EXTENT OF UNDIAGNOSED HIV-INFECTION IN HOSPITALIZED-PATIENTS - ASSESSMENT BY LINKAGE OF SEROPREVALENCE AND SURVEILLANCE METHODS

Routine screening of hospitalized patients for human immunodeficiency virus (HIV) infection has been suggested as a method to identify undia gnosed HIV infection. To evaluate HIV seroprevalence and rates and ris k factors for undiagnosed HN infection among inpatients at an urban ho spital, we linked a blinded HIV seroprevalence survey with data from a health care systemwide HIV surveillance registry. Consecutive nonobst etric adult inpatients admitted over four months had remnant samples o f serum and plasma obtained and demographic and clinical data abstract ed from hospital registration and outpatient encounter billing files. After linkage with the HIV registry, patient data were assigned a stud y code, individual identifiers were removed, and specimens were tested for HIV-1 antibody. Of 2,825 eligible patients, 155 (5.5%) were HIV-s eropositive: 139 (90%) with known infection and 16 (10%) with previous ly undiagnosed infection. Of those with previously undiagnosed infecti on, eight (5%) were newly diagnosed during hospitalization and eight ( 5%) remained undetected following hospitalization. For HIV-seropositiv e patients, previously undiagnosed infection was significantly more co mmon among those with no use of the health care system in the past yea r than those with recent outpatient or inpatient visits (41.7% versus 4.6%, odds ratio [OR] = 14.9, 95% confidence intervals [CI] = 4.7, 47. 1). Despite a relatively high hospital HIV seroprevalence, the rate of undiagnosed infection was low, suggesting that the percentage of the HIV epidemic remaining ''undetected'' may be smaller in some settings than suggested by previous studies of hospitalized patients. Linkage o f surveillance data to blinded seroprevalence studies can be of value in estimating this ''undetected'' percentage and in evaluating the pot ential yield of routine HIV testing programs.