Association Between US News & World Report Medical School Ranking and Physician Opioid Prescribing for New Low Back Pain, 2011–2014

INTRODUCTION

Prior research has found significant physician variation in opioid prescribing.¹ Some research has found that primary care physicians (PCPs) who attended a higher-ranked medical school wrote significantly fewer opioid prescriptions overall than PCPs who attended a lower-ranked school.² Whether an association persists between medical school ranking and opioid prescribing when examining a more clinically defined scenario—patients who develop new low back pain—is unknown.

METHODS

We performed analyses using 2010–2014 claims data for a random 20% sample of Medicare beneficiaries. We included beneficiaries aged 66 and over continuously enrolled in Medicare Parts A, B, and D the year prior, year of, and year subsequent to their episode of new low back pain with no opioid exposure in the prior 365 days. We excluded patients with history of cancer or a hospice claim. The outcome was opioid use in the subsequent 365 days, defined both as a binary variable for ever received and as a continuous variable for the number of morphine equivalents (MEs) received (including zeros for patients who received no opioids). We converted opioids to MEs using standard conversion tables.³

We focused on PCPs, the greatest prescribers of prescription opioids.² We defined PCPs as physicians in general practice, family practice, internal medicine, geriatric medicine, and preventive medicine. We obtained medical school attended from the Physician Compare National file.⁴ We obtained US News & World Report “Best Medical Schools: Research Rankings” from Schnell and Currie,² who averaged a school’s rankings from 2010 to 2017 and re-ranked schools according to this average (1 for highest average ranking, 2 for second highest, etc.); 92 schools were ranked.

We estimated a multivariable regression (linear probability model for opioid receipt; linear for MEs) of each outcome as a function of ranking (entered categorically). We included physician age and experience, along with current zip code fixed effects to compare physicians of different medical school ranking within the same zip code. We controlled for patient age, sex, race/ethnicity, Elixhauser comorbidity score,⁵ dual eligibility for Medicaid, and originally being in Medicare for disability; we chose covariates consistent with prior opioid literature.¹ We clustered standard errors at the physician level. We tested for a monotonic trend by re-estimating the regression model using categorical ranking as a continuous variable.⁶ We obtained study approval from the National Bureau of Economic Research, where the data are housed.

RESULTS

The sample included 93,739 patients with new low back pain cared for by 32,102 physicians (Table 1). There was an average of about 1000 patients per ranked medical school. In total, 9.1% of all patients received an opioid and 13.1% of those with two or more visits for new back pain received an opioid. The average number of MEs was 72.0 for all patients and 111 for those with two or more visits. We did not find a statistically significant association between medical school ranking and either receipt of opioid or MEs received, either for all patients with new back pain or for those with two or more visits (Table 2); p values for trend ranged from 0.50 to 0.83 across the 4 specifications (results not shown). Using a model with ranking entered linearly, we can rule out clinically significant associations; for example, we can rule out an association of a one-unit change in ranking with a greater than 0.01 percentage point change in opioid receipt at the 5% level (results not shown). Our results were substantively unchanged when including a ranking-squared variable, including HRR fixed effects, excluding zeros (when examining MEs), and estimating a Poisson model (when examining MEs).

Table 1 Sample Patient Characteristics, 2011–2014

Table 2 The Association Between Medical School Ranking and Receipt of Opioids for New Low Back Pain, 2011–2014

DISCUSSION

We did not find an association between the ranking of the medical school attended by PCPs and their opioid-prescribing patterns for new low back pain among a national sample of Medicare patients. Medical school ranking may not be a useful proxy for identifying higher opioid-prescribing physicians. Results are limited to low back pain in the elderly Medicare population and may not generalize to younger populations. Another limitation is our data had few physician characteristics that we could control for. Differences from prior literature may be due to our focus on a more clinically defined scenario, use of an outcome that is defined per patient, inclusion of zip code fixed effects, adjustment for patient covariates, or focus on an older patient population.