Abstract / Summary
Background Physical abuse causes morbidity and mortality in young children, yet Latin American emergency data are scarce. Objective To describe characteristics associated with a multidisciplinary child-protection panel's classification and estimate SPUTOVAMO-R sensitivity relative to ICD-10 coding and panel classification. Participants and Setting Children aged 1 month–18 years with emergency department records at a pediatric referral hospital in Bogotá, Colombia (2015–2017), coded with physical-abuse ICD-10 diagnoses. Methods Retrospective cross-sectional study. High- and low-probability classifications were compared using Fisher's exact test and odds ratios (OR) with 95% confidence intervals (CI). Results Of 322 coded children, 59 (18.3%) were classified as high probability. An internally inconsistent account (OR 34.7, 95% CI 12.0–124.0), history inconsistent with injury (OR 18.0, 8.3–40.7), and disclosure by the child (OR 56.1, 12.7–516.7) were strongly associated with high-probability classification. Absence of an adult witness was not associated among children aged 0–4 years and was associated with low-probability classification in older children (OR 0.08, 0.00–0.58). All four high-probability children with high-specificity fractures, and none of three low-probability children, had a history inconsistent with injury. SPUTOVAMO-R sensitivity was 34.2% (110/322; 95% CI 29.0–39.6) relative to ICD-10 coding and 91.5% (54/59; 81.3–97.2) relative to the panel. Conclusions ICD-10 abuse coding corresponded poorly with panel classification. Panel classification was most strongly associated with history coherence. SPUTOVAMO-R was negative in two-thirds of coded children and should be evaluated in unselected populations.