Improving Consistency in Classifying Child Maltreatment for Sudden Unexpected Infant Deaths

After consultation with CDR program managers and epidemiologists, child welfare professionals, and using the American Academy of Pediatrics (AAP) evidence-based safe sleep recommendations¹ as a guide, we identified key variables in the NFR-CRS that might be considered when CDR teams are classifying sleep-related SUID as exposure to hazards or neglect. These variables included infant sleep hazards, supervisor impairment at the time of incident, and whether child protective services (CPS) substantiated the death (Table 1).

The role of CM in the child’s death is documented in Section I5 of the NFR-CRS. The first question in this section, I5a, asks, “Did child abuse, neglect, poor or absent supervision or exposure to hazards cause or contribute to the child’s death?” The response options are yes or probable, no, and unknown. When no or unknown is selected, the rest of the section is skipped. Selecting yes or probable leads to a follow-up question where child abuse, child neglect, poor or absent supervision, or exposure to hazards is chosen. Poor or absent supervision and exposure to hazards are included as options for when the team determines the parent or caregiver failed to appropriately supervise or exposed the child to a hazard, but they do not feel the circumstances rise to the level of neglect. Exposure to hazards and neglect are the most relevant options when classifying CM for sleep-related SUID. For this analysis, we created 3 bivariate outcome variables: exposure to hazards (ExpHaz) compared with no CM, neglect compared with no CM (neglect versus no CM), and neglect compared with exposure to hazards (neglect versus ExpHaz).

Deaths meeting the criteria for sleep-related SUID (Table 2) in the NFR-CRS between 2004 and 2018 were identified. To identify the variables associated with classification of child neglect, or exposure to hazards in the NFR-CRS, we compared the infant sleep hazards, supervisor impairment, and CPS substantiation of the death with each of the 3 outcome variables. Initial comparisons used the χ² statistic to determine statistical significance.

Although many variables were examined initially, the ultimate goal was to identify the fewest variables with the strongest associations, that focus on objective depictions of the circumstances of the SUID. To that end, univariable logistic regression was used to assess the strength of the associations identified in the initial comparisons. Odd ratios (OR) and 95% confidence intervals (CI) were calculated. Finally, multivariable logistic regression was used to eliminate highly correlated variables that did not contribute to model fit and confirm the variables that distinguished between neglect and exposure to hazards. Guidelines were drafted based on the results of the regression analyses.

Race and ethnicity are social constructs, not biological categories. Recognizing that racial and ethnically minoritized populations are overrepresented throughout the child welfare system, compared with the general population,^2–4 we sought to reduce this potential bias by controlling for race in the multivariable logistic regression analysis. Race and ethnicity, obtained from the death certificate and documented in the NFR-CRS, were combined into 1 categorical race and ethnicity variable: White, non-Hispanic; Black, non-Hispanic; Hispanic, any race; additional races, non-Hispanic. This race and ethnicity variable was included in each multivariable logistic regression model as a control variable.

We asked CDR program coordinators from 3 states to help test the guidelines and provide feedback on their utility for increasing consistency in classifying sleep-related SUID. Testing involved inviting CDR team members to a meeting where, for a sample of sleep-related SUID that occurred in their state during 2021, we presented the responses entered into the NFR-CRS for each of the data elements in the guidelines. We then displayed how the death would be classified based on the guidelines and how it was originally classified in the NFR-CRS. For each death reviewed we invited comments and asked whether they thought the guidelines missed any details that would influence their decision on how they would classify the death. Discussion and comments were documented and reviewed later for consideration of guideline revision.

We identified 25 797 infant deaths in the NFR-CRS that met the criteria for sleep-related SUID that occurred between 2004 and 2018. CDR teams determined that CM contributed to 13 034 (51%) of these sleep-related SUID, with 1938 (15%) classified as neglect and 8963 (69%) as exposure to hazards. In the initial exploratory analysis of Table 1 variables and the 3 outcomes, most of the variables were statistically significant (X² < 0.01) for ExpHaz versus NoCM and Neglect versus NoCM. For the outcome comparing Neglect versus ExpHaz, there were few significant associations among the infant sleep variables.

