Risk factors for the recurrence of Severe Maternal Morbidity in first and second births in California, 1997–2020

Abstract

Objective:

To examine the recurrence risk of SMM and SMM subtypes among specific subgroups, such as age, race/ethnicity, educational status, and insurance status.

Study Design:

We used vital records (live birth and fetal death certificates) from 1997 to 2020 that were linked to maternal hospital discharge records to identify 1,989,104 first and second birth pairs. The outcomes, SMM and non-transfusion SMM (nt-SMM), were identified using ICD-9/10 codes for 21 indicators and further categorized into 8 subtypes based on the organ systems impacted (cardiac SMM, renal SMM, respiratory SMM, hemorrhage SMM, sepsis SMM, other obstetric SMM, other medical SMM, and transfusion SMM). We used sequentially adjusted modified Poisson regression model with bootstrapped errors to estimate the recurrences of composite SMM,SMM subtypes, and SMM indicators between first and second birth. Risk ratios were stratified by the following: age, education, insurance, interpregnancy interval, nativity, plurality, race/ethnicity, and the Expanded Obstetric Comorbidity Index.

Results:

The recurrence risk ratio of overall SMM and nt-SMM between first and second births were 3.4 (95% CI: 2.9–4.1) and 3.7 (95% CI: 2.6–5.3), respectively, adjusted for sociodemographic and clinical factors. Among the SMM subtypes, the adjusted risk of recurrence was particularly elevated among individuals who experienced other medical SMM (RR: 119, 95% CI: 30–267) and cardiac SMM (RR: 32.6, 95% CI: 6.6–88.4). In stratified analyses, recurrence risk ratios were highest among individuals with higher education, private insurance, singleton pregnancies, and lower co-morbidity scores (all groups with lower absolute prevalence of SMM). Hispanic populations had a notably lower recurrence risk ratios for SMM compared with White, Black and Asian subgroups. For nt-SMM, Black individuals were the only group that had both higher absolute prevalence and higher recurrence risk ratios for nt-SMM (relative to other racial and ethnic subgroups).

Conclusion:

The recurrence risk of SMM and nt-SMM varies by SMM subtypes as well as among sociodemographic subgroups. This recurrence risk remains elevated after adjustment for sociodemographic and clinical factors. We identify subgroups among whom additional counseling and monitoring may be warranted after an initial SMM occurrence, if additional pregnancies are desired. Some groups that typically have a lower absolute prevalence of SMM had higher recurrence risk ratios relative to their counterparts, which suggests the importance of close monitoring for the potential recurrence of SMM, regardless of baseline prevalence based on known risk factors.

Introduction

Severe maternal morbidity (SMM) comprises several unexpected adverse outcomes during or after pregnancy (e.g., sepsis) with significant consequences to maternal health.^1–4 Rates of SMM in the United States have increased over time, with approximately 2% of individuals who gave birth in 2021 experiencing SMM.^4–6 Equally concerning are the prevalent social disparities in SMM and maternal mortality along axes of historical marginalization such as race, ethnicity, and socioeconomic position.^7–13

The risk of an outcome recurring in subsequent pregnancies is well-documented for maternal outcomes such as preterm birth, preeclampsia, and peripartum cardiomyopathy.^14–19 Understanding recurrence risks, i.e., the risk of experiencing SMM in subsequent pregnancies given prior SMM, provides opportunities for preventing adverse events in future pregnancies. Recent studies report SMM recurrence risk ratios ranging from three-fold to sixteen-fold from index birth to subsequent birth.^20–24 One prior study reported that cardiac complications and uterine rupture (SMM subtypes) were associated with the highest risk of SMM in the subsequent birth, relative to no SMM in first birth.²³ Notably, no prior studies examine the recurrence of SMM subtypes (e.g., risk of cardiac complications in second birth, given cardiac complications in first birth) or how the risk of recurrence differs among specific subgroups.

