Increasing the provision of physical activity advice by healthcare professionals
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
The aim of this review is to assess the effects of interventions aimed at increasing the provision of physical activity advice by HCPs.
Background
The benefits of physical activity in the prevention and management of myriad diseases are well established. The World Health Organization (WHO) recommends a minimum 150 minutes of moderate or 75 minutes of vigorous physical activity per week in bouts of at least 10 minutes for adults (WHO 2010). The Exercise is Medicine initiative, which originated in the United States and is now experiencing global uptake, encourages primary care providers to include physical activity in their care planning for patients and provides resources to this end (ACSM). Similar strategies have been proposed in Europe (WHO 2015).
A number of interventions have been implemented and studied in the primary care environment for encouraging physical activity. Both counselling and physical activity prescription have been shown to be effective in increasing patient levels of physical activity (Frank 2010). A prior Cochrane systematic review examined a variety of interventions for promoting exercise (Hillsdon 2005), including group exercise sessions, workshops, face‐to‐face counselling, phone calls, and written materials; overall the effect was positive and moderate. The role of interdisciplinary teams in the delivery of physical activity interventions are increasingly the focus of studies (Tulloch 2006).
Physical activity interventions in primary care appear to have similar costs as pharmaceutical interventions (Garrett 2011); and exercise may be a viable alternative to drug therapy in many cases (Naci 2013). An estimate of the number needed to treat for an additional beneficial outcome (NNTB) to achieve physical activity recommendations using some combination of advice, written materials and physical activity referral is 12 sedentary adults. This can be compared to NNTB estimates of 50 to 120 for smoking cessation advice (Orrow 2013). Many common medical and surgical interventions have NNTBs significantly higher than 12.
Unfortunately, physical activity is not reliably addressed by healthcare professionals (HCPs) in routine practice settings. As of 2010, it was estimated that only 32.4% of adults in the United States were advised by an HCP over a 12‐month period to begin or continue an exercise routine (Barnes 2012). Although studies are limited, these findings do seem to be consistent globally. Only 37% of Korean primary care providers indicated that they collected physical activity data from patients “always” or “almost always” (Cho 2003). Australian general practitioners reported counselling approximately 30% of their patients about physical activity per week (Eakin 2004). Furthermore, even when physical activity is addressed, the advice provided may not be meaningful or actionable. A national survey in the United States showed that among patients who receive physical activity advice, only 38% are supported in developing a specific activity plan and only 42% receive follow‐up support (Glasgow 2001). Direct observation in Ohio family practices in the USA found that when physical activity counselling was performed, the mean time spent was only 47 seconds (Podl 1999).
A 2012 systematic review found that the most common barriers to physician provision of physical activity advice were lack of time, lack of knowledge or training in physical activity counselling, and a lack of self‐efficacy in changing patient behaviour, in that order (Hebert 2012). For instance, a survey of Canadian family physicians indicated that 17% of family physicians did not feel qualified to offer exercise advice and only 4.9% had any specialized knowledge of exercise science or counselling (Kennedy 2003). Additional barriers include lack of economic incentives, lack of protocols for prescribing physical activity, and organizational barriers (Hebert 2012). Relatedly, in a 2015 systematic review of factors influencing whether primary care providers promote physical activity, the availability of intervention materials such as educational booklets and physical activity prescription aids was emphasized (Huijg 2015). Tools that can be adapted to the specific patient and provider were more likely to be used and providers were also far more likely to utilize an intervention perceived as evidence‐based.
Description of the condition
Sedentary lifestyle and inactivity are a prevalent public health concern. In England, it is estimated that just 37% of men and 25% of women meet recommended levels of physical activity (Stamatakis 2005). In Canada, it is estimated that only 17% of men and 14% of women meet guidelines (Colley 2011). A US report from the Centers for Disease Control and Prevention estimates that 23.1% of men and 15.8% of women meet physical activity guidelines (Schoenborn 2013).
