Infant maltreatment-related mortality in Alaska: Correcting the count and using birth certificates to predict mortality

Objectives: To accurately count the number of infant maltreatment-related fatalities and to use information from the birth certificates to predict infant maltreatment-related deaths. Methods: A population-based retrospective cohort study of infants born in Alaska for the years 1992 through 2005 was conducted. Risk factor variables were ascertained from birth record data and linked with a multi-source data collection system (Alaska Surveillance of Child Abuse and Neglect [SCAN]) to identify maltreatment-related fatalities. Maltreatment-related mortality was classified as 1) confirmed, that is, infant homicide as documented by vital records, 2) probable, as determined by an expert multidisciplinary infant death review committee, and 3) suspect, as determined by the review committee and supported through additional data sources. Results: During the 12-year study period, 143,025 births and 955 infant deaths occurred. Among the 955 deaths, SCAN identified 133 (13.9%) total maltreatment-related fatalities including 22 confirmed, 52 probable, and 59 suspect cases (respective maltreatment-specific infant mortality rates of 0.2, 0.6, and 0.9 per 1000 live births). During multivariate analysis, the strongest risk factors for any infant maltreatment mortality were maternal prenatal substance use (population attributable risk percent [PAR%], 25.2%), one or more other children in the family (PAR%, 34.8%), and having an unmarried mother with (PAR%, 32.4) or without (PAR%, 33.8) the father's name on the birth certificate. Compared to infants with none of these 3 risk factors, the relative risk of maltreatment-related mortality among infants with all 3 plus a father's name identified was 22.2 (95% confidence interval [CI], 8.6-76.7) and for those with all 3 risk factors plus no father's name 79.4 (95% CI, 29.1-283.4). Conclusions: Infant maltreatment-related mortality in Alaska is underestimated using only vital records. Three variables obtainable from birth certificates (or potentially other sources) were highly predictive of maltreatment mortality risk and could be used by clinicians and public health officials for early screening and intervention. (Contains 4 tables and 1 figure.)