(104) Association of physician characteristics and their competency on prescribing of non-steroidal anti-inflammatory drugs (NSAIDs) among patients with congestive heart failure (CHF)

Publication Number: 765

Background: Acute exacerbations of CHF are a leading cause of hospitalization. NSAIDs are commonly used to treat pain from arthritis, but are potentially inappropriate medications (PIM) for CHF patients due to increased exacerbation risk. Physician sex and age frequently predict PIM use. Deficiencies in communication ability and clinical knowledge are hypothesized mechanisms, but these have not been investigated.

Objectives: To disentangle the relationship between physician age, sex, their competencies, and NSAID prescribing to CHF patients.

Methods: We obtained competency assessment scores for the cohort of International Medical Graduates (IMGs) in the US who completed their certification exam in 1998-2004. We linked these IMGs to US Medicare patients they encountered in 2014-2015. We identified patients living with CHF and arthritis, and analyzed their first visit to a study physician between July 2014-Nov 2015. The visit physician's age and sex were the exposures of interest. Excluding patients with an NSAID dispensation in the 6-month baseline period, our outcome was incident NSAID use within 30 days of the visit. With patient as unit-of-analysis, logistic regression models under the Generalized Estimating Equations (GEE) framework was used to estimate the association of physician age and sex on incident NSAID use, adjusting for their competencies, practice characteristics, and patient characteristics.

Results: 10,430 IMGs performed an evaluation & management visit on 287,361 CHF patients. 1,486 (14.3%) IMGs prescribed an incident NSAID to 2,437 (0.9%) patients within 30 days of the visit. In bivariable analysis, physician age was a significant predictor of NSAID prescribing (OR per 5-year increase: 1.19 [95%CI 1.14-1.25]), but not sex. Higher clinical competency and communication scores were associated with decreased odds of NSAID prescribing, but these effects were attenuated after adjusting for physician and patient factors. Older (OR per 5-year: 1.08 [95%CI 1.03-1.13]) and male (OR 1.15 [95%CI 1.01-1.30]) physicians had increased odds of NSAID prescribing after adjustment. 1.1% of patients received an incident NSAID from another physician after the visit, of which 10% were from orthopedic surgeons. These prescribers were more likely than the visit physician to have prescribed to the patient at baseline.

Conclusions: Our preliminary results suggest that the type of prior relationship between physicians and patients likely influences the decision to prescribe an NSAID. Our next step is to characterize these relationships (e.g. by prescription volume), which we hypothesize would modify the effects of physician age, sex, and competency on NSAID prescribing. We will also adjust for presence of an injury or surgical procedure during the baseline period.