The Future of Maternal and Child Health Data in the United States

Pregnancy complications and poor birth outcomes are leading causes of morbidity and mortality for mothers and children in the United States. These outcomes account for more than 40% of neonatal deaths¹ and recently have led to significant increases in childhood morbidity.²

To ensure the health of mothers, children, and families during and beyond the perinatal period, we must have robust data sources from which we can make inferences to effectively shape policy and promote evidence-based interventions. In this issue of AJPH, the article titled “The Pregnancy Risk Assessment Monitoring System (PRAMS): Overview of Design and Methodology,” by Shulman et al. (p. 1305), underscores the importance of having public health surveillance systems for monitoring perinatal health outcomes. PRAMS data continue to serve as a critical resource to evaluate public health programs, track trends, and monitor emerging health issues.

Perinatal data sources are commonly limited by incomplete coverage of pregnancies and births, lack information on social determinants, or do not adequately capture critical exposure periods. To propel the field of maternal and child health forward, enhancements are needed in the design, analysis, and applications of perinatal health data sources for research, policy, and practice. Key opportunities for improvement include the application of theory, data linkage, and sound methodological approaches to these data resources (Figure 1).

FIGURE 1—

Links Between Maternal and Child Health Theory, Data Linkage, and Data Analytics

Note. “Theory” depicts a social-ecological framework indicating the types of factors that affect maternal and child health. The macrosystem consists of cultural contexts and societal beliefs; the exosystem encompasses factors that do not affect individuals directly but may do so indirectly (e.g., health care system factors); the microsystem refers to individuals’ immediate surroundings (e.g., family environment); and individual factors are factors internal to individuals (e.g., health behaviors). The chronosystem refers to the patterning of environmental events and transitions across the life course, including sociohistorical circumstances. Data linkage refers to connecting multiple data sources that relate to the same individual or dyad (or triad). Advanced analytics can be used to gain insights into maternal and child health data.

APPLICATION OF THEORY

Conceptualizing how key determinants accumulate across the life course is necessary to fully operationalize and investigate root causes of perinatal health outcomes. Therefore, the development and enhancement of perinatal data systems should be guided by key theories, including a life course developmental perspective, social-ecological systems theory,³ the multiple determinants of health perspective, and frameworks of perinatal health.⁴

Notably, PRAMS applies a life course developmental approach to data collection by including questions about the preconception and postpartum periods. However, we must go further upstream; mounting evidence shows that many exposures and experiences that affect birth outcomes occur even earlier in life.⁵ For example, researchers may consider how exposures to stressful life events in childhood and young adulthood affect birth outcomes. Similarly, we need to apply a life course approach to understanding child and adolescent health. For instance, to understand the long-term impact of having a low birthweight, we need data sources that follow children beyond the postpartum and early childhood periods.

As informed by social-ecological systems theory and the multiple determinants of health perspective, it is hypothesized that perinatal health outcomes are produced by various determinants, including individual-, family-, and community-level factors. As such, PRAMS and other perinatal data systems should incorporate measures or link to data sets that would allow for examination of multiple domains of determinants (e.g., environmental, psychological, and biological) and time periods in a woman’s life (e.g., preconception, including childhood, adolescence, and interconception periods; pregnancy; and postpartum) that are uniquely and cumulatively associated with health outcomes.

These exposures, contexts, and timing all affect women’s “health capital” at conception. Health capital is conceptualized as the culmination of biological, psychological, and social experiences, exposures, and resources across the life course and generations. Maternal health capital is viewed as a lens through which exposures contribute to a spectrum of health outcomes, such that women with more positive health capital will be less likely to experience poor obstetric and health outcomes attributable to such exposures.⁶

DATA LINKAGE

Data linkage offers a promising avenue for building the comprehensive data sets needed to understand the multiple determinants of perinatal health outcomes. Analyses aiming to address community and health policy factors associated with perinatal health outcomes must incorporate longitudinal information about family-, community-, state-, and federal-level variables. Assembling comprehensive longitudinal data resources may be beyond the mandate of an individual surveillance system but might be achieved by linking surveillance data to administrative, health, and environmental data sources covering a broader range of determinants of perinatal health outcomes.

As surveillance systems develop, consideration of potential data linkages should be incorporated into data plans and protocols. It is particularly important to provide linkages around family structures capturing mother–child–father triads and potentially wider social support networks. Secure data linkages can range from data use agreements to solutions employing technologies aimed at linking while securely preserving the privacy of participants. PRAMS data have been linked to Medicaid claims, child welfare records, vital records, and Head Start data. Perinatal data could be further enhanced by linking to other domains’ data sets, including housing, education, health policy, and environment. Data linkage is critical for leveraging resources across government agencies and potentially incorporating private data sources. Data linkage also allows for creation of measures of latent constructs (e.g., neighborhood disadvantage) to better understand the role of the community context and resources thought to influence perinatal health outcomes.

STATISTICAL AND OTHER ANALYTIC APPROACHES

There is a need for application of current and sophisticated methodological approaches to analyses of perinatal data. Although multilevel and longitudinal data analyses are not new approaches, these methods could be used more frequently in perinatal research. Multilevel analyses that allow for examination of family, community, and health policy data are critical in understanding the role of these factors in outcomes. Other methods such as machine learning, in particular application of cognitive computing, could provide new insights into existing data, unstructured data, or published literature.

For example, through use of cognitive computing tools, a large body of unstructured policy documents could be read efficiently and comprehensively via natural language processing. Linkage of such data to surveillance data would also allow for examination of the effects of policies on outcomes. For instance, state policymakers would be able to assess the impact of policy changes in criminalization of maternal substance use on access to treatment and, ultimately, the incidence of neonatal abstinence syndrome. Moreover, such tools could be used to develop clinical risk assessment measures to further enhance screening, treatment, and outcomes for highly prevalent conditions such as maternal behavioral disorders.

KEY CHALLENGES AND PROPOSED SOLUTIONS

There are several potential challenges to these recommendations. Application of theory to the development of study designs, data architecture, data collection, and analyses is often not economically or administratively feasible. Data linkage presents an excellent, cost-effective opportunity to put theory into practice through linkage of data sets that allow for examination of hypothesized associations. However, deterministic matching often is not possible, and thus there is the potential for exclusion of unmatched dyads that might be systematically different from matched dyads. A solution is to use probabilistic matching to allow inclusion of more dyads in the sample and reduce possible selection bias⁷ or to incorporate data linkage strategies in data plans and protocols. Finally, implementation of such recommendations at the local, state, and federal levels may be challenging in terms of fiscal and human resources.

CONCLUSIONS

Perinatal surveillance systems are critical for ensuring the health and well-being of mothers and children in the United States. Further enhancements, including application of theory, data linkages, and methodological approaches, stand to propel the maternal and child health field forward in developing data resources that will help identify root causes of perinatal health outcomes and inform clinical practice and health policies. Ultimately, these systems will provide the data needed to understand how to improve quality of life and care for women, children, and their families and the communities in which they live.