change in total morbidity rates (which included injuries other than fall‐related injuries). Again, this study did not present a definition of injurious falls. We downgraded the certainty of the evidence to very low. There were no reasons to upgrade the certainty of the evidence for this non‐randomised trial design (CBA), and we downgraded one level for imprecision because the effect estimate (for all age groups) included the possibility of benefit and no benefit. In addition, we noted inconsistency with the findings from the RCT‐derived evidence that we could not explain. See summary of findings Table 3.

Number of people experiencing one or more fall‐related fractures

One cluster RCT, Xia 2009, and three non‐randomised trials, Poulstrup 2000; Svanström 1996; Ytterstad 1996, reported data on fall‐related fractures; we did not include data from Svanström 1996 and Ytterstad 1996 in the summary of findings table.

Xia 2009 reported fall‐related fracture data. Based on the information in the study report, we could not be certain if these data were the number of people who had one fall‐related fracture or the number of people who had at least one fall‐related fracture. We used the adjusted sample sizes as described above, as well as the number of events in each group for this outcome, in order to account for the clustering effect in this study design. We found no evidence of a difference in fall‐related fractures, but we could not be certain whether the imprecision in this effect estimate was driven by the small sample size in this adjusted analysis (RR 0.55, 95% CI 0.17 to 1.85; very low‐certainty evidence; Analysis 1.5). We downgraded the certainty of the evidence by two levels for serious risk of bias because the study included high and unclear risk of bias, and one level for imprecision. See summary of findings Table 2.

Poulstrup 2000 reported little or no difference in the number of fractures that were prevented during the 18‐month follow‐up time period between the regions that received a multicomponent falls prevention intervention and the control regions (14% prevented fractures, 95% CI −9% to 37%; 1 study; 24,365 participants). These data represented all fracture types reported in a separate injury register at the involved hospitals and represent the number of fractures rather than the number of people who experienced one or more fall‐related fractures. The effect estimate was calculated using logistic regression analysis controlling for age, gender, and marital status. We downgraded the certainty of the evidence to very low. There were no reasons to upgrade the certainty of the evidence for this non‐randomised trial design (CBA), and we downgraded one level for imprecision because the effect estimate included the possibility of benefit and no benefit. See summary of findings Table 3.

Poulstrup 2000 also reported data for lower extremity fractures. The study authors reported little or no difference between groups for lower extremity fractures in men (0.1% prevented fractures, 95% CI −0.4% to 0.6%). However, the study authors noted that the multicomponent falls prevention intervention prevented more lower extremity fractures in women (46% prevented fractures, 95% CI 8% to 84%). It is likely that this large difference in prevented fractures for women influenced the overall effect for lower extremity fractures (33% prevented fractures, 95% CI 3% to 63%).

Svanström 1996 reported incidence rates of fall‐related femoral fractures per 1000 residents rather than the number of people who had fall‐related fractures. For completeness, we included these data in the review, and calculated rate ratios using the incidence rate and population data for male and female residents; data were reported separately by sex in the study report. We found that there was no evidence of a difference in fall‐related femoral fractures according to whether female residents (RaR 0.91, 95% CI 0.70 to 1.17; Analysis 1.6) or male residents (RaR 0.66, 95% CI 0.40 to 1.11; Analysis 1.6) were exposed to the Lidköping Accident Prevention programme. Data were collected from hospital discharge records, and included 6970 residents exposed to the falls prevention programme and 51,036 residents in the whole county (control group); these data were collected during a seven‐year study period. Although we did not calculate comparative data for the whole of Sweden, the study authors similarly reported no evidence of a difference in incidence rates of femoral fractures. This study used a non‐randomised trial design (CBA), and there were no reasons to upgrade the certainty of the evidence, which we judged to be very low. We also downgraded the certainty of the evidence by one level for indirectness because these data provided indirect evidence for this outcome, and we noted that the control group included residents from the intervention area. Because this was indirect evidence for this outcome, we did not include it in the summary of findings table.

Ytterstad 1996 similarly reported incidence rates of fall‐related fractures per 1000 residents rather than the number of people who had fall‐related fractures. For completeness, we also included these data in the review. We calculated rate ratios using data from the study report for all residents in the area exposed to the fall prevention intervention. The postintervention rate of fall‐related fractures was lower than pre‐intervention rates (RaR 1.17, 95% CI 1.00 to 1.38; Analysis 1.7). These data were collected from an injury database at emergency departments in the intervention area with a population of 14,850 residents 65 years of age or older at the postintervention time point (after five years). Fall‐related fracture data were not reported for the control group area. This study used a non‐randomised trial design (CBA), and there were no reasons to upgrade the certainty of the evidence, which we judged to be very low. We also downgraded the certainty of the evidence by one level for indirectness because these data provided indirect evidence for this outcome; for this reason we did not include it in the summary of findings table.

Number of people experiencing one or more falls resulting in hospital admission

One cluster RCT, Larsen 2005, and one non‐randomised trial, Kempton 2000, reported data for falls resulting in hospital admission. We did not include data for Kempton 2000 in the summary of findings table.

Whilst Larsen 2005 reported falls leading to hospital admission, we could not be certain whether the data accounted for people having more than one fall. As this study report also included risk ratios that were calculated to account for age and marital status, we selected these data for the review. However, we note that the study authors did not indicate whether these data were adjusted for the clustering effect, and therefore we advise caution in the interpretation of these data. Data were reported separately according to sex, using the intention‐to‐treat principle, and reported RRs did not include 95% CIs. In this study, there was no evidence of a difference when either female or male residents were exposed to an "Environmental and Health Program" (RR 0.96 and RR 1.07, respectively), or when female or male residents were exposed to both the "Environmental and Health Program" and the "Calcium and Vitamin D Program" (RR 0.90 and RR 1.14, respectively). These data were collected from the Danish Hospital Registration Database after a 42‐month period; the population size exposed to these two intervention programmes or in the control group was 7179 residents. We judged the certainty of the evidence to be very low. We downgraded the evidence by two levels for very serious risk of bias because this cluster RCT had high and unclear risk of bias. We also downgraded by one level for imprecision because we could not determine the level of imprecision in these data, which were reported without CIs, and may not have been adjusted for the clustering effect. See summary of findings Table 2.

