Theme: Clinical Research Year: 2024
Background:
People who use drugs (PWUD) are at high risk of acquiring HIV, HBV and HCV. Recent HIV outbreaks among PWUD, preceded by undetected HCV outbreaks, show the need for timely outbreak detection and identification of high-risk individuals. We aimed to develop an automated system for real-time outbreak detection and individual risk assessment for HIV, HBV and HCV among PWUD.
Methods:
Various data sources provided by the Greek Organization against Drugs (OKANA) within the “Hippocrates” project were combined to monitor new weekly diagnoses per infection and region. To detect outbreak onset and its subsequent states (i.e., decline and/or post-epidemic state), we applied different methods including control charts, a new regression-based method and a novel approach using hidden Markov model (HMM-PI). Performance was prospectively assessed with sensitivity, specificity and timeliness (i.e., difference between the start of each state and the first alarm after its onset) on a previous HIV outbreak among PWUD in Athens, Greece. To predict one-year infection risk, a Cox model was constructed using risk factors selected by Random Survival Forest among demographic, socio-economic, behavioural data and drug types. Performance was evaluated on new data with time-dependent area under curve (AUC).
Results:
To detect outbreak onset, all methods achieved excellent performance (sensitivity>95%, specificity>85%, no delay in alarm signal). To detect decline, HMM-PI had the highest performance (sensitivity 84%, specificity 89%, timeliness 15 weeks). To detect post-epidemic state, methods’ performance was unsatisfactory. The risk assessment tool contained nine factors: age, stable housing, unemployment, education level, age at first injection, syringe sharing, primary method of use being injection, injecting and daily use in the past month; yielding satisfactory AUC (>70%).
Conclusion:
An automated system that monitors new HIV, HBV and HCV diagnoses among PWUD and provides alert for high-risk individuals can benefit PWUD and other vulnerable populations susceptible to these infections.