Self-reported Health Related Quality of Life as a Predictor of Mortality in Diabetes: The Importance of Heterogeneity

Several studies have shown that self-reported health related quality of life (HRQoL) can be used as an additional predictor of mortality, highlighting the possibility that self-reported questionnaires may unveil symptoms that otherwise go unreported in the healthcare system. We analyze the relationship between health-related quality of life and mortality in patients with type-2 diabetes. None of the existing studies in this area however have tested to our knowledge the validity of the relationship between HRQoL and mortality in the context of unobserved patient heterogeneity. To understand the consequence of unexplained frailty we compare predictions from a semi-parametric model with and without frailty after controlling for demographic characteristics, clinical markers, age, history of complications, and behavioral variables (exercise and smoking status). Accounting for heterogeneity translates into a bigger effect of HRQoL on survival. The average difference in baseline hazards across questionnaire responders is 45.9%. While the differences in coefficients between the model with and without frailty are small, when the difference in the estimated mortality against observed mortality is translated into life years, the difference in risk compounds to as much as a one-year difference in life expectancy over a 20 year horizon for the average patient between the model with mean frailty and that without frailty. The implication of this work is important as accounting for heterogeneity has a tangible impact on life expectancy predictions and thus on possible cost effectiveness considerations.