A more granular understanding of who is most at risk of SMM recurrence would improve patient counseling and aid in developing effective prevention strategies. Using a large, population-based, longitudinal dataset of linked births in California, we examined SMM recurrence relative risk, in line with prior work, for specific SMM subtypes and among specific subgroups known to have a higher prevalence of SMM, such as by age group, race/ethnicity and socioeconomic position.²⁵

Materials and Methods

Data:

Data from the California Department of Public Health was used for this population-based study, captures over 97% of births in California. We used vital records (live birth and fetal death certificates) from 1997 to 2020 that were linked to maternal hospital discharge records (prenatal, delivery, and postpartum), with 98% successful linkage.^26,27

Study Population:

Starting with births in California between 1997 and 2020, we excluded births with unsuccessful linkages (N=536,969). We retained a single birth record for non-singleton births (N=187,455) and excluded births that were not were not a first-second birth pair (N=7,281,045; some of these births entered the study population as second births). Individuals who did not have a linked sequential birth were excluded (N=2,455,035), resulting in 2,112,901 pairs of first and second births. Finally, records with missing data from either the first or second birth for any of the following covariates (each of which had <5 % missing) were excluded: maternal age, education, insurance, nativity, and race/ethnicity. The final study population consisted of 1,989,104 linked birth pairs (Figure 1).

Figure 1::

Cohort Selection

^a Covariates that were missing from either birth excluded

Study Variables:

We defined severe maternal morbidity (SMM) as occurring during pregnancy, at the birth hospitalization, and up to 42 days postpartum. SMM in first birth was the exposure and SMM in second birth was the outcome.^28,29 Severe maternal morbidity is a composite outcome identified using ICD-9/10 codes from maternal records for 21 indicators identified by the Centers for Disease Control and Prevention (Table A.1).^1,30,31 We also considered non-transfusion SMM (nt-SMM), defined as SMM cases with the exclusion of cases where blood transfusion was the only SMM indicator. Since information on the number of units of blood transfused was not available to estimate whether it was a severe event, nt-SMM serves as a conservative estimation of true SMM.^31,32 Additionally, we considered SMM subtypes at first birth as exposures by grouping the 21 indicators by the organ systems impacted (Table 1), and individuals SMM indicators with ≥20 events in birth 2, given occurrence in birth 1.

Table 1.

SMM subtypes used in this study^aA,23

Condition	Indicators of SMM included
Cardiac SMM	Acute myocardial infarction, aneurysm, cardiac arrest/ventricular fibrillation, conversion of cardiac rhythm, heart failure/arrest during surgery or procedure, acute heart failure
Renal SMM	Acute renal failure
Respiratory SMM	Adult respiratory distress syndrome, pulmonary edema, temporary tracheostomy, ventilation
Hemorrhage SMM	Disseminated intravascular coagulation, shock, hysterectomy
Sepsis SMM	Sepsis
Other Obstetric SMM	Amniotic fluid embolism, eclampsia, severe anesthesia complications, air and thrombotic embolism
Other medical SMM	Puerperal cerebrovascular disorders, sickle cell disease with crisis
Transfusion SMM	Blood products transfusion

^aSee Table A.1 for diagnosis and procedure codes used to define the indicators of SMM

Based on prior work, we considered covariates from the first and second births (obtained from vital records unless stated otherwise).^24,25,33 We considered the following variables as social factors: educational attainment (high school or less, some college, undergraduate degree, postgraduate degree) and insurance at birth (private, public, uninsured, other) as proxies for socioeconomic position, nativity (US-born, foreign-born) as a proxy for unique risk profiles given the known association between SMM and nativity^34,35 and self-reported race/ethnicity (American Indian or Alaska Native, Asian American, Hispanic, Non-Hispanic White, Non-Hispanic Black, Pacific Islander, other, mixed race) as socially constructed a proxy for exposure to structural racism^36–40. Additionally, we considered the following obstetric risk factors: maternal age, interpregnancy interval ([IPI] calculated using birth dates: <6 months, 6–11 months, 12–17 months, 18–59 months, ≥60 months), plurality (singleton or multiple birth), and the Expanded Obstetric Comorbidity Index (from vital records and maternal discharge data, developed to account for maternal health conditions related to SMM, e.g., pregestational diabetes).⁴¹ Age and the Expanded Obstetric Comorbidity Index were adjusted for as continuous variables, but analyzed as a categorical variable in stratified analyses for ease of interpretation

Analysis:

We descriptively present the distribution of covariates by exposure status (SMM in first birth) and overall. Next, we calculated the unadjusted prevalence and adjusted risk ratios (RR) with relative risk regression using Poisson estimating equations. Given that these models tend to overestimate standard errors, we bootstrapped errors, running them 200 times with a 10% sample size and an 80% sample rate; we present the CI from 2.5 and 97.5 percentile of RR. We calculated recurrence for both composite SMM between first birth and second, SMM subtypes and individual SMM indicators (e.g., risk of renal SMM in second birth given renal SMM in first birth). We adopted a sequential adjustment approach whereby models were progressively adjusted for variables based on how proximal or distal they were to the occurrence of SMM in second births: model 0 - unadjusted; model 1 - adjusted for confounders (from first birth, where applicable): maternal age, education, insurance, nativity, and race/ethnicity; model 2 - further adjusted for the Expanded Obstetric Comorbidity Index in first birth. We additionally present the unadjusted and fully adjusted models (models 0 and 2) stratified by age, education, insurance, IPI, nativity, plurality, and race/ethnicity (from second birth, where applicable) as well as by Expanded Obstetric Comorbidity Index (from both births); for bootstrapping, a sample size of 50% was used (80% for smallest strata).

