The economic impact of untreated maternal mental health conditions in Texas

We conducted a systematic literature search following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to obtain measures of association on the relationship between exposure to untreated MMHCs and maternal and child outcomes [7]. We applied three key inclusion criteria to guide our selection of articles: (1) studies had to use adequate controls to verify that the reported effects occurred to MMHC exposure, rather than to potentially confounding factors (either by adjusting for confounding variables or by using a valid matched comparison design); (2) outcomes had to have relevance to the one-, two-, and five-year time frames of the model; and (3) the study had to quantify outcomes in monetary terms. We sought to identify credible, statistically significant measures of association for use in our model; we excluded papers reporting only statistically non-significant results. We retained papers that met quality standards and had significant results for some outcomes of interests and nonsignificant results for others. We considered studies that sufficiently adjusted for confounding variables or used a matched comparison group design rigorous enough to have generated credible measures of association and excluded estimates from studies that did not meet these design and analysis criteria. We started with the articles from Luca et al. (2020), and used the Ovid MEDLINE, CINAHL, Cochrane, APA PsycInfo, and Scopus databases and key search terms to identify newer articles published in peer-reviewed journals. To include the most current and relevant evidence, we restricted articles to those published from 2008 to 2020, prioritizing articles with Texas-specific estimates where possible. We supplemented these articles with references from other publications focusing on the costs and impacts of untreated MMHCs and grey literature.

We identified 111 relevant articles that (1) controlled for confounding variables or used a matched comparison group design, (2) focused on outcomes occurring between conception and the child’s fifth birthday, and (3) estimated associations in numeric or monetary terms. Supplementary Figure 1 and Table 1 contain more information on our literature search strategy and results, and Supplementary Table 2 focuses on the measures of association we found through the literature search. For relationships where more than one measure of association was identified, we calculated the unweighted average and used this in our model. We did not weight estimates from each study by population size.

We estimated the prevalence of MMHCs in Texas and the prevalence or incidence of each outcome in the Texas population. In selecting estimates, we prioritized recency and relevance to pregnant and postpartum women and children ages 0 to 5. For example, the baseline rates of absenteeism, presenteeism, and unemployment apply to reproductive-age women or women with young children. Supplementary Table 3 contains the MMHC prevalence estimates we used in the main model and subgroup analyses.

To calculate the total cost of MMHCs in Texas, we estimated the incremental cost of each outcome attributable to exposure to untreated MMHCs. We estimated baseline costs using peer-reviewed literature and data from the Centers for Disease Control and Prevention (CDC), Bureau of Labor Statistics (BLS), and other government agencies. We collected granular estimates of direct and indirect costs from the literature, accounting for both immediate and downstream costs, and excluded overlapping cost components to avoid overstating costs. We standardized costs to annual units and converted costs to 2019 dollars using the medical component of the consumer price index (CPI) for medical costs and the CPI for all urban consumers, less food and energy, for non-medical costs. Supplementary Table 4 provides details on incremental cost estimates, the interpretation of these estimates, and data sources.

Using Microsoft Excel, we collated estimates and developed a cost-of-illness model to quantify the costs of MMHCs among Texas mothers giving birth in 2019 and their children (Supplementary Table 5). For each outcome, we collected a range of measures of association from studies identified in our literature review. Next, we converted the measures of association to percentage point changes in the likelihood of each outcome given exposure to untreated MMHCs.

We then calculated the expected rate of each outcome among women with MMHCs and their children by adding together the baseline rate of each outcome and the measures of association. For example, the measure of association for cesarean sections represents the incremental risk of a woman with an MMHC needing a cesarean section compared to a woman without MMHCs. The sum of the baseline rate and this estimate represents the incremental risk of a woman with an untreated MMHC needing a cesarean section. We calculated the annual incremental cost of each outcome given exposure to untreated MMHCs by multiplying together the incremental risk of the outcome, the expected number of women with MMHCs, and the per capita cost of each outcome.

Our model assumes modifiable associations of MMHCs with key outcomes; in other words, treatment for MMHCs would eliminate the excess risk of adverse outcomes. Our model reflects potential complex causal pathways from MMHCs to adverse outcomes as represented in published literature.

We projected costs through five years postpartum by assuming that:

We then projected costs through five years postpartum by calculating the percentage change between the December 2018 and December 2019 medical CPIs. Based on the percentage change, we assumed that medical costs increased by 4.53% annually beyond Year 0 [9]. Finally, following a recommendation by the U.S. Preventive Services Task Force, we discounted costs at an annual rate of 3%. This reflects the lower present economic value of expenses incurred in the future [10]. We summed together discounted costs to estimate the total cost of each outcome.

