Author ORCID Identifier

0000-0002-8722-8141

Date of Award

Spring 5-17-2019

Degree Type

Thesis

Degree Name

Master of Public Health (MPH)

Department

Public Health

First Advisor

Matthew Hayat

Second Advisor

Harry Heiman

Third Advisor

Sheryl Strasser

Abstract

INTRODUCTION: Research studies on the general working population (aged 18-64) have found associations between access to paid sick leave (PSL) and increased uptake of preventive health screenings, but few such studies have been done on the fastest growing segment of American workers: those aged 55 and older, and rarely treating paid sick leave as anything but a “yes/no” exposure. AIM: The aims of this work are to determine the associations between access to PSL and preventive health behaviors (blood cholesterol test, flu shot, seeing the dentist) and depression among a nationally representative sample of workers over the age of 51.

METHODS: Using cross-sectional data from the 2014 Health and Retirement Study, we examined associations between PSL and respective outcomes while exploring and comparing techniques to adjust for confounding. PSL was considered as either dichotomous (yes/no) or trichotomous (zero days, 1-9 days, 10+ days) exposure. Two confounder pools were identified, on the basis of previous literature and bivariate associations within this study: covariates that may predict PSL availability, and covariates that may confound the relationship between PSL availability and a given study outcome (e.g. social support as a covariate that would confound the association between PSL availability and depression). Propensity score models were developed and different adjustments explored in multivariable logistic regression models. For each study outcome, a range of models to address confounding were run, including 1) traditional multivariable logistic models adjusting for both covariates that predict PSL availability and covariates specific to the paid sick leave and specific study outcome, and 2) multivariable logistic models adjusting for only the propensity score as a continuous covariate and additional select covariates likely to confound the paid sick leave and specific study outcome.

RESULTS: After adjusting for confounders via models that used traditional multivariable logistic regression and multivariable logistic regression incorporating the propensity score as a continuous covariate and health insurance as an additional covariate, the association between PSL availability (1+ days versus zero days) and receiving a cholesterol test was characterized by estimated odds ratios (OR) of 1.39 (95% CI: 0.99, 2.94) and 1.37 (95% CI: 0.98, 1.92), respectively. After adjusting for confounders via traditional multivariable logistic regression and multivariable logistic regression adjusting for the propensity score and dental insurance, participants with 1+ days PSL had 1.3 times the odds of reporting a dental visit (95% CI: 0.95, 1.79), and 1.22 times the odds of reporting a dental visit (0.90, 1.65), respectively, compared to those who had zero days. Although not statistically significant, trends in the measures of association between PSL as a trichotomous exposure and cholesterol testing and dental visits indicated that participants with higher levels of PSL had higher odds of the outcome than individuals in the next lower category of paid sick leave availability (e.g. using traditional multivariable logistic regression, participants who had 1-9 days PSL had 1.2 times the odds of reporting a cholesterol test compared to those who had zero (95% CI: 0.81, 1.78); by that same model, participants who had 10+ days PSL had 1.6 times the odds of reporting a cholesterol test compared to those who had zero (95% CI: 1.08, 2.36)). Similarly, examination of the relationship between PSL and depression indicated an inverse relationship, though this also was not statistically significant. No associations were found between paid sick leave availability and receiving a flu shot.

DISCUSSION: This study provides preliminary evidence that paid sick leave access might be associated with increased cholesterol tests and dental visits among older workers. These findings may have implications for employment policy as well as public health practice. However, more research is needed to examine how access to paid sick leave affects these health outcomes over time before stronger causal statements can be made.

Available for download on Friday, April 17, 2020

Share

COinS