April 18, 2020
An Exploratory Analysis of the Data
The most obvious examples of the influence of occupation on lifestyle are physical activity and sleep schedule. So, from the government's trove of survey data on 22 standard occupational categories, I downloaded the following statistics for each one: the prevalence of Major Depression (MD) as well as indicators of physical activity, sleep, drug abuse (including smoking), and economic insecurity. (I could not find decent data on diet by occupation). I then regressed depression rates against the four lifestyle factors. Here are the operational definitions of the variables:
Rate of Major Depression (MD): = gender adjusted percentage of workers in an occupational category who experienced a major depressive episode in the last 12 months. MD is defined in accordance with psychiatry's diagnostic manual. Data compiled from single and multi-year NSDUH surveys.
Total Physical Activity (PA): = Work-time Physical Activity (WTPA) + Leisure-time Physical Activity (LTPA). Both are measured in terms of energy expenditure expressed in METs per week: WTPA is calculated from a MET analysis of the specific tasks in an occupational category. The MET value of LTPA is calculated from the time and intensity of self-reported exercise.
Sleep Disturbance (SD): = percentage of employees working irregular hours, e.g., rotating/split shifts and on-demand hours. Irregular work hours undermine the first principles of sleep hygiene, which is to rise and retire at the same times every day, and to expose oneself to morning daylight in order to stabilize circadian rhythm.
Serious Drug Abuse (DA): = rate of unintentional drug overdose deaths in an occupational category. (I had to use overdose data because published estimates of "substance abuse disorder" by occupation were not available. Also unavailable were intentional ODs ).
Economic Insecurity (EI):= prevalence of involuntary part-time work in an occupation. (Under-employment is a stress factor in depression).
So that the results could be summarized on one chart, I combined the four lifestyle factors into a single variable - the Noxious Lifestyle Index. The plot clearly shows that the prevalence of MD is significantly higher in occupations conducive to low physical activity, disturbed sleep, drug use, and insecure employment -- i.e., a noxious lifestyle.
Results of Regression Analysis
The four independent variables explain 73% of the variation in MD among occupations. According to the Beta coefficients, the strongest impact on MD rates is delivered by Total Physical Activity. But most of its effect can be attributed to the Work-Time component; LTPA contributes very little. That's because LTPA, which accounts for only 10% of total METs, is negatively related to WTPA (r= -0.7). In other words, construction workers don't hit the gym nearly as much as economists. (The insignificant effect of LTPA in this case does not contradict the positive findings of RCTs which measure change in depressive symptoms at the individual level).
Contrary to expectations, the Beta for sleep disturbance is the weakest of the four. That's probably because the proxy for Sleep Disturbance - irregular work hours - is a rather crude measure. However, research shows that Economic Insecurity and Drug Abuse also effect sleep quality. So, part of the strong Betas for those two factors can be attributed to the effect of sleep not captured by irregular work hours.
By the way, cigarette smoking was initially included in the regression; but when controlling for Drug Abuse, its correlation with MD collapses to zero.
Outliers
Relative to the regression line on the chart, Legal Occupations and Healthcare Practitioners exhibit much lower than expected MD rates. The most likely reason is that the risk of a depressive episode declines sharply with age, and those two occupations contain the lowest percentage of adults under 25. By the way, Healthcare Practitioners score on the high side of Noxious Lifestyle because of their above-average drug overdose rate and irregular work hours.
On the other hand, the MD risk is much higher that predicted in the notoriously diverse category of Arts, Entertainment and Media (AEM). One explanation is that actors, artists, athletes, and journalists tend to be employed temporarily and episodically. This kind of uncertainty is not adequately captured by my single indicator of Economic Insecurity. Also, AEM's drug abuse problem is probably understated for the same reason: my indicator - drug overdose rate - excludes intentional overdoses. It turns out that AEM's suicide rate tops the list. In sum, better measurement of two variables would have positioned AEM further to the right on the Noxious Lifestyle axis.
Comments