Mental Health Within 24 Months After Delivery Among Women with Common Pregnancy Conditions

Abstract

Objective:

The aim of this study is to estimate the risk of a new mental health diagnosis within the first 24 months postpartum among women with common pregnancy conditions, overall and by rurality.

Materials and Methods:

This longitudinal population-based study used the Maine Health Data Organization's All-Payer Claims Data to estimate the cumulative risk of a new mental health disorder diagnosis in the first 24 months postpartum among women with deliveries during 2007–2019 and who did not have a mental health diagnosis before pregnancy. Cox models were used to estimate hazard ratios for common pregnancy conditions (prenatal depression, gestational diabetes [GDM], and hypertensive disorders of pregnancy [HDP]) on the new diagnosis of five mental health conditions, separately. Models were adjusted for maternal demographics and pregnancy characteristics.

Results:

Of the 123,125 deliveries, the cumulative risk of being diagnosed in the first 24 months postpartum with depression was 28%, anxiety 25%, bipolar disorder 3%, post-traumatic stress disorder 6%, and schizophrenia/psychotic disorder 1%. Women with prenatal depression were at higher risk of having a postpartum mental health diagnosis compared with women without prenatal depression (adjusted hazard ratios [aHRs] ranged from 2.5 [for anxiety] to 4.1 [for postpartum depression]). Risk of having postpartum depression was modestly higher among women with HDP, as was the risk of postpartum bipolar disorder among those with GDM. Findings were generally similar between women living in rural versus urban areas.

Conclusions:

Effective interventions to prevent, screen, and treat mental health conditions among women with pregnancy complications for an extended time postpartum are warranted.

Introduction

Postpartum mental health disorders are common, serious, and can adversely affect the health and well-being of women, infants, and children.^1,2 Not only is the prevalence of perinatal mental health conditions increasing^2,3 but the COVID-19 pandemic is also accelerating this trend.⁴

While several studies have estimated the risk at delivery hospitalization of mental health disorders among those with rare but serious pregnancy complications such as severe maternal morbidity^3,5 and emergency cesarean section,⁶ studies are lacking on the association between common pregnancy conditions and mental health within the first 2 years postpartum.

This is unfortunate because collectively common pregnancy conditions affect more than one in five women and are becoming more prevalent,^7,8 and the first few years postpartum are a critical window of opportunity for treatment and prevention of mental health conditions. This window can also represent the interpregnancy interval when optimal health can be achieved before another pregnancy begins.⁹

Certain subgroups may be at higher risk of developing mental health disorders postpartum, such as women living in rural areas in the United States. Rural women have a higher risk of perinatal depression, which may be due to factors such as educational and economic deprivation, which are more common among rural women.¹⁰

Rural women also face distinct barriers to care for treatment of mental health conditions, such as stigma and lack of local mental health care providers.^11–13 In addition, rural women are at increased risk for many adverse health outcomes compared with women living in urban areas, including pregnancy complications¹⁴ and maternal and infant mortality.¹⁵

Therefore, the aim of our study was to estimate the risk of being diagnosed with a new mental health disorder in the first 24 months postpartum among women with common pregnancy conditions (prenatal depression, hypertensive disorders of pregnancy [HDP], and gestational diabetes [GDM]) and who did not have a mental health diagnosis before pregnancy.

We hypothesized that these three pregnancy conditions could increase the risk for mental health diagnoses postpartum due to the added stress from demands of the pregnancy complication itself (e.g., additional visits, tests and procedures, medications, and behavior change efforts) or its sequelae (e.g., preterm birth, fetal distress, neonatal intensive care admission).^16,17

We also sought to estimate whether the associations differed between women living in rural and urban areas. Our study population was a contemporary cohort of women in Maine, which is the state in the United States with the highest percentage of residents living in rural areas¹⁸ and the state with the third highest percentage of adults reporting a serious mental illness in the past year.¹⁹

Materials and Methods

Data source

We used data from the Maine Health Data Organization's All-Payer Claims Data (APCD), 2006–2021.²⁰ The APCD include service dates, diagnoses, procedures, payments, and provider and geographical information for medical, dental, and pharmacy claims paid for Maine residents by insurance companies licensed in Maine, Maine Medicaid, and Medicare. APCD must meet internal quality standards before being released for analysis.

