ABSTRACT
Hundreds of thousands of previously untested sexual assault kits (SAKs) have been uncovered in police property storage facilities across the United States, representing a national failure in institutional response to sexual assault. Faced with this discovery, jurisdictions must now decide if and how they should test these kits. Some stakeholders have suggested prioritizing kits for testing by victim, offender, or assault characteristics, based on the belief that these characteristics can predict the likely utility of DNA testing. However, little research has examined the empirical merits of such prioritization. To address this gap in the literature and inform SAK testing policies, we randomly sampled 900 previously untested SAKs from Detroit, MI. The sampled SAKs were submitted for DNA testing, and eligible DNA profiles were entered into Combined DNA Index System (CODIS), the federal DNA database. Police records associated with each SAK were coded for victim, offender, and assault characteristics, and logistic regression analyses were conducted to test whether these characteristics predict which SAKs yield DNA profiles that match (“hit”) to other criminal offenses in CODIS. Testing this sample of previously-untested SAKs produced a substantial number of CODIS hits, but few of the tested variables were significant predictors of CODIS hit rate. These findings suggest that testing all previously-unsubmitted kits may generate information that is useful to the criminal justice system, while also potentially addressing the institutional betrayal victims experienced when their kits were ignored.
Notes
1 CODIS hits can be further differentiated as “offender” or “forensic” hits depending on whether the DNA profile is collected from an offender (“offender hit”) or a crime scene, including a SAK (“forensic hit”). In this article, our references to CODIS hits include both offender and forensic hits.
2 The raw unconditional rate is computed from the outcome variable before list-wise deletion of SAKs with missing data on predictor variables used in the model.
3 These results are similar to the results from the stranger-perpetrated model. We ran a combined model (i.e., stranger-perpetrated and nonstranger perpetrated samples together) to directly test the moderating influence of victim-offender relationship (VOR), and found that VOR was nonsignificant and the combined model produced similar results to the models separated by VOR. Because jurisdictions faced with stockpiles of previously unsubmitted SAKs often evaluate the utility of screening kits by victim and assault characteristics separately for stranger and nonstranger perpetrated SAKs, we present models separated by VOR to increase the utility of our data to practitioners and other stakeholders.