Using medicare claims to estimate risk-adjusted performance of Pennsylvania trauma centers.
PLOS Digit Health
; 2(6): e0000263, 2023 Jun.
Article
in En
| MEDLINE
| ID: mdl-37267229
ABSTRACT
Trauma centers use registry data to benchmark performance using a standardized risk adjustment model. Our objective was to utilize national claims to develop a risk adjustment model applicable across all hospitals, regardless of designation or registry participation. Patients from 2013-14 Pennsylvania Trauma Outcomes Study (PTOS) registry data were probabilistically matched to Medicare claims using demographic and injury characteristics. Pairwise comparisons established facility linkages and matching was then repeated within facilities to link records. Registry models were estimated using GLM and compared with five claims-based LASSO models demographics, clinical characteristics, diagnosis codes, procedures codes, and combined demographics/clinical characteristics. Area under the curve and correlation with registry model probability of death were calculated for each linked and out-of-sample cohort. From 29 facilities, a cohort comprising 16,418 patients were linked between datasets. Patients were similarly distributed median age 82 (PTOS IQR 74-87 vs. Medicare IQR 75-88); non-white 6.2% (PTOS) vs. 5.8% (Medicare). The registry model AUC was 0.86 (0.84-0.87). Diagnosis and procedure codes models performed poorest. The demographics/clinical characteristics model achieved an AUC = 0.84 (0.83-0.86) and Spearman = 0.62 with registry data. Claims data can be leveraged to create models that accurately measure the performance of hospitals that treat trauma patients.
Full text:
1
Collection:
01-internacional
Database:
MEDLINE
Type of study:
Etiology_studies
/
Prognostic_studies
/
Risk_factors_studies
Language:
En
Journal:
PLOS Digit Health
Year:
2023
Type:
Article
Affiliation country:
United States