Maine Medicaid has used a fee-for-service model to pay providers since the late 1990s; this means that claims are collected and submitted in a standardized way for each service provided, an advantage over other states that use managed care arrangements.²¹ Claims containing diagnosis codes related to substance use disorder were redacted from the Maine APCD beginning with data requests in 2017²²; for our analysis, this redaction affected claims with service delivery dates after January 2014.

Identification of unique deliveries

Deliveries during the period 2007–2019 were identified using International Classification of Diseases, Clinical Modification (ICD-CM), diagnosis and procedure codes, current procedural terminology (CPT) codes, and the Medicare Severity Diagnosis-Related Group (MS-DRG) classification system. We used a previously published crosswalked ICD-9/10 delivery code list from the Alliance for Innovation on Maternal Health (version 6-27-20).²³

This code list includes codes related to live birth and stillbirth, with exclusion codes for induced and spontaneous abortions, ectopic pregnancies, and molar pregnancies. We added delivery-related CPT procedure codes to further identify deliveries²⁴ and used an algorithm that we developed to identify the likely delivery date for each pregnancy (available upon request).

We restricted our analysis to women continuously enrolled in health insurance for a minimum period from approximately the fourth month of pregnancy through at least 2 months postpartum, with a 1-month gap allowed. This enrollment criterion was consistent with previous studies^25–31 and was necessary to ascertain conditions diagnosed during pregnancy, in addition to at least some postpartum information on morbidity.

Insurance enrollment information in the APCD tracks an individual's enrollment status even if they change insurance providers. We use the terms ‘women’ and ‘maternal’ throughout the article to refer to persons included in our analysis. However, we acknowledge that not all people who become pregnant identify as women or meet the definition of maternal (i.e., relating as a mother).

To estimate start and end dates of each pregnancy, we used a validated algorithm for calculating the gestational length in ICD-9 data^32,33 and then subtracted gestational length from the delivery date. For ICD-10 data, we used Z3A codes to calculate gestational length; Z3A codes accompanied 97% of the ICD-10 era deliveries we identified.

Common pregnancy conditions

We focused on three common pregnancy conditions: prenatal depression, HDP, and GDM. We selected these three conditions because they are common among U.S. women^34–36 and there are well-established national screening recommendations and diagnostic criteria for all three conditions (U.S. Preventive Services Task Force recommendations).³⁷

Prenatal depression was identified using an ICD-9/10 code list from the Mental Health Research Network,³⁸ modified to exclude codes for depressive disorders in full remission and codes for unspecified mental disorders (648.4 and O99.34), which could erroneously capture not only depression but also the mental health disorders we examined postpartum.^39,40

Prenatal depression was defined as at least one claim with a qualifying diagnosis code for service delivery dates between 6 weeks' gestation and delivery date. HDP were identified using a method implemented by the Vaccine Safety Datalink using ICD-9 codes,⁴¹ which were then crosswalked to the corresponding ICD-10 diagnosis codes. This method defined HDP as at least one inpatient claim or at least two claims (for separate service delivery dates) with a qualifying diagnosis between 20 weeks' gestation and 6 weeks postpartum. GDM was identified using ICD-9/10 diagnosis code lists based on previous studies.^42–45 GDM was defined as two or more claims (for separate service delivery dates) with a qualifying diagnosis code between 20 weeks' gestation and the delivery date. A schematic showing time frame definitions for the common pregnancy conditions examined in our analysis can be found in Supplementary Appendix Figure S1, and diagnosis codes for each pregnancy condition are listed in Supplementary Appendix Table S1.

Postpartum mental health disorders

We examined five mental health disorders diagnosed in the first 24 months postpartum: depression, anxiety, post-traumatic stress disorder (PTSD), bipolar disorder, and schizophrenia/other psychotic disorders. These five disorders represent the most common specific mental health diagnoses among adult inpatient stays related to mental health in the United States.^46,47

For postpartum depression, we used the same code list as used for prenatal depression (Supplementary Appendix Table S1); for the other conditions, we used previously published code lists to identify each condition (Supplementary Appendix Table S2).^46–49 For each condition, we identified the date of first diagnosis from delivery date through the first 24 months postpartum.

