Evaluating Time-Space Methodologies to Detect Clusters of HIV Transmission: a Comparison of Advanced Methods in Washington State, 2010-2022

Background Identification of HIV transmission clusters is a key activity under the “Respond” pillar of the United States’ Ending the HIV Epidemic initiative, but the most common method for detecting time-space clusters has low predictive value for future diagnoses in Washington state. We compared different methodologies to the current CDC standard to identify alternative techniques for guiding public health outbreak response locally. Setting Washington state, 2010-2022 Methods Using Washington state HIV surveillance data, we applied four methods of detecting anomalies in time series (CDC time-space cluster detection criteria, SaTScan purely temporal, Cumulative Sum [CUSUM], and Log-CUSUM) to the monthly HIV diagnoses at the county level and evaluated predictive ability for increases in diagnoses using micro and macro rate ratios (number of new diagnoses in a region 1 to 12 months after cluster detection relative to a baseline the year before). Results There were 5,335 new diagnoses in Washington state between 2010 and 2022. The CUSUM method detected 57 clusters and had the highest macro (3.7) and micro (1.7) rate ratios in the month after cluster detection. Over a twelve month period, the log-CUSUM had the highest predictive value (54 clusters, macro 1.7, micro 1.3). Conclusion The CUSUM methods showed superior ability to identify regions of sustained increases in HIV diagnoses and should be considered for HIV cluster detection and response activities.