Workers’ Activity Profiles Associated With Predicted 10‐Year Cardiovascular Disease Risk

Abstract

Background

There is a need to explore common activity patterns undertaken by workers and the association between these activity profiles and cardiovascular disease (CVD). This study explored the number and type of distinct profiles of activity patterns among workers and the association between these profiles and predicted 10‐year risk for a first atherosclerotic CVD event.

Methods and Results

Distinct activity patterns from a cross‐section of workers’ accelerometer data were sampled from Canadian Health Measures Survey participants (5 cycles, 2007–2017) and identified using hierarchical cluster analysis techniques. Covariates included accelerometer wear time, work factors, sociodemographic factors, clinical markers, and lifestyle variables. Associations between activity profiles and high atherosclerotic CVD risk >10% were estimated using robust Poisson regression models. Six distinct activity profiles were identified from 8909 workers. Compared with the “lowest activity” profile, individuals in the “highest activity” and “moderate evening activity” profiles were at 42% lower risk (relative risk [RR], 0.58; 95% CI, 0.47, 0.70) and 33% lower risk (RR, 0.67; 95% CI, 0.44, 0.87) of predicted 10‐year atherosclerotic CVD risk of >10%, respectively. “Moderate activity” and “fluctuations of moderate activity” profiles were also associated with lower risk estimates, whereas the “high daytime activity” profile was not statistically different to the reference profile.

Conclusions

Workers accumulating physical activity throughout the day and during recreational hours were found to have optimal CVD risk profiles. Workers accumulating physical activity only during daytime work hours were not associated with reduced CVD risk. Findings can inform alternative strategies to conferring the cardiovascular benefits of physical activity among workers. Large prospective studies are needed to confirm these findings.

Nonstandard Abbreviations and Acronyms

CHMS

Canadian Health Measures Survey

PA

physical activity

Clinical Perspective

What Is New?

• Compared with workers with the “lowest activity” profile, those with the “highest activity” profile, which was likely during work, active transportation, and recreational time, had a lower predicted 10‐year risk of cardiovascular disease.

What Are the Clinical Implications?

• Strategies promoting physical activity outside work hours may confer greater cardiovascular benefits than during daytime work hours.

Regular physical activity (PA) protects against chronic disease and premature death and, in particular, against cardiovascular disease (CVD). ¹ , ² , ³ Recently, the World Health Organization recommended that adults accumulate a weekly average of 150 to 300 minutes of moderate intensity or ≥75 minutes of vigorous PA, or an equivalent combination. ⁴ This guidance reflects compelling evidence supporting the accumulation of PA throughout the day in contrast to previous recommendations requiring PA to be accumulated over sessions of at least 10 minutes for health benefits. ⁵ , ⁶ , ⁷

To meet PA guidelines, the majority of the working population must find time to accumulate PA before, after, or outside work hours. ⁸ Evidence shows that occupational tasks can affect PA, with workers who report long work hours, lower skill discretion, or physically or psychologically demanding jobs being less likely to accumulate moderate or vigorous intensity PA during recreation. ⁹ , ¹⁰ , ¹¹ In addition, work requiring high occupational PA and with little opportunity for rest (eg, some manual labor jobs) has been linked to increased CVD risk, referred to as a “PA health paradox.” ¹² Considering the different ways workers can accumulate PA, there is an evidence gap in understanding the typical patterns of how PA is accumulated and if patterns of PA are associated with cardiovascular benefits.

This study’s objectives were to describe the distinct patterns (ie, clusters) of PA accrued by workers and identify PA clusters that are associated with a lower likelihood of a high 10‐year predicted risk for a first CVD event.

Methods

Design, Setting, and Population

This study analyzed the Canadian Health Measures Survey (CHMS), an ongoing cross‐sectional survey conducted by Statistics Canada that collects self‐reported surveys, accelerometry, and measured health biomarker information from the Canadian household‐dwelling population aged 6 to 79 years living in the provinces and territories. ¹³ Participants living in rural areas more than 100 km from a mobile collection site (or urban areas more than 50 km from a collection site) are excluded. The included population covers ≈96% of the Canadian population. ¹³ Data from 5 consecutive cycles (2007–2009, 2009–2011, 2012–2013, 2014–2015, and 2016–2017) were used. The combined response rate of the household survey and clinical components ranged from 49% to 56%. ¹³ The data set for this study is held confidentially by Statistics Canada. Because of the sensitive nature of the data collected for this study, requests for access to the data set from qualified researchers trained in human subject confidentiality protocols may be sent to Statistics Canada at statcan.dad-apu-dad-uta.statcan@statcan.gc.ca.

