Combining Data Sources to Produce Nationally Representative Estimates of Hospital Encounter Characteristics
Conference
64th ISI World Statistics Congress
Format: IPS Abstract
Keywords: propensity
Session: IPS 312 - Innovative statistical methods for large-scale surveys
Wednesday 19 July 2 p.m. - 3:40 p.m. (Canada/Eastern)
Abstract
The 2020 National Hospital Care Survey (NHCS) is a stratified random sample of US hospitals, conducted by the Centers for Disease Control and Prevention’s National Center for Health Statistics (NCHS). Hospitals responding to NHCS provide nearly complete records of patient encounters over the entire 2020 calendar year, making the data extraordinarily valuable for understanding US hospital care utilization and informing health care policy. NHCS is subject to hospital-level nonresponse that reduces available sample sizes and potentially biases results, due to differential response rates across hospital types. Accordingly, NCHS and NORC at the University of Chicago have collaborated to enhance NHCS encounter data with additional data sources, reducing potential biases. The additional data sources include a proprietary commercial hospital encounter data source, treated as a nonprobability sample with unknown hospital participation propensities, as well as nationally representative hospital care benchmarks from the Healthcare Cost and Utilization Project (HCUP). The enhanced data can be used to create nationally representative estimates of hospital encounter characteristics. The enhanced data will also serve as the basis for additional data products, including a weighted public use file and experimental synthetic data products. We will describe our data enhancement approach along with methodological challenges and preliminary results.