It is estimated that 31.1% of the global population of adults is sedentary (Hallal 2012). There is substantial variation based upon geographic region: 27.5% in Africa, 43.3% in the Americas, 43.2% in the eastern Mediterranean, 34.8% in Europe, 17% in southeast Asia and 33.7% in the western Pacific (Hallal 2012). There is also significant disparity between men and women: women’s inactivity is estimated at 33.9%, while men’s inactivity is 27.9% (Hallal 2012). Additionally, individuals living in higher income countries are more likely to be inactive than those living in lower‐income countries (Hallal 2012).
Description of the intervention
Studies have tested a variety of interventions seeking to encourage HCPs to more reliably deliver physical activity advice, including changes to the organization of care related to physical activity. Some possible examples may include training, economic incentives, and audit and feedback for practitioners. Systemic interventions will also be considered, for instance the creation of resources for physical activity promotion and tools to assist providers in screening and intervening with regard to sedentarism.
How the intervention might work
Interventions can be directed toward changing the behaviours of HCPs with regard to providing patients with combinations of advice, resources and tools to increase levels of physical activity. Previous systematic reviews identified a number of barriers for provider behaviour change with respect to provision of physical activity advice, including attitudes, knowledge/skills, organizational and professional support, community resources, and competing priorities (Hebert 2012; Huijg 2015). To be successful, interventions will likely need to address many of these barriers.
Why it is important to do this review
The current body of literature includes clear guidelines regarding physical activity and strategies that could increase patients’ physical activity levels, but evidence suggests that HCPs do not reliably implement these recommendations in practice. The effectiveness of interventions aimed at HCPs to encourage them to implement these strategies in practice has yet to be determined. Numerous barriers interfere with intentions and abilities of HCPs to provide physical activity advice, but it is not known if some or all of these barriers must be addressed in order to affect meaningful behaviour change. Our aim is to quantify the effects of specific interventions aimed at HCP behaviour change in delivering physical activity recommendation. Our hope is that this will allow future research to identify and build upon best practices when developing interventions aimed at HCPs for increasing physical activity levels of patients.
Objectives
The aim of this review is to assess the effects of interventions aimed at increasing the provision of physical activity advice by HCPs.
Methods
Criteria for considering studies for this review
Types of studies
We will consider randomized trials, non‐randomized trials, interrupted time‐series studies and repeated‐measures studies that assess interventions aimed at increasing the provision of physical activity advice by HCPs (excluding students). Initial searches have shown very small numbers of randomized trials in this area, and consequently the range of eligible study designs has been broadened and weighed against the possibility of publication bias introduced by broader study‐type inclusion.
We will only include cluster‐randomised trials and non‐randomised cluster trials with at least two intervention sites and two control sites and only include Interrupted time‐series and repeated‐measures studies that have a clearly defined point in time when the intervention occurred and at least three data points before and three after the intervention.
Types of participants
Participants receiving the intervention will be any HCP with the ability to provide physical activity screening, counselling, advice, or prescriptions as part of their routine care. We will include interventions directed toward qualified physician‐trainees (i.e. registrars, fellows, or residents) but will exclude students. These participants may function in a healthcare inpatient, outpatient or community‐based setting.
Types of interventions
We will consider any interventions offered to HCPs (e.g. training, tools, incentives) aiming to increase the frequency/duration of physical activity advice they deliver to their usual patients. The individuals/group of HCPs at whom the intervention is targeted will be noted. The intervention could include organizational or patient‐oriented features, or both, but to meet eligibility it must have a component targeting behaviour change by HCPs during available opportunities (that is all situations — in‐person, phone, mail, email, text — during which an HCP interacts with a patient). We will use the EPOC taxonomy for categorization of interventions (EPOC 2015). Whenever possible, the duration and intensity of the intervention will be considered.
Interventions that do not explicitly address physical activity level will be excluded. Studies with only patient‐oriented interventions (e.g. public health promotional campaigns), and studies testing organizational changes that completely bypass usual healthcare providers will not be included. We will exclude studies focusing on time‐limited physical activity as treatment for acute problems (e.g. physiotherapy for back pain). Studies that include multiple lifestyle interventions (e.g. smoking cessation, weight loss and physical activity) will be evaluated only if there is clear delineation of the interventions for physical activity and outcomes for physical activity.
