Real-World Estimates of HIV Incidence From Sexual Health Clinic EMRs: Feasibility, Limitations, and Implications for Surveillance.

BACKGROUND

Timely and accurate estimates of HIV incidence are essential for program monitoring and planning clinical studies. Real-world evidence from electronic medical records (EMRs) in sexual health clinics may provide a pragmatic alternative to resource-intensive surveys and surveillance.

SETTING

We conducted a retrospective cohort study across 5 health centers in the United States, South Africa, and Thailand to assess the feasibility of estimating HIV incidence using EMR data.

METHODS

Clients aged 16-64 years assigned male at birth who reported ever having sex with someone assigned male at birth and had at least 1 HIV test between July 1, 2022, and December 31, 2023, were included. Deidentified data were extracted and subcohorts created considering 1 versus more than 1 HIV test during the study period. Incidence rates were calculated within the multiple-test (MT) cohort using person-time at risk; diagnosis proportions were calculated for all cohorts.

RESULTS

All sites met data quality thresholds for essential demographic and clinical variables. A total of 25,801 clients were included. Diagnosis proportions in single-test cohorts were higher (2.9%-12.6%) than in MT cohorts (0%-0.53%). Incidence rates among MT cohorts ranged from 0.22 to 0.72 per 100 person-years, substantially lower than expected, likely due to PrEP use and selection bias toward health-engaged clients.

CONCLUSION

EMR-based HIV incidence estimation is feasible but limited by data variability and selection bias thus incidence estimates from MT cohorts may not reflect true community incidence. Diagnoses data may be used for incidence modeling where there is sufficient sample size.