Table 3.
Pooled positive predictive values for NAS, FAS, CM and IPV
| Family violence outcome | Indicator and age cut-off in years | Studies | Ref std. cases/indicator positive | Pooled PPV (95% CI) |
Between-study heterogeneity | |
| (95% prediction interval)* | I2 % | |||||
| Prenatal | ||||||
| NAS | NAS primary diagnosis | 16 | 3030/3796 | 80.9 (71.0 to 87.9) | (32.3–97.4) | 97.4 |
| Newborn affected by drugs | 7 | 384/2202 | 27.2 (16.8 to 40.8) | (6.5–66.7) | 95.0 | |
| FAS | FAS primary diagnosis | 7 | 777/2687 | 39.3 (25.3 to 55.4) | (9.7–79.6) | 97.6 |
| Childhood | ||||||
| CM | CM primary diagnoses ≤18 | 19 | 3090/3717 | 87.8 (83.4 to 91.2) | (63.5–96.8) | 92.1 |
| Abusive head trauma ≤4† | 4 | 437/477 | 91.6 (88.8 to 93.8) | (83.8–96.7) | 36.7 | |
| Traumatic brain injury ≤2 | 7 | 410/1770 | 22.9 (15.3 to 32.9) | (7.0–53.9) | 94.2 | |
| Skull fractures ≤4 | 10 | 603/2872 | 22.1 (18.5 to 26.2) | (13.2–34.7) | 79.7 | |
| Subdural haematoma/haemorrhage ≤3† | 5 | 118/211 | 55.9 (49.2 to 62.5) | (49.2–62.5) | 0.0 | |
| Retinal haemorrhages ≤4† | 4 | 60/74 | 81.1 (70.5 to 88.5) | (70.5–88.5) | 0.0 | |
| Upper limb fractures ≤3 | 4 | 249/298 | 38.5 (11.3 to 75.6) | (1.6–96.1) | 94.9 | |
| Lower limb fractures ≤3 | 6 | 90/272 | 24.0 (12.5 to 41.1) | (4.7–66.8) | 82.0 | |
| Rib fractures ≤4 | 4 | 90/126 | 88.3 (55.2 to 97.9) | (17.7–99.6) | 84.7 | |
| Multiple burns ≤2 | 3 | 70/261 | 22.5 (8.4 to 47.7) | (3.3-.70.9) | 84.3 | |
| Assaults ≤19 | 6 | 218/331 | 77.0 (46.6 to 92.7) | (19.1–97.9) | 97.1 | |
| Assaults/CM/adversity ≤19 | 6 | 1893/4033 | 55.9 (37.0 to 73.3) | (14.4–90.5) | 98.6 | |
| Women predominantly of reproductive age | ||||||
| IPV | IPV primary diagnosis ≤55 | 3 | 1946/2438 | 86.1 (72.2 to 93.6) | (53.0–97.1) | 96.3 |
| Ocular injuries ≤50† | 4 | 22/668 | 3.3 (2.2 to 5.0) | (2.2–5.0) | 0.0 | |
| Facial fractures ≤50 | 4 | 58/1550 | 11.1 (3.0 to 33.9) | (0.6–70.5) | 94.2 | |
| Upper body contusions ≤50 | 3 | 85/325 | 26.5 (18.1 to 37.0) | (13.7–44.9) | 64.9 | |
| Assaults ≤65 | 10 | 6073/20919 | 31.6 (22.3 to 42.7) | (9.2–67.9) | 99.4 | |
Each pooled positive predictive value represents the overall logit transformed estimate by indicator and outcome. Pooled estimates are based on random effects meta-analyses, unless otherwise specified. Individual forest plots are shown in the supplement.
*A 95% prediction interval is a measure of between-study variation and approximates where the PPV is to be expected for 95% of similar future studies.
†Estimates based on a fixed-effects meta-analysis due to lower between-study heterogeneity (ie, I 2 <50%).
CM, child maltreatment; EHR, electronic health record; FAS, fetal alcohol syndrome; IPV, intimate partner violence; NAS, neonatal abstinence syndrome; PPV, positive predictive value; Ref std, reference standard.