Kempton 2000 reported the rate of fall‐related hospital admissions rather than the number of people who had at least one fall‐related hospital admission. For completeness, we have included these data in the review. The study authors reported a reduction in hospital admissions amongst residents who were exposed to the SOYF falls prevention programme (RaR 0.80, 95% CI 0.76 to 0.84). These data were collected from hospital admission records at the end of a four‐year study period in the intervention and control group areas (with an estimated population of 141,183 residents). This rate ratio accounted for differences in resident age and was described as a conservative estimate, as data were incomplete for the final year of follow‐up. This study used a non‐randomised trial design (CBA), and there were no reasons to upgrade the certainty of the evidence, which we judged to be very low. We also downgraded the certainty of the evidence by one level for indirectness because these data provided indirect evidence for this outcome; for this reason we did not include it in the summary of findings table.

Cost‐effectiveness of interventions

Two non‐randomised trials reported data on cost‐effectiveness (Kempton 2000; Ytterstad 1996).

Kempton 2000 conducted a cost‐benefit evaluation for the SOYF intervention, which was delivered between 1992 and 1996. The cost‐benefit analysis used two estimates of savings, first comparing the cost of hospital admissions in the intervention area against a control area of similar demographic characteristics, then comparing hospital use in the intervention region with the state of New South Wales as a whole. The programme's estimated total direct costs were AUD 781,829. The methods yielded overall net benefits from AUD 5.4 million (for avoided hospital admissions) to AUD 16.9 million (for all avoided direct/indirect costs). The average overall benefit‐cost ratio for the programme was 20.6:1. In 1996, the standardised cost ratio (SCR) was (SCR 87.18, 95% CI 84.6 to 89.8).

Ytterstad 1996 analysed the short‐term hospital costs for fall‐related fractures in private homes using eight years of hospital data up to July 1985 in the intervention area in Norway. The study authors noted rate reductions in the intervention area between baseline and the intervention period for hospital admissions (16.1%), hospital bed‐days (16.7%), and operations related to falls (35.1%).

Both of these studies used non‐randomised trial designs, and there were no reasons to upgrade the certainty of the evidence, which we judged to be very low. We also downgraded the certainty of the evidence by one level for indirectness. Whilst we recognise that the population and interventions were eligible for this review, we believe the time at which these studies were conducted means that economic analyses are less reliable because of other changes in the healthcare settings, and this may impact the directness of these results. See summary of findings Table 2.

Discussion

Summary of main results

We included nine studies (two cluster RCTs and seven non‐randomised trials); of the seven non‐randomised trials, five were CBA study designs and two were CITS designs. All studies evaluated population‐based falls prevention interventions, and the geographical regions ranged in size from small municipalities (with approximately 1800 people in the intervention area) to cities (with 79,425 residents in the intervention area).

We found studies evaluating two different types of interventions: medication and nutrition (one study) and multicomponent interventions (nine studies).

Medication and nutrition

One multi‐arm study included an intervention in which all people living in the intervention area were offered free‐of‐charge daily supplements of calcium and vitamin D₃ (the "Calcium and Vitamin D Program"). We were unsure of the findings in this single study. Although the study authors reported that female residents in the intervention area had fewer fall‐related hospital admissions than those in the control area (who were exposed to no fall prevention interventions), the evidence was of very low certainty. The study authors also reported no evidence of a difference in fall‐related hospital admissions amongst male residents. None of our other review outcomes were reported in this study.

Multicomponent interventions

All of the included studies evaluated multicomponent interventions for falls prevention. We used the ProFaNE taxonomy to categorise the different components of the interventions in this review, and note that components broadly included exercise, medication, environment (home and community‐level), social environment (staff training), knowledge/education. However, each study used different components, and the approach to delivery varied amongst studies.

Again, we were unsure of the findings from these studies because the certainty of the evidence for all outcomes was very low. We did not combine data from studies because of the differences between study designs and the differences in quantitative presentation of data.

For rate of falls, there was very low‐certainty evidence from one cluster RCT (with lower rates in the intervention area) and two CBAs that reported no evidence of a difference in rate of falls between the intervention and control group areas (these data were reported for rate of falls inside and outside the home in one study). There was also very low‐certainty evidence for the number of fallers, again with fewer fallers in the intervention area in a cluster RCT, and no evidence of a difference in the number of fallers in two CBAs. This same cluster RCT also found that fewer people in the intervention area had one or more fall‐related injuries, with no evidence of a difference in another CBA (very low‐certainty evidence). There was very low‐certainty evidence of little or no difference in the number of people having one or more fall‐related fractures (reported in a cluster RCT), with a similar finding in a CITS. In a multi‐arm study, there was very low‐certainty evidence of no difference in the number of female or male residents with falls leading to hospital admission after either a multicomponent intervention ("Environmental and Health Program") or a combination of this programme and the "Calcium and Vitamin D Program". Two CBAs reported an economic analysis, with one study reporting a saving in hospital admissions and indirect and direct costs, and another reporting cost reductions for hospital admissions, hospital bed‐days, and fall‐related operations; savings were in the intervention areas in both studies.

No studies reported adverse events.