In the univariable logistic regression analyses, infant sleep hazards, including sleeping on a surface not considered safe for infant sleep per the AAP safe sleep recommendations¹ (eg, adult bed, couch, inclined sitting device), objects in the sleep environment (eg, humans, animals, soft bedding), and infant sleep position, were strongly associated with classification of both exposure to hazards and neglect, as were supervisor impairment and substantiation of the death. However, there were some important differences across outcomes. Specifically, sleeping on an unsafe surface and having objects in the sleep environment were strongly associated with and similar across the outcomes ExpHaz versus no CM and neglect versus no CM. ORs ranged from 1.7 (95% CI 1.4–2.1) for inclined sitting device to OR 3.1 (95% CI 2.9–3.3) for sleeping on an adult bed or mattress, and from OR 2.8 (95% CI 2.4–3.3) for pillows to OR 5.0 (95% CI 4.2–5.9) for humans or animals in the sleep environment. The position the infant was placed to sleep was also associated with these 2 outcomes but less strongly (OR 1.3, 95% CI 1.2–1.4). Infant sleep hazards were not associated with neglect versus ExpHaz with all ORs in the range of OR 0.9 to 1.2 with 95% CIs ranging from 0.6 to 1.7. The nonsleep environment hazards, supervisor impairment, and CPS substantiation of the death were strongly associated with each outcome in the univariable analyses with ORs and 95% CIs ranging from OR 1.7 (95% CI 1.5–1.8) for ExpHaz versus no CM to OR 4.8 (95% CI 4.2–5.4) for neglect versus no CM.

The multivariable logistic regression results show a similar pattern to the univariable results while controlling for child race (Table 3). The outcomes ExpHaz versus no CM and neglect versus no CM have similar odds ratios among the sleep hazard variables but neglect versus no CM has stronger associations for supervisor impairment and CPS substantiation. The neglect versus ExpHaz results document that the odds of sleep hazards for sleep-related SUID classified as neglect were no different than those classified as exposure to hazards with ORs ranging from 0.8 to 1.3 (95% CI 0.4–3.0). Rather, the distinguishing characteristics of a sleep-related SUID being classified as neglect are the supervisor was impaired (OR 2.3, 95% CI 2.0–2.7) and CPS substantiated the death (OR 1.8, 95% CI 1.5–2.0).

Based on these findings, we developed a 3-step guideline for classifying sleep-related SUID in section I5 of the NFR-CRS (Fig 1). The first step is to determine if the death meets the sleep-related SUID criteria (Table 2). Step 2 focuses on sleep environment hazards for classification of exposure to hazards in section I5. If a death does not meet any of the sleep hazard criteria, the response to question I5a is no. If any of the sleep hazards are present, the guideline directs the user to go to step 3 to determine if either of the criteria for neglect are met. If so, they are directed to select yes or probable in I5a and child neglect as the reason. If neglect criteria are not met, then they are directed to select yes or probable and exposure to hazards in I5a.

We met with the CDR coordinator and multidisciplinary CDR team members in 3 states to test the guidelines and obtain feedback on the perceived usefulness of the guidelines for improving consistency in classification of sleep-related SUID. At each meeting, after a brief introduction to the project background and goals, we reviewed the sleep hazard, supervisor impairment, and CPS substantiation information for a sample of 15 to 20 sleep-related SUID that occurred in their state during 2021. The CDR coordinator read additional information about the circumstances of the death if any were documented in the NFR-CRS narrative.

The ensuing discussions helped to identify where additional details were needed in the instructions for teams using the guidelines. In general, the classification of exposure to hazards in step 2 was straightforward and no additional criteria were requested by the teams. The classification of neglect elicited the most discussion. The primary themes that evolved from the 3 meetings were questions related to supervisor impairment, conflating absent or inadequate supervision with impairment, whether parental exhaustion was an impairment, why impairment by illness or disability was included, and the role of intent in the supervisor’s actions. During these discussions the goal of reducing inequitable and inconsistent classification by focusing the guidelines on objective measures related to the circumstances at the time of death were reiterated. After discussions and detailed explanations of the criteria for classifying neglect, no changes in the guidelines were requested during these meetings.

Following recommendations from CDR team members, child welfare professionals, the American Academy of Pediatrics,¹ and analysis of existing NFR-CRS data on the circumstances of sleep-related SUID, guidelines to improve the consistency of CM classification in section I5 of the NFR-CRS were developed. These guidelines prompt classification of exposure to hazards based on presence of hazards in the sleep environment, whereas classification of neglect is based on supervisor impairment and child welfare action, in addition to sleep environment hazards.

