Recent advances in AIDS-related therapies have delayed the onset of AIDS-de
fining illnesses and reduced the usefulness of AIDS surveillance in assessi
ng the incidence of early HIV disease and estimating future needs of the HI
V-infected population. These changes have prompted renewed interest in expa
nding surveillance to include HIV and have engendered national debate on wh
ether an HIV surveillance system should be based on reports of the names of
infected individuals or employ non-name-based data codes. In 1994, the sta
te of Maryland implemented a program to require HIV surveillance by unique
identifier (UI) patient code. This evaluation of Maryland's program found t
hat when complete, the 12-digit UI number provided a virtually unduplicated
count 99.8% unique, was 99.9% unique with only the last four digits of the
U.S. government Social Security Number (SSN), date of birth (DOB), and rac
e, and 77.7% unique if the last four digits of the SSN were missing. Health
care providers were willing to create the UI, with DOB and gender present
98.3% and 98.8% of the time, race was complete 84.1% and last four digits o
f SSN were complete 72.4%. The overall completeness of reporting for HIV te
sts was 87.8% and 84.8%, respectively, using different methodologies. Evide
nce from the Maryland UI evaluation demonstrates that a non-name-based syst
em can provide accurate, timely and valid data concerning the scope of the
HIV epidemic, without the creation of state-wide name-based registry.