The infant health effects of doulas: Leveraging big data and machine learning to inform cost-effective targeting.

Doula services represent an underutilized maternal and child health intervention with the potential to improve outcomes through the provision of physical, emotional, and informational support. However, there is limited evidence of the infant health effects of doulas despite well-established connections between maternal and infant health. Moreover, because the availability of doulas is limited and often not covered by insurers, existing evidence leaves unclear if or how doula services should be allocated to achieve the greatest improvements in outcomes. We use unique data and machine learning to develop accurate predictive models of infant health and doula service participation. We then combine these predictive models within the double machine learning method to estimate the effects of doula services. We show that while doula services reduce risk on average, the benefits of doula services increase as the risk of negative infant health outcomes increases. We compare these benefits to the costs of doula services under alternative allocation schemes and show that leveraging the risk predictions dramatically increases the cost effectiveness of doula services. Our results show the potential of big data and novel analytic methods to provide cost-effective support to those at greatest risk of poor outcomes.

Understanding the rise in overdose deaths involving opioids and non-opioid prescription drugs in the United States.

BACKGROUND: Previous studies demonstrate that the reformulation of OxyContin in the U.S. in 2010 induced substitution to illicit opioids, causing illicit opioid markets to grow disproportionately fast in states more exposed to the reformulation. In this paper, we examine if this shift to the illicit market also led to a rise in polysubstance overdose deaths involving non-opioid prescription drugs, including gabapentinoids and "Z-drugs" and, separately, benzodiazepines. METHODS: Using a difference-in-differences framework, the relationship between exposure to reformulation and overdose death rates including specific substances was studied in each year from 1999 to 2020 while accounting for fixed differences across states, common nationwide shocks, and state-level differences in pain reliever misuse prior to reformulation. Exposure to reformulation was measured as the pre-reformulation rate of OxyContin misuse. RESULTS: Exposure to reformulation predicted growth in overdose deaths involving gabapentinoids and Z-drugs. There is less evidence that it predicted growth in overdose deaths involving benzodiazepines. However, for all substances, there is strong evidence that pre-reformulation OxyContin misuse rates predicted post-reformulation growth in overdose deaths concurrently involving synthetic opioids. DISCUSSION: The opioid crisis has changed in radical ways. This study links a major supply-side intervention to the increase in polysubstance overdose deaths involving non-opioid prescription drugs, specifically gabapentinoids and Z-drugs.

Trends in overdose deaths involving gabapentinoids and Z-drugs in the United States.

INTRODUCTION: As policies have been implemented to limit access to prescription opioids, other drugs have been prescribed off-label, sometimes concurrently with opioids, to manage pain. There are concerns about the use of gabapentinoids and "Z-drugs" with opioids. As the opioid crisis transitions to illicit opioids and polysubstance use, little work quantifies the concurrent involvement of non-opioid prescription drugs and illicit opioids in overdose deaths. METHODS: Data from the census of deaths in the United States for 1999-2020 were used to understand trends in deaths involving gabapentinoids/Z-drugs and opioid co-involvement. These trends were studied overall and by sex, race, age, and education. RESULTS: Per capita overdose deaths involving gabapentinoids/Z-drugs increased almost continuously since 1999, averaging 15.8% annual growth. This rate increased to 32% in 2020, primarily due to overdoses involving synthetic opioids. Women typically had higher rates of overdose deaths involving both opioids and gabapentinoids/Z-drugs, though this disparity disappeared in 2020. White Americans and American Indians/Alaskan Natives historically had higher rates than other racial groups; however, Black Americans experienced over 60% annual growth in recent years. Low education groups have been disproportionately impacted. The age incidence tends to be older than overdoses involving opioids more generally. CONCLUSION: Overdose deaths involving opioids and gabapentinoids/Z-drugs have tended to disproportionately affect women and older age groups compared to all opioid-involved overdoses. As deaths involving synthetic opioids likely reflect use of illicitly-obtained opioids, there may be less of a role for policies targeting the concurrent prescribing of gabapentinoids/Z-drugs with opioids to reduce these deaths.