Maternal characteristics

Rural–urban designation of maternal residence at the time of delivery was classified using ZIP code of residence. These subcounty classifications are based on the U.S. Department of Agriculture Rural–Urban Commuting Area (RUCA) Codes, which take into account population density, urbanization, and daily commuting patterns,⁵⁰ and are more accurate measures of rural residency than county-level designations.⁵¹

Women were categorized into two mutually exclusive groups: metropolitan (“urban”) versus large rural, small rural, and isolated rural areas (“rural”) using the New England RUCA code classification scheme.⁵²

Socioeconomic factors (including race/ethnicity) and access to care measures were not available in the APCD and thus were estimated by linking the APCD records to publicly available community-level information. ZIP code-level data on the median percentage of residents living below the federal poverty level, of non-White race/ethnicity (as Maine's population is 93% non-Hispanic White), and who were adults with less than college educational attainment were identified from the American Community Survey's 5-year ZIP code files for all communities in Maine.⁵³

To assess access to care, information on the number of general practice and medical specialty physicians per capita from the Area Health Resource Files was linked by county Federal Information Processing Standard codes to APCD records.⁵⁴

Insurance status was assessed using the APCD eligibility file. Women were classified as Medicaid insured if they were enrolled in Medicaid during their delivery month⁵⁵; otherwise, they were classified as insured by commercial insurance or Medicare based on delivery month enrollment information using a hierarchy that prioritized commercial insurance.

Prepregnancy health conditions

Prepregnancy health conditions were identified by scanning claims for service dates during the year before the start of pregnancy through the first trimester (Supplementary Appendix Fig. S1). Pre-existing depression was defined as any depression diagnosis (Supplementary Appendix Table S1) on claims during the year before the start of pregnancy through 6 weeks' gestation.

Pre-existing hypertension was defined as any hypertension diagnosis on claims before the start of pregnancy through 20 weeks' gestation, using a code list published by the Centers for Disease Control and Prevention.⁵⁶ Pre-existing diabetes was defined as any diabetes diagnosis on claims before the start of pregnancy through 20 weeks' gestation using an ICD-9/10 code list from a claims-based algorithm validation study.⁴⁴

Pre-existing mental health conditions were identified for each of the five mental health disorder diagnoses (code lists shown in Supplementary Appendix Table S2) using claims during the year before the start of pregnancy.

Statistical analysis

We tabulated maternal and pregnancy characteristics for women by pregnancy condition. We also tabulated these characteristics by insurance coverage window to assess our study's potential for introducing information bias (covariate misclassification) by not restricting the analysis to women with longer durations of continuous insurance coverage around the time of pregnancy, which presumably would have led to more complete covariate information from claims data.

To estimate the cumulative risk for each of the five mental health conditions diagnosed in the first 24 months postpartum, we used Cox proportional hazards models stratified by each pregnancy condition (prenatal depression, HDP, and GDM). Models excluded deliveries to women with any prepregnancy diagnosis of the mental health condition under examination, as well as deliveries with multifetal gestations, implausible gestational length, or time to next pregnancy, or those missing information needed for multivariable modeling (Supplementary Appendix Fig. S1). We exclude multifetal gestations because of the unique mental health challenges experienced by parents of multiples.⁵⁷

Observations were censored upon loss of health insurance, start of next pregnancy (leading to a live birth or stillbirth), or at 24 months, whichever came first. Our models estimated parameters under an independent working assumption and used a robust sandwich covariance matrix estimate to account for the intracluster dependence due to more than one delivery to the same woman.

To estimate aHRs, we used Cox proportional hazards models, with pregnancy condition as the exposure variable. Adjustment sets were determined a priori and included factors that could have caused, separately, both the pregnancy condition and the postpartum mental health condition: maternal age; insurance type; prepregnancy medical conditions, obesity, and smoking; and geographic-level socioeconomic and health care access factors.

We did not include as factors events that could have been consequences of the pregnancy condition (e.g., cesarean section or stillbirth as a result of HDP); however, as a further adjusted analysis, we did include adjustment for common comorbid pregnancy conditions. We also did not include as factors mental health conditions diagnosed during pregnancy, as these could have been along the causal pathway from the pregnancy complication to the postpartum mental health diagnosis.

To see if results varied by rurality of maternal residence, we repeated the modeling after stratification by rural versus urban residency. To assess the proportional hazards assumption and visualize the survival curves, we used inverse probability of treatment-weighted Cox proportional hazards models.⁵⁸

The models examining the effect of prenatal depression on postpartum depression excluded women with depression diagnoses before pregnancy by design, resulting in a de facto examination of new-onset prenatal depression only. As a sensitivity analysis, to examine prenatal depression overall, whether or not it was new-onset prenatal depression, we added back in women with a prepregnancy depression diagnosis and then included this factor in the covariate adjustment set.

