This study assesses Sexual Assault Kit (SAK) evidence selection in the development of a SAK Evidence Machine-Learning Model (SAK-ML Model).
This research study addresses the gap in research on Sexual Assault Kit (SAK) evidence selection protocols to establish best practice guidelines for SAK evidence selection for analysis and also explore the development of a Sexual Assault Kit evidence Machine Learning Model (SAK-ML Model) software program. The goal of this study was to extract and analyze information related to SAK evidence collection and analysis to inform practice and policy. This study highlights the benefits of data collection and analysis from sexual assault medical forensic examinations (SAMFE) forms and SAK testing outcomes. Results were obtained by tracking SAKs from evidence collection, data from SAMFEs, DNA analysis results, and data from publicly funded laboratories. Information gleaned from evaluating DNA analysis findings have significant practice and policy implications. The study had two purposes: to evaluate decision-making protocols on DNA evidence contained in SAKs to develop research-based guidelines regarding which swabs and how many swabs should be tested by crime lab (Part 1); and to develop, implement and evaluate a machine learning statistical model, SAK-ML Model to guide forensic scientists within publicly funded forensic laboratories on the selection of the most probative SAK swabs to analyze (Part 2). The findings from this study have significant implications for practice and policy recommendations for SAK evidence collection and analysis and, therefore, implications for criminal justice in the investigation and prosecution of sexual assault cases. The end-point product of this study was to develop a machine learning model to guide decision-making in the selection of SAK evidence/swabs for analysis. By aggregating de-identified data across disciplines, the researchers aim to develop greater collaboration within communities and improve criminal justice outcomes for survivors.