Low-Value Service Use in Provider Organizations

Abstract

Objective: To assess whether provider organizations exhibit distinct profiles of low-value service provision.

Data sources: 2007-2011 Medicare fee-for-service claims and enrollment data.

Study design: Use of 31 services that provide minimal clinical benefit was measured for 4,039,733 beneficiaries served by 3,137 provider organizations. Variation across organizations, persistence within organizations over time, and correlations in use of different types of low-value services within organizations were estimated via multilevel modeling, with adjustment for beneficiary sociodemographic and clinical characteristics.

Principal findings: Organizations provided 45.6 low-value services per 100 beneficiaries on average, with considerable variation across organizations (90th/10th percentile ratio, 1.78; 95 percent CI, 1.72-1.84), including substantial between-organization variation within hospital referral regions (90th/10th percentile ratio, 1.66; 95 percent CI, 1.60-1.71). Low-value service use within organizations was highly correlated over time (r, 0.98; 95 percent CI, 0.97-0.99) and positively correlated between 13 of 15 pairs of service categories (average r, 0.26; 95 percent CI, 0.24-0.28), with the greatest correlation between low-value imaging and low-value cardiovascular testing and procedures (r, 0.54).

Conclusions: Use of low-value services in provider organizations exhibited substantial variation, high persistence, and modest consistency across service types. These findings are consistent with organizations shaping the practice patterns of affiliated physicians.

Keywords: Low value; Medicare; efficiency; quality measurement.

Figure 1

Distribution of Adjusted Rates of Low‐Value Services

1. Notes: Estimates are predicted modes derived from random intercept model estimation of organizations' total composite low‐value service use, using 2007–2011 data. The model adjusts for overdispersion of observed performance, beneficiary sociodemographic and clinical characteristics, local area economic characteristics, year, and the number of patients qualifying for component measure denominator. [Color figure can be viewed at wileyonlinelibrary.com]