Marc S. Mitchell

Sedentary Time and Its Association With Risk for Disease Incidence, Mortality, and Hospitalization in Adults: A Systematic Review and Meta-analysis

Adults are advised to accumulate at least 150 minutes of weekly physical activity in bouts of 10 minutes or more (1). The intensity of such habitual physical activity has been found to be a key characteristic of primary and secondary health prevention, with an established preventive role in cardiovascular disease, type 2 diabetes, obesity, and some cancer types (2, 3). Despite the health-enhancing benefits of physical activity, this alone may not be enough to reduce the risk for disease and illness. Population-based studies have found that more than one half of an average persons waking day involves sedentary activities ubiquitously associated with prolonged sitting, such as watching television and using the computer (4). This lifestyle trend is particularly worrisome because studies suggest that long periods of sitting have deleterious health effects independent of adults meeting physical activity guidelines (57). Moreover, physical activity and sedentary behaviors may be mutually exclusive. For example, some persons who achieve their recommended physical activity targets may be highly sedentary throughout the remainder of their waking hours, whereas others who may not regularly participate in physical activity may be nonsedentary because of their leisure activities, workplace environments, or both (8). Although studies and subgroups of systematic reviews have explored the independent association between sedentary behaviors and outcomes after adjustment for physical activity, the magnitude and consistency of such associations and the manner by which they change according to the level of participation in physical activity remain unclear (911). The objective of this meta-analysis was to quantitatively evaluate the association between sedentary time and health outcomes independent of physical activity participation among adult populations. We hypothesized that sedentary time would be independently associated with both cardiovascular and noncardiovascular outcomes after adjusting for participation in physical activity but that the relative hazards associated with sedentary times would be attenuated in those who participate in higher levels of physical activity compared with lower levels (10). Methods Data Sources and Searches The Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines were followed in the conduct and reporting of this meta-analysis (9). Published studies on the association between sedentary behavior and various health outcomes were identified and cross-checked by 2 reviewers through a systematic search of the MEDLINE, PubMed, EMBASE, CINAHL, Cochrane Library, Web of Knowledge, and Google Scholar databases. The health outcomes included all-cause mortality, cardiovascular disease incidence (including diabetes), cardiovascular disease mortality, cancer incidence, cancer mortality, and all-cause hospitalizations. Searches were restricted to English-language primary research articles through August 2014 with no publication date limitations (Supplement). The following keywords were applied to the search: (exercise OR physical activity OR habitual physical activity) AND (sedentar* OR inactivity OR television OR sitting) AND (survival OR morbidity OR mortality OR disease OR hospital* OR utilization). References from relevant publications and review articles were hand-searched to supplement the electronic searches. A broad and comprehensive search strategy was chosen to encompass the range of outcomes associated with sedentary behavior among different populations or settings and variations in the operational definition of leisure-time sedentary behavior. Supplement. Search Strategy Supplement. Original Version (PDF) Study Selection The inclusion criteria were primary research studies that assessed sedentary behavior in adult participants as a distinct predictor variable, independent of physical activity and correlated to at least 1 health outcome. We broadly defined sedentary behavior as a distinct class of waking behaviors characterized by little physical movement and low-energy expenditure (1.5 metabolic equivalents), including sitting, television watching, and reclined posture (11). We allowed for studies that assessed the effects of varying intensities of physical activity, provided that they also correlated a measure of sedentary behavior with an outcome. We excluded studies that assessed nonadult populations (such as children and youth), those that did not adjust for physical activity in their statistical regression models or only assessed sedentary behavior as a reference category to the effects of physical activity, and those that measured sedentary behavior as the lowest category of daily or weekly physical activity. Data Extraction and Quality Assessment Data were extracted from all articles that met selection criteria and deemed appropriate for detailed review by 3 authors. If several articles of