Variation in the infant health effects of the women, infants, and children program by predicted risk using novel machine learning methods.

The Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) has an extensive literature documenting positive effects on infant health outcomes, specifically preterm birth, low birthweight, small size for gestational age, and infant mortality. However, existing studies focus on average effects for these relatively infrequent outcomes, thus providing no evidence for how WIC affects those at greatest risk of negative infant health outcomes. Our study focuses on documenting how WIC's infant health effects vary by level of risk. In doing so, we leverage a uniquely rich database describing maternal and infant outcomes and risk factors. Additionally, we use high dimensional data to generate predictions of risk and combine these predictions with the novel double machine learning method to stratify the effects of WIC by predicted risk. Our estimates of WIC's average treatment effects align with those in the existing literature. More importantly, we document significant variation in the effects of WIC on infant health by predicted risk level. Our results show that WIC is most beneficial among those at greatest risk of poor outcomes.

Trends in the Concurrent Prescription of Opioids and Gabapentin in the US, 2006 to 2018.

This cohort study examines trends in concurrent gabapentin and opioid prescribing in the US, overall and by prescriber, patient, and county characteristics, between 2006 and 2018.

Trends in Out-of-Pocket Costs for Naloxone by Drug Brand and Payer in the US, 2010-2018.

Importance: Improving access to naloxone is a critical component of the nation's strategy to curb fatal overdoses in the opioid crisis. Standing or protocol orders, prescriptive authority laws, and immunity provisions have been passed by states to expand access, but less attention has been given to potential financial barriers to naloxone access. Objective: To assess trends in out-of-pocket (OOP) costs for naloxone and examine variation in OOP costs by drug brand and payer. Design, Setting, and Participants: This observational study analyzed US naloxone claims data from Symphony Health and associated OOP costs for individuals filling naloxone prescriptions by drug brand and payer between January 1, 2010, to December 31, 2018. The data were analyzed from March 31, 2021, to April 12, 2022. Main Outcomes and Measures: The main measures were trends in annual number of naloxone claims (overall, by payer, and by drug brand) and mean annual OOP costs per claim (overall, by payer, and by drug brand). Results: Of 719 612 naloxone claims (172 894 generic naloxone, 501 568 Narcan, and 45 150 Evzio) for 2010 through 2018, the number of naloxone claims among insured patients began rapidly increasing after 2014; at the same time, the mean OOP cost of naloxone increased dramatically among the uninsured population. Comparing 2014 with 2018, the mean OOP cost of naloxone decreased by 26% among those with insurance but increased by 506% among uninsured patients. For the uninsured population, the impediment of cost was even larger for certain brands of the drug. In 2016, the mean OOP cost for Evzio among uninsured patients rose to $2136.37 (a 2429% increase relative to 2015) compared with the mean cost of generic naloxone, $72.88, and the cost of Narcan in its first year, $87.95. Throughout the period, the mean OOP costs paid by uninsured patients were higher for Evzio at $1089.17 (95% CI, $884.17-$1294.17) compared with $73.62 (95% CI, $69.24-$78.00) for Narcan and $67.99 (95% CI, $61.42-$74.56) for generic naloxone. Conclusions and Relevance: In this observational study, the findings indicated that the OOP cost of naloxone had been an increasingly substantial barrier to naloxone access for uninsured patients, potentially limiting use among this population, which constituted approximately 20% of adults with opioid use disorder.

Tackling Persistent and Large Disparities in Birth Outcomes in Allegheny County, Pennsylvania.

