Concordance of HIV Surveillance Data and Medical Care: What do CD4 and Viral Loads Fail to Tell Us About Linkage to HIV Care?

BACKGROUND:  Mandatory comprehensive HIV lab reporting allows public health officials to monitor linkage to care following HIV diagnosis. Population-level HIV surveillance data collected by state and local health departments are being used to measure progress towards goals in the National HIV/AIDS Strategy, including prompt linkage to care. Typically, the surveillance definition of timely linkage to care is a CD4 or VL drawn within 91 days of positive diagnostic test and received by the HIV Surveillance Registry (HSR). This approach however may over-represent actual linkage because the first CD4/VL may be part of a routine diagnostic work-up done before initiating care. To determine the accuracy of this proxy measure, we validated it against the medical record (MR).

METHODS:  Newly HIV-diagnosed persons reported to the New York City (NYC) HSR by a high-volume HIV-diagnosing facility (>20/year) with on-site HIV care (N=24) were selected if the first lab was reported within 12 months of diagnosis. All such persons were considered linked-to-care per HSR. A health department HIV physician specialist performed all MR reviews. Patients whose care initiation per HSR was confirmed by MR were compared to the remainder of selected patients on key characteristics, including mortality one-year post diagnosis. Of 3,658 new 2009 NYC HIV diagnoses, 37.9% (N=1,385) occurred at the sites selected, and among this group 77.2% (N=1,069) had HSR evidence of linkage within 12 months. Of these, 287 were excluded after MR review, leaving 782 patients from 22 sites for analysis.

RESULTS:  Using the MR as gold standard, 79.9% (n=625) of the HSR group had an actual initial care visit (medical-visit group) and 157 did not (no-medical-visit group). Compared to the medical-visit group, the no-medical-visit group (N=157) was significantly more likely to be diagnosed during an inpatient admission (73.2% vs. 31.5%, p<0.01), have their first CD4/VL within 7 days of diagnosis (82.8% vs. 51.7%, p<0.01), and die within a year (17.8% vs. 1.4%, p<0.01).

CONCLUSIONS:  Use of surveillance CD4/VL values as a proxy for HIV care initiation overestimated receipt of care by 25%. This overestimate is likely driven by the inclusion of CD4/VLs drawn as part of the HIV diagnostic work-up. Patients misclassified as being linked-to-care had significantly greater mortality at one-year, underscoring the importance of refining this proxy measure. Therefore, when timely linkage-to-care is measured using HSR data, we recommend that only CD4/VLs drawn 8-91 days after diagnosis be considered proxies for HIV care initiation.