The impact of delayed reporting on forensic evidence recovery in anal sexual assault cases: A retrospective study

TL;DR

This study shows that examining victims of anal sexual assault sooner increases chances of finding physical evidence, and delayed exams can still support allegations if documented properly.

Contribution

The study introduces a Random Forest classifier to predict physical findings in delayed anal assault cases and emphasizes the importance of timely forensic exams.

Findings

Earlier examination (≤7 days) significantly increases the likelihood of detecting acute physical findings (OR 6.13).

The Random Forest model showed moderate discrimination with a CV ROC AUC of 0.781 and out-of-fold ROC AUC of 0.762.

Assault-to-examination interval was the most important predictor of physical findings by permutation importance (43.3%).

Abstract

Delayed presentation after sexual assault reduces the probability of detecting injuries or recovering biological evidence. Evidence specific to anal-route allegations and physical findings is comparatively limited.

We retrospectively reviewed 222 consecutive cases alleging anal-route sexual assault seen at a university forensic clinic in Türkiye (2010–2014). Examinations included standardized external anogenital inspection under magnification (colposcopy) and full-body assessment; anoscopy was not routinely available. The primary outcome was acute physical findings (composite of anal/perianal and, when present, parallel genital injuries, plus extragenital traumatic lesions). Bivariate associations used Pearson’s χ² (with df reported; Monte Carlo where appropriate). As a prespecified sensitivity analysis, a Random Forest classifier was trained using pre-examination demographics and incident-context variables with stratified 5-fold cross-validation, reporting ROC AUC, PR-AUC, accuracy, Brier score, and permutation feature importance.

Most victims were female (69.4%) and <18 years (60.8%). The probability of detecting acute findings was higher with earlier examination (e.g., ≤ 7 days vs. later: OR 6.13, χ² = 88.36, p < 0.001). The Random Forest showed moderate discrimination (CV ROC AUC 0.781 ± 0.077; PR-AUC 0.604 ± 0.119; accuracy 0.739 ± 0.079), with out-of-fold ROC AUC 0.762 (bootstrap 95% CI 0.692–0.827) and Brier 0.186. Assault-to-examination interval was the dominant contributor by permutation importance (group share 43.3%), followed by age metrics, prior anal intercourse count, and contextual factors. Findings were directionally similar when the outcome was any physical finding (genital or extragenital).

Earlier examination markedly increases the likelihood of detecting physical findings in alleged anal sexual assaults. Negative examinations after delays are forensically compatible with the allegation and should be documented with precise timelines and standardized magnified imaging. Streamlined, trauma-informed pathways to facilitate timely reporting may improve evidentiary outcomes and reduce secondary victimization.

Figures