To understand the sensitivity of our cost estimates to variation in prevalence, cost, and measures of association, we conducted one-way deterministic sensitivity analyses, calculating differences from the main model as we varied each parameter from the lowest to highest value identified in the literature. (If only one estimate of a parameter appeared in the literature, we used the upper and lower bounds of the 95% confidence interval as the lowest and highest values).

The total estimated cost of exposure to untreated MMHCs in Texas was $2.2 billion for the 2019 birth cohort, accounting for costs incurred in the six years between conception and five years post-delivery (Table 1). Assuming that 49,816 out of 377,397 Texas mothers (13.2%) who gave birth in 2019 had an untreated MMHC, these conditions cost more than $44,000 per affected mother–child pair over the six model years [1, 11]. This cost is higher than the national average of $32,000 that we calculated in Luca et al. (2020) Nearly half of these costs are incurred in Year 0 and are attributable to obstetric complications, pre-term birth, suboptimal breastfeeding, and SIDS.

Overall, 55% of costs are attributable to maternal outcomes. The highest-cost maternal outcomes over the six model years were productivity losses ($610.2 million), non-obstetric health expenditures ($445.1 million), and obstetric-specific health expenditures for excess cases of preeclampsia ($36.8 million), cesarean section ($55.2 million), and long peripartum inpatient stays ($31.3 million). The remaining 45% of costs are attributable to child outcomes. The highest-cost child outcomes over the six model years were child behavioral and developmental disorders ($556 million), preterm births ($372 million), and asthma ($33 million). Although missed well-child care visits could partially offset these costs in the short term (-$11.2 million), they are likely to lead to higher costs in the longer term due to worse child health.

The total estimated cost of exposure to untreated MMHCs among Texas Medicaid for Pregnant Women enrollees giving birth in 2019 was $962 million, accounting for costs incurred in the six years between conception and five years postpartum (Table 2). Assuming that 30,833 out of 179,264 enrolled mothers (17.2%) who gave birth in 2019 had an untreated MMHC, these conditions cost more than $31,000 per affected mother–child pair over the six model years [1, 12]. Medicaid is separate from other programs, including SNAP, WIC, and Temporary Assistance for Needy Families (TANF), and does not pay for productivity losses, suicide, or SIDS, so we set these costs to $0 in this subgroup analysis.

In this population, 35% of costs are attributable to maternal outcomes, with non-obstetric health expenditures ($276 million) yielding the highest cost. The remaining 65% of costs are attributable to child outcomes, with child behavioral and developmental disorders ($345 million) and preterm birth ($231 million representing the largest costs. Historically, Medicaid income eligibility limits in Texas have been low [13, 14], and coverage through Texas Medicaid for Pregnant Women has lapsed after 60 days postpartum, though Texas lawmakers are currently working to extend coverage through a full year postpartum [15]. As a result, a majority of these costs are borne by the health care system (for uninsured or underinsured women and their children) or by insurers. Supplementary Table 6 provides details on the point estimates and data sources we used for this subgroup analysis.

We estimated costs for non-Hispanic White, non-Hispanic Black, and Hispanic mothers to examine potential differences in how MMHCs impact Texas mothers of different racial and ethnic backgrounds and their children. Although we aimed to assess potential differences among women of other backgrounds, we found that stratified data are most consistently available for these racial and ethnic groups. We assumed that MMHCs have the same impact on outcomes across racial and ethnic groups and used measures of association from the main model in our subgroup models. Differences in costs across racial and ethnic groups appeared driven primarily by differences in the prevalence of MMHCs across these groups.

We accounted for differences in population size, as well as the prevalence and societal and health outcomes of MMHCs and found that untreated MMHCs are most prevalent among non-Hispanic Black mothers (18.2%), followed by Hispanic mothers (12%) and non-Hispanic White mothers (11.4%) [1]. These health disparities lead to higher societal costs per non-Hispanic Black mother–child pair (nearly $62,000) than for Hispanic and non-Hispanic White mother–child pairs (about $43,000). Table 3 presents a breakdown of costs by racial and ethnic group, Supplementary Tables 7, 8 and 9 present more detailed results, and Supplementary Table 10 presents model inputs and data sources.