OBJECTIVES: Based in Allegheny County, a coalition of local stakeholders took note of the region's infant mortality rates, particularly the stark disparities observed by race, and established a vision to reduce infant mortality in the region. The group undertook a multi-faceted effort to (1) develop predictive models of infant mortality risk; (2) evaluate the effectiveness of available interventions; and (3) combine these tools in order to tailor intervention referrals based on maternal risk profiles. With this effort, the coalition sought to address the apparent disconnect between the region's robust maternal and child health care system and relatively poor birth and infant outcomes and racial disparities. METHODS: The effort started with the integration of data from a variety of sources into an integrated database built specifically for this research effort covering the period 2003 to 2013. With the database, researchers linked each individual's data across multiple data sources, including the Allegheny County Health Department, the University of Pittsburgh Medical Center, the Allegheny County Department of Human Services Data Warehouse, and individual programs. With these data, we used a standard method for comparing outcomes and measuring the racial disparity between Black and white infants that involved calculating a ratio by dividing the rate or percentage for Black infants by the rate or percentage for white infants. RESULTS: Overall, the results showed that between 2003 and 2013 in Allegheny County disparities were more pronounced for infant mortality (3.25) than low birthweight (1.88) or preterm birth (1.49). Among the different potential causes of infant mortality, the most pronounced disparity was for SIDS (1.78). Among maternal health factors, pre-pregnancy obesity and gestational diabetes had the highest infant mortality disparity. The low birthweight disparity was similar and lower than the infant mortality disparity across all of the maternal health factors, while the preterm birth disparity was even lower. For the maternal behavioral and contextual factors, the infant mortality disparity ranged from 1.5 to 2.3. CONCLUSION: The 11-year span of data reported in the IMPreSIv database and the breadth of intervention data included allowed us to report granular information on birth outcomes within Allegheny County over this time period. The database also allowed us to summarize the various factors associated with the range of birth outcomes and describe the participation rates in the medical and community setting interventions. Against this backdrop of pronounced disparities in birth outcomes across a range of factors, we examined the effectiveness of interventions for women with different risk factors (e.g. substance use disorders) in order to develop a tool to facilitate individualized referrals to the interventions that will help the most for a specific risk profile.

Female Participation in Household Decision Making and the Justification of Wife Beating in Bangladesh.

We examined female participation in household decision making and its association with the justification of wife beating in Bangladesh. We used nationally representative data from the 2014 Bangladesh Demographic and Health Survey. Our sample consisted of currently married women of age 15 to 49 years (n = 16,463). Chi-square tests and multilevel logistic regression models were performed. Approximately 84% of women in the survey were participants in at least one household decision, and 72% reported that wife beating is not justified in any circumstance. Women who reported their participation in at least one type of household decision less frequently reported that wife beating could be justified than those who did not participate in any household decisions (adjusted odds ratio = 1.49; 95% confidence interval = [1.25, 1.78]). In addition to participation in household decision making, other factors including age at first marriage, females' and their husbands' education, religion, parity, contraceptive use, and socioeconomic status were associated with the justification of wife beating. The results indicate that female participation in household decision making is significantly associated with the justification of wife beating in Bangladesh. Further study is needed, but the results suggest that policy makers should consider interventions proven to empower women and lead to increased participation in decision making as methods that may reduce domestic violence against women.

Justification of Wife Beating and Utilization of Antenatal and Delivery Care in Bangladesh.

We examined the association between women's justification of wife beating and their utilization of professional antenatal and delivery care in Bangladesh. We used data which describes a nationally representative sample of currently married women aged 15 to 49 years (n = 3,449). Services from medically trained providers were considered professional antenatal care (ANC) and delivery services. Women's attitudes toward wife beating in five circumstances (if a woman goes out without telling her husband, neglects children, argues with her husband, refuses to have sexual intercourse with her husband, and burns food) were used to describe the justification of wife beating. Chi-square tests and multilevel logistic regression analyses were performed; 74% of the women would not justify wife beating, 65% attended ≥1 ANC visits, 25% attended ≥4 ANC visits, and 44% utilized professional delivery care. Women who would not justify wife beating were more likely to utilize ≥1 ANC visits (adjusted odds ratio [AOR]: 1.89; 95% confidence interval [CI]: [1.26, 2.81]), ≥4 ANC visits (AOR: 1.14; 95% CI: [0.76, 1.71]), and professional delivery care (AOR: 1.93; 95% CI: [1.31, 2.85]). Older age, women's and husband's higher education, lower parity, urban residence (except for ≥1 ANC visits), and higher socioeconomic statuses including divisional differences were significant confounders for increased utilization of both professional ANC and delivery care. In addition, older age at marriage and current unemployment were also associated with increased utilization of delivery care services. This association between women's justification of wife beating and their utilization of professional antenatal and delivery care services has potential implications for maternal and child health policy in Bangladesh where intimate partner violence is commonplace, and societal norms teach women to obey their husbands and accept submissive roles. Public policy should aim to create awareness among women about the negative impact of justifying wife beating, and accepting intimate partner violence on their own and children's health.

