Accuracy of physician self-report of Spanish language proficiency

As health systems strive to meet the needs of linguistically diverse patient populations, determining a physician's non-English language proficiency is becoming increasingly important. However, brief, validated measures are lacking. To determine if any of four self-reported measures of physician Spanish language proficiency are useful measures of fluency in Spanish. Physician self-report of Spanish proficiency was compared to Spanish-speaking patients' report of their physicians' language proficiency. 110 Spanish-speaking patients and their 46 physicians in two public hospital clinics with professional interpreters available. Physicians rated their Spanish fluency with four items: one general fluency question, two clinically specific questions, and one question on interpreter use. Patients were asked if their doctor speaks Spanish ("yes/no"). Concordance, sensitivity, specificity, and positive and negative predictive values (PPV, NPV) were calculated for each of the items, and receiver operating (ROC) curves were used to compare performance characteristics. Concordance between physician and patient reports of physician Spanish proficiency ranged from 84 to 91%. The PPV for each of the four items ranged from 91 to 99%, the NPV from 60 to 90%, and the area under their ROC curves from 90 to 95%. The general fluency question gave the best combination of PPV and NPV, and the item on holding sensitive discussions had the highest PPV, 99%. Physicians who reported fluency as "fair" were as likely to have patients report they did not speak Spanish as that they did. Physician self-report of Spanish language proficiency is highly correlated with patient report, except when physicians report "fair" general fluency. In settings where no financial or other incentives are linked to language skills, simple questions may be a useful way to assess physician language proficiency.