the same study were found, then data were extracted from the most recently published article. Details of individual studies were collected and characterized on the basis of authors or year of publication; study design; sample size or characteristics (age and sex); data collection methods; study outcomes; study limitations; and hazard ratios (HRs), odds ratios, or relative risk ratios (and their associated 95% CIs or SEs). We restricted studies reporting health outcomes to those with direct associations with death, disease incidence (that is, risk for disease in a given period), and health service use (that is, change in health service use) outcomes. This led to the exclusion of studies reporting indirect surrogate outcomes with inconsistent clinical end points and cutoffs (such as insulin sensitivity, quality of life, activities of daily living, metabolic biomarkers, the metabolic syndrome, and weight gain). Our studys primary exposure was overall sedentary or sitting time (hours per week or hours per day). Studies reporting information on total screen time (television or computer screen use), television viewing time, and metabolic equivalents (hours per week) were also abstracted when information on the primary exposure was unavailable. We assessed articles for quality on the basis of methods used by Proper and colleagues (12). Their quality assessment tool had been previously validated (face and content) and evaluated to limit the risk of bias from study participation, study attrition, measurement of prognostic factors, measurement of and controlling for confounding variables, measurement of outcomes, and analysis approaches (13, 14). Each study was evaluated according to a standardized set of predefined criteria consisting of 15 items (Table 1) (15). The use of the original quality assessment tool was expanded to permit and score nonprospective studies. The items of the tool assessed study quality within the domains of study population, study attrition, data collection, and data analysis. Each quality criterion was rated as positive, negative, or unknown. As with other meta-analyses, we required positive quality criteria of 8 items or more to be included in our study (12, 16). Two reviewers independently scored each article for quality. Any scoring inconsistencies were discussed with an additional reviewer. Scores from each reviewer were averaged to attain a final quality score assessment and verified by a single reviewer. When such data were available, we also considered whether the effects of prolonged bouts of sedentary time were modified by the highest or lowest reported participation in physical activity (herein termed as joint effects). Table 1. Criteria List for the Assessment of the Quality of Prospective and Nonprospective Studies* Data Synthesis and Analysis All meta-analyses were done using Comprehensive Meta-analysis, version 2 (Biostat), and the metafor package of R (R Foundation for Statistical Computing) (17, 18). Odds ratios, relative risk ratios, and HRs with associated 95% CIs were collected from studies for each outcome, if available. We considered relative risk ratios to be equal to HRs, and when only odds ratios were provided, they were approximated to relative risk ratios in which we used the assumption of rare events according to methods described and demonstrated elsewhere (19, 20). When studies presented several statistical risk-adjustment models, we only considered relative risk ratios associated with the statistical models that contained the fewest number of additional covariates beyond physical activity to enhance comparability across studies. Adjustment for physical activity (rather than moderate to vigorous physical activity) allowed for a broader range of studies, some of which may not have specified the intensity of physical activity in regression models. KnappHartung small sample estimation was used to pool the analysis of the overall effect size for each outcome. Studies that separately presented results for men and women were combined using a fixed-effects model. We received a 79% response rate from authors we had contacted to provide additional statistical information for our meta-analysis (11 out of 14). Potential modifying effects of physical activity on sedentary time were examined by comparing the statistical effect sizes of any studies that reported the longest period of sedentary time with the highest and lowest duration and intensity of physical activity. Statistical heterogeneity was assessed using the Cochran Q statistic and the I 2 statistic of the proportion of total variation because of heterogeneity (21). When we saw substantial heterogeneity, we considered a KnappHartung modified random-effects model (22). For the summary estimate, a P value less than 0.05 was considered statistically significant. The potential for small study effects, such as publication bias, was explored graphically using funnel plots through the Egger test of asymmetry and quantitatively by the Egger linear regression method (23). We also did a sensitivit

