Predicting Sexual Assault Perpetration in the U.S. Army Using Administrative Data

Abstract

Introduction The Department of Defense uses a universal prevention framework for sexual assault prevention, with each branch implementing their own branch-wide programs. Intensive interventions exist, but would be cost-effective only if targeted at high-risk personnel. This study developed actuarial models to identify male U.S. Army soldiers at high risk of administratively-recorded sexual assault perpetration. Methods This study investigated administratively-recorded sexual assault perpetration among the 821,807 male Army soldiers serving 2004–2009. Other temporally prior administrative data were used as predictors. Penalized discrete-time (person-month) survival analysis (conducted in 2016) was used to select the smallest possible number of stable predictors to maximize number of sexual assaults among the 5% of soldiers with highest predicted risk of perpetration (top-ventile concentration of risk [COR]). Separate models were developed for assaults against non-family and intra-family adults and minors. Results 4,640 male soldiers were found to be perpetrators against non-family adults, 1,384 against non-family minors, 380 against intra-family adults, and 335 against intra-family minors. Top-ventile COR was 16.2–20.2% predicting perpetration against non-family adults and minors and 34.2–65.1% against intra-family adults and minors. Final predictors consisted largely of measures of prior crime involvement and the presence-treatment of mental disorders. Conclusions Administrative data can be used to develop actuarial models that identify a high proportion of sexual assault perpetrators. If a system is developed to routinely consolidate administrative predictors, predictions could be generated periodically to identify those in need of preventive intervention. Whether this would be cost-effective, though, would depend on intervention costs, effectiveness, and competing risks.

A Machine Learning Approach to Identify NIH-Funded Applied Prevention Research

“…For example, ML algorithms can predict adverse birth outcomes, 19 and identify soldiers at risk for sexual assault perpetration. 20 Rose 21 used a similar ensemble ML approach to predict mortality risk in an elderly California population. ML approaches have also been used by researchers to study the etiology of depression, 22 predict recurrence of complex diseases, 23 and detect interactions between risk factors for complex diseases.…”

Improving risk prediction accuracy for new soldiers in the U.S. Army by adding self-report survey data to administrative data

“…As reported in previous publications, separate composite risk scores for each outcome were developed based on models from either the STARRS Historical Administrative Data System (HADS) [ 8 , 9 , 12 ] or the NSS [ 13 ]. The details of building the models that generated these scores are reported in the original papers and will not be repeated here other than to say that they involved the use of iterative machine learning methods [ 20 ] with internal cross-validation to predict the outcomes over a one-month risk horizon in a discrete-time person-month data array [ 21 ].…”

“…Results of performance-based neurocognitive tests administered in conjunction with the NSS were also included in the NSS models [ 22 ]. More detailed descriptions of the HADS and NSS predictors, the final form of each model (i.e., the variables that were ultimately selected for inclusion by the algorithms), and predictive performance are presented in the original reports [ 8 , 9 , 12 , 13 ].…”

The Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS): progress toward understanding suicide among soldiers

“…Machine learning models have also been developed to predict the following: suicides after outpatient mental health specialty visits [75], sexual violence victimization [76] and perpetration [77], non-familial major physical violent crime perpetration [78], and minor violent crime perpetration [79], as well as diverse negative outcomes among new soldiers in the NSS [80]. The high concentration of risk obtained through these machine learning analyses suggests that such models can be useful in targeting soldiers with the highest need for preventive interventions, although final determination requires careful weighing of intervention costs, effectiveness, and competing risks.…”