Overall completeness and applicability of evidence

We included only population‐based studies in which communities were allocated to a falls prevention intervention or to no intervention. Data were reported in all studies for people who were at least 60 years of age. However, there was generally insufficient information about residents in the intervention area, and we could not easily discern whether populations were similar. For example, it is likely that some communities may include differences in socio‐economic status of residents, as well as differences in mean ages and residential status. Although some studies excluded people who were living in institutional settings, others included the whole population of older adults in their data. In addition, the data in this review are limited to a few countries in which studies were conducted, and there may be geographical differences which impact fall hazards. We were unable to explore these differences and to examine their impact on the findings in this review. We therefore cannot be certain whether these studies are applicable to all populations. It should also be noted that we deliberately excluded studies that targeted people at higher risk of falling, therefore this review does not provide evidence for this specific group.

In addition, intervention characteristics were often dissimilar. Although we considered the ProFaNE taxonomy to identify intervention descriptors (Lamb 2011), the level of detail in study reports was often insufficient. We also note that all of the included studies were published before this taxonomy. As all studies included multicomponent interventions, we could not determine whether one particular intervention element was responsible for any changes in falls data. Understanding the mechanisms by which interventions work is of considerable interest; however, an exploration of this was beyond the scope of this review. All of the included studies were conducted at least 16 years ago; we expect that since then there have likely been changes in public health awareness, as well as environmental health and safety improvements, and community environments. However, we anticipate that many of the interventions are still applicable to many public settings.

Certainty of the evidence

We judged all of the evidence in this review to be of very low certainty. All the evidence for multicomponent interventions included studies that used non‐randomised trial designs and, as agreed a priori, we judged evidence from these studies to be of at least low certainty. We found no reasons to increase (or upgrade) the level of certainty in these non‐randomised trials. Using EPHPP to assess risk of bias in these studies (EPHPP 2010), we judged that most were methodologically weak. In addition, we found that the cluster RCTs also included some high and unclear risk of bias that could have impacted outcome data (risks of recall bias due to methods used to report data, and important baseline differences between intervention and control group communities). We note here that although we assessed the risk of recall bias in this review, findings in Rapp 2014 indicate that recall bias may not impact outcome. We also noted that the findings for most outcomes differed between studies: one cluster RCT noted improved outcomes in the intervention areas which were not replicated in the non‐randomised studies. We were unable to explore these differences, and could not be certain whether they were explained by variations in intervention components or other factors. Because of lack of information in one study (CIs and information about analysis of data with a cluster RCT design), we could not be certain of the level of precision, and therefore downgraded evidence from this study for possible imprecision. We also noted imprecise findings in some evidence in which the intervention included the possibility of benefit and no benefit.

Some studies reported data that did not directly match our review outcome criteria. For completeness, we have reported these data in the review and used GRADE to downgrade the certainty of the evidence for indirectness. However, because other studies reported these data according to our criteria, we did not include these indirect data in our summary of findings table.

We did not assess the risk of publication bias in these studies, and our GRADE judgements do not account for this possibility.

Potential biases in the review process

We included a comprehensive search of the published literature in 2020. Due to limited resources, we focused on running top‐up searches in December 2023 in the three main databases (CENTRAL, MEDLINE, and Embase). We independently assessed study eligibility from full‐text reports, extracted data, and assessed the risk of bias in the included studies before reaching consensus together or with one other review author. During the initial screening process, we made a pragmatic decision (because the searches yielded a large volume of reports) to use a modified screening process, having one review author first rule out obvious excluded studies based on titles alone. Although this was less robust than using two review authors, we conducted sample checking to evaluate eligibility decisions.

We had intended to only include studies if there were at least two intervention sites and two control sites because we thought that this would improve the diversity of populations. However, we found that the size of study sites varied widely between studies and that limiting the review by the number of sites did not address this diversity issue (i.e. a study with several intervention sites may have a small population which is not diverse, whilst a study with a single intervention site may have a large diverse population). We therefore opted in the review to include studies even if they included only one intervention and one control site. Although this was a change to the protocol, we believe that it provided a better summary of the available data for population‐based falls prevention interventions.

We recognise that this review included some decision‐making that could be challenged. For example, we found it difficult to describe the non‐randomised trials in the review, and used Reeves 2017 as a guide to provide transparency. However, we found that even this tool did not easily match some of the study designs in our included studies. We also found that the risk of bias tool for the non‐randomised trials was not a perfect design for population‐level studies in which communities rather than individuals were allocated to groups. Because of the large differences in study designs, we gave additional consideration to our decisions in order to provide consistency across studies; other review authors may reach different risk of bias judgements.

Agreements and disagreements with other studies or reviews

Although there are reviews on falls prevention interventions in older people, these target components of falls prevention interventions when delivered to individuals rather than to whole communities (Clemson 2023; Hopewell 2018; Sherrington 2019). Hopewell 2018 distinguishes between multifactorial and multicomponent interventions: “A multifactorial intervention is one in which the selection of falls prevention interventions (such as exercise, home‐hazard modification or medication review) prescribed or provided to each individual is matched to their risk‐of‐falls profile, which is assessed beforehand. This individually‐tailored intervention means that after receiving an assessment of known risk factors for falling, individuals are likely to received different combinations of interventions: i.e. one person may receive supervised exercise and home‐hazard modification whereas another may receive home‐hazard modification and medication review. Multiple component interventions are those where people receive a fixed combination of two or more fall prevention interventions selected from different categories of intervention (e.g. exercises, medication review, environment/assistive technology)”. Clearly, multifactorial interventions are at odds with how we define population‐based interventions, as they require individualised risk assessments, but for completeness and accuracy of comparison we include them here. We are not aware of any comparable reviews of population‐based falls prevention interventions.

