Activity Level as a Mortality Predictor in a Population Sample after Typical Underwriting Exclusions and Laboratory Scoring

Abstract

Objectives.- To quantify the effect of physical activity on the mortality rates of healthy individuals in a population sample, after controlling for other sources of mortality risk. Background.- The widespread availability of activity monitors has spurred life insurance companies to consider incorporating such data into their underwriting practices. Studies have shown that sedentary lifestyles are associated with poor health outcomes and higher risks of death. The aim of this paper is to investigate how well certain measures of activity predict mortality when controlled for other known predictors of mortality including a multivariate laboratory based risk score. Methods.- Data were obtained from the National Health and Nutrition Examination Survey (NHANES) for the years 1999 through 2014. Laboratory and biometric data were scored for mortality risk using a previously developed proprietary algorithm (CRL SmartScore). Data on activity were obtained from the NHANES questionnaires pertaining to activity. In a second analysis, data were obtained from pedometers worn for 1 week by NHANES participants (years 2003-2004, and 2005-2006 only). Before analysis, cases were selected based on commonly used life insurance underwriting criteria to remove from consideration those who have major health issues, which would ordinarily preclude an offer of life insurance. Results.-In fully-adjusted Cox model which included survey-based MET*hours per day as a 3-level categorical variable, the moderate and minimal levels of activity were associated with hazard ratios of 1.15 (95% CI: 1.04-1.28) and 1.38 (95% CI: 1.23-1.56), respectively, when compared to the highest level of activity. When treated as a continuous variable, the fully adjusted model the HR for MET*hours per day was 0.91 (95% CI: 0.87-0.95). In fully adjusted models using pedometer data, the percentage of wear time spent sedentary was associated with mortality (HR: 1.19, 95% CI: 1.09-1.31), while average counts per minute were negatively associated with mortality (HR: 0.82, CI: 0.75-0.90). Conclusions.-It is clear from these results that high proportions of sedentary time are associated with increased mortality, whether the sedentary time is quantified via questionnaire or pedometer. Because both laboratory scores and activity levels remain significant in Cox models where both are included, these factors are largely independent, indicating that they are measuring distinct influences on the risk of mortality.

Keywords: Mortality; activity; laboratory testing; wearables ..