‘Will walk for groceries’: Acceptability of financial health incentives among Canadian cardiac rehabilitation patients

Financial health incentives, such as paying people to exercise, remain controversial despite widespread implementation. This focus group study explored the acceptability of incentives among a sample of Canadian cardiac rehabilitation (CR) patients (n = 15). Focus groups were conducted between March and April 2013 until further sampling ceased to produce new analytical concepts. A thematic analysis approach was adopted in analysing the data. Three broad themes emerged from the focus groups. First, ethical concerns were prominent. Half of participants disagreed with the incentive approach believing that it was unfair, unnecessary or a waste of limited resources. Second, ethical concerns were mitigated in considering a range of incentive features including type, size and source. Specifically, privately sponsored (not government funded) health-promoting voucher-based incentives (e.g. grocery or gym vouchers) were perceived to be highly acceptable. Third, if designed like this, then financial incentives were considered potentially effective in motivating behaviour change and in reducing economic barriers to exercise participation. Overall, the majority of participants welcomed incentives if ethical concerns were addressed through thoughtful incentive programme design. The results of this focus group study will inform the design of a financial health incentive feasibility RCT to promote post-CR programme exercise compliance in this population.

Development of the Health Incentive Program Questionnaire (HIP-Q) in a cardiac rehabilitation population.

The purpose of this study was to develop a questionnaire to facilitate the design of acceptable financial health incentive programs. A multiphase psychometric questionnaire development method was used. Theoretical and literature reviews and three focus groups generated a pool of content areas and items. New items were developed to ensure adequate content coverage. Field testing was conducted with a convenience sample of cardiac rehabilitation (CR) patients (n = 59) to establish face and construct validity (p = 0.021) and reliability (intraclass coefficients = 0.42–0.87). The final questionnaire is comprised of 23 items. This questionnaire builds on previous attempts to explore acceptability by sampling a wider range of instrumental and affective attitudes and by measuring the effect of program features on the likelihood of incentive program participation. Future research is now needed to examine whether tailoring incentives to preferences assessed by the questionnaire improves uptake and effectiveness.

The Feasibility of Financial Incentives to Increase Exercise Among Canadian Cardiac Rehabilitation Patients.

PURPOSE: To examine the feasibility of conducting a randomized controlled trial investigating the effectiveness of financial incentives for exercise self-monitoring in cardiac rehabilitation (CR). METHODS: A 12-week, 2 parallel-arm, single-blind feasibility study design was employed. A volunteer sample of CR program graduates was randomly assigned to an exercise self-monitoring intervention only (control; n = 14; mean age ± SD, 62.7 ± 14.6 years), or an exercise self-monitoring plus incentives approach (incentive; n = 13; mean age ± SD, 63.6 ± 11.8 years). Control group participants received a “home-based” exercise self-monitoring program following CR program completion (exercise diaries could be submitted online or by mail). Incentive group participants received the “home-based” program, plus voucher-based incentives for exercise diary submissions (

Evaluating the Carrot Rewards App, a Population-Level Incentive-Based Intervention Promoting Step Counts Across Two Canadian Provinces: A Quasi-Experimental Study (Preprint)

2 per day). A range of feasibility outcomes is presented, including recruitment and retention rates, and intervention acceptability. Data for the proposed primary outcome of a definitive trial, aerobic fitness, are also reported. RESULTS: Seventy-four CR graduates were potentially eligible to participate, 27 were enrolled (36.5% recruitment rate; twice the expected rate), and 5 were lost to followup (80% retention). Intervention acceptability was high with three-quarters of participants indicating that they would likely sign up for an incentive program at baseline. While group differences in exercise self-monitoring (the incentive “target”) were not observed, modest but nonsignificant changes in aerobic fitness were noted with fitness increasing by 0.23 mL·kg−1·min−1 among incentive participants and decreasing by 0.68 mL·kg−1·min−1 among controls. CONCLUSION: This preliminary study demonstrates the feasibility of studying incentives in a CR context, and the potential for incentives to be readily accepted in the broader context of the Canadian health care system.