Although the model excluded deliveries to women with any prepregnancy diagnosis of the mental health condition under examination, it did include women who had a diagnosis during pregnancy of the mental health condition under examination. As a sensitivity analysis, we reran the models, further adjusting for mental health diagnosis during pregnancy and, separately, excluding women with any diagnosis during pregnancy of the mental health condition under examination.

The Institutional Review Board at University of Southern Maine determined this study to be exempt from human subjects review because it used secondary data through a data use agreement.

Results

A total of 123,125 unique deliveries were included in our analysis, representing 89,672 unique women (Supplementary Appendix Fig. S2). The prevalence of prenatal depression was 17%, HDP 12%, and GDM 9%; 68% of pregnancies had none of these three conditions (Table 1). The prevalence of each of these conditions increased from 2007 to 2019: from 13% to 30% for prenatal depression, 11% to 17% for HDP, and 7% to 11% for GDM (Fig. 1).

Table 1.

Maternal and Pregnancy Characteristics, by Pregnancy Condition, Among Deliveries in Maine 2007–2019, n = 123,125

	TOTAL		Prenatal depression (6 weeks' gestation to delivery)		Hypertensive disorders of pregnancy (20 weeks' gestation to 6 weeks postpartum)		Gestational diabetes (20 weeks' gestation to delivery)		None
	N	%	N	%	N	%	N	%	N	%
Total	123125	100.0	21220	100.0	15313	100.0	10459	100.0	83172	100.0
%, row	100.0			17.2		12.4		8.5		67.6
Maternal age at delivery, years
Missing	52	0.0	11	0.1	a		a		34	0.0
15–19	7631	6.2	1698	8.0	885	5.8	233	2.2	5142	6.2
20–24	28920	23.5	5520	26.0	3377	22.1	1487	14.2	19825	23.8
25–29	36761	29.9	5998	28.3	4540	29.7	2929	28.0	25230	30.3
30–34	31244	25.4	4818	22.7	3963	25.9	3290	31.5	21171	25.5
35+	18517	15.0	3175	15.0	2539	16.6	2519	24.1	11770	14.2
Multifetal gestation
Yes	2163	1.8	403	1.9	526	3.4	255	2.4	1188	1.4
Stillbirth
Yes	744	0.6	174	0.8	78	0.5	42	0.4	492	0.6
C-section
Yes	36417	29.6	6904	32.5	6323	41.3	4476	42.8	22002	26.5
Gestational age, weeks
At least 37	112986	91.8	18983	89.5	13263	86.6	9380	89.7	77331	93.0
20–<37	10086	8.2	2228	10.5	2050	13.4	1079	10.3	5797	7.0
<20	53	0.0	a	0.0	0	0.0	0	0.0	44	0.1
Delivery number in dataset
1	82458	67.0	12950	61.0	11571	75.6	6830	65.3	55835	67.1
2 or more	40667	33.0	8270	39.0	3742	24.4	3629	34.7	27337	32.9
Insurance coverage
Medicaid	68169	55.4	15165	71.5	7759	50.7	5611	53.7	44014	52.9
Private	54649	44.4	5944	28.0	7506	49.0	4808	46.0	39004	46.9
Medicare	307	0.2	111	0.5	48	0.3	40	0.4	154	0.2
Loss of health insurance postpartum, months
1–5	24177	19.6	3525	16.6	3009	19.7	1920	18.4	16901	20.3
6–11	10805	8.8	1594	7.5	1253	8.2	826	7.9	7644	9.2
12–23	17165	13.9	2613	12.3	2096	13.7	1274	12.2	11985	14.4
24 or later	70978	57.6	13488	63.6	8955	58.5	6439	61.6	46642	56.1
Rurality, ZIP code-level categorization
Missing	186	0.2	22	0.1	15	0.1	14	0.1	142	0.2
Out of state/international/military	527	0.4	120	0.6	56	0.4	38	0.4	341	0.4
Metro: 1, 1.1	43334	35.2	8359	39.4	5572	36.4	3686	35.2	28391	34.1
Large rural: 2, 2.1, 3, 4, 4.1, 5, 5.1, 6	43966	35.7	7030	33.1	5570	36.4	3730	35.7	30072	36.2
Small rural: 7, 7.1, 7.2, 8, 8.1, 8.2, 9, 10.1, 10.2, 10.3	17779	14.4	3126	14.7	2130	13.9	1561	14.9	11998	14.4
Isolated rural: 10	17333	14.1	2563	12.1	1970	12.9	1430	13.7	12228	14.7
Any pre-existing hypertension
Yes	9045	7.3	2161	10.2	5580	36.4	1522	14.6	2197	2.6
Any pre-existing diabetes
Yes	3493	2.8	968	4.6	924	6.0	854	8.2	746	0.9
Any pre-existing depression
Yes	20275	16.5	10127	47.7	2728	17.8	2008	19.2	8198	9.9
Any anxiety disorder before pregnancy
Yes	15913	12.9	9751	46.0	2513	16.4	1486	14.2	4948	6.0
Any PTSD before pregnancy
Yes	5250	4.3	2759	13.0	585	3.8	527	5.0	2068	2.5
Any bipolar disorder before pregnancy
Yes	3753	3.0	1701	8.0	438	2.9	390	3.7	1659	2.0
Any schizophrenia or other psychotic disorder before pregnancy
Yes	513	0.4	282	1.3	52	0.3	49	0.5	196	0.2
ZIP code-level data (2011, 2019), mean, %
Below poverty level	122702	12.9	21149	13.2	15272	12.7	10429	13.2	82870	12.8
Non-White race	122728	5.6	21150	6.0	15273	5.6	10432	5.6	82891	5.6
Adults with less than college education	122683	71.5	21144	71.2	15270	71.5	10430	72.0	82855	71.4
County-level data (2010–2021), mean
Ratio of population to primary care physicians	122806	1077.3	21167	1063.3	15284	1070.3	10440	1078.6	82939	1080.6
Ratio of population to mental health providers	122805	1852.8	21167	1558.0	15284	1736.7	10440	1819.7	82938	1925.5
Uninsured adults, %	122806	13.0	21167	13.0	15284	12.9	10440	13.1	82939	13.0
Adult smoking, %	122803	19.1	21167	18.6	15284	18.9	10440	19.3	82936	19.2
Adult obesity, %	122806	26.9	21167	27.2	15284	27.0	10440	27.3	82939	26.9

