Sistemas de citas iniciadas por el paciente para adultos con enfermedades crónicas en la atención secundaria
Resumen
Antecedentes
El incumplimiento de las citas ambulatorias en los hospitales es un problema informado comúnmente en los servicios de salud de todo el mundo; por ejemplo, para el National Health Service (NHS) del Reino Unido tiene un coste de millones de libras cada año y puede causar dificultades de operación y programación en todo el mundo. En 2002 la Organización Mundial de la Salud (OMS) publicó un informe en el que se destacaba la necesidad de contar con un modelo de atención que satisficiera más fácilmente las necesidades de los pacientes con enfermedades crónicas. Los sistemas de citas iniciadas por el paciente pueden satisfacer esta necesidad y al mismo tiempo mejorar la eficiencia de las citas en el hospital.
Objetivos
Evaluar los efectos de los sistemas de citas iniciadas por el paciente en comparación con los sistemas de citas coordinadas por el especialista para pacientes con afecciones crónicas o recurrentes tratados en la atención secundaria.
Métodos de búsqueda
Se hicieron búsquedas en el Registro Cochrane Central de Ensayos Controlados (Cochrane Central Register of Controlled Trials, CENTRAL), MEDLINE, Embase y en otras seis bases de datos. Se estableció contacto con los autores de los estudios identificados y se realizó una búsqueda de citas hacia adelante y hacia atrás. Se buscaron investigaciones actuales o en curso en dos registros de ensayos. Las búsquedas se realizaron el 13 de marzo 2019.
Criterios de selección
Se incluyeron ensayos aleatorizados, publicados y no publicados en cualquier idioma, que compararan el uso de sistemas de citas iniciadas por el paciente con sistemas de citas coordinadas por el especialista para adultos con afecciones crónicas o recurrentes tratados en la atención secundaria cuando informaban de uno o más de los siguientes resultados: medidas físicas del estado de salud o la actividad de la enfermedad (incluidos los daños), la calidad de vida, la utilización o el coste de los servicios, los efectos adversos, la satisfacción del paciente o del médico, o las fallas del «sistema».
Obtención y análisis de los datos
Dos autores de la revisión examinaron de forma independiente todas las referencias en la etapa de selección de títulos/resúmenes y en la etapa de selección del texto completo utilizando criterios de inclusión preestablecidos. Los desacuerdos se resolvieron mediante discusión. Dos autores de la revisión completaron de forma independiente la extracción de datos de todos los estudios incluidos. Se discutieron y se resolvieron las discrepancias con un tercer autor de la revisión. Cuando fue necesario, se estableció contacto con los autores de los trabajos incluidos para proporcionarles más información. Dos autores de la revisión evaluaron de forma independiente el riesgo de sesgo mediante la herramienta «Riesgo de sesgo» de Cochrane Effective Practice and Organisation of Care, resolviendo cualquier discrepancia con un tercer autor de la revisión. Dos autores de la revisión, de forma independiente, evaluaron la certeza de la evidencia según los criterios GRADE.
Resultados principales
Los 17 ensayos aleatorizados incluidos (3854 participantes; media de edad de 41 a 76 años; seguimiento de 12 a 72 meses) abarcaron seis afecciones de salud generales: cáncer, artritis reumatoide, asma, enfermedad pulmonar obstructiva crónica, psoriasis y enfermedad intestinal inflamatoria. La certeza de la evidencia mediante los criterios GRADE fue principalmente de baja a muy baja. Los resultados sugieren que las clínicas iniciadas por los pacientes pueden lograr poca o ninguna diferencia en cuanto a la ansiedad (odds‐ratio [OR] 0,87; intervalo de confianza [IC] del 95%: 0,68 a 1,12; 5 estudios, 1019 participantes; evidencia de certeza baja) o la depresión (OR 0,79; IC del 95%: 0,51 a 1,23; 6 estudios, 1835 participantes; evidencia de certeza baja) en comparación con el sistema de citas coordinado por un especialista. Los resultados también sugieren que las clínicas iniciadas por los pacientes pueden lograr poca o ninguna diferencia en la calidad de vida (diferencia de medias estandarizada [DME] 0,12, IC del 95%: 0,00 a 0,25; 7 estudios, 1486 participantes; evidencia de certeza baja) en comparación con el sistema de citas coordinado por el especialista. Los resultados de la utilización de servicios (contactos) sugieren que puede haber poca o ninguna diferencia en la utilización de servicios en términos de contactos entre los grupos de citas iniciadas por el paciente y de citas coordinadas por el especialista; sin embargo, el efecto no es seguro, ya que el cociente de tasas osciló entre 0,68 y 3,83 en los estudios (mediana del cociente de tasas 1,11; intercuartil [RIC] 0,93 a 1,37; 15 estudios, 3348 participantes; evidencia de certeza baja). No se sabe si la utilización de los servicios (costes) se reduce en los grupos de consulta iniciados por el paciente en comparación con los grupos de consulta coordinados por el especialista (8 estudios, 2235 participantes; evidencia de certeza muy baja). Los resultados sugieren que los eventos adversos como las recaídas en algunas afecciones (enfermedad intestinal inflamatoria y cáncer) pueden presentar poca o ninguna reducción en el grupo de citas iniciadas por el paciente en comparación con el grupo de citas coordinadas por el especialista (DM ‐0,20; IC del 95%: ‐0,54 a 0,14; 3 estudios, 888 participantes; evidencia de certeza baja). Los resultados no son claros en cuanto a las diferencias que la intervención puede lograr en la satisfacción de los pacientes (DME 0,05; IC del 95%: ‐0,41 a 0,52; 2 estudios, 375 participantes) debido a que la certeza de la evidencia es baja, ya que cada estudio utilizó diferentes preguntas para recopilar sus datos en diferentes puntos temporales y en diferentes afecciones de salud. Algunas áreas de riesgo de sesgo en todos los estudios incluidos fueron consistentemente altas (por ejemplo, para el cegamiento de los participantes y el personal y el cegamiento de la evaluación de resultados, otras áreas fueron en gran parte de bajo riesgo de sesgo o se vieron afectadas por el informe deficiente que dio lugar a que la evaluación fuera poco clara).
Conclusiones de los autores
Los sistemas de citas iniciados por el paciente pueden tener poco o ningún efecto en la ansiedad, la depresión y la calidad de vida del paciente en comparación con los sistemas de citas coordinados por el especialista. Otros aspectos del estado de la enfermedad y la experiencia también revelan una diferencia escasa o nula entre los sistemas de citas iniciadas por el paciente y las coordinadas por el especialista. Los sistemas de citas iniciados por el paciente pueden tener poco o ningún efecto en la utilización de los servicios en términos del contacto con el servicio y hay incertidumbre sobre los costes en comparación con los sistemas de citas coordinados por los especialistas. Los sistemas de citas iniciadas por el paciente pueden tener poco o ningún efecto sobre los eventos adversos, como las recaídas o la satisfacción del paciente, en comparación con los sistemas de citas coordinadas por el especialista.
PICOs
Resumen en términos sencillos
Citas iniciadas por el paciente para personas con enfermedades crónicas tratadas en entornos hospitalarios ambulatorios
¿Cuál era el objetivo de esta revisión?
El objetivo era establecer si las citas iniciadas por el paciente (citas solicitadas por el paciente) para los pacientes con enfermedades crónicas y recurrentes es una mejor manera de gestionar la atención en los entornos hospitalarios ambulatorios que las citas estándar programadas por el especialista. Los investigadores encontraron 17 estudios para analizar.
Mensajes clave
En general, esta revisión aporta principalmente evidencia de calidad baja de que los sistemas de citas iniciadas por el paciente pueden tener poco o ningún impacto en la ansiedad, la depresión, la calidad de vida, los eventos adversos y la satisfacción del paciente. Este sistema también puede tener poco o ningún impacto en el contacto que tienen los pacientes con los servicios, pero el impacto en los costes del servicio no está claro. Hay cierta variabilidad en la forma en que los pacientes experimentan la atención.
¿Qué se estudia en la revisión?
Las citas de seguimiento programadas por un especialista del hospital se utilizan actualmente para los pacientes con enfermedades crónicas como (pero no limitadas a) la artritis reumatoide, la enfermedad intestinal inflamatoria y el cáncer. Este sistema suele dar lugar a que los pacientes falten a las citas, a que no puedan acceder a la atención cuando la necesitan y a la acumulación de pacientes que esperan ser diagnosticados o tratados por primera vez. La posibilidad de permitir a los pacientes acceder a consejos y citas médicas cuando lo necesiten (citas iniciadas por el paciente) podría ayudar a aliviar estos problemas y las citas a las que no se asiste. El sistema de atención iniciado por el paciente también puede dar lugar a una mayor satisfacción de los pacientes con un servicio más conveniente, y puede reducir los costes para los pacientes y los proveedores de servicios. En esta revisión se comparan los sistemas de citas iniciadas por el paciente que le permiten llamar por teléfono a una enfermera especializada en una línea de ayuda, donde puede discutir su consulta primero, y concertar una cita con el especialista reservado según sea necesario, versus los sistemas de citas coordinadas por el especialista. Se deseaba saber si estos sistemas son seguros para que los pacientes los utilicen, si los pacientes y los médicos los encuentran apropiados y satisfactorios para su uso, cómo pueden afectar la salud física y mental de los pacientes, y cómo pueden repercutir en los recursos y los costes de los servicios de salud.
¿Cuáles son los principales resultados de la revisión?
La revisión identificó 17 estudios. Los estudios abarcaron seis enfermedades: cáncer (siete estudios), artritis reumatoide (cuatro), enfermedades digestivas (tres), asma (uno), psoriasis (uno) y enfermedad pulmonar obstructiva coronaria (uno). La mayoría de los estudios se realizaron en el Reino Unido (nueve), pero también estuvieron representados Suecia (tres), Dinamarca (tres), Finlandia (uno) y los Países Bajos (uno). La media de edad de los 3854 pacientes oscilaba entre 41 y 76 años y la mayoría eran mujeres. Los resultados sugieren que los sistemas de citas iniciadas por el paciente pueden tener poca o ninguna diferencia en cuanto a la ansiedad, la depresión y la calidad de vida del paciente en comparación con el sistema de citas dirigido por el especialista (evidencia de calidad baja debido al riesgo alto de sesgo y a la variación de los resultados). Otros aspectos del estado de la enfermedad y la experiencia también revelan una diferencia escasa o nula entre los sistemas de citas iniciadas por el paciente y las coordinadas por el especialista. Los resultados de la utilización de los servicios (contacto con los servicios y el personal sanitario) sugieren que puede haber poca o ninguna diferencia en los contactos de servicio entre los servicios iniciados por el paciente y los coordinados por el especialista (la evidencia de calidad baja debido a los diferentes niveles de contacto informados en los estudios hacen que sea difícil de evaluar). No se sabe si la utilización de los servicios (costes de los servicios o del personal) se reduce en los grupos de consulta iniciados por los pacientes en comparación con los grupos de consulta coordinados por el especialista, ya que la calidad de esta evidencia es muy baja (debido al riesgo de sesgo y a la variabilidad de las monedas y los niveles de costes informados en los distintos estudios). Los resultados sugieren que puede haber poco o ningún impacto en los eventos adversos como las recaídas en algunas afecciones (cáncer o enfermedad intestinal inflamatoria) en el grupo de citas iniciadas por el paciente en comparación con el grupo de citas coordinadas por el especialista (evidencia de calidad baja debido a la inconsistencia y la precisión de los estudios en el informe y la medición de las recaídas). Los resultados sugieren que puede haber poca o ninguna repercusión en la satisfacción de los pacientes (evidencia de calidad baja, ya que cada estudio utilizó diferentes preguntas para reunir sus datos en diferentes momentos y en diferentes enfermedades). No todos los estudios informaron sobre sus fuentes de financiación, pero en los que se informaron, la mayoría fueron financiados por organizaciones sin fines de lucro. Un estudio (sobre el asma) fue financiado por una compañía farmacéutica.