To examine only workers, the sample was restricted to CHMS respondents aged ≥18 years who worked the week before responding (survey question: “last week, did you work at a job or businesses? Please include part‐time jobs, seasonal work, contract work, self‐employment, baby‐sitting and any other paid work, regardless of the number of hours worked.”). CHMS participants were asked to report working in the past week instead of the current week as this aligned with other CHMS survey questions concerning activities conducted in the previous week. There were 10 582 participants who met the eligibility criteria and among these individuals, 8909 had valid accelerometer data. The study was approved by the University of Toronto’s Health Sciences Research Ethics Board (REB #00037753). No informed consent was required.

Daily Physical Activity

Participants were asked to wear an Actical accelerometer (Philips Respironics, Bend, OR) on their right hip during waking hours for 7 consecutive days starting at a date scheduled soon after completing the survey component of the CHMS. The Actical was previously validated to measure PA in adults. ¹⁴ Each device measured time‐stamped acceleration in all directions. For this study, respondents were required to have at least 4 days with ≥10 hours of accelerometer wear time. Device‐measured activity counts were summed over 1‐minute intervals. The total activity undertaken was derived by calculating the average counts per minute for 10‐minute periods over the course of a 24‐hour day. Weekday and weekend counts were examined separately. Over 10‐minute periods, low PA was defined as accumulating <1000 counts, moderate PA defined as 1000 to 2000 counts, and high PA defined as >2000 counts. Figure 1 outlines the accelerometer data processing steps.

Figure 1. Accelerometer data processing steps.

CHMS indicates Canadian Health Measures Survey.

To supplement the accelerometer‐derived activity profiles, we obtained self‐reported PA from the household questionnaire of CHMS cycles 4 and 5. Respondents were asked to report their weekly time, of at least 10 minutes, in the following activities: “recreational PA,” “active transportation,” and “other PA” (eg, while working, in or around the home). For each PA question, the proportion of respondents reporting any time in a PA domain, and total time per week in each PA domain were examined.

Outcome: Predicted 10‐year ASCVD Risk

The study outcome was the predicted 10‐year risk for a first atherosclerotic CVD event (ASCVD) using pooled cohort equations developed by the American College of Cardiology/American Heart Association Task Force on Practical Guidelines. ¹⁵ The risk for developing a first ASCVD event was defined as fatal or nonfatal myocardial infarction or coronary heart disease, or fatal or nonfatal stroke, over a 10‐year period among people free from ASCVD at the time of CHMS participation. ASCVD risk scores have been found to be accurate at estimating predicted 10‐year ASCVD risk regardless of PA level. ¹⁶ ASCVD risk scores were generated from sex‐ and race‐stratified proportional hazards models that controlled for age, treated or untreated systolic blood pressure, total cholesterol, high‐density lipoprotein cholesterol, current smoking status, and diabetes diagnosis. For the main analysis, we dichotomized the predicted ASCVD risk at ≤10% and >10%, which is the upper risk threshold recommended by the US Preventive Services Task Force for low‐to‐moderate dose statins to reduce the probability of CVD. ¹⁷ A lower outcome threshold of predicted ASCVD risk of >7.5%, which is the American College of Cardiology/American Heart Association recommendation for initiating statin therapy, ¹⁸ was explored in a sensitivity analysis.

Covariates

Sociodemographic, health, and lifestyle variables included age, educational attainment, household income adequacy, marital status, cohabitating with dependent children, body mass index (BMI based on measured height and weight), and regularity of alcohol consumption. Work‐related variables included usual hours worked per week, work stress, and employed or self‐employed status. Furthermore, average ratings of physical demands of work and typical body position on the job (mostly stationary eg, sitting or standing; or mostly involving dynamic activities eg, frequent walking) were imputed from a validated job exposure matrix of occupation titles. ¹⁹ , ²⁰ , ²¹ Seasonality effects were based on whether accelerometers were worn during colder (November to March) or warmer (April to October) months. Accelerometer wear time was also included.

Statistical Analysis

Descriptive statistics examined the distribution of sample characteristics. Chi‐square and ANOVA tests examined how variables differed across different PA patterns.