The three Cochrane Reviews found similar results to one another for rate of falls (Clemson 2023; Hopewell 2018; Sherrington 2019). Hopewell 2018 found low‐certainty evidence that multifactorial falls prevention interventions may reduce the rate of falls by 23%, and moderate‐certainty evidence that multiple component interventions probably reduce falls by 26%. Sherrington 2019 assessed the effects of exercise interventions in community‐dwelling older people, and found high‐certainty evidence that exercise interventions of any type reduce the rate of falls by 23%. Clemson 2023 assessed the effects of environmental interventions in community‐dwelling older people and found that eliminating fall hazards in the home probably reduces the rate of falls by 26% (moderate‐certainty evidence).

The level of intervention detail in the studies included in our review prevents direct comparison with these Cochrane Reviews; however, all studies included multicomponent interventions, most of which included environmental components, and some that also included exercise components. One of our included studies also reported a reduction in the rate of falls after a multicomponent intervention, although the effect was much greater (65% reduction in Xia 2009). Other studies included in our review reported no evidence of a difference in the rate of falls, and we could not determine whether this was because of unknown confounders or other methodological issues in these non‐randomised trials, rather than because the interventions were allocated at the population level.

Hopewell 2018 reported little or no difference for multifactorial interventions in the risk of falling (number of fallers), but an 18% reduction in risk of falls for multiple component interventions (moderate‐certainty evidence). Sherrington 2019 reported a 15% risk reduction for one or more falls after exercise interventions (high‐certainty evidence), and Clemson 2023 reported moderate‐certainty evidence of an 11% risk reduction for one or more falls after home fall hazards were modified. In our review, there were inconsistent findings across the included studies. There was a large effect in a cluster RCT (66% reduction) after the intervention, but no evidence of a difference in a non‐randomised trial. For fall‐related fractures, Hopewell 2018 reports a 27% reduction for multifactorial interventions (low‐certainty evidence), which is comparable to the report by Sherrington 2019 of a reduction of 27% after exercise interventions (low‐certainty evidence). However, the very low‐certainty evidence for fall‐related fractures in Hopewell 2018 means that they were uncertain of the effects of multiple component interventions on the risk of fall‐related fractures, and Clemson 2023 found little or no difference in the risk of fall‐related fractures. Likewise, we found no evidence of a difference in fall‐related fractures in a cluster RCT and a non‐randomised trial evaluating multicomponent interventions at a population level.

Overall, the findings in our review are very uncertain, which limits the reliability of any comparisons with the findings in these other Cochrane Reviews.

Figure 1

Study flow diagram.

Ir a la figura de la revisiónAbrir en una pestaña nueva

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included cluster RCT.

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.1

Comparison 1: Multicomponent falls prevention interventions versus control, Outcome 1: Rate of falls (postintervention data per 1000 persons per year)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.2

Comparison 1: Multicomponent falls prevention interventions versus control, Outcome 2: Number of fallers

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.3

Comparison 1: Multicomponent falls prevention interventions versus control, Outcome 3: Number of fallers (postintervention data per 1000 persons per year)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.4

Comparison 1: Multicomponent falls prevention interventions versus control, Outcome 4: Number of people who experienced 1 or more fall‐related injuries

Ir a la figura de la revisiónAbrir en una pestaña nueva

$Comparison 1: Multicomponent falls prevention interventions versus control, Outcome 5: Number of people who experienced 1 or more fall‐related fractures$

Analysis 1.5

Comparison 1: Multicomponent falls prevention interventions versus control, Outcome 5: Number of people who experienced 1 or more fall‐related fractures

Ir a la figura de la revisiónAbrir en una pestaña nueva

$Comparison 1: Multicomponent falls prevention interventions versus control, Outcome 6: Rate of fall‐related femoral fractures$

Analysis 1.6

Comparison 1: Multicomponent falls prevention interventions versus control, Outcome 6: Rate of fall‐related femoral fractures

Ir a la figura de la revisiónAbrir en una pestaña nueva

$Comparison 1: Multicomponent falls prevention interventions versus control, Outcome 7: Rate of fall‐related fractures (any type of fracture)$

Analysis 1.7

Comparison 1: Multicomponent falls prevention interventions versus control, Outcome 7: Rate of fall‐related fractures (any type of fracture)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Summary of findings 1. Medication or nutrition fall prevention interventions versus control: evidence from RCTs

Outcomes	Impact of the intervention	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Population: community‐dwelling older adults ≥ 65 years of age Settings: communities Intervention: free‐of‐charge daily supplement of calcium carbonate and vitamin D₃ Comparison: no falls prevention intervention
Rate of falls	Not estimable	‐	‐	No studies reported this outcome.
Number of fallers	Not estimable	‐	‐	No studies reported this outcome.
Number of people experiencing 1 or more fall‐related injuries	Not estimable	‐	‐	No studies reported this outcome.
Number of people experiencing 1 or more fall‐related fractures	Not estimable	‐	‐	No studies reported this outcome.
Number of people experiencing 1 or more falls resulting in hospital admission Measured using Danish Hospital Registration Database Follow‐up: fall data collected during 42‐month study period	Female residents exposed to a "Calcium and Vitamin D" falls prevention programme had fewer fall‐related hospital admissions than female residents in the control area (RR 0.89; P < 0.10). For male residents, there was no evidence of a difference between the intervention and control areas (RR 1.08).^a	4542 (1 cluster RCT)	Very low^b
Number of people who experienced 1 or more adverse events	Not estimable	‐	‐	No studies reported this outcome.
Economic analysis	Not estimable	‐	‐	No studies reported this outcome.
Abbreviations: RCT: randomised controlled trial; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aData as reported in the study report. These data were reported separately for female and male residents, and were reported without confidence intervals. In addition, no P value was reported with the effect estimate for male residents. ^bWe downgraded by two levels owing to very serious risk of bias (high and unclear risk of bias). We also downgraded by one level for imprecision because the effect estimates were reported without confidence intervals, and we were unable to determine the degree of imprecision in the data (particularly for male residents).

