To assess the completeness of human immunodeficiency virus (HIV) reporting among hospital inpatients whose records listed diagnostic codes for HIV infection but who did not meet the 1987 AIDS case definition, we conducted a statewide hospital study of admissions between January 1, 1986 and December 31, 1990. Of the 396 HIV-infected hospital inpatients identified, 313 (79%) had been reported to the State HIV Registry. HIV reporting was less complete for patients who were older and/or were blood product recipients. Of the 313 reported patients, 189 (60%) had been reported prior to their first hospital admission. Temporal improvements were noted in the completeness of HIV reporting among the hospital patients (1986: 65%; 1987: 81%; 1988: 64%; 1989: 82%; 1990: 86%; Chi square for linear trend 9.6, p < 0.01) and prior to their first hospital admission (1986: 31%; 1987: 34%; 1988: 49%; 1989: 64%; 1990: 72%; Chi square for linear trend 26.6; p < 0.01). Women were more likely than men to be reported prior rather than during or after their first hospital admission (71% vs. 55%; p < 0.01). Of the 155 patients with CD4+T-lymphocyte test results, 41 had CD4+counts <200 mm3and met the 1993 but not the 1987 AIDS case definition. In South Carolina most (79%) diagnosed, hospitalized, HIV-infected patients had been reported to the State HIV Registry, with improvements in reporting occurring over time. Findings suggest that the 1993 AIDS case definition will improve our ability to monitor severe morbidity related to HIV.