There was a maximum of 1008 PA data points for each respondent, based on six 10‐minute recordings every hour for 24 hours over 7 consecutive days. The accelerometer data were highly dimensional, noisy, discontinuous, and nonindependent. The distance between all PA time points of every pair of respondents in the sample was calculated using the dynamic time warping method to account for the correlations between temporal series by considering the overall shape. Hierarchical cluster analysis was applied on the dynamic time warping distance matrix. Ward’s minimum variance method was used to identify the strongest clustering structures. A combination of the elbow, average silhouette, and gap statistics methods was assessed to determine the optimal number of clusters. R version 4.1.0 was used for the cluster analysis.

Robust Poisson regression models were used to examine associations between activity clusters and the likelihood of a 10‐year ASCVD risk of >10%. Individuals were excluded from modeling if they had a preexisting or prior CVD diagnosis or were pregnant. Model 1 examined crude associations; Model 2 adjusted for accelerometer wear time and in addition, subsequent partial models also included the separate adjustment for sociodemographic, or work‐related, or lifestyle, or BMI variables. All models were age standardized and a final model adjusted for all covariates. To examine the role of BMI as a confounder or mediator, the influence on effect size of removing BMI compared with its inclusion in the final model was examined. Effect estimates are reported as relative risks (RRs) and 95% CIs. All estimates were weighted using sampling weights to account for the probability of selection and nonresponse. Variance estimates around each estimate were adjusted using 500 bootstrap replicate weights to account for the clustered design of the CHMS. ²² Tests were 2 sided and P<0.05 was considered statistically significant. The descriptive and regression analysis was conducted using SAS version 9.4.

Results

Weighted characteristics are shown in Table 1. Overall, 8909 individuals were analyzed. The sociodemographic profile of the analytical sample versus the excluded sample (working sample with <4 days of valid accelerometer data) was similar, except for average age, marital status, and educational attainment. The analytical sample was on average, slightly older (42 years versus 39 years), had a greater proportion of married individuals (66% versus 58%), and more postsecondary educated (69% versus 60%). The distributions of average blood pressure, cholesterol ratio, waist circumference, and BMI were also similar for the included and excluded samples.

Table 1.

Characteristics of Participants Included and Excluded in the Final Analytical Sample (Survey Weights Applied)

	Analytical sample	Excluded sample*	Chi‐square test on nonmissing values
	N=8909	N=1673	χ2	P value
Sociodemographic variables
Sex, %			4.0	0.05
Women	46.8	58.1
Men	53.2	41.9
Age, y, mean (SD)	42.4 (0.3)	38.5 (0.6)		<0.001
Marital status, %			24.2	<0.0001
Married/common‐law relationship	65.7	57.8
Widowed/separated/divorced	9.0	6.3
Single/never married	25.3	35.9
Educational attainment, %			34.3	<0.0001
Less than high school education	7.2	10.7
High school diploma	21.6	22.1
Some postsecondary education	2.7	7.3
Postsecondary education	68.5	59.9
Have children (<12 y) living at home, %			3.1	0.08
No	59.9	64.0
Yes	40.1	36.0
Income adequacy, %			5.0	0.08
Lowest/lower middle‐income group	12.2	14.7
Upper middle‐income group	25.1	27.8
Highest income group	62.7	57.5
Work variables
Employment status, %			0.7	0.41
Employed	82.1	80.5
Self‐employed	17.9	19.5

