Grant: Travel Awards - Institute Community Support

Background: Accurate mortality risk prediction is vital to patients, physicians and health system planners. Laboratory test results can be used to predict mortality and are usually ordered by physicians to diagnose health issues; therefore, those having tests are usually sicker. Consequently, having a test ordered may predict outcomes. This bias impacts methods for dealing with missing data, all of which assume that data missingness is not associated with outcomes. Clinical experience and existing evidence indicate that this assumption is not true for lab tests. My doctoral research will address the fundamental issue of missing data in lab tests using linked health administrative data held at ICES. The Ontario Laboratory Information System (OLIS), which captures all physician ordered lab tests in Ontario will be linked to other data holdings at ICES including the Canadian Longitudinal Study of Ageing (CLSA). Overarching objective: To identify optimal approaches for incorporating population-level laboratory tests, including non-random missing data, into population mortality risk estimation. Specific objectives: (1) Describe laboratory testing practices across a provincial healthcare system; (2) Develop methodology to incorporate both laboratory test order status and test results into mortality risk prediction; (3) Derive and validate a laboratory test risk score to predict health service use and life expectancy; (4) Develop methodology to informatively impute missing laboratory test values for an entire population; (5) Conduct a simulation study of the proposed methods compared to naïve missing data methods Implications: This research will create a methodological framework to leverage complex laboratory data to improve the accuracy of population-level mortality and morbidity risk prediction using routinely collected health data. The findings will support population health risk assessment and health system planning, evaluation and management.