The following is a summary of “Development and Internal Validation of an Emergency Department Admission Intensity Measure Using Data From a National Group”, published in the September 2023 issue of Emergency Medicine by Oskvarek, et al.

Researchers created and analyzed variations in an ED admission intensity metric for value-based payment models. Included in the metric are emergency department diagnoses for which evidence-based protocols exist and for which admission decisions are left to the discretion of individual doctors. Using their knowledge and experience, three emergency physicians chose International Classification of Diseases (ICD)-10 codes based on their perceived validity. 

The opinions of a separate technical panel were sought. Investigators used descriptive plots, multilevel linear probability models, and intraclass correlation coefficients (ICC) on data from a nationwide sample (2018-2019) to evaluate and measure stability at the physician and facility level each quarter. From a pool of 23,590 possible ICD-10 measure codes, 535 were ultimately chosen. The median facility-quarter admission rate was 26.1% (95% CI 24.5–27.7) among 127 EDs. The inter-class correlation coefficient (ICC) was 0.91, and the standard deviations (SDs) between and within facilities were 9.2 (95% CI 8.2 to 10.5) and 2.9 (95% CI 2.7 to 3.0), respectively. More than three-thirds (749/961) of ED quarters were within the norm of their institution. 

The average quarterly rate was 29.1% (95% CI: 28.6-29.6) among 2,398 doctors. 220 doctors (9.2%) had a consistently higher admission rate than average, while one hundred and ninety-three doctors (8.0%) had an admission rate consistently lower than normal, yielding an ICC of 0.58. The quarterly admission rates for both facilities and physicians show that this set of ICD-10 diagnoses is stable and has good face validity. The metric can be used in value-based models to track facility admission rates and consistently identify high and low admitters within facilities so that admission variance may be better managed.

Source: sciencedirect.com/science/article/abs/pii/S019606442201335X