Excluding deliveries among women without health insurance during pregnancy and the first 2 months postpartum.

^a Suppressed cell count per data use agreement with Maine Health Data Organization.

PTSD, post-traumatic stress disorder.

FIG. 1.

Prevalence of prenatal depression, hypertensive disorders of pregnancy, and gestational diabetes among deliveries in Maine 2007–2019, n = 123,125. ICD, International Classification of Diseases, Clinical Modification.

Several characteristics were more commonly observed among pregnancies with prenatal depression, HDP, and GDM, compared with pregnancies not complicated by these conditions, such as cesarean section, preterm birth, pre-existing hypertension, pre-existing diabetes, and pre-existing mental health conditions (Table 1). Characteristics of pregnancies differed minimally by insurance coverage window (Supplementary Appendix Table S3), suggesting that our study was at low risk of bias due to the continuous enrollment window that we selected for study inclusion.

The cumulative risk of new diagnoses for mental health conditions in the first 24 months postpartum was 28% for depression, 25% for anxiety, 3% for bipolar disorder, 6% for PTSD, and 0.6% for schizophrenia or other psychotic disorders and was generally higher for pregnancies complicated by prenatal depression, HDP, and GDM compared with pregnancies without these conditions (Table 2). The cumulative risk of a new diagnosis in the first 24 months postpartum increased over the study years for anxiety and PTSD (Supplementary Appendix Fig. S3).

Table 2.

Cumulative Risk of New Diagnosis in the First 24 Months Postpartum for Depression, Anxiety, Post-traumatic Stress Disorder, Bipolar disorder, and Schizophrenia, by Exposure Status, Among Deliveries in Maine 2007–2019, N = 123,125

