The Potential Risk of Shift Work Leading to Diabetes: A Logistic Regression Analysis of an Open Data Set

The Potential Risk of Shift Work Leading to Diabetes: A Logistic Regression Analysis of an Open Data Set Sarah J. Diekman, MD, JD, MS, (sdiekman1@jhu.edu) Occupational Environmental Medicine Resident, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD Introduction: The body of sleep epidemiology studies was scant leading into the 1980s. After the 1980s, tens of thousands of studies of sleep epidemiology have emerged. The increased interest correlated in part with a public interest in reducing motor vehicle accidents. As the public started to see accidents that were associated with sleep deprivation, the public wanted to know more about sleep and motor vehicles. Another reason for the increased interest in sleep is that the general public was finding sleep problems to be common and this resulted in pressure on the scientific community to find out why. Over the past 30 years, international data shows that the average night sleep has decreased by 18min per night (Ferrie et al.). Currently, the CDC recommends that adult sleep 7 or more hours per each 24-hour period to maximize health and wellbeing.(CDC Data and Statistics Sleep and Sleep Disorders) Sleep disturbance has been implicated in nearly every chronic disease: cardiovascular disease, hypertension, inflammation, obesity, diabetes and impaired glucose tolerance, and psychiatric disorders, such as anxiety and depression (Ferrie et al.). OSHA’s community recommendations includes strategies to reduce the major environmental risk of work induced sleep deprivation. OSHA recommends that the workplace consider a normal work shift be a work period of no more than eight consecutive hours during the day, five days a week with at least an eight-hour rest in between shifts. Periods that allow for less sleep should be considered to be unusual (Extended Unusual Work Shifts). State, local, and federal governments have various laws addressing sleep and motor vehicle operation. Professional drivers are most often affected by these laws. Further the airline industry also regulates sleep. These laws are a balance of the business interest for workers to be more productive, the workers interest to make an adequate pay-check, and the interest of public safety (Åkerstedt and Wright). Research Objectives: The aim of this study was to determine whether an association exists between shift work and diabetes. The target population was workers affected by swing shifts. There was no data source that directly 2021 American College of Legal Medicine JOURNAL OF LEGAL MEDICINE 2021, VOL. 41, NO. S1, 15–17 https://doi.org/10.1080/01947648.2021.1914475 inventoried this population, so sleep below the CDC recommended 7 hours was used as a surrogate for the shortened sleep seen in shift work. The actual study population was the general population of India where the researchers conducted their original study. Methods: The hypothesis that was tested was whether there was an association between chronic sleep deprivation and diabetes. The statistical analysis were designed in advance, a priori, in accordance with ethical statistical practice. The data for this study was obtained using Kaggle open data set: Diabetes Dataset 2019 Prediction and Classification Using Machine Learning. This de-identified dataset was collected by Neha Prerna Tigga and Dr. Shruti Garg of the Department of Computer Science and Engineering, BIT Mesra, Ranchi, India. The data was collected from patients in a cross-sectional design. The information was collected from 952 participants who were asked to answer a questionnaire about variables that might lead to diabetes. The data set was de-identified and made public for immediate download on the Kaggle platform. For the immediate study, the public data set was downloaded in CSV format and imported to Stata. Initially the Descriptive statistics were generated using standard Stata graph box commands. This analysis queried the association between diabetes and sleep as a continuous variable. Then the hypothesis that sleep duration below the CDC recommended 7 hours was tested using a logistic regression with sleep as a binary variable (at or above the recommendation verses below the recommendation. The logistic regression was further stratified by age. Results: The results of a box plot of the hours of sleep as a continuous variable between those with and without diabetes, generated a sample median of 7 hours for both groups, reflecting a similar central tendency between groups. The upper fence did indicate a difference between groups for the maximum amount of sleep. A logistic regression was performed on sleep as a continuous variable on the variable of diabetes status. The results showed a non-statistically significant difference (.812) in the odds between groups (p1⁄4 0.160) (95%CI .6078555, 1.085405). The group above 60 years of age sleeping less than the CDC recommended time, showed a statistically significant increase in odds of diabetes compared to the group sleeping the CDC recommended time (p1⁄4 .004) (CI 1.541754, 10.55749). Conclusions: This analysis explored the association between diabetes and chronic sleep durations as a continuous variable and as a binary variable using the 7 hours recommended by the CDC. As a continuous variable no statistically significant association was shown, although there was a trend showing that sleep hours were longer for those without diabetes. Once the sleep variable was converted to a binary variable that differentiated those sleeping less than the CDC recommended time verses those sleeping the 16 ABSTRACT