Sensitivity Analysis.

In addition to adjustment for the Expanded Obstetric Comorbidity Index, we considered adjustment for the non-transfusion Expanded Obstetric Comorbidity Index, as the indices include the same maternal variables but weight them differently.

Per our data use agreement, any cells with <12 individuals were suppressed for data privacy. The study protocol was approved by the state of California’s Committee for the Protection of Human Subjects (17–04-2932) and the Institutional Review Board of Stanford University (43029). Study design, analysis, and reporting followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.⁴² All analyses were performed in SAS 9.4 (Cary, NC).

Results

Population Characteristics

The study sample consisted of 1,989,104 pairs of first and second births, with a prevalence of 12 SMM cases per 1,000 births for first birth (N=23,578; Table 2) and 11 cases of SMM per 1,000 second births (N=22,518). Individuals who experienced SMM in first birth were, at their first birth, more often younger than 20 years (21.3% vs. 19.2%) or older than 40 years (0.9% vs. 0.6%), had lower educational attainment (less than high school: 17.8% vs. 15.5%), were more likely to be publicly insured (40.4% vs. 35.5%), had a shorter IPI (e.g., <6 months: 5.8% vs. 5.1%), and had higher Expanded Obstetric Comorbidity Index scores (≥11: 74.2% vs. 25.9%), compared to individuals who did not experience SMM at first birth. Compared to those without SMM, individuals who experienced SMM at first birth were more often Black (7.9% vs. 5.4%) and Hispanic (44.3% vs. 39.3%). The distribution of characteristics at second birth is presented in Table A.2.

Table 2.

Characteristics of Study population, by SMM status in first birth: California 1997–2020 (N=1,989,104)

Characteristic in first birtha	SMM at first birth		Total (N =1,989,104)
	No (N=1,965,526, 98.8%)	Yes (N=23,578, 1.2%)
Age (years)
<20	376,852 (19.2)	5,031 (21.3)	381,883 (19.2)
20–29	1,036,119 (52.7)	11,855 (50.3)	1,047,974 (52.7)
30–39	541,640 (27.6)	6,488 (27.5)	548,128 (27.6)
≥40	10,915 (0.6)	204 (0.9)	11,119 (0.6)
Education
Less than high school	304,822 (15.5)	4,190 (17.8)	309,012 (15.5)
High school graduate	518,841 (26.4)	6,365 (27.0)	525,206 (26.4)
Some college	481,201 (24.5)	5,870 (24.9)	487,071 (24.5)
Undergraduate degree	392,072 (19.9)	4,266 (18.1)	396,338 (19.9)
Postgraduate degree	268,590 (13.7)	2,887 (12.2)	271,477 (13.6)
Insurance
Private	1,228,015 (62.5)	13,580 (57.6)	1,241,595 (62.4)
Public	698,646 (35.5)	9,537 (40.4)	708,183 (35.6)
Uninsured/Other	38,865 (2.0)	461 (2.0)	39,326 (2.0)
IPI
<6 months	97,576 (5.1)	1,356 (5.8)	98,932 (5.1)
6–11 months	236,449 (12.3)	2,884 (12.4)	239,333 (12.3)
12–17 months	307,955 (16.0)	3,572 (15.4)	311,527 (16.0)
18–59 months	994,886 (51.8)	11,959 (51.5)	1,006,845 (51.7)
60+ months	285,586 (14.9)	3,433 (14.8)	289,019 (14.9)
Nativity
Foreign-born	514,511 (26.2)	6,132 (26.0)	520,643 (26.2)
US-born	1,451,015 (73.8)	17,446 (74.0)	1,468,461 (73.8)
Plurality
Singletons	1,953,087 (99.4)	22,970 (97.4)	1,976,057 (99.3)
Multiple	12,439 (0.6)	608 (2.6)	13,047 (0.7)
Race/ethnicity
AIAN	7,371 (0.4)	86 (0.4)	7,457 (0.4)
Asian American	282,710 (14.4)	3,333 (14.1)	286,043 (14.4)
Hispanic	784,565 (39.9)	10,470 (44.4)	795,035 (40.0)
Non-Hispanic White	734,578 (37.4)	7,164 (30.4)	741,742 (37.3)
Non-Hispanic Black	106,676 (5.4)	1,873 (7.9)	108,549 (5.5)
Pacific Islander	8,168 (0.4)	113 (0.5)	8,281 (0.4)
Other	1,037 (0.1)	-b	-b
Mixed race	40,421 (2.1)	531 (2.3)	40,952 (2.1)
Expanded Obstetric Comorbidity Indexc
Low	1,500,305 (76.3)	13,818 (58.6)	1,514,123 (76.1)
High	465,221 (23.7)	9,760 (41.4)	474,981 (23.9)