	TOTAL (N = 123,125)			Prenatal depression (6 weeks' gestation to delivery) (n = 21,220)			Hypertensive disorders of pregnancy (20 weeks' gestation to 6 weeks postpartum) (n = 15,313)			Gestational diabetes (20 weeks' gestation to delivery) (n = 10,459)
Postpartum mental health diagnosis	Observations in Model	Number of events	%a	Number exposed	%a among exposed	%a among unexposed	Number exposed	%a among exposed	%a among unexposed	Number exposed	%a among exposed	%a among unexposed
Depression
Yes	99,825	18,359	27.7	10,755	84.8	22.5	12,039	30.5	27.4	8,164	29.4	27.6
Anxiety
Yes	103,936	16,589	25.1	14,073	62.4	20.1	12,541	27.8	24.7	8,654	27.2	24.9
Bipolar disorder
Yes	115,804	2,793	3.3	18,915	9.0	2.2	14,218	3.1	3.2	9,716	3.8	3.3
PTSD
Yes	114,358	4,938	6.1	17,886	17.0	4.1	14,080	5.7	6.2	9,583	6.6	6.0
Schizophrenia or other psychotic disorder
Yes	118,949	542	0.6	20,276	1.7	0.4	14,589	0.7	0.6	10,048	0.8	0.6

Each model excluded records with any diagnosis before pregnancy of the mental health disorder being examined postpartum; records with implausible gestational age (<20 weeks), missing rural versus urban residence, and multifetal gestation (e.g., twin pregnancy); those without evidence of insurance in the month of delivery; those with implausible time to next pregnancy (<60 days); and those without health insurance during pregnancy and the first 2 months postpartum.

^a Cumulative risk by 24 months. Censoring events were loss of health insurance coverage or start of the next pregnancy, whichever was earlier.

aHRs for new diagnoses of mental health conditions within the first 24 months postpartum among women with prenatal depression were 4.0 (95% confidence interval [CI]: 3.9–4.2) for postpartum depression (nearly the same as that for new-onset prenatal depression, 4.1 [95% CI: 3.9–4.2]), 2.5 (95% CI: 2.4–2.6) for anxiety, 2.7 (95% CI: 2.4–2.9) for bipolar disorder, 2.7 (95% CI: 2.6–2.9) for PTSD, and 2.8 (95% CI: 2.3–3.5) for schizophrenia and other psychotic disorders (Figs. 2 and 3).

FIG. 2.

Adjusted hazard ratio of new diagnosis of mental health condition among women with prenatal depression, hypertensive disorders of pregnancy, and gestational diabetes, separately, deliveries in Maine 2007–2019, n = 123,125. Above, in black triangles: models adjusted for maternal age at time of delivery, prepregnancy depression, prepregnancy hypertension, prepregnancy diabetes, obesity, smoking, nulliparity, year of delivery, Medicaid coverage during pregnancy, common comorbid pregnancy conditions (depression, hypertensive disorders of pregnancy, and gestational diabetes), county-level measures (ratio of population to primary care providers and ratio of population to mental health providers), and ZIP code-level measures (% population non-White, % adults with no bachelor's degree, and % population in poverty). Models did not adjust for race/ethnicity because the variable was not available in the dataset. Each model excluded records with any diagnosis before pregnancy of the mental health disorder being examined postpartum; records with missing preterm birth, missing rural residence, and multiples (e.g., twin pregnancy); those without evidence of insurance at the time of delivery; those with implausible time to next pregnancy (<60 days); and those without health insurance during pregnancy and the first 2 months postpartum. For prenatal depression, “depression (new-onset)” models excluded women with evidence of depression before pregnancy and did not adjust for prepregnancy depression, and “depression” models included women with evidence of depression before pregnancy and adjusted for prepregnancy depression. Censoring events were loss of health insurance coverage, start of the next pregnancy, or 24 months of follow-up, whichever was earlier. PTSD, post-traumatic stress disorder.

FIG. 3.

Weighted–adjusted^a cumulative risk curves for new diagnosis of mental health condition among women with and without prenatal depression, deliveries in Maine 2007–2019, n = 123,125. (a) Postpartum depression. (b) Postpartum anxiety. (c) Postpartum bipolar disorder. (d) Postpartum PTSD. (e) Postpartum schizophrenia/psychotic disorders. ^aModels adjusted for maternal age at time of delivery, prepregnancy depression, prepregnancy hypertension, prepregnancy diabetes, obesity, smoking, nulliparity, year of delivery, Medicaid coverage during pregnancy, county-level measures (ratio of population to primary care providers and ratio of population to mental health providers), ZIP code-level measures (% population non-White, % adults with no bachelor's degree, and % population in poverty), and comorbid pregnancy conditions (hypertensive disorders of pregnancy and gestational diabetes). Models did not adjust for race/ethnicity because the variable was not available in the dataset.

