Many persons with HIV infection do not receive consistent ambulatory medical care and are excluded from studies of patients in medical care. However, these hard-to-reach groups are important to study because they may be in greatest need of services. This study compared the sociodemographic, clinical, and health care utilization characteristics of a multisite sample of HIV-positive persons who were hard to reach with a nationally representative cohort of persons with HIV infection who were receiving care from known HIV providers in the United States and examined whether the independent correlates of low ambulatory utilization differed between the 2 samples. We compared sociodemographic, clinical, and health care utilization characteristics in 2 samples of adults with HIV infection: 1286 persons from 16 sites across the United States interviewed in 2001-2002 for the Targeted HIV Outreach and Intervention Initiative (Outreach), a study of underserved persons targeted for supportive outreach services; and 2267 persons from the HIV Costs and Services Utilization Study (HCSUS), a probability sample of persons receiving care who were interviewed in 1998. We conducted logistic regression analyses to identify differences between the 2 samples in sociodemographic and clinical associations with ambulatory medical visits. Compared with the HCSUS sample, the Outreach sample had notably greater proportions of black respondents (59% vs. 32%, P = 0.0001), Hispanics (20% vs. 16%), Spanish-speakers (9% vs. 2%, P = 0.02), those with low socioeconomic status (annual income < Dollars 10,000 75% vs. 45%, P = 0.0001), the unemployed, and persons with homelessness, no insurance, and heroin or cocaine use (58% vs. 47%, P = 0.05). They also were more likely to have fewer than 2 ambulatory visits (26% vs. 16%, P = 0.0001), more likely to have emergency room visits or hospitalizations in the prior 6 months, and less likely to be on antiretroviral treatment (82% vs. 58%, P = 0.0001). Nearly all these differences persisted after stratifying for level of ambulatory utilization (fewer than 2 vs. 2 or more in the last 6 months). In multivariate analysis, several variables showed significantly different associations in the 2 samples (interacted) with low ambulatory care utilization. The variables with significant interactions (P values for interaction shown below) had very different adjusted odds ratios (and 95% confidence intervals) for low ambulatory care utilization: age greater than 50 (Outreach 0.55 [0.35-0.88], HCSUS 1.17 [0.65-2.11)], P = 0.05), Hispanic ethnicity (Outreach 0.81 [0.39-1.69], HCSUS 2.34 [1.56-3.52], P = 0.02), low income (Outreach 0.73 [0.56-0.96], HCSUS 1.35 [1.04-1.75], P = 0.002), and heavy alcohol use (Outreach 1.74 [1.23-2.45], HCSUS 1.00 [0.73-1.37], P = 0.02). Having CD4 count less than 50 was associated with elevated odds of low ambulatory medical visits in the Outreach sample (1.53 [1.00-2.36], P = 0.05). Compared with HCSUS, the Outreach sample had far greater proportions of traditionally vulnerable groups, and were less likely to be in care if they had low CD4 counts. Furthermore, heavy alcohol use was only associated with low ambulatory utilization in Outreach. Generalizing from in care populations may not be warranted, while addressing heavy alcohol use may be effective at improving utilization of care for hard-to-reach HIV-positive populations.