Summary of findings 1. Medication or nutrition fall prevention interventions versus control: evidence from RCTs

Ir a la tabla de la revisión

Summary of findings 2. Multicomponent fall prevention interventions versus control: evidence from RCTs

Outcome	Impact of the intervention (findings as reported by study authors, unless specified otherwise)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Population: community‐dwelling older adults at least 60 years of age Settings: communities Intervention: multicomponent falls prevention interventions (details of components in each study are described in footnotes) Comparison: no falls prevention intervention
Rate of falls Measured using questionnaire data (self‐reported) Follow‐up: end of study follow‐up at 18 months	In a cluster RCT,^a the rate of falls was lower in the intervention area than in the control area (RaR 0.356, 95% CI 0.253 to 0.501).	1422 (1 cluster RCT)	⊕⊝⊝⊝ Very low^b
Number of fallers Measured using questionnaire data (self‐reported) Follow‐up: end of study follow‐up at 18 months	In a cluster RCT,^a fewer people had falls in the intervention area than in the control area (RR 0.34, 95% CI 0.19 to 0.62).^c	1422^d (1 cluster RCT)	⊕⊝⊝⊝ Very low^e
Number of people experiencing 1 or more fall‐related injuries Measured using questionnaire data (self‐reported) Follow‐up: end of study follow‐up at 18 months	In a cluster RCT,^a fewer people had injurious falls in the intervention area than in the control area (RR 0.39, 95% CI 0.20 to 0.77).^c	1422^d (1 cluster RCT)	⊕⊝⊝⊝ Very low^e
Number of people experiencing 1 or more fall‐related fractures Measured using questionnaire data (self‐reported) Follow‐up: end of study follow‐up at 18 months	In a cluster RCT,^a there was no evidence of a difference between the intervention and control group areas in fall‐related fractures (RR 0.55, 95% CI 0.17 to 1.85).^c	1422^d (1 cluster RCT)	⊕⊕⊝⊝ Very low^e
Number of people experiencing 1 or more falls resulting in hospital admission Measured using hospital records Follow‐up: falls data collected during 42‐month study period	In a cluster RCT^f evaluating an "Environment and Health" programme, there was no evidence of a difference between the intervention and control areas in number of females (RR 0.96) or males (RR 1.07) with falls leading to hospital admission. In the same cluster RCT, evaluating this intervention in combination with a nutrition and medication intervention ("Calcium and Vitamin D" programme), there was also no evidence of a difference between the intervention and control areas for females (RR 0.90) and males (RR 1.14).	7179 (1 cluster RCT)	⊕⊝⊝⊝ Very low^g
Adverse events	‐	‐	‐	No studies reported this outcome.
Cost‐effectiveness (economic analysis)	‐	‐	‐	No studies reported this outcome.
Abbreviations: CI: confidence interval; RaR: rate ratio; RCT: randomised controlled trial; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aInterventions in this study included: education programme (to reduce risk of falls, and including information on diet and exercise); home hazard assessment; modification of community settings (removing obstacles on pavements, roads, lawns; installing handrails). ^bDowngraded two levels for very serious risk of bias, and one level for imprecision because we could not be certain whether the effect estimate included an adjustment for clustering (and therefore may represent an overestimation of the true effect). ^cCalculated using Review Manager 2020 from data in the study report (Review Manager 2020). ^dIncluded data from a sample of the whole target population. To account for clustering, we calculated effect sample sizes to use in our analysis. ^eDowngraded two levels for very serious risk of bias, and one level for imprecision because the sample size for this population‐level study was small. ^fInterventions included: an "Environmental and Health Program" with home safety inspection, ways to avoid falls, health and dietary correction; a "Calcium and Vitamin D Program" in which residents were offered free‐of‐charge daily supplements; or a combination of both programmes. ^gWe downgraded by two levels for very serious risk of bias because this cluster RCT had high and unclear risk of bias. We also downgraded by one level for imprecision because we could not determine the level of imprecision in these data, which were reported without CIs and may not have been adjusted for the clustering effect.

Summary of findings 2. Multicomponent fall prevention interventions versus control: evidence from RCTs

Ir a la tabla de la revisión

Summary of findings 3. Multicomponent fall prevention interventions versus control: evidence from non‐randomised trials