Abbreviations: AIAN American Indian or Alaska Native, IPI interpregnancy interval, SMM severe maternal morbidity, US United States

^aFor age, education, insurance, and Expanded Obstetric Comorbidity Index – race/ethnicity and nativity are assumed to be unchanged between births.

^bCell value suppressed due to data privacy restrictions.

^cPresented here categorically as low (first to third quartile) and high (fourth quartile).

Recurrence risk of SMM, SMM subtypes, and SMM indicators

The unadjusted recurrence RR of overall SMM between first and second births was 5.5 (95% CI: 4.8–6.6; Table 3). Adjustment for sociodemographic characteristics minimally changed the relative risk (RR: 5.3, 95% CI: 4.6–6.3), while adjustment for the Expanded Obstetric Comorbidity Index substantially attenuated recurrence risk (RR: 3.4, 95% CI: 2.9–4.1). Similarly, the fully adjusted recurrence risk ratios for nt-SMM was 3.7 (95% CI: 2.6–5.3). Among the SMM subtypes, adjusted recurrence risk ratios ranged from 4.4 (transfusion SMM, 95% CI 4.0–6.5) to 32.6 (cardiac SMM, 95% CI 6.6–88.4) and 119 (other medical SMM; 95% CI 30–267). For SMM subtypes, adjustment for the Expanded Obstetric Comorbidity Index instead of the nt-SMM Expanded Obstetric Comorbidity Index resulted in similar estimates, with the recurrence risk ratio for renal SMM being the most different (RR adjusted for Expanded Obstetric Comorbidity Index: 5.9, 95% CI: 2.2–10.3 vs. nt-SMM Expanded Obstetric Comorbidity Index: 5.0, 95% CI: 1.9–9.0; Table A.3).

Table 3.

Prevalence (per thousand) and Recurrence Risk Ratios for SMM and SMM subtypes at first and second birth: California 1997–2020 (N=1,989,104)