Confidence intervals excluded the null for the aHR for HDP and postpartum depression (aHR = 1.1, 95% CI: 1.1–1.2) and for GDM and bipolar disorder (aHR = 1.3, 95% CI: 1.1–1.4). Results were similar without adjustment for common comorbid pregnancy conditions. After repeating the modeling stratified by rural versus urban residency, findings were generally similar between residency types; however, the aHRs for prenatal depression and postpartum depression were stronger for pregnancies among women living in urban areas and weaker for prenatal depression and schizophrenia and other psychotic disorders (Supplementary Appendix Table S4).

After repeating the modeling, further adjusting for any diagnosis during pregnancy of the mental health disorder under examination, estimates for the aHRs for prenatal depression and anxiety, bipolar disorder, PTSD, and schizophrenia and other psychotic disorders were attenuated by 14%–30%, but still statistically significant and above 1 (e.g., greatest relative change was for postpartum anxiety [from 2.5 to 1.8]). Estimates for aHRs for HDP and GDM were not substantially different. Findings were similar when we excluded women with any diagnosis during pregnancy of the mental health disorder under examination.

Discussion

Principal findings

Using the Maine Health Data Organization's APCD on over 100,000 deliveries during the past 15 years, we found that there were substantial risks of new diagnoses of depression and anxiety in the first 24 months postpartum among women without these conditions before pregnancy (28% and 25%, respectively). We also found that postpartum diagnoses for bipolar disorder and PTSD were not uncommon (3% and 6%, respectively) and that diagnoses for PTSD increased over time (from 6% to 11%).

The prevalence of each of the three pregnancy conditions we examined (prenatal depression, HDP, and GDM) increased from 2007 to 2019. Women with prenatal depression were 2.5–4.1 times more likely to have mental health conditions diagnosed postpartum compared with women without prenatal depression. Women with HDP and GDM were at elevated risk for depression and bipolar disorder diagnoses, respectively, but the associations were modest. We did not find consistent differences by maternal rurality of residence, which was one of the aims of our analysis.

Results in the context of what is known

The increasing trends we found for prenatal depression, HDP, and GDM are consistent with those reported in national studies. Increases may be due to better screening for these conditions, particularly after national screening recommendations were published in the last decade, and the changing health profiles of childbearing women in the United States.^7,37 Increasing trends in mental health conditions diagnosed postpartum, such as anxiety and PTSD, may also be due to better detection or because these conditions are more common among more recent birth cohorts due to societal circumstances.²

Prenatal depression is widely acknowledged as a risk factor for postpartum depression^35,39 and has been associated with increased risks for postpartum diagnosis of PTSD and anxiety.^2,59 Our findings showing that women with prenatal depression are also at elevated risk for bipolar disorder and schizophrenia/psychotic disorder are new, but plausible given the co-occurrence of depression and other mental health conditions.

Severe forms of HDP, such as pre-eclampsia, have been associated with postpartum depression, anxiety, and PTSD, while less severe forms of HDP, such as gestational hypertension, have shown mixed associations.^16,60–62 This may be due to the birth-related trauma experienced by women with pre-eclampsia, who often undergo preterm emergency c-section and other medical interventions that may lead to lingering psychological effects.⁶³

Although not well studied, GDM has only been found to be consistently associated with an increased risk of postpartum depression.⁶⁴ Yet, type 2 diabetes, a downstream consequence of GDM, is associated with bipolar disorder and schizophrenia in the nonobstetric population.^65,66 Our finding that GDM was associated with a nearly 30% higher risk of bipolar disorder, after adjustment for demographics and pregnancy characteristics, is novel and requires further study. This association might possibly be due to reverse causation as there is some evidence that women with bipolar disorder are at increased risk for GDM, although we excluded women with prior bipolar disorder to reduce this possibility.⁶⁷

Our study's findings showing generally similar associations between common pregnancy conditions and mental health diagnoses in the first 24 months postpartum for rural versus urban women are somewhat unexpected. Rural women in the United States are at increased risk for adverse health outcomes and face distinct barriers to mental health treatment compared with urban women.^11,12,15

Yet, we found that rural women appeared to show a weaker association between prenatal depression and postpartum depression. This could be due to higher rates of postpartum depression diagnoses among urban women, perhaps because of their increased access to medical care and mental health providers. The stronger association between prenatal depression and schizophrenia and other psychotic disorders among rural versus urban women is harder to explain and may require further investigation.