	N=8229	N=1673	χ2	P value
Employment status, %			0.7	0.41
Full time	82.1	80.5
Part time	17.9	19.5
Hours worked per week, mean (SD)	39.1 (0.28)	39.7 (0.7)		0.46
Self‐perceived work stress, %			5.8	0.21
Not at all	7.9	9.1
Not very	19.5	22.0
A bit	41.5	36.7
Quite a bit	25.5	24.8
Extremely	5.6	7.4
Health variables
Smoking status, %			5.7	0.02
Nonsmoker	80.0	75.0
Smoker	20.0	25.0
Type of drinker, %			2.3	0.51
Regular drinker	73.0	75.6
Occasional drinker	13.7	12.8
Former drinker	8.2	6.5
Never drank	5.1	5.1
Body mass index, %			4.5	0.10
Underweight/normal weight (18.5–24.9 kg/m2)	39.5	42.6
Overweight (15.0–29.9 kg/m2)	35.9	31.0
Obese (30.0 kg/m2 and over)	24.6	26.4
Waist circumference norms † , %			16.3	0.01
Needs improvement	23.2	23.2
Fair/good	16.0	12.3
Very good	21.4	18.0
Excellent	39.4	46.5
Blood pressure (BP) norms			2.0	0.37
Within acceptable range (<120/80 mm/Hg)	85.9	85.1
High end of acceptable range (120–130/80–89 mm Hg)	7.7	9.6
Above acceptable range/high (>130/>90 mm Hg)	6.4	5.3
Average systolic BP, mean (SD)	111.7 (0.3)	110.3 (0.7)		0.05
Average diastolic BP, mean (SD)	74.4 (0.2)	71.6 (0.4)		0.07
Total/high‐density lipoprotein cholesterol ratio, mean (SD)	3.8 (0.1)	3.9 (0.1)		0.53
Diabetes, %			0.6	0.44
No	96.0	96.6
Yes	4.0	3.4
Outcome–ASCVD risk score ‡ , %
≥10%	40.0
<10%	60.0
>7.5%	46.5
≤7.5%	53.5

ASCVD indicates atherosclerotic cardiovascular disease.

^*Excluded sample based on incomplete accelerometer data (working sample with <4 days of valid accelerometer data).

^† Waist circumference norms: low risk (men, 93.9 cm or less; women, 79.9 cm or less), increased risk (men, 94.0 to 101.9 cm; women, 80.0 to 87.9 cm), and high risk (men, 102.0 cm or more; women, 88.0 cm or more).

^‡ ASCVD risk, 10‐year risk for a first atherosclerotic cardiovascular disease event.

Physical Activity Profiles of Each Cluster

Six distinct accelerometer‐measured activity clusters were identified (Figure 2) and their activity profiles are described in Table 2 (additional cluster characteristics are provided in Table S1). Self‐reported domain‐specific PA characteristics from CHMS cycles 4 and 5 are also reported in the following descriptions.

Figure 2. Free‐living activity clusters from working participants (n=8909)^*.

^*Plots show accelerometer activity counts averaged every 10 minutes. Weekday and weekend counts shown separately. Dark shade denotes nighttime hours and light shade denotes daytime hours.

Table 2.

Characteristics of Activity Clusters (Survey Weights Applied)

	Cluster 1	Cluster 2	Cluster 3	Cluster 4	Cluster 5	Cluster 6
	Moderate consistent activity	Lowest activity	Fluctuating moderate activity	High daytime activity	Moderate evening activity	Highest activity
No.	3219	2808	1194	713	225	750
Mean activity counts per min during a weekday (SD)	114.0 (1.2)	53.4 (0.9)	126.3 (1.8)	221.8 (5.3)	196.2 (7.3)	212.1 (5.9)
Daytime h (SD)	183.2 (1.0)	84.9 (0.7)	198.8 (2.8)	357.2 (5.2)	232.5 (6.7)	318.9 (8.6)
Nighttime h (SD)	44.7 (0.7)	21.9 (0.5)	53.8 (1.9)	86.4 (3.5)	160.0 (6.9)	105.4 (5.9)
Mean activity counts per min during a weekend day (SD)	97.6 (1.6)	61.0 (1.8)	119.8 (8.5)	150.8 (6.0)	148.7 (10.1)	161.8 (6.4)
Daytime h (SD)	160.8 (1.5)	99.2 (1.4)	192.9 (7.6)	245.7 (5.8)	190.5 (8.9)	271.4 (9.6)
Nighttime h (SD)	34.4 (0.7)	22.8 (0.8)	46.7 (7.2)	55.9 (3.6)	107.0 (6.6)	52.3 (3.1)
Physically demanding job, %
High (requires handling of loads up to 20 kg)	30.8	16.8	11.9	26.6	28.7	13.5
Low (requires handling of loads up to 10 kg)	69.2	83.2	88.1	73.4	71.3	86.5
Stationary job, %
No (dynamic activities, eg, frequent walking)	67.7	52.0	45.2	57.6	81.8	47.0
Physically demanding job, %
High (requires handling of loads up to 20 kg)	30.8	16.8	11.9	26.6	28.7	13.5
Self‐reported PA*
Total min of recreational PA per wk, mean (SD)	81.1 (7.5)	87.0 (8.0)	152.3 (17.7)	124.3 (15.9)	88.2 (33.1)	179.7 (17.7)
Proportion reporting recreational PA per wk, %	52%	54%	79%	69%	44%	84%
Total min of other PA per wk, mean, (SD)	278.6 (32.1)	156.4 (22.7)	99.9 (19.2)	190.7 (55.5)	443.3 (188.0)	103.5 (25.1)
Proportion reporting other PA per wk, %	76%	66%	71%	64%	56%	71%
Total min of active transportation per wk, mean (SD)	81.9 (20.1)	47.4 (5.5)	91.7 (11.0)	118.8 (24.0)	41.6 (16.7)	76.6 (15.1)
Proportion reporting active transportation per wk, %	35%	30%	56%	54%	33%	50%