Outcome	Impact of the intervention (findings as reported by study authors, unless specified otherwise)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Population: community‐dwelling older adults at least 60 years of age Settings: communities Intervention: multicomponent falls prevention interventions (details of components in each study are described in footnotes) Comparison: no falls prevention intervention
Rate of falls Measured using questionnaire data (self‐reported) or hospital records Follow‐up: falls data collected during study periods ranging from 14 months to 4 years	In a CBA,^a the reduction in rate of falls in the intervention group was not statistically significant (0.066 falls/person/year; P = 0.14). In another CBA,^b there was no evidence of a difference in rate of falls inside the home (RaR 1.07, 95% CI 0.39 to 2.99) or outside the home (RaR 0.91, 95% CI 0.61 to 1.37).^c	4197^d (2 non‐randomised trials)	⊕⊝⊝⊝ Very low^e
Number of fallers Measured using questionnaire data (self‐reported) Follow‐up: falls data collected during study periods ranging from 2 to 4 years	In a CBA,^a there was no evidence of a difference between the intervention and control areas in the number of fallers (OR 0.95, 95% CI 0.79 to 1.15). In another CBA,^f there was no evidence of a difference between the intervention and control areas in the number of fallers (RR 1.03, 95% CI 0.81 to 1.31).^c	3047^d (2 non‐randomised trials)	⊕⊝⊝⊝ Very low^e
Number of people experiencing 1 or more fall‐related injuries Measured using healthcare centre records Follow‐up: falls data collected during study period (5 years)	In a CBA,^g there was no evidence of a difference between the intervention and control areas in the number of people having injurious falls (OR 0.89, 95% CI 0.77 to 1.03).	67,300^h (1 non‐randomised trial)	⊕⊝⊝⊝ Very low^e
Number of people experiencing 1 or more fall‐related fractures Measured using hospital injury record system Follow‐up: falls data collected during 18‐month study period	In a CITS,ⁱ there was no evidence of a difference in the number of fractures that were prevented as a result of the intervention (14% prevented fractures in intervention group, 95% CI 9% more fractures to 37% fewer fractures).	24,365 (1 non‐randomised trial)	⊕⊝⊝⊝ Very low^j
Number of people experiencing 1 or more falls resulting in hospital admission Measured using hospital records Follow‐up: falls data collected during 42‐month study period	‐	‐	‐	No studies reported direct evidence for this outcome.
Adverse events	‐	‐	‐	No studies reported this outcome.
Cost‐effectiveness (economic analysis) Measured using healthcare records Follow‐up: falls data collected during study periods ranging from 4 to 8 years	A CBA^a reported a cost‐benefit in favour of the intervention with savings for avoided hospital admissions and indirect/direct costs (SCR 87.18, 95% CI 84.6 to 89.8). Another CBA^k reported cost reductions in favour of the intervention for hospital admissions (16.1%), hospital bed‐days (16.7%), and operations related to falls (35.1%).	163,683 (2 non‐randomised trials)	⊕⊝⊝⊝ Very low^l
CBA: controlled before‐and‐after study (a non‐randomised trial design); CI: confidence interval; CITS: controlled interrupted time‐series (a non‐randomised trial design); OR: odds ratio; RaR: rate ratio; RR: risk ratio; SCR: standardised cost ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aInterventions included: footwear, vision, physical activity, balance and gait, medication use, chronic conditions, plus home and public environmental hazards modification. ^bInterventions included: traffic safety, balance training, physical activity, home safety hazard modification, home modification, safe pavement (removing obstacles from pavements in the community), and medication use. ^cCalculated using Review Manager 2020 from data in the study report (Review Manager 2020). ^dIncluded data from a sample of the whole target population. ^eThere were no reasons to upgrade the certainty of the evidence from the non‐randomised trials. In addition, we downgraded the certainty of the evidence for imprecision. We also noted that the findings were inconsistent with the findings from the randomised controlled trial‐derived evidence. ^fInterventions included: an educational programme, government involvement with architectural consultations, exercise programmes, risk assessment, dietary and medicine guidance, and prevention of falls risk at home. ^gInterventions included: elimination of environment hazards (e.g. improvements to roads, pavements, street lighting); behavioural safety education and information programmes; injury prevention features in local media; availability of safety products; home modifications; and exercise support. ^hWhole population rather than target population of people aged ≥ 65 years. ⁱInterventions included: information on fall risk factors, and identifying and modifying hazards in the home and surrounding areas. Interventions aimed to reduce physical hazards, age‐debilitating illnesses, psychiatric illnesses, improper use of medication, diet insufficiency, and physical inactivity. ^jThere were no reasons to upgrade the certainty of the evidence from the non‐randomised trials. In addition, we downgraded the certainty of the evidence for imprecision. ^kInterventions included: identifying and remedy of home hazards, promoting environmental safety, health, diet, and lifestyle, and reduction in isolation and inactivity; pensioners providing skilled low‐cost services to improve physical environments in other older people's homes; availability of safety items including spiking of boots for icy pavements. ^lThere were no reasons to upgrade the certainty of the evidence from these non‐randomised trials. In addition, we downgraded by one level for indirectness. Although the population and interventions were eligible for this review, we believe the time at which these studies were conducted meant that economic analyses are less reliable because of other changes in the healthcare settings, and this may impact the directness of these results.

Summary of findings 3. Multicomponent fall prevention interventions versus control: evidence from non‐randomised trials

