Work

Ethnic disparities in the predictive validity of forensic risk assessment instruments: a systematic review and meta-analysis

Year: 2026
Type: review
Abstract: SUMMARY Background It remains unclear whether the predictive validity of specific risk assessment instruments used in forensic populations differs between ethnic groups. Prior systematic reviews each examined a single instrument or a single ethnic contrast, which precludes comparison of instruments under a uniform methodology. Methods We conducted a pre-registered systematic review and meta-analysis (PROSPERO CRD 420251246825) in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA-2020) guideline and assessed risk of bias with the Prediction model Risk Of Bias ASsessment Tool (PROBAST). We searched PubMed, PsycINFO, EMBASE, MEDLINE, Web of Science, and Google Scholar until 1 November 2025 and manually screened reference lists of included studies. We included only studies reporting discrimination (area under the receiver operating characteristic curve, AUC, with 95% confidence interval) for recidivism across different ethnic groups. We excluded overlapping samples and risk instruments with only one eligible study per ethnic group. Differences between ethnic groups were tested formally, and the involvement of instrument developers was examined in sensitivity analyses. Findings Three risk instruments met inclusion criteria: the Youth Level of Service/Case Management Inventory (YLS/CMI), the Structured Assessment of Violence Risk in Youth (SAVRY), and the Static-99R. We included 27 studies with a total of 47,579 participants. Across all three instruments and all ethnic groups, discriminative accuracy was moderate. Differences between ethnic groups were not significant for the SAVRY. Accuracy was higher in Asian than in White and Indigenous youth for the YLS/CMI (0.76 vs 0.65; p ≤ .008) and higher in Asian than in Indigenous and Hispanic men for the Static-99R (0.72 vs 0.64; p < .001); each of these differences was driven by a single large study. Static-99R estimates in Indigenous and Hispanic men (both 0.64) were lower than in White men (0.73; p = .04) but not than in Black men (0.71), and these differences did not survive correction for multiple comparisons. Eleven of the thirteen within-study ethnic comparisons were not significant. All Static-99R estimates for White and Black men derived from studies co-authored by instrument developers. Findings were similar across outcome definitions (conviction vs. charge/re-arrest), interrater reliability and follow-up duration. Interpretation The YLS/CMI, SAVRY and Static-99R showed broadly comparable discriminative accuracy across the ethnic groups studied. For the Static-99R, lower estimates in Indigenous and Hispanic men and one significant within-study difference do not establish a disparity, but do not allow one to be excluded either. Comparable discrimination does not imply comparable calibration. Predictive validity cannot be assumed when an instrument is applied outside the sociocultural and justice-system context in which it was developed. Independent local validation studies that report both discrimination and calibration are needed.
Language: English
Field: Psychology
Open Access status: gold
Estimated APC paid: $2,120
OpenAlex ID: W7158445501
FWCI:
Cites: 43
Cited by:
Related to: 10