Variations in the diagnosis and quality of care for pain by primary care physicians: Results from a factorial experiment

Pain management decisions have been shown to vary by sociodemographic characteristics of patients and providers, and differences exist in the prevalence of diagnosed pain conditions by patient race/ethnicity, sex, age, and socioeconomic status (SES). We conducted a randomized factorial experiment to disentangle characteristics such as patient SES and race while testing the hypothesis that specific patient, physician, or practice organizational factors influence clinical decision-making for pain conditions. From 2010-2012, 192 primary care physicians from six U.S. states were purposefully enrolled to fill design cells of gender and experience. Physician subjects viewed two clinically-authentic videos of patients (actors) presenting with pain either as undiagnosed sciatica symptoms or diagnosed knee osteoarthritis. Patient gender, SES, and race (white, black, Hispanic), and physician gender and experience (<10 vs. >=10 years in practice) systematically varied, permitting estimation of unconfounded effects. Analysis of variance was used for outcomes of physician decision-making, including diagnosis of sciatica and pain management (test ordering, behavioral counseling, prescriptions, referrals). Results showed that 93.7% of physicians appropriately diagnosed sciatica, but there was considerable variation in pain treatment decisions largely unexplained by patient race, gender or provider gender. The treatment of lower vs. higher SES patients with sciatica symptoms were generally similar, e.g., x-ray (55% vs. 49%, P=0.4), MRI (32% vs. 35%, P=0.7), exercise counseling (33% vs. 32%, P=0.9), and referrals (20 vs. 23%, P=0.6). However, lower SES patients less frequently received narcotics for sciatica (52.1% vs. SES 68.7%, P=0.01). Patient race had no impact on management of the sciatica patient, yet for the osteoarthritis patient, whites were more likely to receive narcotics (47%, vs. blacks 27%, Hispanics 33%; P=0.03; no significant interaction with SES). Gender had no effects. Physicians in practice <10 y (vs. >10 y) ordered fewer tests, particularly basic lab work or urinalysis, were more likely to prescribe NSAIDs for pain relief, and to provide advice on lifestyle, particularly exercise (P<=.01). MRI for osteoarthritis pain was more commonly used by physicians who reported not using clinical practice guidelines (21% vs. 13%, P=0.04). Overall test ordering decreased as organizational emphasis on business or profits increased. The finding that physicians' length of time in practice, as well as organization emphasis on business, influences pain management decisions indicates a need for the systematic implementation of quality measures. Policy-makers should also develop methods to assure that narcotic analgesics, when appropriate, are prescribed and monitored for patients of various races and socioeconomic levels.