Prior research has documented the challenge of classifying child deaths as neglect-related by CDR teams and identified intent and previous neglectful behavior as attributes that influence their classification of neglect.⁵ However, intent is rarely known, and subjective judgements of intent might be implied by CDR team members based on bias and inconsistently documented circumstances. To reduce the potential for past CPS involvement to bias CM classification, the guidelines we developed are based solely on objective circumstances of the death documented during the review. Social and demographic characteristics that might subjectively influence classification, such as poverty, prior CPS history, race, or ethnicity, were purposely excluded in an effort to improve equitable classification.

During pilot testing, none of the teams we met with had questions about the variables used to determine exposure to hazards; all the questions and discussions were focused on the classification of neglect. Although there was an occasional question about the use of substantiation by child welfare as a criterion for neglect classification, all teams agreed that it was important to include this for consistency across the NFR-CRS data. The supervisor impairment criteria dominated the discussions. Specifically, the teams asked if supervisor exhaustion should be considered, particularly if it was related to poor judgement, such as staying up late into the night playing video games rather than exhaustion typical for caregivers of infants. The inclusion of some measure of inappropriate supervision was also discussed by all 3 teams. One team questioned why impairment caused by disability and illness were included, questioning whether impairment should be restricted to a choice the supervisor made, eg, choosing to drink or take drugs.

In response to these questions, we explained that including supervisor exhaustion would require an effort to distinguish between exhaustion that typically occurs when caring for a young infant and exhaustion caused by choices the caregiver made, such as staying out late with friends. This distinction is highly judgmental and would require that information on all the caregiver’s activities leading up to the death are consistently known and documented during the review for every sleep-related SUID. The goal of the guidelines is to reduce subjectivity and judgement and base classification on objective measures as much as possible. We also reiterated that these are guidelines and as such, teams can classify these deaths based on all the facts presented during the review and consensus of the team. However, we continued to emphasize that focusing on objective measures would result in more consistent classification within and across CDR teams.

Inappropriate (or absent) supervision was brought up by all 3 teams. Our response and explanation focused on the fact that there is no consensus on appropriate supervision of children across the age spectrum, except that children less than 6 years old require constant supervision most of the time.⁶ However, it is not expected that parents are constantly watching their infants as they sleep. Given that these guidelines are for the classification of infant deaths in the sleep environment, the focus is on hazards in the sleep environment, and the role that supervisor impairment might play in creating a hazardous sleep environment. If an infant is placed in a safe environment to sleep, a lapse in supervision will not make the sleep environment less safe.

In response to one team’s concern about including supervisor impairment because of illness or disability, we explained that, when documented in the NFR-CRS, being impaired by illness or disability refers to a physical illness, mental illness, or condition that renders a person incapable of safely caring for an infant at the time of the incident. This is not intended to place blame. Rather, noting impairment by illness or disability is critical to inform prevention, such as making sure systems are designed to better support parents and caregivers with an illness or disability.

Guidelines to improve consistency in the classification of CM among sleep-related SUID were developed. Classification of exposure to hazards is based on presence of sleep environment hazards, whereas neglect is based on supervisor impairment and child welfare action. Testing the guidelines with CDR teams in 3 states provided helpful insight into components of the guidelines that needed additional clarification. The guidelines will encourage consistent CM classification based on incident circumstances, rather than an inequitable classification influenced by social or demographic factors. The guidelines were released for use by all CDR teams in March 2024, and their impact will be assessed in the 2 to 3 years after widespread implementation. Data quality improvements and the increased consistency in classifying sleep-related SUID generated by these guidelines will aid in a better understanding of how behavior on the part of parent or supervisor contributes to sleep-related SUID. This increased consistency will facilitate development of specific, focused prevention recommendations for clinical pediatricians as well as pediatricians and others serving on CDR teams.

The authors thank all the CDR team members who participated in the pilot testing phase of guideline development, and the CDR Coordinators who assembled these team members for the pilot meetings: Jill Munger (SD), Sarah Blezinger (TX), and Kate Jankovsky (CO).

CDR

child death review

CM

child maltreatment

NFR-CRS

National Fatality Review-Case Reporting System

SUID

sudden unexpected infant death