Risk for Severe Intimate Partner Violence in Nairobi's Informal Settlements: Tailoring the Danger Assessment to Kenya.

Introduction: Understanding the risk for severe intimate partner violence (IPV) can help women and providers assess danger. The validated, widely used Danger Assessment (DA) developed for this purpose has not been tested in a low- and middle-income country (LMIC). We tailored the DA to Nairobi, Kenya, and prospectively evaluated baseline danger against severe IPV at 3-month follow-up.

Methods: We used data from the myPlan Kenya trial conducted in 3 informal settlements in Nairobi, Kenya, from 2017 to 2018. DA items were refined through formative and pilot phases, yielding minor wording modifications. Quantitative analyses prospectively evaluated baseline DA against severe IPV at 3-month follow-up to understand the predictive effect of the (1) original 20-item DA, (2) 16-item Kenya-DA (highest relative risk ratios [RRR] with severe IPV), and (3) 16-item Kenya-DA weighted (weighting based on strength of RRRs). Diagnostic criteria, including C-statistics, sensitivity, specificity, receiver operating characteristic curve, and area under the curve, were examined; logistic regressions quantified the odds of each metric predicting severe IPV at follow-up.

Results: The original 20-item DA produced the highest specificity (75.41%) and lowest sensitivity (57.14%), resulting in the overall lowest C-statistic. Compared to the 16-item Kenya-DA, the Kenya-DA weighted produced slightly higher sensitivity (66.67% vs. 64.29%) and specificity (77.05% vs. 72.13%), resulting in the highest C-statistic (0.78 vs. 0.75). All versions successfully predicted severe IPV at 3-month follow-up (original DA: odds ratio [OR]=1.26, 95% confidence interval [CI]=1.12, 1.41, P<.001; Kenya-DA: OR=1.33, 95% CI=1.16, 1.53, P<.001; Kenya-DA weighted: OR=1.19, 95% CI=1.10, 1.28, P<.001). Several factors identified as homicide risk factors in other settings were not prospectively associated with severe IPV.

Conclusion: Within a high-danger LMIC context, all 3 DA configurations performed well diagnostically. We recommend the 16-item Kenya-DA given the value for simplicity and field implementation, whereas the Kenya-DA weighted can add accuracy for research purposes.