PA indicates physical activity.

^*Self‐reported PA was available only for Canadian Health Measures Survey cycles 4 and 5.

Cluster 1: “Moderate Activity” (N=3219)

Mostly moderate levels of PA on both weekdays and weekends. Moderate levels of PA started at ≈8 am on weekdays and noon on weekends, continuing to ≈5 pm, and steadily decreasing over the evening.

This cluster had the highest proportion of self‐reported time in other PA. The proportion reporting recreational PA was low and comparable to clusters 3 and 6.

Cluster 2: “Lowest Activity” (N=2808)

Mostly low levels of PA on both weekdays and weekends. Low levels of PA starting at ≈8 am on weekdays and slightly later in the weekend, continuing to ≈7 pm on weekdays and 5 pm on weekends, and steadily decreasing over the evening.

This cluster consisted of the lowest proportion self‐reporting active transportation. Time spent in recreational PA was low, with the proportion similar to clusters 1, and 5, whereas time spent in other PA was low to moderate.

Cluster 3: “Fluctuations of Moderate Activity” (N=1194)

Fluctuating levels of moderate PA on both weekdays and weekends, with moderate activity starting at ≈8 am on weekdays and weekends, fluctuating to highest levels of moderate activity at 9 am, noon, and 5 pm on weekdays, and fluctuating to highest levels at 11 am, 2 pm, and 4 pm on weekends, and steadily decreasing for the rest of the evening.

This cluster included the second highest proportion of self‐respondents with recreational PA and weekly recreational PA time. Similar to cluster 2 and 6, the proportion of respondents and weekly time in other PA was low, whereas the proportion of respondents reporting active transportation were similar to clusters 2, 4, and 6 and weekly time spent was the second highest among clusters.

Cluster 4: “High Daytime Activity” (N=713)

Most PA occurred from ≈8 am on weekdays and 10 am on weekends, with high levels of activity occurring from 9 am to 6 pm on weekdays and 10 am to 5 pm on weekends. Activity levels were higher on weekdays compared with weekends.

This cluster reported the highest proportion and most time spent in other PA and time spent in active transportation. The proportion and time spent in recreational PA were third highest among clusters.

Cluster 5: “Moderate Evening Activity” (N=225)

PA occurred from ≈8 am on weekdays and 10 am on weekends, with activity levels on weekdays steadily increasing to moderate levels up to midnight and then steadily decreasing in the early morning up to 5 am. Activity on weekends also continued up to midnight but peaked at 4 pm and then gradually decreased.

This cluster self‐reported the most weekly time spent in other PA and active transportation. The proportion reporting recreational PA was lowest among clusters and the time spent was second lowest.

Cluster 6: “Highest Activity” (N=750)

Individuals engaged in the highest levels of activity, with fluctuations of high and moderate activity occurring between 8 am and 11 am, 12:30 pm to 2 pm, and 5 pm to 7 pm on weekdays, with activity levels steadily decreasing afterwards. Weekends followed a similar activity pattern but with a peak of highest activity at ≈10 am and fluctuations of moderate‐to‐high activity from noon to 4 pm.

This cluster self‐reported the most time and proportion of individuals engaging in recreational PA. Individuals in this cluster also reported among the highest clusters for weekly active transportation time but reported lower levels of other PA similar to clusters 2, 3, and 4.