Capsule Commentary on Misra-Hebert et al., Financial Incentives and Diabetes Disease Control in Employees: A Retrospective Cohort Analysis

Background The Carrot Rewards app was developed as part of an innovative public-private partnership to reward Canadians with loyalty points, exchangeable for retail goods, travel rewards, and groceries for engaging in healthy behaviors such as walking. Objective This study examined whether a multicomponent intervention including goal setting, graded tasks, biofeedback, and very small incentives tied to daily step goal achievement (assessed by built-in smartphone accelerometers) could increase physical activity in two Canadian provinces, British Columbia (BC) and Newfoundland and Labrador (NL). Methods This 12-week, quasi-experimental (single group pre-post) study included 78,882 participants; 44.39% (35,014/78,882) enrolled in the Carrot Rewards “Steps” walking program during the recruitment period (June 13–July 10, 2016). During the 2-week baseline (or “run-in”) period, we calculated participants’ mean steps per day. Thereafter, participants earned incentives in the form of loyalty points (worth Can

Financial Incentives for Exercise Adherence in Adults Systematic Review and Meta-Analysis

0.04 ) every day they reached their personalized daily step goal (ie, baseline mean+1000 steps=first daily step goal level). Participants earned additional points (Can

A randomized controlled trial testing an Internet delivered cost–benefit approach to weight loss maintenance

0.40) for meeting their step goal 10+ nonconsecutive times in a 14-day period (called a “Step Up Challenge”). Participants could earn up to Can

Leveraging Legacies: Will the Toronto 2015 Pan Am Games really benefit public health?

5.00 during the 12-week evaluation period. Upon meeting the 10-day contingency, participants could increase their daily goal by 500 steps, aiming to gradually increase the daily step number by 3000. Only participants with ≥5 valid days (days with step counts: 1000-40,000) during the baseline period were included in the analysis (n=32,229).The primary study outcome was mean steps per day (by week), analyzed using linear mixed-effects models. Results The mean age of 32,229 participants with valid baseline data was 33.7 (SD 11.6) years; 66.11% (21,306/32,229) were female. The mean daily step count at baseline was 6511.22. Over half of users (16,336/32,229, 50.69%) were categorized as “physically inactive,” accumulating <5000 daily steps at baseline. Results from mixed-effects models revealed statistically significant increases in mean daily step counts when comparing baseline with each study week (P<.001). Compared with baseline, participants walked 115.70 more steps (95% CI 74.59 to 156.81; P<.001) at study week 12. BC and NL users classified as “high engagers” (app engagement above sample median; 15,511/32,229, 48.13%) walked 738.70 (95% CI 673.81 to 803.54; P<.001) and 346.00 (95% CI 239.26 to 452.74; P<.001) more steps, respectively. Physically inactive, high engagers (7022/32,229, 21.08%) averaged an increase of 1224.66 steps per day (95% CI 1160.69 to 1288.63; P<.001). Effect sizes were modest. Conclusions Providing very small but immediate rewards for personalized daily step goal achievement as part of a multicomponent intervention increased daily step counts on a population scale, especially for physically inactive individuals and individuals who engaged more with the walking program. Positive effects in both BC and NL provide evidence of replicability.

Do Incentives Promote Action Planning in a Web-based Walking Intervention?

A dearth of randomized trials testing incentives for chronic disease self-management makes this retrospective analysis of propensity-matched diabetic employees and nonemployees particularly valuable. The authors found that incentives increased wellness program participation amongst employees (7 to 50 %), although those with poorer glycemic control (HbA1c >9 %) were less likely to participate. The authors suggest that more immediate rewards (not delayed insurance reimbursements) may have stimulated greater participation, and suggest that in the future these could take the form of Bwaived^ copayments. The literature, however, suggests that monetary rewards contingent on health behaviors are more effective than subsidies alone. In addition, more information about the incentive program would have provided greater design insight. For instance, it was not clear how employees were being rewarded, and for what exactly. Did employees receive lump-sum cash payments, or were rewards rolled into paychecks? Did program Bparticipation^ mean enrollment only, 50 % attendance? Fuller reporting of incentive program features using published checklists would have helped identify specific areas for improvement. At the population level, incentives stimulated a clinically meaningful 0.33-point reduction in HbA1c, though this took 5 years to achieve.We suggest that, in the context of an already engaged patient population with near optimal glycemic control at baseline (7.38 %), larger incentives worth