Tuesday, June 14, 2016: 3:00 PM
401 (Fisher-Bennett Hall)

Author(s): Jose J. Escarce; Susan Paddock; John L Adams; Lane F Burgette; Cheryl Damberg

Discussant: Erin Strumpf

Studies have found weak correlation between the costs and quality of care provided by health care providers. Consequently, many contemporary report cards provide consumers information on provider quality and costs, to help them identify high “value” providers that offer high quality at low costs, and payers are increasingly introducing value-based payment models.

High-value physician organizations (medical groups and independent practice associations) may be more efficient than others – i.e., better at using resources to produce high quality and, presumably, better patient outcomes. Studies have assessed cost, quality, and efficiency in nursing homes and hospitals. To date, however, there are no such studies of physician organizations.

In this study, we estimate a production function for quality for physician organizations and use stochastic frontier estimation to assess variation across physician organizations in efficiency.

We obtained data on HMO and POS patients treated by 129 physician organizations in California from the Integrated Healthcare Association. The data included number of patients; total cost of care per patient, adjusted for age, gender, ICD-9-CM diagnoses and geographic area (range, $2500-$5800); and two quality measures: a 23-item composite of care processes and a 6-item composite of patient experience. We weighted the two measures equally to obtain a single quality measure (range, 0.50-0.75 on 0-1 scale). We used stochastic frontier estimation to estimate the following production function:

Q = KCαNε

where Q is average quality of care, K is a constant, C is average total cost of care per patient, N is number of patients, and ε is efficiency.

The correlation between quality and cost per patient was 0.27, but a relationship was hardly discernible on the scatterplot. The estimation yielded α=0.15 (95% CI, -.06,.34), consistent with diminishing quality returns to higher costs. Additionally, we estimated β=0.14 (95% CI, -.08,.34), suggesting that quality depends mainly on cost per patient, with no scale effects. Most strikingly, efficiency scores ranged from 65% to 98%. All physician organizations providing the highest quality (>0.70) had cost per patient exceeding $3500 and efficiency scores exceeding 90%. Conversely, numerous physician organizations had cost per patient exceeding $3500 yet offered low quality (<0.60); these organizations had efficiency scores below 80%. Sensitivity analyses using the process quality measure alone gave similar results.

We assessed the validity of our efficiency scores by correlating them with rates of use of antibiotics for acute bronchitis (r=-.21), imaging for lower back pain (r=-.36), and antibiotics for common colds (r=-.55) (three clinical measures of overutilization); emergency department visits (r=-.62); and meaningful use of health information technology (HIT) (r=.41), which we hypothesized increases efficiency.

Both high efficiency and moderate to high levels of spending were necessary for high quality. We also found substantial variation in efficiency across physician organizations, that inefficiency is associated with higher rates of “unnecessary” care, and that meaningful use of HIT may improve efficiency. Additional research is needed to identify the strategies used by high-efficiency physician organizations.