State-Level Variation in Infant Death Identification and Classification

Background and Purpose Vital statistics are the primary source for tracking infant mortality in the U.S. Accurate classification of infant deaths is important for identifying preventable fatalities, including those related to abuse and neglect. However, this classification often relies on local investigative practices and policies, which vary widely across jurisdictions. Inconsistent use of autopsies, variations in death certification processes, and unresolved classification (e.g., pending investigation) may obscure the true burden of maltreatment-related infant deaths. In addition, recent increases in sudden unexpected infant death (SUID) rates, along with ongoing ambiguity in how SUID is defined and certified, highlight the need to better understand how classification practices vary across states. Using 18 years of national vital statistics data, this study addresses two questions: (1) How do states vary in their manner-of-death classification for all infant deaths and for those with SUID–related underlying causes? and (2) How do states vary in autopsy completion rates, and in what ways do these differences influence the manner-of-death classification?

Methods: This study used the 2005-2022 restricted-use Period Linked Birth/Infant Death data from the National Center for Health Statistics (NCHS) to examine all infant (< 1 year old) deaths (N = 426,773). We examined state-level variations in the manner-of-death classification (natural, accidental, homicide, undetermined, pending investigation, and not specified) and completion of autopsies, overall and among deaths with a SUID-related cause code (ICD-10 codes R95 [sudden infant death syndrome, or SIDS], R99 [unspecified cause], and W75 [accidental suffocation]). We used multinomial logistic regression models to examine the relationship between autopsy completion and manner-of-death classification.

Results: There was substantial variation in manner-of-death classification and autopsy status across states. The percentage of all infant deaths classified with a “not specified” manner ranged from 0.1% to 80.0% across states. Among infant deaths classified as “natural”, the share with an autopsy also varied widely, from 83.0% in California to 11.3% in Oklahoma. Similar variability was found among deaths with SUID-related underlying causes. Nearly all deaths caused by accidental suffocation (W75) were classified as accidental (85.7% to 100%). In contrast, some states predominantly classified SIDS (R95) as natural deaths (e.g., North Carolina: 90.3%), while others deemed them undetermined (e.g., District of Columbia: 90.0%). Unspecified causes (R99) were most often classified as undetermined (e.g., Alaska: 98.4%) or pending investigation (e.g., West Virginia: 74.4%). SUID-related causes without an autopsy were 1.9 times more likely to be classified as “not specified” than “natural”, compared to those with an autopsy.

Conclusions and Implications: Findings suggest that infant deaths related to maltreatment, prenatal substance exposure, or unsafe environments may be substantially undercounted in vital records due to state-level variation in autopsy practice and manner-of-death classification. These variations threaten the reliability of national surveillance and obscure opportunities for prevention. Strengthening coordination among medical examiners, child welfare systems, and public health agencies may help ensure more accurate reporting of infant deaths nationwide.