Predictive value of indicators for identifying child maltreatment and intimate partner violence
A systematic review and meta-analysis evaluating the accuracy of coded electronic health records for family violence.
Published in Archives of Disease in Childhood
Read the full peer-reviewed systematic review assessing the validity of coded indicators.
Intimate partner violence (IPV) and child maltreatment (CM) represent forms of family violence that often go undetected by support services, despite World Health Organization (WHO) recommendations to improve monitoring efforts. CM and IPV refer to any act that causes biopsychosocial harm to a child, a future child, or a partner. In the UK, statutory definitions of CM also include conditions resulting from prenatal neglect, such as fetal alcohol syndrome (FAS) and neonatal abstinence syndrome (NAS).
Assessing health records manually for detailed information on family violence is time-consuming and resource-intensive. Consequently, clinical services and researchers are increasingly utilising routinely coded electronic health records (EHRs) to assess family violence. Coded EHRs offer the potential for longitudinal population-based assessments, automated early warning systems, and the identification of high-risk populations at a relatively low cost.
However, the validity of these coded indicators has remained uncertain due to reported quality issues and a lack of external validation. To address this, a comprehensive systematic review and meta-analysis was conducted to estimate the positive predictive values (PPVs) of coded indicators for different forms of family violence based on independent reference standards.
Key findings
The researchers searched 18 electronic databases and 20 selected journals, adhering to PRISMA-DTA and MOOSE guidelines. The final meta-analysis included 65 cross-sectional and 23 longitudinal studies, encompassing 20 distinct clinical indicators and evaluating 3,875,183 individuals across 11 different countries.
- High predictive value: The pooled PPV for general child maltreatment (0-18 years) was 87.8%, and for IPV among women (12-50 years) was 86.1%.
- Prenatal neglect indicators: The pooled PPV of primary diagnoses for neonatal abstinence syndrome (NAS) was 80.9%, while fetal alcohol syndrome (FAS) was 39.3%.
- Variation in physical injury indicators: Specific CM indicators showed variation; the PPV for rib fractures was notably high at 88.3%, whereas multiple burn injuries in children under 5 years yielded a PPV of 19.6%.
- Injury-related IPV: General injury-related presentations of IPV provided relatively lower predictive values compared to explicit diagnostic codes.
- Coding accuracy: The proportion of misclassifications (false positives) resulting strictly from coding errors averaged 2.1%. However, in assault-coded cases among women, 28.0% lacked recorded perpetrator information in the underlying medical charts.
Lead author Dr Shabeer Syed (UCL Great Ormond Street Institute of Child Health), said:
“Our analysis demonstrates the utility of using routinely coded medical data to evaluate services for at-risk groups. The consistently high positive predictive values for explicit child maltreatment and IPV indicators in electronic health records suggest significant potential for the early identification and support of vulnerable individuals.”
Clinical implications
The findings indicate that utilising EHRs to identify family violence can facilitate improved, targeted care for at-risk individuals.
- Coded indicators of family violence have the potential to be integrated into computerized clinical decision support systems, helping to automatically flag potential at-risk cases to clinicians.
- Linking the EHRs of family members enables a "think-family" approach, aiding practitioners in identifying vulnerable children through maternal records, or vice versa.
- Coded physical injury patterns may serve as a broader measure to identify high-risk groups requiring secondary assessment.
Study limitations
- Predictive estimates varied considerably depending on the specific indicator and outcome, revealing substantial heterogeneity across studies.
- The complexity of definitively identifying CM and IPV in clinical practice means potential missed diagnoses or misclassifications remain a factor in EHR data.
- While automated EHR systems offer benefits for family violence identification, careful consideration of potential unintended consequences—such as stigma, legal repercussions, compromised patient trust, and reduced help-seeking behaviour—is required prior to implementation.