¿Cuán actualizada está esta revisión?
Los autores de la revisión buscaron estudios que se habían publicado hasta marzo 2019.
Authors' conclusions
Summary of findings
| Patient‐initiated appointment systems compared with consultant‐led appointment systems for people with chronic conditions | ||||
| Patient or population: adults with a chronic or recurrent condition Settings: secondary care Intervention: patient‐initiated appointment systems Comparison: consultant‐led appointment systems | ||||
| Outcomes | Relative effect | No of participants | Certainty of the evidence | Comments |
| Patient outcomes: anxiety (HADS)a 12–24 months | OR 0.87 (0.68 to 1.12) | 1019 | ⊕⊕⊝⊝ | 7 studies reported anxiety, 2 could not be included in the analysis due to data available, but findings appeared consistent. Heterogeneity in health condition (cancer, psoriasis and RA), participants and follow‐up. |
| Patient outcomes: depression (HADS)a 9–24 months | OR 0.79 (0.51 to 1.23) | 1835 | ⊕⊕⊝⊝ | 7 studies reported depression, 1 could not be included in the analysis due to data available, but findings appeared consistent. Heterogeneity in health condition (cancer, RA, psoriasis and IBD), participants and follow‐up. |
| Patient outcomes: quality of life (different scales used across studies) 12–18 months | SMD 0.12 (0.00 to 0.25) | 1486 | ⊕⊕⊝⊝ | 12 studies reported quality of life, 5 could not be included due to data available, but where some data were available findings appeared consistent. Heterogeneity in health condition (asthma, IBD, psoriasis, RA and cancer) and tools used to measure quality of life. |
| Service utilisation (contacts) 12 months | Contact rate ratio ranged from 0.68 to 3.83 (median rate ratio 1.11, IQR 0.93 to 1.37) | 3348 (15) | ⊕⊕⊝⊝ | Studies included different elements of service contact in their data, and may expect contact to vary across health condition. |
| Service utilisation (costs) per patient per year | Studies reported the intervention group may have had higher or lower costs than the control group. | 2235 (8) | ⊕⊝⊝⊝ | Studies included different elements of service costs in their data, data were presented in different currencies and it may be expected that service costs vary across health condition. |
| Adverse events – relapse 12–60 months | MD –0.20 (–0.54 to 0.14) | 888 (3) | ⊕⊕⊝⊝ | 5 studies reported adverse events as relapse or recurrence, 2 could not be included in the analysis due to data available but findings appeared consistent. Each study used different questions/definitions to collect their data at difference time points and across different health conditions (inflammatory bowel disease and cancer). |
| Patient satisfaction 6–72 months | SMD 0.05 (–0.41 to 0.52) | 375 (2) | ⊕⊕⊝⊝ | 12 studies reported patient satisfaction, 10 could not be included in the analysis due to data available but findings appeared broadly consistent. Each study used different questions to collect their data at different time points and across different health conditions. No studies reported on clinician satisfaction. |
| CI: confidence interval; HADS: Hospital Anxiety and Depression Scale; IBD: inflammatory bowel disease; IQR: interquartile range; MD: mean difference; OR: odds ratio; RA: rheumatoid arthritis; SMD: standardised mean difference. | ||||
| GRADE Working Group grades of evidence | ||||
| aAnxiety and depression (HADS) was pooled as dichotomous data using the score of 10 as the cut point as identified in two of the papers, continuous data were converted to dichotomous data as no other data could be retrieved from the papers presenting dichotomous data for these outcomes. | ||||
Background
Chronic disease has a wide impact worldwide. The World Health Organization (WHO) reported that 63% of all deaths are from chronic disease (WHO 2011).
Missed hospital appointments are commonly reported in healthcare services around the world; for example, they cost the National Health Service (NHS) in the UK millions of pounds every year (HES 2010) (estimated GBP 225 million in 2012/2013; National Audit Office 2014). This unnecessary cost and a change in public expectations has brought into question the efficiencies of outpatient appointment scheduling systems, particularly in chronic conditions. Alternative appointment systems have been explored, for example, the Expert Patient Programme (in the UK) was specifically aimed at people with long‐term conditions (DoH 2001), early discharge with rapid access to care for patients with chronic obstructive pulmonary disease (COPD) in Ireland (Lawlor 2009), self‐management training for people with COPD in Canada (Bourbeau 2003), specialist nurse provision for people with inflammatory bowel disease (IBD) in Belgium (Coenen 2017). Alternative forms of appointment scheduling may not be appropriate for all healthcare areas, but those managing people who have long‐term or chronic conditions may see some benefits.
In 2002, the WHO published a report highlighting the need for a model of care that more readily meets the needs of people with chronic conditions (WHO 2002). The authors suggested that innovations that build on evidence‐based decision‐making, have a population and quality focus, and are flexible to the needs and demands of the patient population should do well in improving the management of chronic conditions.
More recently, both the NHS Long Term Plan (NHS 2019) and a report on outpatient care by the Royal College of Physicians (RCP 2018) have concluded that current, more traditional models of outpatient care are unstainable as outpatient appointments have doubled in the last 10 years. The reports highlight how outpatient care needs to be updated to be more flexible and cost effective to meet the needs of both patients and the NHS.
Description of the condition
Chronic conditions, defined as "diseases of long duration and generally slow progression" (WHO 2013), include rheumatoid arthritis, asthma, cancer and diabetes. People are faced with an opportunity to manage their condition but not cure it. Traditionally, people with these conditions are managed by the clinician through regularly scheduled appointments (e.g. one to four times per year) at outpatient clinics (Kirwan 1991; Probert 1993). These appointments often occur at a time when a person is feeling relatively well, with little action taken as a result. Conversely, when symptoms recur or suddenly worsen, it may be difficult to obtain immediate urgent appointments. Subsequently, people are often unable to get help during periods of exacerbated disease due to the number of regular follow‐up appointments also scheduled in the system. In some cases, conditions are managed in primary care and several studies reported the success of similar systems in this setting (Liu 2010; Robinson 2010; Rose 2011).
Description of the intervention
A patient‐initiated appointment system (PIAS) enables patients to make appointments at times when they feel they cannot manage their condition or where something has unexpectedly changed. The system does not completely replace the need for a scheduled follow‐up appointment, but the new system could reduce the number of follow‐up appointments, for example, to every one to two years.
How the intervention might work
The PIAS could free up clinician time, therefore making the service more flexible for urgent appointments, while still being able to deliver an acceptable standard of care. Using this type of service may also mean that the numbers of missed appointments are reduced (and therefore financial and resource costs are reduced too) as most patients will be attending because they need or choose to, and not just because the appointment is thought to be a requirement. There is a potential risk in situations where the patient fails to request an appointment at the time of relapse or escalation of their condition, and symptoms become worse, possibly critically. This risk is more likely when the appointment systems do not include a 'safety net' appointment system (an appointment which is scheduled by the clinician/researcher for a certain time to ensure the patient is using the system correctly) or when clinicians are unable to select appropriate patients for the PIAS pathway. In addition, there are elements of preventive health care or patient education that occur during a routine appointment that are not addressed during a patient‐initiated appointment. This risk can be minimised by incorporating an appropriate checklist into a 'safety net' appointment. Several studies have explored the effectiveness of PIAS in primary care (Liu 2010; Robinson 2010; Rose 2011). The results of some of these studies suggest that patient initiation of care results in improvements in satisfaction with a reduced cost for care delivery (Berkhof 2014; Robinson 2001).
Why it is important to do this review
There are several Cochrane Reviews that have considered alternative methods to improve attendance to appointments (Car 2012; Reda 2012); however, none of these reviews has looked at the impact of PIAS in secondary care. With the increasing focus on healthcare efficiencies and the increasing emphasis on enabling people to manage their own conditions (NHS 2019; Nuffield 2011; WHO 2002), determining the benefits and harms of PIAS in secondary care is crucial to understanding their worth for both healthcare systems and patients.
Objectives
To assess the effects of patient‐initiated appointment systems compared with consultant‐led appointment systems in people with chronic or recurrent conditions managed in secondary care.
Methods
Criteria for considering studies for this review
Types of studies
This review included randomised trials, including cluster‐randomised trials (published and unpublished and in any language) that compared PIAS with consultant‐led appointment systems. We did not include a broader range of study designs, as randomised trial designs are able to inform on causal relationships more reliably than non‐randomised designs; this may be particularly important with this topic with the progressive and fluctuating nature of long‐term conditions. Furthermore, we were confident there was sufficient randomised trial evidence on this topic to make evidence‐based recommendations.
Types of participants
We included adults (18 years of age or older) diagnosed with any chronic or recurrent condition, which was managed in secondary care (we did not search for named conditions, the conditions included in this review are those we found evidence for).
Types of interventions
A PIAS, established in the secondary care system, where appointments can be initiated by the patient whenever they require support from a relevant health professional to manage their ongoing condition. The appointments must not have been used for the purposes of diagnosis.
The comparator was a consultant‐led appointment system where patients were given a scheduled appointment to see the relevant health professional (usually a consultant) in secondary care, whether or not they required support. Other appointments were only made in case of an emergency or crisis.
The intervention described here correlates with the Effective Practice and Organisation of Care (EPOC) taxonomy of intervention co‐ordination of care and management of care processes subcategories: care pathways; disease management and PIAS (EPOC 2015).
Note: PIAS within a research trial may include a 'safety net' appointment where patients do receive a scheduled appointment with a consultant as part of the research project to make sure participation is safe.
Types of outcome measures
We included studies that reported one or more of the following outcomes. Studies were not excluded based on outcome.
Primary outcomes
-
Patient outcomes, such as anxiety, depression, quality of life and other health/disease status measurements (such as disease activity, disability, self‐efficacy, pain and other disease specific outcomes).
-
Service utilisation (contacts) – contact rates with the healthcare system including missed appointments measured through health records.
-
Service utilisation (costs).
-
Adverse events.
Secondary outcomes
-
Other service utilisation (other involved health service professionals including those outside of secondary care).
-
Patient satisfaction.
-
Clinician (consultant or specialist nurse) satisfaction.
-
Failures of the 'system' (e.g. how long people were on the PIAS pathway but not using it correctly before the clinical team needed to reinstigate regular follow‐up clinics).
Search methods for identification of studies
Electronic searches
We searched the Cochrane Database of Systematic Reviews (CDSR) and the Database of Abstracts of Reviews of Effects (DARE) for primary studies included in related systematic reviews. We searched the following databases on 13 March 2019:
-
Cochrane Central Register of Controlled Trials (CENTRAL) via Cochrane Library (2019, Issue 3);
-
MEDLINE via OvidSP (1946 to 12 March 2019);
-
Embase via OvidSP (1974 to 12 March 2019);
-
PsycINFO via OvidSP (1806 to 12 March 2019);
-
Health Technology Assessment (HTA) via CRD database (13 March 2019);
-
NHS Economic Evaluation Database (NHS EED) via CRD database (13 March 2019);
-
CINAHL via EBSCOhost (1937 to 13 March 2019);
-
Health Management Information Consortium (HMIC) via OvidSP (1979 to 13 March 2019);
-
ProQuest Dissertations & Theses Global via ProQuest (2015 to 13 March 2019).
Search strategies were comprised of keywords and controlled vocabulary terms. We applied no language or time limits. Search strategies for all databases can be found in Appendix 1.