Generally the comparison will be ‘usual care’. The authors recognize that the definition of 'usual care' will likely be broad and varied in different studies. Ideally this 'usual care' will involve a standardized treatment regimen. We will endeavour to describe the specific definitions of 'usual care' in the individual studies identified.
We will also consider studies that compare multiple different approaches to increase rates of physical activity counselling by HCPs. These approaches will be defined based upon EPOC taxonomy and compared against others in the same category (EPOC 2015). Approaches of different categories will also be compared.
Types of outcome measures
Described below
Primary outcomes
Primary outcomes will be provider‐level 'quality of care' measures, including the following indicators.
-
Proportion of patients assessed for activity level (indicated by a percentage of patients in a year).
-
Proportion of patients receiving a recommendation regarding physical activity (indicated by a percentage of patients in a year).
-
Proportion of available opportunities in which HCPs offer a recommendation regarding physical activity (indicated by a percentage of opportunities in a year).
-
Proportion of patients referred regarding physical activity (indicated by a percentage of patients in a year).
-
Proportion of available opportunities in which HCPs make a referral regarding physical activity (indicated by a percentage of opportunities in a year).
-
Proportion of patients provided resources regarding physical activity (indicated by a percentage of patients in a year).
-
Proportion of available opportunities in which HCPs offer resources regarding physical activity (indicated by a percentage of opportunities in a year).
'Available opportunities' will be defined as all situations (in‐person, phone, mail, email, text) during which an HCP interacts with a patient.
For each main (dichotomous) outcome (i.e. proportion receiving/conducting the specific desired practice), the absolute difference and relative effect will be summarized. These outcomes will be included in a 'Summary of findings' table.
Secondary outcomes
Secondary outcomes will be 'patient outcomes', such as the following 'health behaviour' indicators...
-
quantity of physical activity (self‐reported);
-
quantity of physical activity (measured)
...and 'health status' indicators:
-
improved health indicators such as weight reduction, decreased blood pressure, decreased fasting blood sugar or HbA1c, glycaemic control, improved cardiovascular fitness. (Operational definitions for each of these outcomes are described below);
-
adverse events such as injuries sustained during physical activity (sprains, strains, fractures), cardiac events sustained during physical activity, and worsening pain related to activity.
We will also attempt to capture 'resource use' indicators:
-
cost of intervention;
-
time involved in delivering intervention;
-
mean duration of interaction with patients regarding physical activity (indicated by the average number of minutes per interaction).
Reporting of the outcomes listed here will not be an inclusion criteria for the review. Studies that only report secondary outcomes will not be included in the review.
Search methods for identification of studies
We will search for potentially relevant study reports from the electronic databases described in the Electronic searches section and other resources (Searching other resources).
Electronic searches
The EPOC Information Specialist (IS), in consultation with the review authors, has developed a sensitive search strategy designed to retrieve trials and studies from electronic bibliographic databases. We will search the Cochrane Database of Systematic Reviews (CDSR) and the Database of Abstracts of Reviews of Effects (DARE) for related systematic reviews and the following databases from inception to date of search for primary studies:
-
Cochrane EPOC Group Specialised Register;
-
Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library;
-
MEDLINE In‐Process and other non‐indexed citations, OvidSP;
-
Embase, OvidSP;
-
CINAHL (Cumulative Index to Nursing and Allied Health Literature), EBSCOhost.
The search strategies comprise keywords and controlled vocabulary terms. Language limits will not be applied. All databases will be searched from database start date to date of search.
Two methodology search filters will be used to limit retrieval to appropriate study designs: a modified version of the Cochrane Highly Sensitive Search Strategy (sensitivity‐ and precision‐maximizing version – 2008 revision) (Lefebvre 2011), to identify randomized trials (cf. the Cochrane Handbook for Systematic Reviews of Interventions 6.4d, Higgins 2011); and an EPOC methodology filter to identify non‐randomised trial designs.
The search strategy is shown in Appendix 1. We will use the EPOC search log template to track database names, time span, provider, number of results identified and date for when the searches were run (EPOC 2014).
