Assessment of Overuse of Medical Tests and Treatments at US Hospitals Using Medicare Claims

Key Points Question What hospital characteristics are associated with overuse of health care services in the US? Findings In this cross-sectional study of 1 325 256 services performed at 3351 hospitals, we found that hospitals in the South, for-profit hospitals, and nonteaching hospitals were associated with the highest rates of overuse. Meaning Variation within specific hospital types and regions may uncover opportunities for targeted interventions to address overuse.


eTable 1. Details of the Service Capacity Filter Rules for Each Service
We used the following rules to decide whether a hospital had the necessary equipment/capacity to carry out a specific service. Hospitals with capacity but zero overuse were true 'zeroes', while hospitals with zero capacity were treated as missing. The capacity filter for hysterectomy, carotid endarterectomy, coronary artery stenting and spinal fusion were the denominator definitions of these services (an annual average of at least one count of the procedure without any condition qualifiers).

Service
Capacity filter rule

Pulmonary artery catheterization
An annual average (2015)(2016)(2017)  *Corresponds to two low-value service metrics: EEG for headache and EEG for syncope.

eTable 2. Details of the Low-Value Service Algorithms and Any Modifications
We made the following updates to the published overuse/low-value service indicators of Schwarz and Segal. For services with syncope or headache, we required that this diagnosis is the reason for admission or that there was no history in the claims for this diagnosis. We updated the spinal fusion indicator to exclude cervical spinal fusions, and allowed disc degenerative disorders as a potential indicator of overuse. For coronary stents for stable disease, we used additional exclusion criteria of ICD-9-CM codes for myocardial infarctions and unstable angina within the current claim and any recent claim (14 days), rather than the original specific definition which used the indication date for myocardial infarction provided in the Chronic Conditions Warehouse comorbidity table.
Some of the Schwartz indicators flagged services using the AMA's Current Procedural Terminology (CPT) codes in carrier and outpatient claims, and we instead found these codes used in outpatient and inpatient claims. For coronary stents procedures, we flagged services using CPT codes in outpatient claims, or if the CPT codes in the inpatient Arthroscopic debridement/ chondroplasty of the knee with diagnosis of osteoarthritis or chondromalacia in the procedure claim.
All women with hysterectomy excluding a malignancy diagnosis. source

Original Updates
Pulmonary Artery Catheterization in the ICU (Schwartz 2015. Specific version.
Pulmonary artery catheterization for monitoring purposes during an inpatient stay that involved an ICU but not a surgical DRG. Exclude claims that involved pulmonary hypertension, cardiac tamponade, or preop assessment diagnoses.
-eFigure 1. Distribution of the Overuse Rates for Hospitals (Cohort A, N = 2,415) The fitted Beta distribution used for the reliability adjustment of rates is shown using the red line. Overuse ratios are measured as the count of low-value services by either A) all patients at the hospital B) all patients at the hospital with the service and C) all patients at the hospital with the condition. IVC: inferior vena caval filter; CEA: carotid endarterectomy; EEG: electroencephalography; img.: imaging. eFigure 3. Scree-Plot for K-Means Cluster Analysis of (n ≥ 7) Service Overuse Rates The scree-plot for the K-means cluster analysis. K is the number of clusters, while the y-axis shows the sum of squared distances between the points in each cluster.

eFigure 4. Cluster Visualization of Results From K-Means Analysis
Visualization of the four clusters in two dimensions, which are the first and second factors from a principal components analysis (PCA) applied to the thirteen overuse scores (note these first two factors explain 37.39% of the variability in the data). Hospital characteristics are reported if the within cluster proportion is largely different from the overall cohort proportion. Rows are ordered by the most significant hospital characteristic within each cluster. For a hospital characteristic/cluster comparison to be included in the table, this difference was significant at a 5% significance level and the effect size was non-trivial (Cohen's h value was greater than 0.2). The effect size criteria was included as the p-value from the significance test is sensitive to large sample sizes.