Lifetime economic impact of the burden of childhood stunting attributable to maternal psychosocial risk factors in 137 low/middle-income countries.

INTRODUCTION: The first 1000 days of life is a period of great potential and vulnerability. In particular, physical growth of children can be affected by the lack of access to basic needs as well as psychosocial factors, such as maternal depression. The objectives of the present study are to: (1) quantify the burden of childhood stunting in low/middle-income countries attributable to psychosocial risk factors; and (2) estimate the related lifetime economic costs. METHODS: A comparative risk assessment analysis was performed with data from 137 low/middle-income countries throughout Asia, Latin America and the Caribbean, North Africa and the Middle East, and sub-Saharan Africa. The proportion of stunting prevalence, defined as <-2 SDs from the median height for age according to the WHO Child Growth Standards, and the number of cases attributable to low maternal education, intimate partner violence (IPV), maternal depression and orphanhood were calculated. The joint effect of psychosocial risk factors on stunting was estimated. The economic impact, as reflected in the total future income losses per birth cohort, was examined. RESULTS: Approximately 7.2 million cases of stunting in low/middle-income countries were attributable to psychosocial factors. The leading risk factor was maternal depression with 3.2 million cases attributable. Maternal depression also demonstrated the greatest economic cost at $14.5 billion, followed by low maternal education ($10.0 billion) and IPV ($8.5 billion). The joint cost of these risk factors was $29.3 billion per birth cohort. CONCLUSION: The cost of neglecting these psychosocial risk factors is significant. Improving access to formal secondary school education for girls may offset the risk of maternal depression, IPV and orphanhood. Focusing on maternal depression may play a key role in reducing the burden of stunting. Overall, addressing psychosocial factors among perinatal women can have a significant impact on child growth and well-being in the developing world.

Utilization and quality: How the quality of care influences demand for obstetric care in Nigeria.

This paper examines the association between health facility quality, subjective perceptions, and utilization of obstetric care. We draw on unique survey data from Nigeria describing the quality of care at rural primary health care facilities and the utilization of obstetric care by households in the service areas of these facilities. Constructing a quality index using the detailed survey data, we show that facility quality is positively related to perceptions of quality and utilization. Disaggregating quality into structural, process and outcome dimensions, we find a consistently strong relationship only between utilization and structural measures of quality. The results suggest that efforts to improve quality may involve a trade-off between investing in dimensions that are more easily observed by households, which will influence utilization, and investing in dimensions that are more closely related to outcomes.

Correction: Early Childhood Developmental Status in Low- and Middle-Income Countries: National, Regional, and Global Prevalence Estimates Using Predictive Modelling.

[This corrects the article DOI: 10.1371/journal.pmed.1002034.].

Risk Factors for Childhood Stunting in 137 Developing Countries: A Comparative Risk Assessment Analysis at Global, Regional, and Country Levels.

