Determinants of the productivity and efficiency of proceduralist physician practices: Radiology as a case study

Objectives: Rapid growth in healthcare costs has motivated extensive research in the productivity and efficiency of health care delivery. However, most of this research has focused on hospitals. Physicians have been only a secondary focus, with primary care physicians predominantly studied. As best we can ascertain, the physician practices of procedural specialists have never been studied. Given the extremely rapid growth of the quantity and cost of imaging, identifying means of increasing productivity and efficiency of imaging should be especially valuable and should offer an avenue to slow its cost increase.

Our objective thus is to identify the factors affecting productivity and efficiency of procedural physician practices, such as radiology, and to quantify the effects of each, using relatively sophisticated analytical techniques and richer data than previous studies have had.

Methodology: We estimate a stochastic frontier model to study the effects of (i) technologies used in work production and (ii) practice characteristics, such as work hours, practice size, and output mix on the productivity and efficiency of radiology practices. Using data from 2003 and 2007 surveys of radiology practices, we investigate time trends. We employ a flexible functional form that allows elasticity to vary by practice size and we investigate alternative model specifications for evaluating the determinants of radiology practice (in)efficiency.

Results: Preliminary analysis indicates that inefficiency explains 73% of the variance among radiology practices in the sample, with the average radiology practice capable of increasing the number of procedures performed by 25% relative to the best-practice frontier. We find some production technologies—for example, “templates” (use of standard language in reports) or “nighthawks” (use of external, after-hours teleradiology services) have sizable effects on productivity, with effects on the order of 10-15%. Elasticity of output with respect to annual hours worked appears to be much less than 1—approximately 0.4. Elasticity with respect to practice size also seems less than 1—approximately 0.8—perhaps due to errors in variables. We do not find a time trend beyond that attributable to the diffusion of production technologies.

Conclusions: Use of some production technologies much increases productivity. Their diffusion and adoption presumably underlies the large increase in annual output per radiologist, accompanied by relatively small changes in annual work hours, that we have found in other studies. Adoption of these technologies, and the development of new ones, can permit rapid growth in imaging without requiring increased physician labor, and thus abet cost control.