The total estimated cost of untreated MMHCs among non-Hispanic White mother–child pairs was $599 million for the 2019 birth cohort, accounting for costs incurred over six model years. Assuming that 13,896 of the 121,899 Non-Hispanic White mothers (11.4%) who gave birth in 2019 had an untreated MMHC [1], these conditions cost $43,106 per affected non-Hispanic White mother–child pair over six years. In this population, 56% of costs are attributable to maternal outcomes, with productivity losses ($170 million) and non-obstetric health expenditures ($123 million) yielding the highest costs. The remaining 44% of costs relate to child outcomes, with child behavioral and developmental disorders ($155 million) and preterm birth ($89 million) yielding the highest costs.

The total estimated cost of exposure to untreated MMHCs among non-Hispanic Black mother–child pairs was $521 million. Assuming 8,448 of the 46,420 non-Hispanic Black mothers (18.2%) who gave birth in 2019 had an untreated MMHC, these conditions cost $61,671 per affected non-Hispanic Black mother–child pair over six years. In this population, 65% of these costs relate to maternal outcomes, with productivity losses ($189 million) and non-obstetric health expenditures ($118 million) yielding the highest costs. The remaining 35% of costs relate to child outcomes, with child behavioral and developmental disorders ($94 million) and preterm birth ($74 million) yielding the highest costs.

Finally, the total estimated cost of exposure to untreated MMHCs among Hispanic mother–child pairs was $928 million. Assuming 21,421 of the 178,509 Hispanic mothers (12%) who gave birth in 2019 had an untreated MMHC, these conditions cost $43,322 per affected Hispanic mother–child pair over six years. In this population, 55% of these costs relate to maternal outcomes, with productivity losses ($229 million) and non-obstetric health expenditures ($211 million) yielding the highest costs. The remaining 45% of costs relate to child outcomes, with child behavioral and developmental disorders ($239 million) and preterm birth ($149 million) yielding the highest costs.

The magnitude of the disparities in the prevalence and cost of untreated MMHCs suggests ample room for policy interventions to ensure more equitable access to screening and effective treatment options. In the long run, these interventions could reduce suffering caused by untreated MMHCs as well as reduce costs to employers, the health care system, insurers, and the government.

The sensitivity analysis showed that our model estimated a range of $739 million to $4.3 billion for the 2019 Texas birth cohort through varying all prevalence and impact estimates at once (Fig. 2). The measure of association between MMHCs and productivity losses had the greatest impact on estimated costs ($1.7 billion to $2.7 billion), followed by the MMHC remission rate ($1.9 billion to $2.8 billion) and preterm birth ($1.8 billion to $2.3 billion). Parameters with large ranges in the literature and the costs of these outcomes meant that varying these parameters led to substantial variation in model results.

Other parameters had a more modest impact on model results because of a low baseline rate of the outcome, low costs incurred through five years post-delivery, or narrow range of measures of association.

This study had several limitations. First, our literature review focused on identifying statistically significant estimates for use in our model. Since we did not use estimates that did not reach statistical significance, our model may exclude relevant and important relationships between MMHCs and maternal or child outcomes for which statistically significant associations have not appeared in the literature; publication bias may exacerbate this issue (e.g., nonsignificant results may have a lower likelihood of appearing in the literature than significant findings). Although we focused on immediate associations in the six-year period between conception and five years post-delivery, we recognize that MMHCs can have long-term impacts on mothers and their children. For example, children who develop behavioral and developmental disorders due to exposure to MMHCs could have lower educational attainment. Our model does not account for the increased prevalence of MMHCs arising from the COVID-19 pandemic. We did not account for the impact of MMHCs on nonmaternal caregivers, though we recognize they may also have mental health conditions. In addition, most of the literature on the impacts of untreated MMHCs focused on maternal depression rather than maternal anxiety. If maternal anxiety impacts outcomes independently of maternal depression and a mother has both anxiety and depression, the model may underestimate costs of untreated MMHCs [16]. As a result, our model results should be interpreted as conservative estimates of the true cost. Similarly, our model assumes independent associations between MMHCs and various maternal and child outcomes. Although we conducted a sensitivity analysis to understand the possible range of cost estimates we could have generated based on our inputs, there may still be some imprecision in our estimates. For example, MMHCs remain under-detected, which could contribute to an underestimate in the overall model.

Additionally, we did not account for treatment options and did not distinguish between untreated and inadequately treated MMHCs, so did not make the economic case for treatment, or estimate potential cost savings through treatment of MMHCs. Finally, we did not analyze primary data, but rather used inputs from the literature and secondary data sources.