Searching other resources
We searched for current/ongoing research on the following trial registries:
-
WHO International Clinical Trials Registry Platform (ICTRP; www.who.int/ictrp/en/; searched 15 March 2019);
-
ClinicalTrials.gov, US National Institutes of Health (NIH; clinicaltrials.gov/; searched 15 March 2019).
We also:
-
searched PROSPERO for registered systematic reviews (www.crd.york.ac.uk/prospero/; searched 15 March 2019);
-
reviewed reference lists of all included studies, relevant systematic reviews/primary studies;
-
contacted authors of relevant studies/reviews to clarify reported published information and to seek unpublished results/data;
-
contacted researchers with expertise relevant to the review topic/ EPOC interventions;
-
conducted a cited reference search for all included studies in Scopus (4 April 2019).
Data collection and analysis
Selection of studies
Two review authors (pairs of RW, JT‐C, RA, MR, VG, LA, SB) independently screened all references at the title and abstract stage using prespecified inclusion criteria. We resolved disagreements though discussion. At the full‐text stage, two review authors (pairs of RW, MR, VG, LA, SB) independently screened all references; we discussed disagreements and resolved them with a third review author (JT‐C, RA, MP, or KS). We recorded reasons for exclusion at the full‐text stage and recorded the selection process in sufficient detail to complete a PRISMA flow diagram (Liberati 2009).
Data extraction and management
Two review authors (pairs of RW, RA, MR, VG, LA, SB) independently completed data extraction for all included studies. The data extraction covered details, such as information on the participants, setting, interventions and comparisons, outcomes and study design, and we piloted this before use (Appendix 2; EPOC 2017a). We discussed and resolved discrepancies with a third review author (JTC, MP, or KS) where necessary. We contacted authors of included papers to request clarifications and provide additional data, and where possible, entered the data into Review Manager 5 (Review Manager 2014).
Assessment of risk of bias in included studies
Two review authors (pairs of RW, RA, MR, VG, LA, SB) independently assessed the risk of bias using the Cochrane 'Risk of bias' tool (Higgins 2011). Items in this tool include sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting and other bias (such as bias in study samples prerandomisation that are unaccounted for in analysis or inappropriate administration of the intervention itself). We assessed studies on the basis of having low, unclear or high risk of bias. We discussed and resolved discrepancies with a third review author (JTC, MP or KS) where necessary. We used the 'Risk of bias' tool to inform the analysis and interpretation of the results, but not as criteria for excluding studies from the review.
Measures of treatment effect
We expected a wide range of data (measures of treatment effect). We used mean difference (MD) when studies used the same scales or standardised mean difference (SMD) when studies used different scale with 95% confidence intervals (CIs) to quantify the difference between trial arms for continuous outcomes. We reported odds ratios (OR) with 95% CIs on two analyses where we combined continuous and dichotomous data (anxiety and depression) by calculating the ORs and inputting them into Review Manager 5 to pool the data (Review Manager 2014). Anxiety and depression (using the Hospital Anxiety and Depression Scale (HADS)) were pooled as dichotomous data using the score of 10 as the cut point as identified in two of the papers, continuous data were converted to dichotomous data as no other data could be retrieved from the papers presenting dichotomous data for these outcomes. There were three analyses (rheumatoid arthritis: disability, self‐efficacy and pain) where we used a fixed‐effect model, as the population and measures for those outcomes were comparable. We also used Review Manager 5 to calculate the median effect, interquartile range (IQR) and SMD for continuous data. We had planned to present the risk ratios (RR) of any dichotomous data, but were unable to do so due to the data available being mixed dichotomous and continuous data in any one analysis.
Unit of analysis issues
Where studies had several points of follow‐up, we selected the most common point of follow‐up across studies (or the point of follow‐up closest to the common point of follow‐up) for each outcome reported in a meta‐analysis, otherwise we reported the longest point of follow‐up. Where studies used a cluster‐randomised trial design, but did not allow for clustering in the analysis, we adjusted the standard error (SE) of estimate appropriately based on the mean cluster size (i.e. number of participants per cluster) and assumed plausible value of the intracluster correlation coefficient (ICC) (e.g. if the cluster was the hospital, then we assumed values for the ICC between 0 and 0.3 for patient outcomes).
Dealing with missing data
We contacted authors for clarification and to retrieve more information and data that we could more readily combine.
Assessment of heterogeneity
We quantified heterogeneity across estimates using the I2 statistic. We expected that the type of condition would add to the heterogeneity of the results as well as the content/structure of the PIAS.
Assessment of reporting biases
We did not examine the likelihood of publication bias due to the low number of studies included in the meta‐analyses. Had it been appropriate, we would have used Egger's regression test for asymmetry (metabias command in Stata) and funnel plots.
Data synthesis
Our analyses combined estimates of the effect of PIAS on the outcomes across studies. We used random‐effects meta‐analyses to pool the estimates using the DerSimonian‐Laird method (DerSimonian 1986). We performed the meta‐analysis using Review Manager 5 (Review Manager 2014). Where studies used different continuous measures to quantify a given outcome (e.g. quality of life – 36‐item Short Form (SF‐36), 12‐item Short Form (SF‐12), Inflammatory Bowel Disease Questionnaire (IBDQ), St George's Respiratory Questionnaire (SGRQ), etc.), we pooled the SMD. Where meta‐analysis was not possible, we presented the data in tables to best describe and summarise them. In particular, we could not pool the data on service utilisation due to the varying descriptions of the services used and variation in currency. Instead, the service utilisation tables aim to provide a summary of the results while taking a detailed account of the limitations and descriptions of the data. We calculated the person‐years by multiplying the number of people in the control/intervention arm by the number of years at follow‐up. We calculated the rate ratio (as in Analysis 1.9) by dividing the sum of the (intervention person‐years divided by the intervention total contacts) by the sum of the (control person‐years divided by the control total contacts) = rate ratio (e.g. Brown 2002; (28/6 = 4.67)/(26/4 = 6.5) = 0.72). We also calculated the median of the contact rate ratio to summarise the intervention effects on service utilisation in terms of contact with professionals, but this was not appropriate for the analysis of costs, due to the varying currencies used.
Summary of findings
We created a 'Summary of findings' table for the comparison, PIAS versus consultant‐led appointments, and included the most important outcomes in order to draw conclusions about the certainty of the evidence within the text of the review. We included the primary outcomes of anxiety, depression, quality of life, service utilisation (contacts and costs), and adverse events (relapse) and the secondary outcome of patient satisfaction, as these are all important for decision‐making (clinician satisfaction is reported as a comment).
Two review authors (RW, RA) independently assessed the certainty of the evidence (high, moderate, low, and very low) using the five GRADE considerations (risk of bias, consistency of effect, imprecision, indirectness and publication bias) (Guyatt 2008). We used GRADEpro to develop the GRADE table (GRADEpro GDT), and resolved disagreements on certainty ratings by discussion and provided justification for decisions to downgrade or upgrade the ratings using footnotes in the table to aid readers' understanding of the review. We used plain language statements to report these findings in the review (EPOC 2017b).
Where it was not possible to meta‐analyse the data, we summarised the results in the text.
Subgroup analysis and investigation of heterogeneity
We conducted analyses to investigate whether pooled estimates differed across subgroups such as health condition. We conducted a subgroup analysis by health condition for the outcomes that had three or more studies with useable data (anxiety, depression and quality of life). No other outcomes were eligible for subgroup analysis. We were unable to conduct a subgroup analysis based on age.
Sensitivity analysis
Sensitivity analysis would have been conducted in the case of missing data particularly in regards to where length of follow‐up, attrition rates and units of analysis were missing or unclear. Sensitivity analyses were conducted to check the robustness of the results to potential risk of bias. Studies judged to be of high risk of bias for sequence generation or allocation concealment (or both) were excluded in secondary analyses.
Results
Description of studies
For a full description of included studies, see Characteristics of included studies table, and for excluded studies, see Characteristics of excluded studies table.
Results of the search
Figure 1 presents the search of included studies from the literature searches. We screened 18,917 titles and abstracts for inclusion, of these, we retrieved 50 full‐text articles and finally included 26 references in the review (Figure 1). The 26 references referred to 17 studies with multiple references to four studies, reflecting differing periods of follow‐up and cost‐effectiveness analyses.
Study flow diagram.
Included studies
Study characteristics
We included 17 randomised trials (seven single site and 10 multisite including one cluster randomised) in the review. Nine studies were set in the UK, three in Sweden, three in Denmark, one in Finland and one in the Netherlands.
Population
The populations represented in these studies had a mean age ranging from 41 to 76 years. The studies covered six broad health conditions: cancer (breast: Brown 2002; Gulliford 1997; Kirshbaum 2017; Koinberg 2004; Sheppard 2009; prostate: Helgesen 2000; endometrial: Jeppesen 2018); rheumatoid arthritis (Fredriksson 2016; Goodwin 2016; Hewlett 2000; Primdahl 2012); digestive conditions (IBD: Kennedy 2003; Williams 2000; ulcerative colitis: Robinson 2001); asthma (Lahdensuo 1996); psoriasis (Khoury 2018); and coronary obstructive pulmonary disease (Berkhof 2014). The cancer studies had either men or women in their populations but the other health conditions used mixed populations, although women dominated in almost every mixed study (nine reported 52% to 77% with one study at 26% and one at 36%), which may reflect an increased incidence of these conditions within women. The sample sizes ranged from 61 to 635 participants (total 3854), and length of follow‐up ranged from 10 to 72 months.
Interventions
The interventions all involved patients (or the patients' GP) being able to access disease‐specific advice and care by telephoning their hospital. In most interventions, the patient contacted a specialist nurse who was able to give immediate advice, ask a doctor for further advice to give to the patient and arrange an outpatient appointment if necessary. When enrolling into the intervention, some patients were given personal education by trained nurses (Kirshbaum 2017; Lahdensuo 1996), or doctors (Jeppesen 2018); some were given general written information/guidance (Brown 2002; Hewlett 2000; Kirshbaum 2017); and some received a personalised guided self‐management plan (Kennedy 2003; Robinson 2001). The remaining studies did not report this level of detail, though some written materials may be assumed in the process of enrolling within the healthcare context. Most studies ran a safety net procedure for the duration of the study in the form of an annual hospital review (Berkhof 2014; Goodwin 2016; Khoury 2018; Primdahl 2012); an annual mammogram (Brown 2002; Gulliford 1997; Koinberg 2004; Sheppard 2009); or another review or contact at three months (Hewlett 2000; Fredriksson 2016), four months (Lahdensuo 1996), six months (Helgesen 2000), or 24 months (Williams 2000). Only four studies reported no safety net procedures (Jeppesen 2018; Kennedy 2003; Kirshbaum 2017; Robinson 2001).
Comparisons
The comparator for each study was a consultant‐led clinic appointment with varying appointment schedules across the studies. Some studies reported that those in the comparator condition could receive an emergency appointment if necessary (Fredriksson 2016; Gulliford 1997; Helgesen 2000; Hewlett 2000; Khoury 2018; Primdahl 2012; Williams 2000), but for the remainder the availability of this service was unclear. For full details, see the Characteristics of included studies tables.
Outcomes
The review included the following outcomes (studies were not excluded based on outcome).