Association Between Physical Activity Clusters and Predicted 10‐Year ASCVD Risk

The results of the regression model, adjusted for all covariates, for the likelihood of a predicted 10‐year ASCVD risk of >10% are presented in Figure 3 (results from partial models and sensitivity analyses are in Tables S2 through S5). Compared with the “lowest activity” cluster, at lowest risk were individuals in the “highest activity” and “moderate evening activity” clusters with a 42% (RR, 0.58; 95% CI, 0.47–0.70) and 33% (RR, 0.67; 95% CI, 0.44–0.87) lower risk, respectively. Additionally, individuals in the “fluctuations of moderate activity” cluster were at 27% lower risk (RR, 0.73; 95% CI, 0.63–0.84) and those in the “moderate activity” cluster were at 14% lower risk (RR, 0.86; 95% CI, 0.78–0.96). Individuals in the “high daytime activity” cluster were associated with a nonstatistically significant lower risk (RR, 0.93; 95% CI, 0.77–1.08). Partially adjusted regression models showed mostly consistent results to the final model estimate, although the “high daytime activity” cluster was associated with a statistically significant lower risk when regression models were separately adjusted for accelerometer wear time, workplace factors, BMI (only for 7.5% threshold), or health behaviors when using the 10% and 7.5% ASCVD risk thresholds. Final effect estimates for all clusters after the removal of BMI were not meaningfully different from the model with BMI included, suggesting associations were independent of BMI. The results were consistent when using an ASCVD risk threshold of 7.5%, although risk reductions were lower for each cluster.

Figure 3. Final model effect estimates showing associations between activity clusters and predicted 10‐year ASCVD risk^*†.

^*Values shown are for regression models adjusted for age, educational attainment, household income adequacy, marital status, cohabitating with dependent children, body mass index, alcohol consumption, usual hours worked per week, work stress, employed or self‐employed status, physical demands of work, typical body position on the job, seasonality effects, and accelerometer wear time. ^†ASCVD risk, 10‐year risk for a first atherosclerotic cardiovascular disease event. ASCVD indicates atherosclerotic cardiovascular disease.

Discussion

In this study of 8909 Canadian workers, 6 distinct activity patterns were identified. Most individuals engaged in moderate levels of PA or low levels of PA on both weekdays and weekends. Compared with workers with the “lowest activity”, those with the “highest activity” profile, which was likely during work, active transportation, and recreational time, were at lowest predicted 10‐year ASCVD risk. Workers with “moderate activity” mostly accrued during daytime hours, and “high daytime activity,” had modest or nonstatistically significant associations with a high predicted ASCVD risk, respectively. Results suggest that although high PA is associated with optimal CVD benefits, strategies promoting PA only in daytime work hours may be less effective than those promoting PA outside daytime work hours.

We know of one other study to have profiled workers according to their daily PA. Using compositional latent profile analysis on a Danish worker sample, Gupta et al. reported 4 distinct profiles. ²³ Results were similar to the present study, with most workers defined as being inactive throughout the day and the other groups defined as those with more PA during work hours, more PA outside work hours, and workers engaged in PA throughout the day. ²³ Although we were unable to classify activity patterns according to work schedules as this information was not collected in all CHMS surveys, we expect that most of the sample participated in daytime work as 87% of working respondents in CHMS cycles 1 and 2 reported daytime work hours. ²⁰ Workers in the “high daytime activity” cluster and “highest activity” clusters accumulated the most PA, and these occurred throughout the day, and when cross‐referenced with self‐reported PA, likely during work, active transportation, and recreation time. The third highest accumulation of PA was for the “moderate evening activity” cluster, which might correspond with workers with jobs working in at least some part of their evenings. It is also possible that this group represents some night shift workers engaging in PA while at work, although this cannot be verified. The 2 clusters with a high accumulation of PA during typical daytime work hours (clusters 3 and 6) also comprised the fewest workers. These findings suggest that PA during work hours might not be achievable for most workers, supporting previous research from the United States and Canada showing that most full‐time working adults are physically inactive and sedentary. ²⁴ , ²⁵