BACKGROUND: Stunting affects one-third of children under 5 y old in developing countries, and 14% of childhood deaths are attributable to it. A large number of risk factors for stunting have been identified in epidemiological studies. However, the relative contribution of these risk factors to stunting has not been examined across countries. We estimated the number of stunting cases among children aged 24-35 mo (i.e., at the end of the 1,000 days' period of vulnerability) that are attributable to 18 risk factors in 137 developing countries. METHODS AND FINDINGS: We classified risk factors into five clusters: maternal nutrition and infection, teenage motherhood and short birth intervals, fetal growth restriction (FGR) and preterm birth, child nutrition and infection, and environmental factors. We combined published estimates and individual-level data from population-based surveys to derive risk factor prevalence in each country in 2010 and identified the most recent meta-analysis or conducted de novo reviews to derive effect sizes. We estimated the prevalence of stunting and the number of stunting cases that were attributable to each risk factor and cluster of risk factors by country and region. The leading risk worldwide was FGR, defined as being term and small for gestational age, and 10.8 million cases (95% CI 9.1 million-12.6 million) of stunting (out of 44.1 million) were attributable to it, followed by unimproved sanitation, with 7.2 million (95% CI 6.3 million-8.2 million), and diarrhea with 5.8 million (95% CI 2.4 million-9.2 million). FGR and preterm birth was the leading risk factor cluster in all regions. Environmental risks had the second largest estimated impact on stunting globally and in the South Asia, sub-Saharan Africa, and East Asia and Pacific regions, whereas child nutrition and infection was the second leading cluster of risk factors in other regions. Although extensive, our analysis is limited to risk factors for which effect sizes and country-level exposure data were available. The global nature of the study required approximations (e.g., using exposures estimated among women of reproductive age as a proxy for maternal exposures, or estimating the impact of risk factors on stunting through a mediator rather than directly on stunting). Finally, as is standard in global risk factor analyses, we used the effect size of risk factors on stunting from meta-analyses of epidemiological studies and assumed that proportional effects were fairly similar across countries. CONCLUSIONS: FGR and unimproved sanitation are the leading risk factors for stunting in developing countries. Reducing the burden of stunting requires a paradigm shift from interventions focusing solely on children and infants to those that reach mothers and families and improve their living environment and nutrition.

Schooling and wage income losses due to early-childhood growth faltering in developing countries: national, regional, and global estimates.

BACKGROUND: The growth of >300 million children <5 y old was mildly, moderately, or severely stunted worldwide in 2010. However, national estimates of the human capital and financial losses due to growth faltering in early childhood are not available. OBJECTIVE: We quantified the economic cost of growth faltering in developing countries. DESIGN: We combined the most recent country-level estimates of linear growth delays from the Nutrition Impact Model Study with estimates of returns to education in developing countries to estimate the impact of early-life growth faltering on educational attainment and future incomes. Primary outcomes were total years of educational attainment lost as well as the net present value of future wage earnings lost per child and birth cohort due to growth faltering in 137 developing countries. Bootstrapped standard errors were computed to account for uncertainty in modeling inputs. RESULTS: Our estimates suggest that early-life growth faltering in developing countries caused a total loss of 69.4 million y of educational attainment (95% CI: 41.7 million, 92.6 million y) per birth cohort. Educational attainment losses were largest in South Asia (27.6 million y; 95% CI: 20.0 million, 35.8 million y) as well as in Eastern (10.3 million y; 95% CI: 7.2 million, 12.9 million y) and Western sub-Saharan Africa (8.8 million y; 95% CI: 6.4 million, 11.5 million y). Globally, growth faltering in developing countries caused a total economic cost of $176.8 billion (95% CI: $100.9 billion, $262.6 billion)/birth cohort at nominal exchange rates, and $616.5 billion (95% CI: $365.3 billion, $898.9 billion) at purchasing power parity-adjusted exchange rates. At the regional level, economic costs were largest in South Asia ($46.6 billion; 95% CI: $33.3 billion, $61.1 billion), followed by Latin America ($44.7 billion; 95% CI: $19.2 billion, $74.6 billion) and sub-Saharan Africa ($34.2 billion; 95% CI: $24.4 billion, $45.3 billion). CONCLUSIONS: Our results indicate that the annual cost of early-childhood growth faltering is substantial. Further investment in scaling up effective interventions in this area is urgently needed and likely to yield long run benefits of $3 for every $1 invested.

Early Childhood Developmental Status in Low- and Middle-Income Countries: National, Regional, and Global Prevalence Estimates Using Predictive Modeling.