Primary outcomes
Sixteen studies reported on patient outcomes, such as anxiety (Brown 2002; Helgesen 2000; Hewlett 2000; Kennedy 2003; Khoury 2018; Kirshbaum 2017; Koinberg 2004); depression (Brown 2002; Helgesen 2000; Hewlett 2000; Kennedy 2003; Khoury 2018; Kirshbaum 2017; Koinberg 2004); quality of life (Berkhof 2014; Brown 2002; Gulliford 1997; Hewlett 2000; Kennedy 2003; Khoury 2018; Kirshbaum 2017; Lahdensuo 1996; Primdahl 2012; Robinson 2001; Sheppard 2009; Williams 2000); and other health/disease status measurements such as disease activity (Hewlett 2000; Primdahl 2012), disability (Hewlett 2000; Primdahl 2012), self‐efficacy (Hewlett 2000; Primdahl 2012), pain (Hewlett 2000; Primdahl 2012), and other disease‐specific outcomes (Berkhof 2014; Brown 2002; Fredriksson 2016; Gulliford 1997; Helgesen 2000; Hewlett 2000; Jeppesen 2018; Kennedy 2003; Khoury 2018; Koinberg 2004; Lahdensuo 1996; Primdahl 2012; Robinson 2001; Sheppard 2009; Williams 2000).
Fifteen studies reported on service utilisation, such as contact rates with the healthcare system (Brown 2002; Fredriksson 2016; Goodwin 2016; Gulliford 1997; Helgesen 2000; Hewlett 2000; Jeppesen 2018; Kennedy 2003; Khoury 2018; Koinberg 2004; Lahdensuo 1996; Primdahl 2012; Robinson 2001; Sheppard 2009; Williams 2000), including missed appointments measured through health records (Kennedy 2003; Khoury 2018), and costs (Berkhof 2014; Goodwin 2016; Helgesen 2000; Hewlett 2000; Kennedy 2003; Koinberg 2004; Primdahl 2012; Williams 2000).
Four studies reported on adverse events (Berkhof 2014; Kennedy 2003; Robinson 2001; Sheppard 2009).
Secondary outcomes
No studies reported on other service utilisation other than that which was captured in service utilisation contacts.
Twelve studies reported on patient satisfaction in some way (Brown 2002; Fredriksson 2016; Goodwin 2016; Gulliford 1997; Helgesen 2000; Hewlett 2000; Kennedy 2003; Khoury 2018; Koinberg 2004; Primdahl 2012; Robinson 2001; Williams 2000).
No studies reported on clinician (consultant or specialist nurse) satisfaction.
One study reported on failures of the 'system' such as 'safety net' failure (Hewlett 2000).
Funding source
Of the 17 studies, two did not report their funding source (Brown 2002; Kirshbaum 2017). Of the 15 studies that did report their funding source, 14 were funded by not‐for‐profit organisations such as local government or health groups or health charity organisations, one study was funded by an industry organisation.
Excluded studies
We excluded 23 studies at the full‐text screening stage: 12 because they were not a randomised trial, 10 were the wrong intervention and one had the wrong population. See Characteristics of excluded studies table for full details.
Ongoing studies
We identified one ongoing study, which commenced in 2015 and the findings are yet to be reported. See Characteristics of ongoing studies table for full details.
Risk of bias in included studies
See the 'Risk of bias' tables within the Characteristics of included studies table, Figure 2 for a summary of judgements about each risk of bias item and Figure 3 for a graph of risk of bias items presented as percentages across all included studies.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Allocation
Random sequence generation
Random sequence generation was low risk in 14 studies with methods such as computer‐generated randomisation, numbers pulled randomly from a hat or random number tables (Berkhof 2014; Brown 2002; Fredriksson 2016; Goodwin 2016; Hewlett 2000; Jeppesen 2018; Kennedy 2003; Khoury 2018; Koinberg 2004; Lahdensuo 1996; Primdahl 2012; Robinson 2001; Sheppard 2009; Williams 2000). There were three studies at unclear risk due to the lack of reporting on the random sequence generation (Gulliford 1997; Helgesen 2000; Kirshbaum 2017). No studies were at high risk.
Allocation concealment
Allocation concealment was at low risk in 11 studies where allocations were concealed in sealed envelopes or dealt with by a third party not directly involved in the study, or both (Fredriksson 2016; Helgesen 2000; Hewlett 2000; Jeppesen 2018; Kennedy 2003; Koinberg 2004; Lahdensuo 1996; Primdahl 2012; Robinson 2001; Sheppard 2009; Williams 2000). There were six studies at unclear risk due to the lack of reporting on the allocation concealment (Berkhof 2014; Brown 2002; Goodwin 2016; Gulliford 1997; Khoury 2018; Kirshbaum 2017). No studies were at high risk.
Blinding
Blinding of participants and personnel
No studies were at low risk of blinding of participants and personnel as, due to the nature of the intervention, participants would know which arm of the trial they were in. There was one study at unclear risk due to the lack of reporting on the this area of blinding (Gulliford 1997). Sixteen studies were at high risk due to the obvious nature of the intervention (whether they were given allocated appointments by the consultant or were able to make their own appointments) (Berkhof 2014; Brown 2002; Fredriksson 2016; Goodwin 2016; Helgesen 2000; Hewlett 2000; Jeppesen 2018; Kennedy 2003; Khoury 2018; Kirshbaum 2017; Koinberg 2004; Lahdensuo 1996; Primdahl 2012; Robinson 2001; Sheppard 2009; Williams 2000).
Blinding of outcome assessment
No studies were at low risk of blinding of outcome assessment as many of the outcomes were self‐reported and this may have influenced the interpretation of the results around these outcomes. There were three studies at unclear risk due to the lack of reporting (Fredriksson 2016; Lahdensuo 1996; Robinson 2001). Fourteen studies were at high risk due to the obvious nature of the intervention (whether they were given allocated appointments by the consultant or were able to make their own appointments); this is largely due to the self‐reporting nature of the outcomes of interest by the patients or because the blinding of other outcome assessors was not reported or unclear (Berkhof 2014; Brown 2002; Goodwin 2016; Gulliford 1997; Helgesen 2000; Hewlett 2000; Jeppesen 2018; Kennedy 2003; Khoury 2018; Kirshbaum 2017; Koinberg 2004; Primdahl 2012; Sheppard 2009; Williams 2000).
Incomplete outcome data
Incomplete outcome data was at low risk for nine studies (Berkhof 2014; Fredriksson 2016; Hewlett 2000; Jeppesen 2018; Kennedy 2003; Khoury 2018; Primdahl 2012; Robinson 2001; Williams 2000). Three studies were at unclear risk due to the poor reporting of this outcome in those studies (Gulliford 1997; Kirshbaum 2017; Koinberg 2004). Five studies were at high risk of incomplete outcome data as it was sometimes unclear where the attrition lay or why data was lost or where lost data was not accounted for appropriately in the analysis (Brown 2002; Goodwin 2016; Helgesen 2000; Lahdensuo 1996; Sheppard 2009).
Selective reporting
Selective reporting was at low risk in 11 studies where all the outcomes appeared to be reported for the necessary time points (Berkhof 2014; Brown 2002; Fredriksson 2016; Goodwin 2016; Helgesen 2000; Kennedy 2003; Kirshbaum 2017; Lahdensuo 1996; Primdahl 2012; Sheppard 2009; Williams 2000). Six studies were at high risk, one as it did not report the results of all participants but only those who completed all the assessments and there are no details about those who did not complete all the assessments or why (Hewlett 2000), and five that did not fully report all the outcomes they specified (Gulliford 1997; Jeppesen 2018; Khoury 2018; Koinberg 2004; Robinson 2001).
Other potential sources of bias
In other potential sources of bias, we looked at anything else that could have affected the study results that could not be accounted for above such as problems with study samples prerandomisation or problems with inappropriate administration of the intervention itself. In this area, 11 studies were at low risk where there was no evidence of further bias (Berkhof 2014; Fredriksson 2016; Goodwin 2016; Gulliford 1997; Helgesen 2000; Hewlett 2000; Kennedy 2003; Khoury 2018; Lahdensuo 1996; Primdahl 2012; Robinson 2001). Two studies were at unclear risk where there was a difference in the administration of the intervention but the impact on the results was difficult to establish (Kirshbaum 2017; Williams 2000). Four studies were at high risk of other bias due to these studies having unaccounted for instances of participants declining to participate in part due to not wanting to be randomised – this could be showing some type of characteristic difference in the participants which the study was unable to explore or account for (Brown 2002; Jeppesen 2018; Koinberg 2004; Sheppard 2009). One study also reported difficulties with administering the intervention itself but this was not further described (Koinberg 2004).
Effects of interventions
We were able to perform meta‐analyses across health conditions on three outcomes (anxiety, depression and quality of life). We were also able to perform a series of small meta‐analyses within the studies on rheumatoid arthritis for five outcomes (disease activity, disability, self‐efficacy, pain and patient satisfaction) and within IBD for the outcome of relapses only. The remaining outcomes (primary and secondary) are presented in other data comparison tables. In general, collating of data from studies both within and across health conditions was limited and complex in nature due to the range of tools used to measure the same or similar concepts, the differing time points at which data were collected, the differing units data were presented in, and the differing analyses and data that were reported. These factors, alongside the risk of bias identified across studies, have led us to report the certainty of the evidence (using GRADE) in this review as ranging mainly from low to very low, meaning we are quite uncertain about the estimate of effect on all the outcomes reported (Guyatt 2008).
See summary of findings Table for the main comparison.
Patient‐initiated appointment systems versus consultant‐led appointment systems
Primary outcomes
The studies for each of the meta‐analyses were selected based on the outcome reported, the closest common time point of data available (usually 12 to 24 months) and the availability of the data in the correct format.
Patient outcomes
Anxiety
Seven studies (1780 participants) reported anxiety – four were studies that recruited people with cancer, one recruited people with psoriasis, one recruited people with IBD and one recruited people with rheumatoid arthritis, all used the Hospital Anxiety and Depression Scale (HADS) as their measure. In the HADS scale, lower scores indicate fewer symptoms of anxiety or depression (range 0 to 21). The meta‐analysis was conducted using five of these studies (1019 participants) (Helgesen 2000 at 12 months; Hewlett 2000 at 24 months; Khoury 2018 at 12 months; Kirshbaum 2017 at 12 months; Koinberg 2004 at 18 months). The results suggest PIAS may make little or no difference to patient anxiety, with all studies reporting considerably wide CIs (OR 0.87, 95% CI 0.68 to 1.12; 5 studies, 1019 participants; I2 = 0%; low‐certainty evidence; Analysis 1.1; Figure 4). The GRADE rating of the evidence for anxiety was downgraded twice to low certainty due to serious risk of bias and imprecision. Sensitivity analysis including only studies at low risk of bias for random sequence generation, allocation concealment or for length of follow‐up did not change the results.
Forest plot of comparison: 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, outcome: 1.1 Patient outcomes: anxiety (HADS lower score = better 0–21) (12–24 months).
Brown 2002 and Kennedy 2003 (2 studies, 761 participants) also reported on anxiety using the same scale; however, we were unable to retrieve the necessary data to include them in the meta‐analysis (Kennedy 2003 could not provide separate data for the anxiety outcome but reported no significant difference in the text – no data available). Brown 2002 also reported no significant difference in anxiety scores between the PIAS group and the consultant‐led group (median score: 4 with PIAS versus 6.5 with consultant‐led; P = 0.069 at 12 months).
Depression
Seven studies (1896 participants) reported on depression – four were studies that recruited people with cancer, one recruited people with rheumatoid arthritis, one recruited people with psoriasis and one recruited people with IBD, with all using HADS as their measure (0 to 21 where lower scores indicate fewer symptoms). The meta‐analysis was conducted using six of these studies (1835 participants) (Helgesen 2000 at 12 months; Hewlett 2000 at 24 months; Kennedy 2003 at nine months; Khoury 2018 at 12 months; Kirshbaum 2017 at 12 months; Koinberg 2004 at 18 months). These results suggest PIAS may make little or no difference to patient depression (OR 0.79, 95% CI 0.51 to 1.23; 6 studies, 1175 participants; I2 = 33%; low‐certainty evidence; Analysis 1.2; Figure 5). The GRADE rating of the evidence for depression was downgraded twice to low certainty due to high risk of bias and imprecision. Sensitivity analysis including only studies at low risk of bias for random sequence generation, allocation concealment or length of follow‐up did not change the results.