Searching other resources
We will search the following trial registries with no date or language restrictions for ongoing trials.
-
ClinicalTrials.gov, US National Institutes of Health (NIH) (ClinicalTrials.gov).
-
WHO International Clinical Trials Registry Platform (ICTRP) (WHO ICTRP).
We will also:
-
review the reference lists of all included studies, relevant systematic reviews and primary studies for potentially relevant sources;
-
contact authors of potentially relevant studies to seek unpublished results and clarify reported published information;
-
contact researchers with expertise relevant to the review to discuss other potential sources of information.
Data collection and analysis
Described below.
Selection of studies
Two review authors will independently screen all study reports, including titles and abstracts, identified by the searches. They will reject articles if the title or abstract do not meet the inclusion/exclusion criteria. They will obtain and screen the full text in instances where the study cannot be categorically rejected, with discrepancies resolved by consensus and the assistance of a third review author.
Two review authors will scan reference lists of screened articles for other potentially relevant articles. They will then independently assess the potentially relevant studies for inclusion by reviewing titles and abstracts. They will again obtain and screen the full text in instances where the study cannot be categorically rejected, with discrepancies resolved by consensus and the assistance of a third review author, as necessary.
We will include in the 'Characteristics of excluded studies table' (with the reason for exclusion described) all studies that may initially appear to meet the eligibility criteria but on further inspection do not, as well as those that do not meet all of the criteria but are well known and likely to be thought relevant by some readers.
We will record the selection process in sufficient detail to complete a PRISMA flow diagram.
Data extraction and management
Two review authors will independently extract data from published reports using a standardized data collection form. They will resolve disagreements by consensus or referral to a third reviewer when agreement cannot be reached. No attempt will be made to blind reviewers to either the results of the primary studies or the intervention the subjects received. Data will be compiled using Early Review Organizing Software (EROS).
We will identify information on the following design characteristics.
-
Country of origin.
-
Study design/methodology (particularly statistical analyses used).
-
Type of provider (physician, nurse, therapist).
-
Patient characteristics (demographics, comorbidities).
-
Setting (healthcare inpatient, outpatient and community‐based clinics).
-
Outcome(s).
-
Length of follow‐up.
We will extract data and enter it into Review Manager 5 (RevMan) for the following outcome variables, where reported.
-
Proportion of patients assessed for physical activity status.
-
Proportion of patients receiving recommendation or providers offering recommendation regarding physical activity (i.e. an interaction between patient and HCP during which barriers are reviewed or guidance/recommendations are provided, or both).
-
Proportion of patients referred or providers referring regarding physical activity (i.e. referral to an HCP or facility with expertise or where it is expected that additional advice regarding activity would be provided).
-
Proportion of patients provided resources or providers offering resources regarding physical activity (i.e. written resources such as combinations of prescriptions, handouts and mobile phone applications that may include reminders).
-
Duration of interaction with patients regarding physical activity (i.e. number of minutes).
-
Quantity of physical activity (minutes per week; metabolic equivalents (METS) per week; proportion reaching > 150 minutes per week; proportion sedentary).
-
Improved health indicators (e.g. weight reduction, decreased blood pressure, decreased fasting blood sugar or HbA1c, glycaemic control, improved cardiovascular fitness).
-
Adverse events (e.g. cardiac events, musculoskeletal injuries, pain).
We will contact study authors by email to clarify any necessary study details.
We will describe any necessary transformation of data along with any assumptions made.
We will describe the procedures used for extracting numeric data from graphs.
Assessment of risk of bias in included studies
Two review authors will independently assess risk of bias for each study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), and the guidance from the EPOC group (EPOC 2016). They will resolve any disagreement by discussion or, in the event of deadlock, by involving a third review author. We will assess the risk of bias according to the following domains.
-
Random sequence generation.
-
Allocation concealment.
-
Blinding of participants and personnel.
-
Blinding of outcome assessment.
-
Incomplete outcome data.
-
Selective outcome reporting.
-
Baseline outcomes measurement.
