Using Existing Data for Representative Hospital Use Estimates
Problem
Severe nonresponse has prevented the NHCS from producing nationally representative estimates on hospital utilization.
The National Hospital Care Survey (NHCS) is designed to provide timely and reliable statistics on hospital care utilization in the U.S. to inform health care policy and serve a variety of research needs. The biggest challenge for the NHCS has been hospital recruitment. NHCS faces stiff competition for participation from other mandatory data collection and surveillance systems, which, in part, contributes to low hospital participation rates. As such, the NHCS has yet to produce nationally representative estimates on hospital utilization.
Because of the desire for NHCS representative estimates from 2020 when the COVID-19 pandemic was overwhelming hospitals, the NCHS wanted to design and develop a methodology for creating nationally representative estimates of hospital utilization using the participating NHCS hospital data and similar hospital data obtained from a commercial vendor. This would provide a means to create a public-use NHCS hospital data file that could be used by researchers examining the effects that COVID-19 had on hospital utilization.
Solution
NORC used external data to create model-based estimates and representative data files.
NORC’s solution involved the design, development, and implementation of a methodology to create a nationally representative file of hospital utilization by using information from the participating NHCS hospitals and the commercial hospital database. Innovative weighting methods were required to permit construction of restricted use and public use data files that improved on the NHCS data while releasing none of the commercial microdata. In addition, NORC’s solution entailed the implementation of a pilot study to design and develop synthetic data using the NHCS collected data and commercial hospital database. The synthetic data file can produce nationally representative estimates of hospital utilization while maintaining confidentiality and statistical integrity.
Result
NORC’s efforts allowed NHCS to release its first-ever data files for inpatient stays and emergency department visits.
NORC’s ability to combine datasets and produce a nationally representative file allowed NCHS to release the 2020 National Hospital Care Survey (NHCS) Public Use Files with data files for inpatient stays and emergency department visits. This is a monumental release and captures many firsts: the first year NHCS has been able to make national estimates, the first time in the survey’s history that a public use file has been released, the first time external data was used to create model-based estimates, and the first time data files have been released in a format ready to be used by the public not only in SAS and Stata, but also in R.
Related Tags
Project Leads
-
F. Jay Breidt
Senior FellowPrincipal Investigator -
Edward Mulrow
Senior Vice President & DirectorProject Director -
Scott Campbell
Senior StatisticianProject Manager -
Dean Resnick
Principal Data ScientistChief Data Scientist
Data & Findings
Presented by Jay Breidt at the following conferences:
- 10/24/2023, 2023 FCSM (Federal Committee on Statistical Methodology) Research and Policy Conference, Hyattsville, MD.
- 07/19/2023, 64th ISI (International Statistical Institute) World Statistics Congress, Ottawa, Canada.
- 06/01/2023, 2023 IISA (International Indian Statistical Association) Annual Conference, Colorado School of Mines, Golden, CO.