Background:  Public health officials are promoting widespread HIV screening to address the persistent, nationwide problem of late HIV diagnosis. We analyzed data from a random sample of US physicians to examine factors associated with screening for HIV infection.

Methods:  In 1999-2000, 7300 physicians were randomly selected from the American Medical Association Master File and mailed a questionnaire. Our outcome of interest is whether providers reported ever screening asymptomatic men or non-pregnant woman for HIV infection. We compared those reporting screening and not reporting screening using t-tests and χ² statistics and created multivariate logistic regression models to examine independent factors associated with HIV screening.

Results:  The response rate was 70.2%, with 4133 questionnaire included in this analysis.  Overall, only 1133 (27.4%) of physicians surveyed reported that they screen either asymptomatic males or non-pregnant females for HIV (9.5%). Family/general practice physicians were most likely to screen (34.9%), followed by internal medicine (21.2%), pediatrics (18.8%), OB/GYN (15.6%), and emergency medicine (9.5%). Statistically significant independent predictors of offering HIV screening were:  female gender (AOR 1.64, 95%CI 1.37 to 1.96); practicing in southern United States (AOR 0.63, 0.50 to 0.79, Northeast reference); being black (AOR 2.28, 1.55 to 3.36) or Hispanic (AOR 1.46, 1.02 to 2.11, white reference); specializing in OB/GYN (AOR 6.34, 3.53 to 9.00), Family/General Practice (AOR 5.51, 3.57 to 8.51), Internal Medicine (AOR 3.75, 2.40 to 5.86), or Pediatrics (AOR 2.44, 1.52 to 3.94, Emergency Medicine reference); practicing in city with >250,000 citizens (AOR 1.37, 1.12 to 1.66); working in private clinic (AOR 0.62, 0.49 to 0.80); and diagnosing an HIV case within past 2 years (AOR 1.73, 1.32 to 2.16, never diagnosed HIV reference). The only attitudinal factor associated with offering HIV screening was usually or always following-up to see if the patient’s sexual partners were notified (AOR 1.43, 1.20 to 1.70).

Conclusions:  Given that few providers reported screening any asymptomatic patients for HIV, the United States appears far from implementing routine HIV screening. Practice setting and physician characteristics, rather than physician attitudes and beliefs about HIV testing and intervention, were associated with conducting HIV screening.  Institutional and structural level changes may prove critical in implementing more widespread HIV screening nationwide.