The comparison of the 10‐year predicted ASCVD risks of clusters where PA was mostly accumulated over the daytime workday compared with clusters where PA was mostly accumulated during (or including) recreational hours/likely nonwork time suggests that recreational PA might be more health enhancing than PA during work hours. This might be because of work‐related PA being too low to improve fitness or too physically strenuous—supporting the PA health paradox. ¹² Compared with the “lowest activity” cluster, the final regression model estimate of the “high daytime activity” cluster had a nonstatistically significant association with a high predicted 10‐year ASCVD risk. In addition, workers in the “moderate activity” cluster, where PA was also likely accumulated during daytime work hours, had a modest risk reduction. A sizable proportion of workers in these profiles also reported physically demanding work and prolonged occupational sitting or standing, which could have attenuated the cardiovascular benefits of PA voluntarily undertaken for health. ¹² , ²⁶ , ²⁷ Compared with the “lowest activity” cluster, the effects of the “high daytime activity” cluster, which are not associated with reduced CVD risk, are different and have nonoverlapping CIs to the “highest activity” cluster, which are associated with reduced CVD risk, despite their similar accumulation of daily activity counts. The results suggest the possibility of a PA paradox, where the health effects of activity are different depending on the context in which it is accumulated. Some studies also suggest that the “high daytime activity” cluster would have a slightly increased ASCVD risk. ²⁸ , ²⁹ , ³⁰ The discrepancy in findings with other studies corresponds with recent review studies that have reported that evidence of the interaction between physically demanding work, recreational PA, and CVD is mixed, and the certainty of evidence is low. ²⁷ , ³¹ This study also addresses some methodological limitations of previous studies examining a PA health paradox by taking a whole‐day approach to assessing the contributions of daily PA accumulation on ASCVD risk and limiting residual confounding by adjusting for working conditions and alcohol consumption. The use of accelerometer‐measured PA helps reduce PA misclassification from self‐reported measures. However, the hip‐worn accelerometers mostly characterize ambulatory activities and might miss some of the unhealthy characteristics of occupational PA, for example, lifting heavy loads and standing. An important future research step is to confirm these inferences through longitudinal studies with device‐based PA measures that can provide a clearer understanding of any PA health paradox and the potential need for public health initiatives to limit or redesign physically demanding work conditions.

Compared with the lowest activity cluster, the direction of effect estimates for all activity clusters (including the clusters where PA was mostly accumulated during daytime work hours) were in the direction of a reduced 10‐year risk for a first ASCVD event, although this effect was not statistically different in the group where PA was accumulated during the day, despite this group having the highest mean activity levels. It is possible that many working adults are benefiting from daily PA, despite being categorized as insufficiently active according to PA guidelines and this is highlighted in previous studies on the cardiometabolic benefits of accumulating daily steps and light‐intensity PA. ³² , ³³ The “moderate evening activity” and “moderate consistent activity” clusters had a higher proportion of lower educated and lower income workers, whereas the highest activity” cluster had the largest proportion of postsecondary educated workers and of the highest income level (see Table S1). Although this finding is in keeping with previous studies showing that those with greater socioeconomic status have fewer barriers to recreational PA, ³⁴ , ³⁵ the links between moderate‐intensity activity clusters and reduced 10‐year ASCVD risk suggest the potential of multiple PA strategies for conferring CVD benefits, particularly among disadvantaged populations at high risk of physical inactivity.

This study has some limitations. First, accelerometry can underestimate activities that involve limited waist activity or carrying heavy loads. ³⁶ , ³⁷ In response, this study controlled for body position and physically demanding work by occupational title to approximate work‐related activities that might otherwise not be captured by accelerometer. Second, this study is cross‐sectional, and temporality between clusters and ASCVD risk cannot be established. Yet, prospective population‐based studies involving accelerometers are rare and this study addresses some limitations of similar studies involving only self‐reported PA. Third, work schedule information and self‐reported domain‐specific PA were collected in only 2 CHMS cycles and it was not possible to confirm from accelerometer time stamps whether PA was accrued at work or outside work hours. Fourth, residual confounding by unmeasured factors such as diet quality and other CVD risk factors may affect the results. Finally, despite adjustments to the CHMS sampling weights to compensate, ³⁸ , ³⁹ estimates may be biased by systematic differences between respondents and nonrespondents.

Conclusions

This study found workers who were active at moderate or high intensities throughout the day or mostly during evening hours had a reduced risk of future CVD compared with the least active workers. Workers with high daytime activity had no statistically significant difference in future risk of CVD compared with the least active workers, despite high mean activity levels. The findings suggest that strategies promoting PA only in daytime work hours may be less effective than those promoting PA outside daytime work hours, which can inform alternative strategies to conferring the cardiovascular benefits of PA in working populations.

Sources of Funding

This work was supported through a project grant from the Canadian Institutes for Health Research (grant number 162244).

Disclosures

None.

Supporting information

Tables S1–S5