Forest plot of comparison: 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, outcome: 1.2 Patient outcomes: depression (HADS lower score = better 0–21) (9–24 months).
Brown 2002 (61 participants) also reported on depression at 12 months using the same scale. We were unable to retrieve the necessary data to include the study in the meta‐analysis but they reported no significant difference in the text between groups (median score: 1 with PIAS versus 2 with consultant‐led; P = 0.232).
Quality of life
Twelve studies (2232 participants) reported quality of life – four recruited people with cancer, three recruited people with IBD, one recruited people with psoriasis, two recruited people with rheumatoid arthritis, one recruited people with COPD and one recruited people with asthma, each using a different measure of quality of life, with the exception of two IBD studies. For the quality of life measures, higher scores indicate better quality of life. The meta‐analysis was conducted using seven of these studies (1486 participants) (Sheppard 2009 Functional Assessment of Cancer Therapy (FACT‐B) plus endocrine subscale at 18 months; Lahdensuo 1996 Part 3 St George's Asthma Questionnaire at 12 months; Robinson 2001 IBDQ at 14 months; Primdahl 2012 EQ‐5D at 12 months; Kennedy 2003 IBDQ at 12 months; Khoury 2018 Dermatology Life Quality Index (DLQI) at 12 months; Kirshbaum 2017 European Organisation for Research and Treatment of Cancer EORTC QLQ‐C30 global scale at 12 months). The results suggest that PIAS may make little or no difference to patient quality of life (SMD 0.12, 95% CI –0.00 to 0.25; 7 studies, 1486 participants; I2 = 29%; low‐certainty evidence; Analysis 1.3; Figure 6). The GRADE rating of the evidence for quality of life was downgraded two levels to low due to serious risk of bias and imprecision. Sensitivity analysis including only studies at low risk of bias for random sequence generation, allocation concealment or length of follow‐up did not change the results.
Forest plot of comparison: 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, outcome: 1.3 Patient outcomes: quality of life (12–18 months).
Five studies (746 participants) also reported quality of life using differing measures and differing time points, and found similar results in that there appears to be little evidence of differences between the PIAS and consultant‐led appointment groups. Brown 2002 reported no significant differences between the PIAS and the consultant‐led group using the EORTC QLQ‐C30 global scale at 12 months (median score: 11 with PIAS versus 10 with consultant‐led; no significant difference). Gulliford 1997 did not report results. Hewlett 2000 measured quality of life using the SF‐36 at 36 months but presented no data. Williams 2000 reported no significant differences on the UK IBDQ or the SF‐36 at 12 months (no total scores provided but all differences in subscales were not significant). Berkhof 2014 reported no significant difference at 24 months on the SF‐36 questionnaire (mean change from baseline: –5.2 with PIAS versus –4.8 with consultant‐led; 95% CI –6.7 to 5.8; P = 0.89).
Other health/disease status measurements
Fifteen studies reported on other health/disease status measurements across all the health conditions; however, data from only two studies were able to be combined in a meta‐analysis. These were in studies that recruited people with rheumatoid arthritis (Hewlett 2000; Primdahl 2012). For two of these studies, we ran small meta‐analyses on outcomes such as disease activity (Analysis 1.4), self‐efficacy (Analysis 1.5), pain (Analysis 1.6) and disability (Analysis 1.7) related to this particular condition. The studies measured disease activity using a visual analogue scale (VAS; 1 mm to 100 mm) and the Disease Activity Scale (DAS) where lower scores are better (a score of 5.1 or more implies active disease, less than 3.2 low disease activity, and less than 2.6 remission; at 24 months). Disease activity is probably slightly reduced in the PIAS group compared with the consultant‐led group (SMD –0.21, 95% CI –0.41 to –0.01; 2 studies, 375 participants; I2 = 0%; moderate‐certainty evidence), but this was unlikely to be clinically significant where for the DAS a persistent reduction in score of 1.2 from the original score is required to allow continuation of treatment (NRAS 2020). Fredriksson 2016 also used DAS to measure disease activity and found no significant differences between the PIAS group and the consultant‐led group at 18 months (mean change from baseline: 0.24 with PIAS versus 0.59 with consultant‐led, 95% CI –0.01 to 0.91; P = 0.055).
The results regarding self‐efficacy measured using the Arthritis Self‐Efficacy Scales (ASES; higher scores are better – range 0 to 100) suggest self‐efficacy is probably slightly increased in PIAS group compared with the consultant‐led group at 24 months (MD 4.95, 95% CI 0.57 to 9.34; 2 studies, 375 participants; I2 = 20%; moderate‐certainty evidence), but that the difference may not be clinically significant (minimal clinically important difference 8.5 on ASES; Park 2019).
The results regarding pain (measured using VAS at 12 months where lower scores are better 1 cm to 10 cm), suggest PIAS probably slightly reduces pain in comparison to consultant‐led appointment groups (MD –0.71, 95% CI –1.17 to –0.26; 2 studies, 375 participants; I2 = 46%; moderate‐certainty evidence), but that the difference may not be clinically significant (minimal clinically important difference –1.19 cm Kitchen 2013). Fredriksson 2016 also measured pain using VAS and found no differences (median change from baseline: –3 (IQR –13 to 5) with PIAS versus –4 (IQR –18 to 8) with consultant‐led; 1 study, 131 participants; P = 0.75).
The results regarding disability (measured using Health Assessment Questionnaire (HAQ); score range 0 to 3) where lower scores are better) suggest there is probably little or no difference between PIAS and consultant‐led appointment groups on disability scores at 12 months (MD –0.06, 95% CI –0.19 to 0.06; 2 studies, 375 participants; I2 = 0%; moderate‐certainty evidence). The minimal clinically important difference on the HAQ is 0.22 (Kitchen 2013). Fredriksson 2016 also measured disability using the HAQ and found no significant differences between the PIAS and the consultant‐led group in the MD since baseline at 18 months (median change from baseline: –0.12 (IQR –0.25 to 0.13) with PIAS versus –0.11 (IQR –0.25 to 0.12) with consultant‐led; P = 0.91). The GRADE of evidence for these studies was downgraded once to moderate due to serious risk of bias in these studies.
The remaining measures of disease‐specific and other outcomes of interest reported by individual studies can be found in Analysis 1.14; Analysis 1.15; Analysis 1.16; Analysis 1.17; Analysis 1.18; and Analysis 1.19). A brief summary of these outcomes follows.
Berkhof 2014 described three additional COPD and respiratory symptom outcomes which reported no differences from baseline to 24‐month follow‐up for the PIAS or the consultant‐led groups, and no difference in the time it took to experience an exacerbation of the condition between the PIAS and consultant‐led groups (Analysis 1.14; P = 0.40).
Seven additional outcomes were reported for cancer (Analysis 1.15). Gulliford 1997 reported that the proportion of participants reassured with the service at 16 months was slightly higher for the consultant‐led group (94%) than the PIAS group (88%); Helgesen 2000 found that the proportion of participants reporting the service inaccessible by telephone at 36 months was slightly higher for the PIAS group than the consultant‐led group (RR 1.18, 95% CI 0.46 to 3.03); Jeppesen 2018 reported fear of cancer recurrence at 10 months was slightly higher in the PIAS group in comparison to the consultant‐led group (P = 0.02), but also that clinical fear of recurrence at 10 months showed no difference between the groups (P = 0.89); Koinberg 2004 reported on the accessibility of the service by telephone at 60 months but suggested contrasting evidence to that of Helgesen 2000 by finding little or no difference between the groups; Sheppard 2009 reported no substantial differences between the PIAS and the consultant‐led groups on fear of recurrence at 18 months (P = 0.066) or the proportion of people feeling isolated (P = 0.245).
Five studies reported additional outcomes for IBD (Analysis 1.16): Kennedy 2003 found participants in the PIAS group compared to the consultant‐led group were more likely to feel enabled after their initial consultation (MD 0.026, 95% CI 0.12 to 1.68) and more likely to make appointments for themselves (MD 2.70, 95% CI 1.63 to 4.46; P < 0.001), with little difference in the delay before starting treatment between the two groups (MD 0.95, 95% CI –0.52 to 0.49). Conversely, Robinson 2001 reported that the time to symptom treatment was substantially longer for the consultant‐led group (mean 49.6 hours) in comparison to the PIAS group (mean 14.8 hours; P < 0.0001), they also reported that the cost to the patient at 14 months was considerably less in the PIAS group (mean GBP 0.86) compared to the consultant‐led group (mean GBP 8.92; P < 0.0001).
Khoury 2018 reported on four additional outcomes for psoriasis (Analysis 1.17). At 12 months, there was little difference between PIAS and consultant‐led groups in missed laboratory tests (MD –0.10, 95% CI –0.20 to 0.18) and disease activity (MD –0.24, 95% CI –0.84 to 0.36). Whereas missed medication collection at 12 months happened more frequently in the PIAS group compared to the consultant‐led group (MD –0.10, 95% CI –0.024 to –0.04) and changes in medication at 12 months happened more frequently in the consultant‐led group (20.6%) in comparison to the PIAS group (6.8%; P < 0.001).
Two studies reported 10 additional outcomes for rheumatoid arthritis (Analysis 1.18). There was no substantial difference for confidence in care, general health, number of swollen joints, number of tender joints and C‐reactive protein levels between baseline and follow‐up at 18 months for the PIAS and consultant‐led groups; however, the number of cases where treatment was changed (P = 0.0002) and the erythrocyte sedimentation rate differed substantially (P = 0.03) between the two groups (Fredriksson 2016). Hewlett 2000 reported that participants in the PIAS group were more likely to be confident with their interventions than those in the consultant‐led group (P < 0.01), but found little difference in the change from baseline in terms of feelings of helplessness in the two groups.
For asthma, Lahdensuo 1996 reported on lung function and found no differences between the PIAS and consultant‐led groups (Analysis 1.19). This study also reported a difference in the number of days off work with participants in the PIAS group taking fewer days off work (mean 2.8 SD 0 to 62) than those in the consultant‐led group (mean 4.8, SD 0 to 27; P = 0.02).
The data available for these outcomes were variable and incomplete, which makes it difficult to draw these results together and interpret them in a way that might be useful for the reader.
Service utilisation (contacts)
Sixteen studies reported service utilisation in terms of appointments or telephone calls (contacts) with various healthcare professionals. The results suggest there may be little or no difference in service utilisation (including hospital appointments, referrals, telephone calls, GP visits, nurse visits, other outpatient contacts, e.g. occupational therapist, physiotherapist, podiatrist, orthopaedic surgeon, orthotics, and tests) in terms of contacts with health professionals/services between the PIAS and consultant‐led appointment groups at 12 to 60 months (range 0.68 to 3.83 across the studies; median rate ratio 1.11, IQR 0.93 to 1.37; 15 studies, 3348 participants; low‐certainty evidence). Each study included different elements of service contact in their data, and there may be some expectation that contact might vary between health conditions (see Analysis 1.9 for details). We downgraded the evidence twice to low certainty due to serious risk of bias and imprecision in these studies. The rate ratio does not account for clusters of events/contacts within the same person. Berkhof 2014 (100 participants) could not be included in Analysis 1.9 as we were unable to obtain the appropriate data but their results suggest the GP was visited less frequently in the PIAS compared to the consultant‐led group (median number of visits: 4 (range 0 to 32) with PIAS versus 5 ( range 0 to 20) with consultant‐led; P = 0.01); the pulmonologist was visited similarly in each group (median number of visits: 3 (range 0 to 17) with PIAS versus 3 (range 0 to 13) with consultant‐led; P = 0.82); and the pulmonary nurse practitioner was visited more often in the PIAS compared to the consultant‐led group (median number of visits: 1 (range 0 to 14) with PIAS versus 0 (range 0 to 4) with consultant‐led; P = 0.003).