Clinical and research implications

Our findings contribute to the growing body of literature highlighting the first few years postpartum as a period of particular vulnerability for women in regard to their mental health. This phase coincides with raising young children, often while managing other responsibilities and stresses, including financial stresses.⁶⁸ The substantial risk of being diagnosed with a mental health disorder during these years highlights the need for implementation of and further research on effective interventions to prevent, screen, and treat mental health conditions among pregnant and postpartum women and women raising young children.^5,15,69,70

To implement interventions, adequate health insurance coverage and primary care provider support are recommended.⁷¹ Loss of health insurance is of particular concern for women insured by Medicaid as coverage postpartum is currently only federally mandated through the first 60 days postpartum⁷² (recent postpartum coverage expansion is at the state level and not federally mandated).

In addition, our study suggests that common pregnancy conditions can increase mental health risks for women postpartum and supports women's access to contraception and abortion services.^71,73,74 This access allows women to avoid pregnancy—or terminate pregnancy once it has begun—thereby avoiding possible downstream adverse mental health effects caused by pregnancy complications.

Strengths and limitations

Our study has important limitations that should be considered when interpreting these findings. First, although Maine-based public and private payers are required to submit their claims to the APCD, a Supreme Court decision in 2016 allowed certain self-funded Employee Retirement Income Security Act health plans to not submit their data.⁷⁵ In Maine, this was estimated to result in ∼10% less of the total population included in the APCD, making our study not truly statewide.²²

In addition, claims containing diagnosis codes or procedure codes related to substance use disorders were excluded from our study for claims data with service dates since January 2014. This exclusion would be expected to result in lower incidence of mental health condition diagnoses as substance use disorder and other mental health diagnoses often co-occur.

Second, Maine's population is 93% non-Hispanic White and has the greatest percentage of residents living in rural areas among all U.S. states.¹⁸ These characteristics likely make our findings less generalizable to the United States overall, but still generalizable to rural areas of the United States with largely White populations.

Third, we restricted our analysis to pregnancies with a minimum length of health insurance coverage so that we could better assess pregnancy and postpartum conditions. This may also limit the generalizability of our findings; however, the decision to not impose eligibility inclusion criteria could lead to misclassification of pregnancy exposures and we found few differences in maternal and pregnancy characteristics compared with pregnancies with longer continuous insurance coverage (Supplementary Appendix Table S3). If we had required a longer minimum length of health insurance coverage, generalizability would have been even more limited.

Finally, our analysis relied on administrative billing codes, which can be imperfect measures of health conditions, and used claims from certain time windows; if women had pre-existing, prenatal, or postpartum health conditions that were not documented in our study data, the missing data could lead to misclassification of our exposure and outcomes as well as residual confounding in our adjusted estimates. In particular, our study's identification of a new mental health diagnosis in the first 24 months postpartum among women without the condition before pregnancy would be flawed if this condition existed before pregnancy, but was not documented in the medical claims we examined.

Strengths of our study include our use of population-based data across 15 years and we were able to follow women across insurance providers. We accounted for variation in length of health insurance coverage by using survival analysis, taking advantage of differing coverage postpartum up to 24 months. We also adjusted for a rich set of covariates, including factors diagnosed prepregnancy and select factors assessed during pregnancy.

Conclusions

In summary, the cumulative risk of being diagnosed with a new mental health condition in the first 2 years following delivery is substantial and particularly elevated among women with prenatal depression. Effective interventions aimed at preventing, screening, and treating mental health conditions among women with pregnancy complications throughout the first 2 years postpartum are warranted.

Authors' Contributions

The authors certify that each author has made substantial contributions to the conception and design, acquisition of data, or analysis and interpretation of data; drafting the article and revising it critically for important intellectual content; and final approval of the version to be published. Each author certifies that they have participated sufficiently in the work to believe in its overall validity and to take public responsibility for appropriate portions of its content.

Author Disclosure Statement

No competing financial interests exist.

Disclaimer

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Funding Information

The research reported in this publication was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health under Award No. R15HD101793.

Supplementary Material

Supplementary Appendix Table S1

Supplementary Appendix Table S2

Supplementary Appendix Table S3

Supplementary Appendix Table S4

Supplementary Appendix Figure S1

Supplementary Appendix Figure S2

Supplementary Appendix Figure S3

Supplementary Appendix Figure S3