-
Baseline characteristics.
-
Other bias e.g. use of validated outcome measures; intention‐to‐treat analysis; final outcome controlled for baseline behaviour; loss of clusters in cluster‐randomised trials; statistical analyses used to account for clustering in cluster‐randomised trials.
In interrupted time‐series studies, we will assess the risk of bias for:
-
intervention independence from other changes;
-
pre‐specification of intervention effect;
-
influence of intervention on data collection;
-
blinding of outcome assessors;
-
incomplete outcome data;
-
selective outcome reporting;
-
appropriateness of analysis including adjusting of estimates of intervention effects;
-
other risks of bias (e.g. seasonality; consideration of changes in secular trends; use of standardized protocols; comparability of groups).
We will judge each potential source of bias as high, low, or unclear and provide a quote from the study report together with a justification for our judgment in the 'Risk of bias' table. Within each study across domains, each outcome (or class of outcomes) will be defined as ‘high risk of bias’ if it demonstrates high risk of bias for one or more domains; or as ‘unclear risk of bias’ if it demonstrates unclear risk of bias for at least one domain. We will summarise the 'Risk of bias' judgments across different studies for each of the domains listed.
We will not assign judgments for blinding of participants to group allocation. This would not be appropriate for studies of this type, as it would be difficult or impossible to offer education, resources or incentives in a blinded manner. Where information on risk of bias relates to unpublished data or correspondence with a trialist, we will note this in the 'Risk of bias' table. We will not exclude studies on the grounds of their risk of bias, but will clearly report the risk of bias when presenting the results of the studies. When considering treatment effects, we will take into account the risk of bias for the studies that contribute to that outcome.
Measures of treatment effect
We anticipate outcome data will be presented as dichotomous, continuous or ordinal. In cases of dichotomous data, risk ratios and risk difference will be calculated. In cases of ordinal data, initial assessment will be made of the validity of the used measurement scale. The ordinal data may be analysed as a continuous outcome, or dichotomized based upon the scale used and existence of well‐defined cut‐off values. In cases of continuous data, we will calculate the mean difference or standardized mean difference if studies assess the same outcome in different ways. For all effect measures, effect estimates will be presented with 95% confidence intervals.
For interrupted time‐series studies, the ARIMA (autoregressive integrated moving average) model will be used for data analysis.
Unit of analysis issues
Cluster trials
If studies do not use appropriate statistical methods to account for clustering of data, but report the intraclass correlation coefficient (ICC) then we will adjust for clustering by inflating the standard error. In the case of unit‐of‐analysis errors where the ICC is not reported, we will contact the study authors in an attempt to obtain additional data necessary to adjust for clustered data. If we are unable to obtain sufficient information to reanalyse the data, we will not report confidence intervals orP values for study outcomes where there is a unit‐of‐analysis error.
Studies with multiple intervention groups
In the case of studies with multiple intervention group arms where multiple comparisons could contribute to a single meta‐analysis, we will include all treatment arms meeting eligibility criteria and combine them into a single treatment group. For dichotomous outcomes, we will add the sample sizes and event numbers of each treatment group. For continuous outcomes, we will combine means and standard deviations using methods outlined in chapter 7.7.3.8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). This offers a simple approach to eliminate unit of analysis issues that might arise if several comparisons from a given study are used in a single meta‐analysis without adjustment and thereby enables random effects meta‐analysis, but it will limit the opportunity to examine the role of the intervention components featured in each of the intervention arms in such studies when exploring heterogeneity.
Dealing with missing data
We will evaluate missing participant data on an available‐case analysis basis as described in Chapter 16.2.2 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We will address missing standard deviations by imputing data from the studies within the same meta‐analysis or from a different meta‐analysis as long as these use the same measurement scale, have the same degree of measurement error and the same time periods (between baseline and final value measurement, as per Chapter 16.1.3.2 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011)). Where statistics essential for analysis are missing (e.g. group means and standard deviations for both groups are not reported) and cannot be calculated from other data, we will attempt to contact the authors to obtain data. We will assume loss of participants that occurred prior to performance of baseline measurements to have no effect on the eventual outcome data of the study. We will assess and discuss losses after the baseline measurements were taken. We will report studies that have more than 30% attrition (i.e. deaths and withdrawals) in text only and exclude them from the meta‐analysis.
