Quality measurement is increasingly important for medical centers. Regulatory organizations and payers are demanding quality care and quality improvement initiatives. Measurement is a foundation for all improvement methodologies but can be costly. Administrative data are used for external reporting, but local stakeholders often question its validity. Providers are particularly concerned given the move toward measurement at the individual provider level. We sought to determine the validity of administrative data in accurately identifying the provider of record at the level of hospitalist or nonhospitalist.
Over a period of 1 month, all patients admitted to University of Chicago Medical Center (UCMC) were screened for the following events: (1) death during hospitalization, (2) 30‐day readmission to UCMC, and (3) transfer to the ICU. Death was measured by expired status at discharge. Thirty‐day readmission and ICU transfers were measured by TSI (UCMC's facility billing system). Two sources of administrative data, TSI and the University of Chicago Practice Group (UCPG) pro‐fee data, were screened for identifying the provider of record. Any patient who was cared for by a hospitalist during any part of their hospital stay was counted as a hospitalist patient. Among patients identified for each of the 3 events, we cross‐tabulated patients identified as hospitalist patients using the 2 data sources to examine discrepancies.
Thirty‐eight deaths, 135 30‐day readmissions, and 148 ICU transfers were identified during the study period. Large differences between TSI and UCPG data were identified. Among the 38 patients who died, TSI identified 2 as being cared for by hospital‐ists, whereas UCPG pro‐fee data identified 0 hospitalist patients. Among the 135 30‐day readmit patients, TSI identified 19 as cared for by hospitalists, whereas UCPG pro‐fee data identified 32. Eleven of these patients were identified by both TSI and UCPG. TSI identified 19 of the 148 iCU transfer patients as hospitalist patients, whereas UCPG pro‐fee data identified 27. Eleven of the ICU transfer patients were identified by both TSI and UCPG.
Administrative data can be an attractive resource for quality improvement projects and are used extensively in external reporting of quality measures. Significant validity issues were identified in this administrative data. UCMC generally reports quality data using TSI, and striking differences would be reported if using UCPG pro‐fee data versus TSI. If these data had been publicly reported, TSI would have reported greater hospitalist mortality and underreported 30‐day readmission by 40% and !CU transfer by 30% than if UCPG pro‐fee data were used. This project highlights the need for a rigorous understanding of the validity of administratively generated data used for institutional quality decisions, public reporting, pay for performance, and regulatory review, particularly when at the individual provider level.
A. Schmitz, none; E. Abbo, none; N. Artz, none; D. Lovinger, none; N. Perovic, none; C. Whelan, none.
To cite this abstract:Whelan C, Schmitz A, Abbo E, Artz N, Lovinger D, Perovic N. Using Administrative Data for Quality Improvement: The Good, the Bad, and the Unknown. Abstract published at Hospital Medicine 2009, May 14-17, Chicago, Ill. Abstract 103. Journal of Hospital Medicine. 2009; 4 (suppl 1). https://www.shmabstracts.com/abstract/using-administrative-data-for-quality-improvement-the-good-the-bad-and-the-unknown/. Accessed March 28, 2020.