Clinician validation of the Medicare measure for potentially avoidable hospital visits after chemotherapies.

BACKGROUND

Medicare's OP-35 measure (the 35th hospital outpatient quality measure) tracks unplanned hospital visits within 30 days of chemotherapy and defines a subset as "potentially avoidable" using approximately 300 diagnosis codes. Despite widespread adoption in policy and oncology quality reporting, the measure has not been validated with clinician review.

METHODS

We identified 14 220 acute hospital visits within 30 days of chemotherapy (2016-2023) from 22 hospitals in 3 health systems (academic, safety-net, community). A 5% stratified random sample of 705 visits underwent blinded review by 3 clinicians, who adjudicated avoidability and assigned a clinical classification for each visit (eg, nonemergent care occurring overnight, nonurgent blood product transfusion, uncontrolled symptoms requiring hospital care). The gold standard definition for an avoidable visit was based on a majority of the clinicians. We assessed the diagnostic characteristics of OP-35 using sensitivity, specificity, accuracy, and area under receiver operator curve (AUROC). We used the clinical classifications to develop a new set of Actionable Categories of avoidability and assessed its diagnostic characteristics.

RESULTS

Clinicians judged 30.2% (213/705) of visits as avoidable. OP-35 classified a similar proportion (30.8%, 217/705), but agreement was low. Sensitivity of OP-35 was 34.7% (95% confidence interval [CI] = 28.2% to 41.6%), specificity 70.9% (95% CI = 66.7% to 74.8%), and accuracy 59.9% (95% CI = 56.2% to 63.6%), with AUROC of 0.53. The Actionable Categories classification performed better: sensitivity 89.7%, specificity 85.2%, accuracy 86.5%, and AUROC 0.87.

CONCLUSIONS

OP-35 showed poor agreement with clinicians for avoidable hospital visits, raising concerns about its clinical validity. An alternative classification system grouping visits into actionable clinical scenarios offered superior diagnostic accuracy.