Valuing survival gains for people with diabetes in developed Chinese population of Hong Kong from 2007 to 2014

Abstract

China has the highest absolute disease burden of diabetes in the world accounting for almost one in three diabetes patients globally. Validated health status measurement tools are essential for assessing quality and value, and thus guide service development. We developed and validated risk prediction models for mortality among the economically developed Chinese population of Hong Kong to estimate the value of survival gains for people with diabetes from 2007-2008 to 2013-2014. We used the Hong Kong Hospital Authority Clinical Management System to identify all patients with diabetes in the public healthcare system from 2007 to 2014. Our dataset included a single unique identifier to allow tracking across datasets and linkage to mortality records during this period. We developed Cox proportional hazards models to predict the 5-year risk of mortality and externally validated the models in a cohort of Singapore patients. We identified 678 750 participants in the Hong Kong derivation cohort as aged 20 years or over with a confirmed diagnosis of type 2 diabetes, and 386 425 participants with type 2 diabetes in the Ministry of Health (Singapore) validation cohort. We calculated the value of survival improvements by estimating the difference in the present discounted value of remaining life for two-time points: the baseline (2007-2008) and final (2013-2014) observation periods. Using the validated 5-year mortality risk score, we estimated the individual modifiable mortality risk at the baseline and final periods, keeping the non-modifiable predictors such as the individual's age and duration of diabetes at the baseline value. We use the risk score to approximate the predicted probability of death by giving one fifth of the predicted probability (0.2 Pjt) to each of the first 5 years, and assuming that all patients surviving beyond year 5 (with probability 1- Pjt) have the same age- and sex-specific remaining life expectancy as a general individual from the population lifetables. We assumed a 3% real annual discount rate and a given value of 1 life-year (from $50 000 to $200 000) in calculating the present discounted value of remaining life in the model simulations. We found there were significant gains in survival achieved in Hong Kong from 2007-2008 to 2013-2014: $24 858.87 at $50 000 life-year value, $49 717.73 at $100 000 life-year value, $99 435.47 at $200 000 life-year value. These individual-level prediction models using routinely recorded clinical data serves as an economic valuation tool for health system performance and in comparative analysis of other economically developed Chinese populations such as Taiwan, Singapore and parts of mainland China. The contribution of health services to the observed survival gains merits further research.