Grant: Operating Grant

Accuracy of the estimating the chronic disease burden is crucial for budgeting and evaluating the performance of strategies. Due to the reliability and validity for surveillance of chronic disease using administrative data, the errors, such as false positive cases exist. The purpose of this study is to develop a method to adjust the biased prevalence of chronic disease for surveillance using health administrative data. Specifically, the objectives are; 1) to develop and validate the methods to adjust the biased prevalence using chart review data; 2) apply the methods to estimate hypertension prevalence overtime using health administrative data. We will use chart review data as a reference standard and link it with administrative data to develop and validate the methods to adjust the biased prevalence of hypertension overtime. We will use optimizing data period and classification modeling method to minimize the error for identifying cases more accurately. Classification modeling is a process, not only using ICD9/10 diagnosis codes but also multiple data features to develop the algorithms to discriminate the individuals as cases and non-cases. We will estimate parameters associated with "true" hypertension cases of chart review data in classification model, and use these parameters to discriminate the individuals in administrative data as case and non-case. The discriminative performance and prediction error of the classification algorithms will be evaluated using the area under receiver operative curve (ROC). We will use the condition of hypertension as an example to illustrate the method. The method will be externally validated using British Columbia data. The performance for the method to identify chronic conditions will also be tested using Alberta EMR data from Canadian Primary Care Sentinel Surveillance Network (CPCSSN). We will then apply the newly developed methods to Alberta health administrative data and estimate the prevalence of hypertension over time.