BACKGROUND: The development of cognitive and socioemotional skills early in life influences later health and well-being. Existing estimates of unmet developmental potential in low- and middle-income countries (LMICs) are based on either measures of physical growth or proxy measures such as poverty. In this paper we aim to directly estimate the number of children in LMICs who would be reported by their caregivers to show low cognitive and/or socioemotional development. METHODS AND FINDINGS: The present paper uses Early Childhood Development Index (ECDI) data collected between 2005 and 2015 from 99,222 3- and 4-y-old children living in 35 LMICs as part of the Multiple Indicator Cluster Survey (MICS) and Demographic and Health Surveys (DHS) programs. First, we estimate the prevalence of low cognitive and/or socioemotional ECDI scores within our MICS/DHS sample. Next, we test a series of ordinary least squares regression models predicting low ECDI scores across our MICS/DHS sample countries based on country-level data from the Human Development Index (HDI) and the Nutrition Impact Model Study. We use cross-validation to select the model with the best predictive validity. We then apply this model to all LMICs to generate country-level estimates of the prevalence of low ECDI scores globally, as well as confidence intervals around these estimates. In the pooled MICS and DHS sample, 14.6% of children had low ECDI scores in the cognitive domain, 26.2% had low socioemotional scores, and 36.8% performed poorly in either or both domains. Country-level prevalence of low cognitive and/or socioemotional scores on the ECDI was best represented by a model using the HDI as a predictor. Applying this model to all LMICs, we estimate that 80.8 million children ages 3 and 4 y (95% CI 48.1 million, 113.6 million) in LMICs experienced low cognitive and/or socioemotional development in 2010, with the largest number of affected children in sub-Saharan Africa (29.4.1 million; 43.8% of children ages 3 and 4 y), followed by South Asia (27.7 million; 37.7%) and the East Asia and Pacific region (15.1 million; 25.9%). Positive associations were found between low development scores and stunting, poverty, male sex, rural residence, and lack of cognitive stimulation. Additional research using more detailed developmental assessments across a larger number of LMICs is needed to address the limitations of the present study. CONCLUSIONS: The number of children globally failing to reach their developmental potential remains large. Additional research is needed to identify the specific causes of poor developmental outcomes in diverse settings, as well as potential context-specific interventions that might promote children's early cognitive and socioemotional well-being.

Early Childhood Development and Schooling Attainment: Longitudinal Evidence from British, Finnish and Philippine Birth Cohorts.

BACKGROUND: While recent literature has highlighted the importance of early childhood development for later life outcomes, comparatively little is known regarding the relative importance of early physical and cognitive development in predicting educational attainment cross-culturally. METHODS: We used prospective data from three birth cohorts: the Northern Finland Birth Cohort of 1986 (NFBC1986), the 1970 British Cohort Study (BCS1970), and the Cebu Longitudinal Health and Nutrition Survey of 1983 (CLHNS) to assess the association of height-for-age z-score (HAZ) and cognitive development measured prior to age 8 with schooling attainment. Multivariate linear regression models were used to estimate baseline and adjusted associations. RESULTS: Both physical and cognitive development were highly predictive of adult educational attainment conditional on parental characteristics. The largest positive associations between physical development and schooling were found in the CLHNS (ß = 0.53, 95%-CI: [0.32, 0.74]) with substantially smaller associations in the BCS1970 (ß = 0.10, 95% CI [0.04, 0.16]) and the NFBC1986 (ß = 0.06, 95% CI [-0.05, 0.16]). Strong associations between cognitive development and educational attainment were found for all three cohorts (NFBC1986: ß = 0.22, 95%-CI: [0.12, 0.31], BCS1970: ß = 0.58, 95%-CI: [0.52, 0.64], CLHNS: ß = 1.08, 95%-CI: [0.88, 1.27]). Models jointly estimating educational associations of physical and cognitive development demonstrated weaker associations for physical development and minimal changes for cognitive development. CONCLUSION: The results indicate that although physical and cognitive early development are both important predictors of educational attainment, cognitive development appears to play a particularly important role. The large degree of heterogeneity in the observed effect sizes suggest that the importance of early life physical growth and cognitive development is highly dependent on socioeconomic and institutional contexts.