Two studies (599 participants) report on missed appointments – one recruited people with IBD (Kennedy 2003), and one recruited people with psoriasis (Khoury 2018). The results suggest missed appointments may be slightly reduced in the PIAS group compared with the consultant‐led group (MD –0.27, 95% CI –0.42 to –0.12; 2 studies, 599 participants; I2 = 0%; low‐certainty evidence; Analysis 1.11). The results were downgraded due to serious risk of bias and imprecision (wide CIs).
Service utilisation (costs)
Eight studies reported service utilisation in terms of financial costs. We do not know if service utilisation (costs) are reduced in the PIAS compared to the consultant‐led appointment groups as there is very low‐certainty evidence (8 studies, 2235 participants). Although six of the eight studies reported reduced costs per patient‐year in the PIAS group compared with the consultant‐led group, the effect is not certain as each study included different elements of service costs in their data, data were presented in different currencies, and there may be some expectation that service costs will vary between health conditions and locations (see Analysis 1.10 for details). The evidence was downgraded three times due to serious risk of bias (due to incomplete data), serious indirectness (as different currencies and levels of costs reported across studies) and serious imprecision.
Adverse events
Seven studies (1819 participants) reported adverse events. Five studies (1202 participants) reported relapse or exacerbations or recurrences of the condition (Berkhof 2014; Kennedy 2003; Koinberg 2004; Robinson 2001; Sheppard 2009). We were able to meta‐analyse data from three studies on relapse occurrence (Kennedy 2003; Koinberg 2004; Robinson 2001). The results suggest that the intervention may make little or no difference to relapses between the PIAS group and the consultant‐led group at 12 to 60 months (MD –0.20, 95% CI –0.54 to 0.14; 3 studies, 888 participants; I2 = 67%; low‐certainty evidence; Analysis 1.12).
Two studies (314 participants) reported relapse within cancer and COPD. Sheppard 2009 reported no significant differences in the percentage of patients with cancer recurrence between the PIAS group and the consultant‐led group at 18 months (rate of recurrence was 4% in both groups). Berkhof 2014 reported no significant difference in the number of patients with at least one exacerbation of COPD between the PIAS group and the consultant‐led group at 24 months (38.8% with PIAS versus 31.4% with consultant‐led; P = 0.44). The GRADE of evidence for these studies was downgraded twice to low certainty due to serious risk of bias and imprecision (wide CIs) in these studies.
Two studies (315 participants) reported other adverse events and use of other drugs to assist the condition (Lahdensuo 1996; Primdahl 2012). Only one study (122 participants) reported other adverse events such as use of antibiotics and use of prednisolone (Lahdensuo 1996). This study found that people with asthma within the PIAS group were significantly less likely to report use of other drugs at 12 months than those in consultant‐led appointment group (adverse events: 0.6 with PIAS versus 2.1 with consultant‐led; 95% CI 0.9 to 2.1; P < 0.0001). Primdahl 2012 report no significant differences in the number of adverse effect alerts between people with rheumatoid arthritis in the PIAS group (mean 1.19) and the consultant‐led group (mean 1.61) at 24 months.
Secondary outcomes
Other service utilisation
None of the studies reported this separately from the 'Service utilisation (contacts)' and 'Service utilisation (costs)' recorded above.
Patient satisfaction
Twelve studies (2950 participants) reported patient satisfaction with (and acceptability of) the intervention (see Analysis 1.13 and Analysis 1.8 where higher scores indicate better satisfaction). The results were unclear about any differences the intervention may make to patient satisfaction as the certainty of the evidence was low. Of the four studies that recruited people with rheumatoid arthritis (796 patients) (Fredriksson 2016; Goodwin 2016; Hewlett 2000; Primdahl 2012), meta analysis was conducted on two studies (375 participants) (Hewlett 2000; Primdahl 2012) and suggest that PIAS had little or no effect on satisfaction compared to the consultant‐led system at 12 to 24 months (SMD 0.05, 95% CI –0.41 to 0.52; 2 studies, 375 participants; I2 = 81%; low‐certainty evidence). A third study (131 participants) (Fredriksson 2016) could not be included in the analysis but also reported little or no difference between the groups. The fourth study (290 participants) Goodwin 2016 also could not be included in the analysis but reported a significant difference (p = 0.002) between the groups with greater satisfaction with the service in the PIAS compared to the consultant‐led system. We downgraded the evidence twice to low certainty due to serious risk of bias and indirectness.
The results across other conditions were also variable (see Analysis 1.13 for more details for each study). For example, of the four studies reporting patient satisfaction in people with cancer (921 patients), one reported that the consultant‐led group had greater satisfaction than the PIAS group though significance was not reported (mean: 122 (SE 91.6) with PIAS versus 130 (SE 98.9) with consultant‐led; 264 participants at 60 months; Koinberg 2004). Of the three studies reporting patient satisfaction in people with IBD, two reported that the PIAS group may have greater satisfaction than the consultant‐led group (Kennedy 2003: mean: 54.6 (SD 8.5) with PIAS versus 53.6 (SD 9.1) with consultant‐led; 95% CI –3.67 to 6.03; P = 0.62; Williams 2000: mean: 69 (SE 85) with PIAS versus 4 (SE 41) with consultant‐led; P < 0.01; 280 participants at 12 to 24 months) but with varying levels of significance. One study reporting patient satisfaction in people with psoriasis reported no difference between the PIAS group and the consultant‐led group (Khoury 2018: mean: 3.6 (SE 0.1) with PIAS versus 3.5 (SE 0.1) with consultant‐led; P = 0.75; 147 participants at 12 months).
Clinician (consultant or specialist nurse) satisfaction
None of the studies reported clinician (consultant or specialist nurse) satisfaction.
Failures of the 'system'
None of the studies reported outcomes regarding failures of the 'system' (e.g. how long participants are on the PIAS pathway but not using it correctly before the clinical team needs to reinstigate regular follow‐up clinics). One study (209 participants) reported safety net failure. Hewlett 2000 defined a safety net failure as an increase of more than 20% of the clinical experience of pain, disease activity or disability as recorded on the relevant scales. This study reported no significant differences in the number of safety net failures between the people with rheumatoid arthritis in the PIAS group (26.5%) and the consultant‐led group (28.9%) at 24 months.
Discussion
Summary of main results
In this review of PIASs, we included 17 studies reported in 26 papers. The studies involved 3854 adults all enrolled in a PIAS or in consultant‐led appointment systems across six conditions: cancer, rheumatoid arthritis, IBD, asthma, psoriasis and COPD. Study follow‐up ranged from 12 to 72 months, with the most commonly reported time points (those used in meta‐analyses) being 12 to 24 months.
The review provides mainly low‐certainty evidence for PIASs on a number of outcomes, as well as some less‐frequently reported outcomes of interest (summary of findings Table for the main comparison). The results of meta‐analyses provide some indication of the likely effect; however, the likelihood that it will be substantially different is high. The results suggest PIAS may have little or no impact on patient anxiety, depression and quality of life. For other health disease status measurements, such as disease activity and pain, results suggest these are probably slightly reduced in the PIAS group compared with the consultant‐led group. There is probably little or no impact on disability between PIAS and consultant‐led groups, and self‐efficacy is probably slightly increased in the PIAS group compared with the consultant‐led group. Other aspects of disease status and experience appear to show little or no difference between PIAS and consultant‐led appointment systems.
The results for service utilisation (contacts) are mixed with no clear indication of reduced or increased service contact. Linked to this, results also suggest missed appointments may be slightly reduced in the PIAS group compared with the consultant‐led group. We do not know if service utilisation (costs) could be reduced in the PIAS compared to the consultant‐led appointment groups due to the low‐certainty evidence and because we did not use a currency convertor to establish a common currency, this decision was made in part due to the fact the studies did not report consistently on the services they included in their costs which also made the evidence less comparable across studies.
The results for adverse events suggest people within the PIAS group may be no more or less likely to report a relapse or recurrence of their condition or other adverse effects compared with the consultant‐led group. Evidence from one study suggests people with asthma within the PIAS group were significantly less likely to report use of other drugs at 12 months than those in consultant‐led appointment group but this is based on low‐certainty evidence.
It is also uncertain whether PIAS improve patient satisfaction because the certainty of evidence is low, each study used different questions to collect their data at different time points and across different health conditions. This may also reflect the perceived severity of the condition or the ease with which symptoms can be identified.
None of the studies reported failures of the system other than safety net failure or clinician (consultant or specialist nurse) satisfaction.
Overall completeness and applicability of evidence
This review highlights a few issues concerning the completeness and applicability of the evidence.
First, we found few published studies investigating this question across differing health conditions. We suspect that much of this type of service change may occur without evaluations being published. This may mean there is some form of publication bias to be considered or may mean that these types of service changes are evolving (or not) within the clinical context with only local internal evaluations being conducted.
Second, we noticed a distinct lack of studies from outside Europe. This might reflect the differing healthcare system structures in place internationally or a lack of reporting of this type of system change. It may also reflect how funding of research trials in the UK and Europe is geared towards randomised trials of service delivery interventions. This means the review may have a good deal of applicability to service designs such as (and similar to) the UK NHS.
Third, while the mean age range of the people involved in these studies may reflect the population who are largely affected by long‐term conditions, it may also reflect bias in whom the clinicians might also trust to know and manage their own condition. This implies that this type of appointment system may not be appropriate for all patients, particularly those with a new diagnosis (although this has not been studied in the research included in this review).
Fourth, variation in the quality of reporting of the intervention makes it difficult to know how to reliably interpret the findings of the review. For example, the results of this review suggest great variability in satisfaction of patients with the intervention. This difference could be due to: poor reporting of the levels of satisfaction between PIAS and consultant‐led appointment groups; a real difference between those with rheumatoid arthritis and those with different types of cancer (suggesting that the PIAS system may not work for all long‐term chronic conditions); different tools/bespoke sets of questions being used to ask patients about their satisfaction with the service; or that particular elements of the intervention have been used or are missing from those studies. The reporting of interventions in detail is important, yet difficult with word restrictions placed on peer review publications and with authors potentially being uncontactable at the time of the review.
Certainty of the evidence
The overall certainty of the evidence included in this review ranged mainly from low to very low according to the GRADE classification (Higgins 2011). This was due to the poor reporting of the studies, but even where authors were contacted, it was not always possible to gain clarification of the study details, leading to a substantial amount of uncertainty within the existing evidence surrounding this intervention. For example, the evidence around anxiety, depression and quality of life was downgraded to low certainty due to the risk of bias identified within studies and the wide CIs presented in the analyses. Service utilisation (contacts) evidence was downgraded to low certainty due to serious indirectness and difficulties assessing the consistency and precision of the results. Service utilisation (costs) evidence was downgraded to very‐low certainty due to risk of bias across studies, serious indirectness and serious imprecision. The evidence around adverse events was downgraded to low certainty due to risk of bias across studies and serious consistency and imprecision and the evidence for patient satisfaction was downgraded to low certainty due to risk of bias, the indirectness related to how satisfaction was measured across studies, and due to consistency and precision being difficult to assess.
This made the details of data analysis and attrition inconsistent and difficult to follow. When clearly reported, random sequence generation and allocation concealment were areas of low risk of bias, but for some studies this remained unclear. Blinding of the participants and personnel would have not been possible with this type of intervention.
Potential biases in the review process
This review reduced the likelihood of potential biases by following the comprehensive guidelines set out in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). The review outcomes were also broad so that we could be sure to capture all the relevant research evidence. The review did not include research conducted with children and did not include studies which referred to patients 'self‐management' without a description of the intervention as in the Description of the intervention.
