OBJECTIVES: To assess the patterns of use of handheld clinical decision support tools by internal medicine residents in clinical settings. METHODS: Eighty-two internal medicine residents were given personal digital assistants (PDAs) containing a suite of clinical decision support (CDS) programs. A tracking program was used to prospectively track program use during the study period, and a follow-up survey regarding self-reported program use was administered after the study period. Patterns of program use from the tracking data were compared to the data from the self-report survey. RESULTS: Sixty-eight residents were followed using the tracking data. Residents used an average of 1.81 CDS programs (SD: 1.57; range, 0-5) per month. Forty-nine residents completed the self-report survey. Residents reported using an average of 3.15 (SD: 1.61) and 3.92 (SD: 1.40) CDS programs during a typical clinic session and inpatient day, respectively. In both inpatient and outpatient settings and for both methods of assessing program use, 2 programs (Epocrates and MedCalc) were used more often than the other programs. No association was observed between age, gender, race, and PGY level with the use of handheld clinical decision support tools for either tracked or self-report data. The self-report data show higher estimates of CDS program use than the tracking data in the clinical setting. CONCLUSIONS: The data show that physicians prefer to use certain handheld CDS tools in clinical settings. Drug references and medical calculators have been consistently used more than clinical prediction rules and diagnostic systems. Self-report survey instruments may overestimate recorded use of CDS programs.