Ir a la tabla de la revisión

Table 1. Included studies: study design, location, and trial size

Study ID	Study design^a	Country	Communities	Target population	Total population size or total size of target group (intervention + control)	Sample size^b
Kempton 2000	CBA	Australia	Intervention: North Coast of New South Wales Control: Queensland Sunshine Coast	Community‐dwelling, ≥ 60 years of age	Total target group: 141,183 Intervention: 79,425 Control: 61,758	2445
Larsen 2005	Cluster RCT	Denmark	3 intervention groups: each in 1 block within Randers municipality Control: 1 block within Randers municipality	Community‐dwelling, ≥ 66 years of age	Total target group: 9605 Intervention Block 1 ("Environmental and Health Program": 2532) Intervention Block 2 ("Calcium and Vitamin D Program"): 2426 Intervention Block 3 (both programmes): 2531 Control Block 4: 2116	N/A
Lindqvist 2001	CBA	Sweden	Intervention: Motala (municipality in Östergötland County) Control: Mjölby (municipality in Östergötland County)	≥ 65 years of age	Total whole population (before intervention): 67,300 Intervention whole population (before intervention): 41,000 Control whole population (before intervention): 25,900	N/A
Poulstrup 2000	CITS	Denmark	Intervention: 5 municipalities of County Council of Vijle Control: 4 municipalities of County Council of Vijle	Community‐dwelling, ≥ 65 years of age	Total target group: 26,221 Intervention: 12,905 Control: 11,460	N/A
Pujiula Blanch 2010	CBA	Spain	Intervention: Salt (health district in Girona) Control: Girona‐4 (health district in Girona)	≥ 70 years of age	Total target group: 3727 Intervention: 2515 Control: 1212	602
Svanström 1996	CITS	Sweden	Intervention: Lidköping (municipality in Skaraborg County) Control: Skaraborg County, and whole of Sweden	≥ 65 years of age	Total target group (first study year): 1,173,698 Intervention (first study year): 6817 Control (first study year): 50,052 (Skaraborg County); 1,116,829 (Sweden)	N/A
Wijlhuizen 2007	CBA	Netherlands	Intervention: Sneek (in Area Health Authority of Fryslân) Control: Harlingen and Heerenveen (2 communities within Area Health Authority of Fryslân)	≥ 65 years of age, living independently	Total target group: 12,769 Intervention: 4369 Control: 8400^c	1752
Xia 2009	Cluster RCT	China	Intervention: 2 residential communities, Shanghai Control: 2 residential communities, Shanghai	≥ 60 years of age	Total target group: 3600^c Intervention: 1800^c Control: 1800^c	1422
Ytterstad 1996	CBA	Norway	Intervention: Harstad Control: 6 municipalities ("close to Harstad"), and Trondheim	≥ 65 years of age	Total whole population: 172,500 Intervention whole population: 22,500 Control whole population: 15,000 (6 municipalities); 135,000 (Trondheim)	N/A
CBA: controlled before‐and‐after study (non‐randomised study design); CITS: controlled interrupted time‐series study (non‐randomised study design); N/A: not applicable; RCT: randomised controlled trial ^aClassification using Reeves 2017 as a guide. ^bSample of target population who were contacted in telephone or questionnaire surveys for some or all outcome data. In some studies, data were collected for all residents, and there were no sample data in these studies. ^cApproximate values, as reported in the study report.

Table 1. Included studies: study design, location, and trial size

Ir a la tabla de la revisión

Table 2. Included studies: intervention characteristics

Study ID	Type of intervention	Duration of intervention (study dates)	Main sites of delivery	Interventions delivered by
Kempton 2000	Multicomponent interventions Footwear, vision, physical activity, balance and gait, medication use, chronic conditions, plus home and public environmental hazards	4 years (1992 to 1995) Baseline data collection in 1991	Community setting (in participants' homes, organisations and other locations in the community)	General practitioners and healthcare workers
Larsen 2005	Medication or nutrition Block 1: Free‐of‐charge daily supplement 1000 mg calcium carbonate or 400 IU (10 μg) vitamin D₃ Multicomponent interventions Block 2: home safety inspection, ways to avoid falls, health and dietary correction Block 3: combined interventions from Block 1 and Block 2	42 months (1995 to 1998) Baseline data collection from 1990 to 1994	Community setting (in participants' homes, organisations and other locations in the community, including the pharmacy)	Trained community nurses
Lindqvist 2001	Multicomponent interventions Elimination of environmental hazards (e.g. improvements to roads, pavements, street lighting), behavioural safety education and information programmes, injury prevention features in local media, availability of safety products, home modifications, and exercise support	5 years (1984 to 1989) Pre‐implementation phase 1983 to 1984	Community setting (in participants' homes, organisations and other locations in the community)	Staff in social care facilities were given additional training.
Poulstrup 2000	Multicomponent interventions Information on fall risk factors, and identifying and correcting hazards in the home and surrounding areas. Interventions aimed to reduce physical hazards, age‐debilitating illnesses, psychiatric illnesses, improper use of medication, diet insufficiency, and physical inactivity	18 months (1985 to 1988)	Community setting (in participants' homes, organisations and other locations in the community)	General practitioners, district nurses, and home helpers
Pujiula Blanch 2010	Multicomponent interventions Educational programme through information pamphlets, media, and conferences; government involvement with architectural consultations; exercise programmes, risk assessment, dietary and medicine guidance, and prevention of fall risks at home	2 years (2001 to 2003)	Community setting (in participants' homes, organisations and other locations in the community)	Primary care team, including professionals and "individuals"
Svanström 1996	Multicomponent interventions Educational programme targeting health hazards and how to reduce risk, environmental changes regarding traffic environment (streetlights control, new cycle paths), 2 newspaper articles which advertised updates on preventive work and increased awareness on how to target risk factors	7 years (1985 to 1992; gradual implementation over each year) Baseline data collection in 1987	Community setting (in participants' homes, organisations and other locations in the community)	District nurse, policymakers, home visitors, service apartment personnel
Wijlhuizen 2007	Multicomponent interventions Traffic safety, balance training, physical activity, home safety hazard, home modification, safe pavement (removing obstacles from pavements in the community), and medication use	14 months (1999 to 2002) Baseline data during 10 months of 1999. Final data in 10 months after 14‐month intervention period	Community setting (in participants' homes, organisations and other locations in the community)	Professionals (types not specified) Trained volunteer peer consultants
Xia 2009	Multicomponent interventions Education programme (to reduce risk of falls, and including information on diet and exercise), home hazard assessment, modification of community settings (removing obstacles on pavements, roads, lawns; handrails)	18 months (2006 to 2007)	Community setting (in participants' homes, organisations and other locations in the community)	A multidisciplinary group including individuals from local government, the community health centre, and other members of the community
Ytterstad 1996	Multicomponent interventions Identifying and remedy of home hazards, promote environmental safety, health, diet, and lifestyle and reduction in isolation and inactivity; pensioners providing skilled low‐cost services to improve physical environments in others' homes; availability of safety items including spiking of boots for icy pavements	8 years (1985 to 1993; intervention was mostly delivered in the latter 5 years of study period)	Community setting (in participants' homes, organisations and other locations in the community)	An injury prevention group comprising representatives from a hospital and several public and private organisations, including "health personnel" Nurses, nurse‐aides, and other home helpers. Physiotherapists. Peer‐support (older people offering skilled work to other older people or delivery of spiked boots)
IU: international unit