Agreements and disagreements with other studies or reviews
The previous reviews in this area attempted to incorporate a wider range of study design, but found only nine studies to include between them and these studies were all based in the UK (Whear 2013a; Whear 2013b). Despite these differences, our review reports similar findings, with uncertainty remaining a strong message throughout.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Forest plot of comparison: 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, outcome: 1.1 Patient outcomes: anxiety (HADS lower score = better 0–21) (12–24 months).
Forest plot of comparison: 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, outcome: 1.2 Patient outcomes: depression (HADS lower score = better 0–21) (9–24 months).
Forest plot of comparison: 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, outcome: 1.3 Patient outcomes: quality of life (12–18 months).
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 1 Patient outcomes: anxiety (HADS lower score = better 0–21) (12–24 months).
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 2 Patient outcomes: depression (HADS lower score = better 0–21) (9–24months).
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 3 Patient outcomes: quality of life (12–18 months).
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 4 Rheumatoid arthritis: disease activity (VAS and DAS 12–24 months).
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 5 Rheumatoid arthritis: self‐efficacy (ASES 12–24 months).
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 6 Rheumatoid arthritis: pain (VAS 12–24 months).
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 7 Rheumatoid arthritis: disability (HAQ 12–24 months).
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 8 Rheumatoid arthritis: patient satisfaction.
| Study | Consultant‐led n | Person‐years | Total contacts | Patient‐initiated n | Person years | Total contacts | Rate ratio | Contacts covered |
| Brown 2002 | 26 | 26 | 4 | 28 | 28 | 6 | 0.72 | Total telephone calls and referrals to hospital |
| Fredriksson 2016 | 60 | 90 | 185 | 61 | 92 | 165 | 0.87 | Consultant appointments only |
| Goodwin 2016 | 152 | 152 | 143 | 138 | 138 | 140 | 0.93 | Telephone contacts, face‐to‐face appointments nurse/consultant |
| Gulliford 1997 | 96 | 128 | 57 | 97 | 129 | 61 | 0.94 | Contacts by telephone, clinic appointments, GP visits |
| Helgesen 2000 | 200 | 600 | 2728 | 200 | 600 | 4028 | 0.68 | Telephone calls, hospital and GP visits, primary care nurse visits, other outpatient contacts |
| Hewlett 2000 | 89 | 178 | 2178 | 93 | 186 | 1955 | 1.16 | Hospital doctor, GP, nurse, occupational therapist, physiotherapist, podiatrist, orthopaedic surgeon, orthotics |
| Jeppesen 2018 | 77 | 64 | 291 | 79 | 66 | 269 | 1.13 | All cancer‐related GP and gynaecologist visits, telephone calls and examinations at gynaecology department |
| Kennedy 2003 | 366 | 366 | 2164 | 285 | 285 | 1166 | 1.45 | GP, disease‐related outpatients, other outpatients |
| Khoury 2018 | 75 | 75 | 409 | 72 | 72 | 197 | 1.99 | Consultations with dermatologist and telephone (not medication collection or tests) |
| Koinberg 2004 | 131 | 655 | 1172 | 133 | 665 | 1023 | 1.16 | Physician and nurse visits, telephone calls, other consultations |
| Lahdensuo 1996 | 59 | 59 | 59 | 56 | 56 | 28 | 2.00 | Outpatient visits only |
| Primdahl 2012 | 97 | 194 | 3430 | 96 | 192 | 3116 | 1.09 | Rheumatologist, nurse GP consultations, GP telephone call |
| Robinson 2001 | 102 | 119 | 344 | 101 | 118 | 89 | 3.83 | Booked clinic appointments |
| Sheppard 2009 | 112 | 168 | 90 | 112 | 168 | 92 | 0.98 | Specialist nurses and GPs contacts |
| Williams 2000 | 78 | 156 | 1062 | 77 | 154 | 1136 | 0.92 | Outpatient visits, day cases, inpatient days, GP surgery/home visits |
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 9 Service utilisation (contacts).
| Study | Consultant‐led n | Mean cost per patient | SD | Mean cost per patient per year | Patient‐initiated n | Mean cost per patient | SD | Mean cost per patient per year | Mean difference (per patient‐year) | Costs covered (excluding medication) (currency) |
| Berkhof 2014 | 51 | 2321 | 3967 | 1160.5 | 49 | 1803 | 2617 | 901.5 | –259 | GP, pulmonologist, pulmonary nurse emergency department, pulmonary ward (EUR) |
| Goodwin 2016 | 152 | 32.21 | 22.15 | 32.21 | 138 | 34.72 | 34.51 | 34.72 | 2.51 | Telephone, face‐to‐face nurse and consultant (GBP) |
| Helgesen 2000 | 200 | 15280 | — | 3056 | 200 | 12873 | — | 2574.6 | –481.4 | Outpatients costs, hospital, nursing home and hospital‐based home care (SEK) |
| Hewlett 2000 | 89 | 624.69 | 339.48 | 312.345 | 93 | 436.94 | 332.76 | 218.47 | –93.88 | Hospital consultants, occupational therapy, physiotherapy, orthoptist, orthopaedic surgeon, chiropodist, transport, including predicted GBP 2000 nurse cost for helpline (GBP) |
| Kennedy 2003 | 366 | 494.99 | — | 494.99 | 285 | 364.78 | — | 364.78 | –130.21 | GP and all outpatient visits, hotel cost per inpatient day, intervention (GBP) |
| Koinberg 2004 | 131 | — | — | 630 | 133 | — | — | 495 | –135 | Nurse, GP, physiotherapist, social worker, prosthetic technicians, plus tests and telephone contacts (EUR) |
| Primdahl 2012 | 97 | 1225 | 522 | 612.5 | 96 | 1085 | 362 | 542.5 | –70 | Face‐to‐face and phone appointments, tests, other primary care services not disclosed (EUR) |
| Williams 2000 | 78 | 688 | — | 344 | 77 | 670 | — | 335 | 47.5 | Outpatient visits, tests, day cases, inpatient days, GP visits, home visits (GBP) |
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 10 Service utilisation (costs).
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 11 Missed appointments.
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 12 Adverse event: relapses (12–60 months).
| Study | Outcome | Patient‐initiated n | Patient‐initiated mean (SE) | Consultant‐led n | Consultant‐led mean (SE) | P value (95% CI) |
| Rheumatoid arthritis | ||||||
| Fredriksson 2016 | Satisfaction with care at 18 months (mean change from baseline) | 64 | 1 | 67 | 1 | 0.47 |
| Goodwin 2016 | Satisfaction with service at 12 months – (mean score) | 138 | 100 | 152 | 80 | 0.002 |
| Inflammatory bowel disease | ||||||
| Kennedy 2003 | Satisfaction with hospital visit (mean score at 12 months from patient diary) | 64 | 54.6 (SD 8.5) | 132 | 53.6 (SD 9.1) | 0.62 (–3.67 to 6.03) |
| Robinson 2001 | Acceptability – preferred PIAS intervention (at 14 months) | 86 | 71 (82) | 85 | 80 (95) | NR |
| Williams 2000 | Acceptability – preferred PIAS intervention at (24 months) | 81 | 69 (85) | 83 | 34 (41) | < 0.01 |
| Cancer | ||||||
| Brown 2002 | Satisfaction with allocated intervention (at 6 months) | 28 | 26 (93) | 24 | 24 (100) | NR |
| Gulliford 1997 | Acceptibility (% preferring intervention) | 97 | 11 | 96 | 6 | NR |
| Helgesen 2000 | Reporting deficient outpatient service (at 36 months) | NR | NR (6.6) | NR | NR (4.9) | NR |
| Koinberg 2004 | Satisfied with medical centre (SaaC at 60 months) | 133 | 122 (91.6) | 131 | 130 (98.9) | ns |
| Psoriasis | ||||||
| Khoury 2018 | Satisfaction with overall care (at 12 months) | 72 | 3.6 (0.1) | 75 | 3.5 (0.1) | P = 0.75 (mean change difference between groups 0.10, 95% CI –0.07 to 0.26) |
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 13 Patient satisfaction.
| Study | Outcome | Patient‐initiated n | Patient‐initiated mean change from baseline (SE) | Consultant‐led n | Consultant‐led mean change from baseline (SE) | P value (95% CI) |
| Berkhof 2014 | Chronic obstructive pulmonary disease symptoms (Clinical COPD Questionnaire range 0–6 lower score is better, minimal clinically important difference (MCID) = 0.4) | 40 | 0.33 (± 0.11) | 29 | 0.53 (± 0.13) | 0.24 (–0.55 to 0.14) |
| Berkhof 2014 | Respiratory symptoms (St George's Respiratory Questionnaire (SGRQ) total score at 24 months, range 0–100 low score indicates better health status, MCID of SGRQ total | 38 | 5.0 (± 2.2) | 30 | 6.4 (± 2.4) | 0.67 (–7.9 to 5.1) |
| Berkhof 2014 | Time to exacerbation (median no. of days) | 49 | 307 (± 61.6) | 51 | 335 (± 60.2) | 0.40 |
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 14 Chronic obstructive pulmonary disease: other outcomes.
| Study | Outcome | Patient‐initiated n | Patient‐initiated (%) | Consultant‐led n | Consultant‐led (%) | P value (95% CI) |
| Gulliford 1997 | % reassured with service (at 16 months) | 97 | 88 | 96 | 94 | NR |
| Gulliford 1997 | — | — | — | — | — | — |
| Helgesen 2000 | Accessibility (% reporting service inaccessible by telephone 36 months) | 200 | 8.6 | 200 | 7.3 | RR 1.18 (0.46 to 3.03) |
| Helgesen 2000 | — | — | — | — | — | — |
| Jeppesen 2018 | Fear of cancer recurrence at 10 months (mean (SD)) | 105 | 43.4 (26.5) | 107 | 40.5 (28.5) | 0.02 |
| Jeppesen 2018 | Clinical fear of recurrence at 10 months (%) | 105 | 17.7 | 107 | 22.1 | 0.89 |
| Koinberg 2004 | Accessibility by telephone at 60 months (%) | 133 | 100 | 131 | 96.5 | ns |
| Koinberg 2004 | — | — | — | — | — | — |
| Sheppard 2009 | Fear of recurrence at 18 months (score from 3 questions ‐ mean) | 107 | 5.6 | 107 | 5.0 | 0.066 |
| Sheppard 2009 | Isolation (% felt isolated) | 107 | 14 | 107 | 9 | 0.245 |
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 15 Cancer: other outcomes.
| Study | Outcome | Patient‐initiated n | Patient‐initiated mean (SD) | Consultant‐led n | Consultant‐led mean (SD) | P value/mean difference (95% CI) |
| Kennedy 2003 | Enablement after initial consultation | 260 | 4.0 (3.9) | 352 | 3.0 (3.9) | 0.026 (0.12 to 1.68) |
| Kennedy 2003 | Delay before starting treatment (% within 1 day) | 81 | 49.7 | 109 | 50.6 | 0.95 (–0.52 to 0.49) |
| Kennedy 2003 | Appointments made by patient (% at least 1 self‐made appointment) | 144 | 43 | 250 | 22 | < 0.001 (1.63 to 4.46) |
| Robinson 2001 | Time to symptom treatment (hours) | 101 | 14.8 | 102 | 49.6 | < 0.0001 |
| Robinson 2001 | Cost to patient (GBP) at 14 months | 101 | 0.86 | 102 | 8.92 | < 0.0001 |
| Robinson 2001 | — | — | — | — | — | — |
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 16 Inflammatory bowel disease: other outcomes.