We will make an attempt to contact all study authors for verification of variables related to methodological quality, classification of the intervention(s), and outcomes data. We will attempt to contact the second author if we are unsuccessful in contacting the first author and will document all communications with study authors.
Assessment of heterogeneity
The review is expected to have substantial heterogeneity due to variation in the context (i.e. characteristics of clinical environment and of patients), differences in intervention characteristics (i.e. characteristics of the different strategies aiming to change HCP behaviours and HCP‐patient interactions), and various measurement tools used to assess outcomes. Heterogeneity will be assessed visually by inspection of forest plots; and statistically using Cochran’s Q test where a P value is used to determine statistical significance (Cochran 1954). The I² statistic will be used to quantify inconsistency across studies (Higgins 2011). The presence of significant heterogeneity will be further explored through pre‐specified subgroup analyses.
Assessment of reporting biases
We will plot trial effect against standard error and present it as a funnel plot, given that sufficient study numbers meet inclusion criteria (i.e. 10 or more) (Sterne 2011). Asymmetry could be caused by a relationship between effect size and sample size or by publication bias, and we will also examine any observed effect for clinical heterogeneity (Sterne 2011).
Data synthesis
We will group studies primarily by study type and we will analyse data from randomised trials, non‐randomised trials, and interrupted time‐series studies separately. We will further categorise studies such that studies with the same or similar outcomes are grouped together and we will summarise the results descriptively by outcome category. We will perform meta‐analysis of dichotomous and continuous outcomes, where appropriate, using Review Manager 5 software (RevMan). If two or more studies report the same or similar outcome, and there are no apparent unit‐of‐analysis errors (or ICC data are available), and no significant heterogeneity is present (I² < 75%) (Higgins 2011), we will calculate summary measures of intervention effect. We will conduct traditional random‐effects meta‐analysis, using standard methods to account for cluster‐randomized trials as relevant. If substantial heterogeneity is found, or we are unable to adjust for unit‐of‐analysis errors in two or more studies with the same or similar outcomes, we will present the results as the median effect and interquartile range. We will compile a 'Summary of findings' table for each primary outcome, containing pooled effect sizes or narrative descriptions, where appropriate. We will use the GRADE approach as outlined in the GRADE handbook to assign levels of confidence (high, moderate, low, and very low) to the findings (GRADEpro), considering the body of evidence for each outcome.
Subgroup analysis and investigation of heterogeneity
We will conduct subgroup analyses to examine the role of various intervention, provider, and patient characteristics. Where median effect sizes were estimated, we will explore potential associations between subgroup characteristics and effect size using a nonparametric Mann‐Whitney rank‐sum test (Shojania 2010).
We will conduct subgroup analyses on the following given sufficient study numbers.
Intervention characteristics
-
Type of intervention (i.e. implementation strategies, provider financial arrangements*, governance arrangements).
-
Number of component interventions (i.e. one‐component intervention compared to interventions with two or more components).
-
Presence or absence of patient co‐intervention.
* financial arrangements may include provision of monetary benefits or rewards with monetary value
HCP characteristics
-
Primary care providers versus specialists.
-
Nursing versus physician versus physiotherapist versus kinesiologist.
Patient characteristics
-
Age (patients < 65 versus ≥ 65 years).
-
Gender (male versus female patients).
-
Comorbidities (presence of chronic diseases versus no chronic disease diagnoses).
Sensitivity analysis
We will conduct sensitivity analyses to assess the robustness of our main findings. Given adequate study numbers, we will explore the influence of the following factors on effect size.
-
Taking account of risk of bias by excluding studies with an unclear or high risk of bias for sequence generation or allocation concealment, or both; or loss to follow‐up of greater than 20%.
-
Excluding any large studies to establish how they impact on the results.
-
Excluding any apparent outlying studies identified through assessment of heterogeneity.
