Physician Practice Pattern Variations in Common Clinical Scenarios Within 5 US Metropolitan Areas

Abstract

Importance: While variations in quality of care have been described between US regions, physician-level practice pattern variations within regions remain poorly understood, notably among specialists.

Objective: To examine within-area physician-level variations in decision-making in common clinical scenarios where guidelines specifying appropriateness or quality of care exist.

Design setting and participants: This cross-sectional study used 2016 through 2019 data from a large nationwide network of commercial insurers, provided by Health Intelligence Company, LLC, within 5 metropolitan statistical areas (MSAs). Physician-level variations in appropriateness and quality of care were measured using 14 common clinical scenarios involving 7 specialties. The measures were constructed using public quality measure definitions, clinical guidelines, and appropriateness criteria from the clinical literature. Physician performance was calculated using a multilevel model adjusted for patient age, sex, risk score, and socioeconomic status with physician random effects. Measure reliability for each physician was calculated using the signal-to-noise approach. Within-MSA variation was calculated between physician quintiles adjusted for patient attributes, with the first quintile denoting highest quality or appropriateness and the fifth quintile reflecting the opposite. Data were analyzed March through October 2021.

Main outcomes and measures: Fourteen measures of quality or appropriateness of care, with 2 measures each in the domains of cardiology, endocrinology, gastroenterology, pulmonology, obstetrics, orthopedics, and neurosurgery.

Results: A total of 8788 physicians were included across the 5 MSAs, and about 2.5 million unique patient-physician pairs were included in the measures. Within the 5 MSAs, on average, patients in the measures were 34.7 to 40.7 years old, 49.1% to 52.3% female, had a mean risk score of 0.8 to 1.0, and more likely to have an employer-sponsored insurance plan that was either self-insured or fully insured (59.8% to 97.6%). Within MSAs, physician-level variations were qualitatively similar across measures. For example, statin therapy in patients with coronary artery disease ranged from 54.3% to 70.9% in the first quintile of cardiologists to 30.5% to 42.6% in the fifth quintile. Upper endoscopy in patients with gastroesophageal reflux disease without alarm symptoms spanned 14.6% to 16.9% in the first quintile of gastroenterologists to 28.2% to 33.8% in the fifth quintile. Among patients with new knee or hip osteoarthritis, 2.1% to 3.4% received arthroscopy in the first quintile of orthopedic surgeons, whereas 25.5% to 30.7% did in the fifth quintile. Appropriate prenatal screening among pregnant patients ranged from 82.6% to 93.6% in the first quintile of obstetricians to 30.9% to 65.7% in the fifth quintile. Within MSAs, adjusted differences between quintiles approximated unadjusted differences. Measure reliability, which can reflect consistency and reproducibility, exceeded 70.0% across nearly all measures in all MSAs.

Conclusions and relevance: In this cross-sectional study of 5 US metropolitan areas, sizeable physician-level practice variations were found across common clinical scenarios and specialties. Understanding the sources of these variations may inform efforts to improve the value of care.

Figure.. Physician-Level Variations in Practice Patterns Across 6 Example Clinical Scenarios in the South Central Metropolitan Statistical Area

Example measures are shown for statin therapy in patients with chronic coronary artery disease (CAD) (A), endoscopy in patients with gastroesophageal reflux disease (GERD) and no alarm symptoms (B), caesarean delivery in patients with low-risk pregnancies (C), any physical therapy (PT) prior to elective hip or knee replacement (D), use of bronchodilator in patients with chronic obstructive pulmonary disease (COPD) (E), and spinal fusion for patients with low back pain (F). Each physician is denoted by a data point and vertical 95% CI. Quintile 1 represents, on average, more appropriate care, and quintile 5 denotes less appropriate care on average.