Using large datasets to measure health status and service use of older adults

Commonly used data sources for measuring health status and service use of older adults include national surveys and secondary data analysis of electronic data sources including healthcare claims data and electronic health records (EHRs). Depending on how the data are generated in EHRs and medical claims, and depending on how long people are observed for, the ability to measure prevalence or incidence of chronic conditions and the ability to measure incidence or a history of acute conditions will vary. Various data types spanning standardized data (diagnostic codes, procedure codes), medication administered or prescribed, unstructured free text such as clinical notes, and clinical assessment data can all be used to measure health status and service use. Different data sources and types of variables have different benefits and limitations depending on how data are generated and the incentives for those recording data (i.e., healthcare providers and billing staff) to be complete. Testing assumptions and exploring the validity of measures can be accomplished by approaches such as comparing agreement of measures (e.g., disease prevalence) across data tables within a data source, comparing agreement with linked data sources, and comparing rates of disease or service use to rates in data sources that have similar populations. Future directions for administrative data such as data linkage and natural language processing will improve the utility of administrative data. The information and concepts are broadly applicable, but for illustrative purposes, examples of how these approaches have been applied to electronic data from administrative records including EHRs and claims data to fill important knowledge gaps and measure health status and quality of care from Australia and the United States are presented.