| Study | Outcome | Patient‐initiated N | Patient‐initiated m (SD) | Consultant‐led N | Consultant‐led m (SD) | p value or mean diff (95% CI) |
| Khoury 2018 | Missed medication collection (at 12mths) | 72 | 11 (8.2) | 75 | 4 (4) | −0·10 (−0·024 to −0·04) |
| Khoury 2018 | Missed laboratory tests (at 12mths) | 72 | 16 (13.8) | 75 | 14 (18) | −0·10 (−0·20 to 0·18) |
| Khoury 2018 | Changes in medication (at 12 mths) | 72 | 6.8% | 75 | 20.6% | p<0.001 |
| Khoury 2018 | Disease activity at 12mths (Psoriasis Area and Severity Index ‐ PASI) | 72 | 0.1 (SE+/‐0.2) | 75 | ‐0.16 (SE+/‐0.2) | −0·24, (−0·84 to 0·36) |
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 17 Psoriasis: other outcomes.
| Study | Outcome | Patient‐initiated n | Patient‐initiated median change from baseline (IQR) | Consultant‐led n | Consultant‐led median change from baseline (IQR) | P value (95% CI) |
| Fredriksson 2016 | Confidence in care at 18 months (VAS 1–100 mm) | 64 | 1 (–2 to 7) | 67 | 0 (–4 to 7) | 0.97 |
| Fredriksson 2016 | General health at 18 months (VAS 1–100 mm) | 64 | –1 (–12 to 8) | 67 | –4 (–17 to 9) | 0.57 |
| Fredriksson 2016 | Number swollen joints at 18 months | 64 | –3 (–5 to 0) | 67 | –1 (–3 to 1) | 0.40 |
| Fredriksson 2016 | Number of tender joints at 18 months | 64 | –1 (–4 to 0) | 67 | –1 (–3 to 0) | 0.81 |
| Fredriksson 2016 | Cases where treatment was changed at 18 months (%) | 64 | 68 | 67 | 48 | 0.0002 |
| Fredriksson 2016 | C‐reactive protein (mg/L) at 18 months | 64 | 0 (–2 to 1) | 67 | 0 (–3 to 1) | 0.70 |
| Fredriksson 2016 | Erythrocyte sedimentation rate (mm/hour) at 18 months | 64 | 4 (–2 to 10) | 67 | 0 (–4 to 5) | 0.03 |
| Hewlett 2000 | Helplessness (AHI 0–30) at 48 months (mean difference from baseline) | 74 | –0.2 | 60 | 1.0 | ns |
| Hewlett 2000 | Confidence with intervention at 48 months (VAS 1–10) | 74 | 0.5 | 60 | –0.6 | < 0.01 |
| Hewlett 2000 | — | — | — | — | — | — |
| Hewlett 2000 | — | — | — | — | — | — |
| Hewlett 2000 | — | — | — | — | — | — |
| Hewlett 2000 | — | — | — | — | — | — |
| Hewlett 2000 | — | — | — | — | — | — |
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 18 Rheumatoid arthritis: other outcomes.
| Study | Outcome | Patient‐initiated n | Patient‐initiated mean (SD) | Consultant‐led n | Consultant‐led mean (SD) | P value (95% CI) |
| Lahdensuo 1996 | Lung function | 56 | NR | 59 | NR | ns |
| Lahdensuo 1996 | Days off work | 56 | 2.8 (0–62) | 59 | 4.8 (0–27) | 0.02 |
Comparison 1 Patient‐initiated appointment systems versus consultant‐led appointment systems, Outcome 19 Asthma: other outcomes.
| Patient‐initiated appointment systems compared with consultant‐led appointment systems for people with chronic conditions | ||||
| Patient or population: adults with a chronic or recurrent condition Settings: secondary care Intervention: patient‐initiated appointment systems Comparison: consultant‐led appointment systems | ||||
| Outcomes | Relative effect | No of participants | Certainty of the evidence | Comments |
| Patient outcomes: anxiety (HADS)a 12–24 months | OR 0.87 (0.68 to 1.12) | 1019 | ⊕⊕⊝⊝ | 7 studies reported anxiety, 2 could not be included in the analysis due to data available, but findings appeared consistent. Heterogeneity in health condition (cancer, psoriasis and RA), participants and follow‐up. |
| Patient outcomes: depression (HADS)a 9–24 months | OR 0.79 (0.51 to 1.23) | 1835 | ⊕⊕⊝⊝ | 7 studies reported depression, 1 could not be included in the analysis due to data available, but findings appeared consistent. Heterogeneity in health condition (cancer, RA, psoriasis and IBD), participants and follow‐up. |
| Patient outcomes: quality of life (different scales used across studies) 12–18 months | SMD 0.12 (0.00 to 0.25) | 1486 | ⊕⊕⊝⊝ | 12 studies reported quality of life, 5 could not be included due to data available, but where some data were available findings appeared consistent. Heterogeneity in health condition (asthma, IBD, psoriasis, RA and cancer) and tools used to measure quality of life. |
| Service utilisation (contacts) 12 months | Contact rate ratio ranged from 0.68 to 3.83 (median rate ratio 1.11, IQR 0.93 to 1.37) | 3348 (15) | ⊕⊕⊝⊝ | Studies included different elements of service contact in their data, and may expect contact to vary across health condition. |
| Service utilisation (costs) per patient per year | Studies reported the intervention group may have had higher or lower costs than the control group. | 2235 (8) | ⊕⊝⊝⊝ | Studies included different elements of service costs in their data, data were presented in different currencies and it may be expected that service costs vary across health condition. |
| Adverse events – relapse 12–60 months | MD –0.20 (–0.54 to 0.14) | 888 (3) | ⊕⊕⊝⊝ | 5 studies reported adverse events as relapse or recurrence, 2 could not be included in the analysis due to data available but findings appeared consistent. Each study used different questions/definitions to collect their data at difference time points and across different health conditions (inflammatory bowel disease and cancer). |
| Patient satisfaction 6–72 months | SMD 0.05 (–0.41 to 0.52) | 375 (2) | ⊕⊕⊝⊝ | 12 studies reported patient satisfaction, 10 could not be included in the analysis due to data available but findings appeared broadly consistent. Each study used different questions to collect their data at different time points and across different health conditions. No studies reported on clinician satisfaction. |
| CI: confidence interval; HADS: Hospital Anxiety and Depression Scale; IBD: inflammatory bowel disease; IQR: interquartile range; MD: mean difference; OR: odds ratio; RA: rheumatoid arthritis; SMD: standardised mean difference. | ||||
| GRADE Working Group grades of evidence | ||||
| aAnxiety and depression (HADS) was pooled as dichotomous data using the score of 10 as the cut point as identified in two of the papers, continuous data were converted to dichotomous data as no other data could be retrieved from the papers presenting dichotomous data for these outcomes. | ||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Patient outcomes: anxiety (HADS lower score = better 0–21) (12–24 months) Show forest plot | 5 | Odds Ratio (Random, 95% CI) | 0.87 [0.68, 1.12] | |
| 1.1 Cancer | 3 | Odds Ratio (Random, 95% CI) | 0.94 [0.55, 1.59] | |
| 1.2 Psoriasis | 1 | Odds Ratio (Random, 95% CI) | 0.89 [0.64, 1.23] | |
| 1.3 Rheumatoid arthritis | 1 | Odds Ratio (Random, 95% CI) | 0.70 [0.36, 1.36] | |
| 2 Patient outcomes: depression (HADS lower score = better 0–21) (9–24months) Show forest plot | 6 | Odds Ratio (Random, 95% CI) | 0.79 [0.51, 1.23] | |
| 2.1 Cancer | 3 | Odds Ratio (Random, 95% CI) | 0.80 [0.23, 2.71] | |
| 2.2 Inflammatory bowel disease | 1 | Odds Ratio (Random, 95% CI) | 0.74 [0.34, 1.63] | |
| 2.3 Psoriasis | 1 | Odds Ratio (Random, 95% CI) | 1.0 [0.72, 1.38] | |
| 2.4 Rheumatoid arthritis | 1 | Odds Ratio (Random, 95% CI) | 0.39 [0.16, 0.96] | |
| 3 Patient outcomes: quality of life (12–18 months) Show forest plot | 7 | 1486 | Std. Mean Difference (IV, Random, 95% CI) | 0.12 [‐0.00, 0.25] |
| 3.1 Cancer | 2 | 296 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.07 [‐0.33, 0.20] |
| 3.2 Asthma | 1 | 115 | Std. Mean Difference (IV, Random, 95% CI) | 0.47 [0.10, 0.84] |
| 3.3 Inflammatory bowel disease | 2 | 735 | Std. Mean Difference (IV, Random, 95% CI) | 0.15 [0.00, 0.29] |
| 3.4 Psoriasis | 1 | 147 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.01 [‐0.33, 0.31] |
| 3.5 Rheumatoid arthritis | 1 | 193 | Std. Mean Difference (IV, Random, 95% CI) | 0.21 [‐0.07, 0.50] |
| 4 Rheumatoid arthritis: disease activity (VAS and DAS 12–24 months) Show forest plot | 2 | 375 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.21 [‐0.41, ‐0.01] |
| 5 Rheumatoid arthritis: self‐efficacy (ASES 12–24 months) Show forest plot | 2 | 375 | Mean Difference (IV, Fixed, 95% CI) | 4.95 [0.57, 9.34] |
| 6 Rheumatoid arthritis: pain (VAS 12–24 months) Show forest plot | 2 | 375 | Mean Difference (IV, Fixed, 95% CI) | ‐0.71 [‐1.17, ‐0.26] |
| 7 Rheumatoid arthritis: disability (HAQ 12–24 months) Show forest plot | 2 | 375 | Mean Difference (IV, Fixed, 95% CI) | ‐0.06 [‐0.19, 0.06] |
| 8 Rheumatoid arthritis: patient satisfaction Show forest plot | 2 | 375 | Std. Mean Difference (IV, Random, 95% CI) | 0.05 [‐0.41, 0.52] |
| 9 Service utilisation (contacts) Show forest plot | Other data | No numeric data | ||
| 10 Service utilisation (costs) Show forest plot | Other data | No numeric data | ||
| 11 Missed appointments Show forest plot | 2 | 599 | Mean Difference (IV, Fixed, 95% CI) | ‐0.27 [‐0.42, ‐0.12] |
| 12 Adverse event: relapses (12–60 months) Show forest plot | 3 | 888 | Mean Difference (IV, Random, 95% CI) | ‐0.20 [‐0.54, 0.14] |
| 12.1 Inflammatory bowel disease | 2 | 624 | Mean Difference (IV, Random, 95% CI) | ‐0.4 [‐0.72, ‐0.08] |
| 12.2 Cancer | 1 | 264 | Mean Difference (IV, Random, 95% CI) | 0.03 [‐0.08, 0.14] |
| 13 Patient satisfaction Show forest plot | Other data | No numeric data | ||
| 13.1 Rheumatoid arthritis | Other data | No numeric data | ||
| 13.2 Inflammatory bowel disease | Other data | No numeric data | ||
| 13.3 Cancer | Other data | No numeric data | ||
| 13.4 Psoriasis | Other data | No numeric data | ||
| 14 Chronic obstructive pulmonary disease: other outcomes Show forest plot | Other data | No numeric data | ||
| 15 Cancer: other outcomes Show forest plot | Other data | No numeric data | ||
| 16 Inflammatory bowel disease: other outcomes Show forest plot | Other data | No numeric data | ||
| 17 Psoriasis: other outcomes Show forest plot | Other data | No numeric data | ||
| 18 Rheumatoid arthritis: other outcomes Show forest plot | Other data | No numeric data | ||
| 19 Asthma: other outcomes Show forest plot | Other data | No numeric data | ||