Table 2. Included studies: intervention characteristics

Ir a la tabla de la revisión

Table 3. Summary and global rating of quality assessment for non‐randomised studies

Study ID	Selection bias	Study design	Confounders	Blinding	Data collection method	Withdrawals and dropouts	Overall
Kempton 2000	MODERATE	MODERATE	MODERATE	WEAK	WEAK	MODERATE	WEAK
Lindqvist 2001	MODERATE	MODERATE	STRONG	WEAK	MODERATE	MODERATE	MODERATE
Poulstrup 2000	MODERATE	MODERATE	MODERATE	WEAK	WEAK	MODERATE	WEAK
Pujiula Blanch 2010	MODERATE	MODERATE	WEAK	WEAK	WEAK	MODERATE	WEAK
Svanström 1996	MODERATE	MODERATE	WEAK	WEAK	STRONG	MODERATE	WEAK
Wijlhuizen 2007	MODERATE	MODERATE	MODERATE	WEAK	WEAK	WEAK	WEAK
Ytterstad 1996	MODERATE	MODERATE	WEAK	WEAK	STRONG	MODERATE	WEAK

Table 3. Summary and global rating of quality assessment for non‐randomised studies

Ir a la tabla de la revisión

Comparison 1. Multicomponent falls prevention interventions versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Rate of falls (postintervention data per 1000 persons per year) Show forest plot	1		Rate Ratio (IV, Fixed, 95% CI)	Totals not selected

1.1.1 Number of falls in and around the home	1		Rate Ratio (IV, Fixed, 95% CI)	Totals not selected
1.1.2 Number of falls outside the home	1		Rate Ratio (IV, Fixed, 95% CI)	Totals not selected
1.2 Number of fallers Show forest plot	2		Risk Ratio (IV, Fixed, 95% CI)	Totals not selected

1.2.1 In a cluster RCT	1		Risk Ratio (IV, Fixed, 95% CI)	Totals not selected
1.2.2 In a non‐randomised trial	1		Risk Ratio (IV, Fixed, 95% CI)	Totals not selected
1.3 Number of fallers (postintervention data per 1000 persons per year) Show forest plot	1		Rate Ratio (IV, Fixed, 95% CI)	Totals not selected

1.3.1 Falls in and around the home	1		Rate Ratio (IV, Fixed, 95% CI)	Totals not selected
1.3.2 Falls outside the home	1		Rate Ratio (IV, Fixed, 95% CI)	Totals not selected
1.4 Number of people who experienced 1 or more fall‐related injuries Show forest plot	1		Risk Ratio (IV, Fixed, 95% CI)	Totals not selected

1.5 Number of people who experienced 1 or more fall‐related fractures Show forest plot	2		Risk Ratio (IV, Fixed, 95% CI)	Totals not selected

1.5.1 In a cluster RCT	1		Risk Ratio (IV, Fixed, 95% CI)	Totals not selected
1.5.2 In a non‐randomised trial	1		Risk Ratio (IV, Fixed, 95% CI)	Totals not selected
1.6 Rate of fall‐related femoral fractures Show forest plot	1		Rate Ratio (IV, Fixed, 95% CI)	Totals not selected

1.6.1 Female residents	1		Rate Ratio (IV, Fixed, 95% CI)	Totals not selected
1.6.2 Male residents	1		Rate Ratio (IV, Fixed, 95% CI)	Totals not selected
1.7 Rate of fall‐related fractures (any type of fracture) Show forest plot	1		Rate Ratio (IV, Fixed, 95% CI)	Totals not selected

Comparison 1. Multicomponent falls prevention interventions versus control

Ir a la tabla de la revisión

	Idioma de la Revisión Cochrane Escoja su idioma de preferencia para las revisiones Cochrane y otros contenidos. Las secciones sin una traducción aparecerán en inglés.

	Idioma de la web Escoja su idioma de preferencia para la web de la Biblioteca Cochrane.

Idioma de la Revisión Cochrane

Idioma de la web

Resumen

Antecedentes

Objetivos

Métodos de búsqueda

Criterios de selección

Obtención y análisis de los datos

Resultados principales

Conclusiones de los autores

PICO

PICO

Population

Intervention

Comparison

Outcome

Resumen en términos sencillos

¿Son útiles las intervenciones poblacionales (dirigidas a comunidades enteras en lugar de a individuos) para prevenir las caídas y las lesiones derivadas en personas mayores?

Resumen visual

Authors' conclusions

Implications for practice

Implications for research

Summary of findings

Background

Description of the condition

Description of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Unit of analysis issues

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

Summary of findings and assessment of the certainty of the evidence

Results

Description of studies

Results of the search

Included studies

Study design

Participants

Interventions

Outcomes

Funding sources

Excluded studies

Ongoing studies

Studies awaiting classification

Risk of bias in included studies

Risk of bias in cluster RCTs

Risk of bias in non‐randomised studies

Effects of interventions

Medication or nutrition interventions

Number of people experiencing one or more falls resulting in hospital admission

Multicomponent interventions

Rate of falls

Number of fallers

Number of people experiencing one or more fall‐related injuries

Number of people experiencing one or more fall‐related fractures

Number of people experiencing one or more falls resulting in hospital admission

Cost‐effectiveness of interventions

Discussion

Summary of main results