TY - JOUR
T1 - Validation of a method for identifying nursing home admissions using administrative claims
AU - Zuckerman, Ilene H.
AU - Sato, Masayo
AU - Hsu, Van Doren
AU - Hernandez, Jose J.
PY - 2007
Y1 - 2007
N2 - Background. Currently there is no standard algorithm to identify whether a subject is residing in a nursing home from administrative claims. Our objective was to develop and validate an algorithm that identifies nursing home admissions at the resident-month level using the MarketScan Medicare Supplemental and Coordination of Benefit (COB) database. Methods. The computer algorithms for identifying nursing home admissions were created by using provider type, place of service, and procedure codes from the 2000 - 2002 MarketScan Medicare COB database. After the algorithms were reviewed and refined, they were compared with a detailed claims review by an expert reviewer. A random sample of 150 subjects from the claims was selected and used for the validity analysis of the algorithms. Contingency table analysis, comparison of mean differences, correlations, and t-test analyses were performed. Percentage agreement, sensitivity, specificity, and Kappa statistics were analyzed. Results. The computer algorithm showed strong agreement with the expert review (99.9%) for identification of the first month of nursing home residence, with high sensitivity (96.7%), specificity (100%) and a Kappa statistic of 0.97. Weighted Pearson correlation coefficient between the algorithm and the expert review was 0.97 (p < 0.0001). Conclusion. A reliable algorithm indicating evidence of nursing home admission was developed and validated from administrative claims data. Our algorithm can be a useful tool to identify patient transitions from and to nursing homes, as well as to screen and monitor for factors associated with nursing home admission and nursing home discharge.
AB - Background. Currently there is no standard algorithm to identify whether a subject is residing in a nursing home from administrative claims. Our objective was to develop and validate an algorithm that identifies nursing home admissions at the resident-month level using the MarketScan Medicare Supplemental and Coordination of Benefit (COB) database. Methods. The computer algorithms for identifying nursing home admissions were created by using provider type, place of service, and procedure codes from the 2000 - 2002 MarketScan Medicare COB database. After the algorithms were reviewed and refined, they were compared with a detailed claims review by an expert reviewer. A random sample of 150 subjects from the claims was selected and used for the validity analysis of the algorithms. Contingency table analysis, comparison of mean differences, correlations, and t-test analyses were performed. Percentage agreement, sensitivity, specificity, and Kappa statistics were analyzed. Results. The computer algorithm showed strong agreement with the expert review (99.9%) for identification of the first month of nursing home residence, with high sensitivity (96.7%), specificity (100%) and a Kappa statistic of 0.97. Weighted Pearson correlation coefficient between the algorithm and the expert review was 0.97 (p < 0.0001). Conclusion. A reliable algorithm indicating evidence of nursing home admission was developed and validated from administrative claims data. Our algorithm can be a useful tool to identify patient transitions from and to nursing homes, as well as to screen and monitor for factors associated with nursing home admission and nursing home discharge.
UR - http://www.scopus.com/inward/record.url?scp=38849138640&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=38849138640&partnerID=8YFLogxK
U2 - 10.1186/1472-6963-7-202
DO - 10.1186/1472-6963-7-202
M3 - Article
C2 - 18070360
AN - SCOPUS:38849138640
SN - 1472-6963
VL - 7
JO - BMC Health Services Research
JF - BMC Health Services Research
M1 - 202
ER -