Event in first birth	Total births	Prevalence of event in second birth, N (Prev)a	RR (95% CI)i
			Model 0b	Model 1c	Model 2d
SMM
No	1,965,526	21,124 (11)		reference
Yes	23,578	1,394 (59)	5.5 (4.8–6.6)	5.3 (4.6–6.3)	3.4 (2.9–4.1)
Non-Transfusion SMMe
No	1,977,463	10,882 (6)		reference
Yes	11,641	429 (37)	6.7 (5.0–8.7)	6.3 (4.7–8.1)	3.7 (2.6–5.3)
SMM subtypes f
Cardiac SMM
No	1,988,084	1,150 (1)		reference
Yes	1,020	48 (47)	81.4 (16.5–176.3)	62.7 (12.0–141.6)	32.6 (6.6–88.4)
Renal SMMg
No	1,988,230	915 (0.5)		reference
Yes	874	12 (14)	29.8 (12.0–50.5)	24.0 (9.5–40.0)	5.9 (2.2–10.3)
Respiratory SMMg
No	1,987,778	1,367 (1)		reference
Yes	1,326	20 (15)	21.9 (12.3–33.2)	18.7 (10.5–28.2)	10.3 (5.6–15.7)
Hemorrhage SMM
No	1,985,179	5,071 (3)		reference
Yes	3,925	136 (35)	13.6 (6.5–22.4)	12.8 (6.2–21.1)	6.6 (3.0–11.8)
Sepsis SMMg
No	1,987,122	1,959 (1)		reference
Yes	1,982	20 (10)	10.2 (5.9–16.2)	9.5 (5.5–14.9)	7.5 (4.3–11.8)
Other obstetric SMMg,h
No	1,985,844	1,789 (1)		reference
Yes	3,260	42 (13)	14.3 (9.0–19.0)	12.7 (8.0–17.0)	10.1 (6.3–13.5)
Other medical SMMi
No	1,988,362	752 (0.4)		reference
Yes	742	65 (88)	232 (55–432)	185 (43–336)	119 (30–267)
Transfusion SMM
No	1,975,593	12,804 (6)		reference
Yes	13,511	736 (54)	8.4 (6.7–10.2)	8.0 (6.4–9.7)	4.4 (4.0–6.5)
SMM indicators (SMM subtype) k
Acute heart failure (included in cardiac SMM subtype)
No	1,988,426	799 (<0.1)		reference
Yes	678	40 (5.9)	147.0 (96.7–216.2)	103.0 (68.3–154.6)	48.2 (30.7–73.6)
Disseminated intravascular coagulation (included in hemorrhage SMM subtype)
No	1,985,424	3761 (0.2)		reference
Yes	3,680	120 (3.3)	17.2 (13.8–20.8)	16.5 (13.2–20.0)	8.4 (6.6–10.2)
Eclampsia (included in other medical SMM subtype)
No	1,986,557	945 (<0.1)		reference
Yes	2,547	27 (1.1)	22.3 (12.0–32.7)	18.6 (10.1–27.5)	14.0 (7.6–20.8)
Puerperal cerebrovascular disorders (included in other obstetric SMM subtype)
No	1,988,435	708 (<0.1)		reference
Yes	669	24 (3.6)	101.0 (52.8–151.7)	95.8 (51.6–142.1)	77.8 (42.2–122.8)
Sickle cell disease with crisis (included in other obstetric SMM subtype)
No	1,989,028	46 (<0.1)		reference
Yes	76	40 (52.6)	23000 (14799–34287)	1563 (980–2704)	739 (469–1315)

Abbreviations: CI confidence interval, nt-SMM non-transfusion SMM, SMM severe maternal morbidity, Prev prevalence; RR risk ratio

^aPrevalence per 1,000 births.

^bModel 0 - Unadjusted

^cModel 1 - adjusted for age, education, insurance, nativity, race/ethnicity from first birth

^dModel 2 – Model 1 + adjusted for Expanded Obstetric Comorbidity Index from first birth

^eAdjusted for nt-SMM Expanded Obstetric Comorbidity Index instead of Expanded Obstetric Comorbidity Index

^fSee Table 1 for SMM indicators included in each subtype.

^gBootstrapping sample size of 80% used given the smaller sample size.

^hIncludes puerperal cerebrovascular disorders and sickle cell disease with crisis.

ⁱIncludes amniotic fluid embolism, eclampsia, severe anesthesia complications, air and thrombotic embolism

^j.Recurrence risk ratios refer to the recurrence of the same outcome in the second pregnancy that was experienced in the first pregnancy.

^kSMM indicators with >20 events of indicator recurrence are presented.

Recurrence risk stratified by subgroups

The adjusted risk ratios for SMM recurrence were highest among the following strata in second birth: age 20 to 39 years, higher levels of educational attainment, private and uninsured/other insurance, IPI <12 months, foreign-born, singleton births, White race/ethnicity, and lower comorbidity index (Table 4). For race/ethnicity, Hispanic individuals had a lower recurrence RR (3.03, 95% CI 2.74–3.38), compared with other subgroups (e.g., RR among Black individuals; 3.68, 95% CI 2.98–4.40). The pattern of results for nt-SMM was similar with a few notable exceptions (Table A.4). The recurrence RRs were similar for private and public insurance, estimates for IPI of <6 months had wide confidence bounds (2.94, 95% CI: 0.96–5.47), and Black individuals were at the highest RR of recurrence (4.90, 95% CI 3.57–6.57).

Table 4.

Prevalence (per thousand) and Risk Ratios for SMM recurrence between first and second birth, stratified by sociodemographic subgroups: California 1997–2020 (N=1,989,104)

Stratifying Variable in second birth	SMM in first birth	Total births (%)	Prevalence of SMM in second birth, N (Prev)a	RR (95% CI)e
				Unadjusted	Adjustedb
Age (years)c
<20	No	72,849 (98.6)	707 (1.0)
	Yes	1,012 (1.4)	44 (4.3)	4.48 (3.30–6.07)	3.05 (2.01–4.69)
20–29	No	949,212 (98.8)	9,438 (1.0)
	Yes	11,608 (1.2)	657 (5.7)	5.69 (5.26–6.16)	3.45 (3.17–3.73)
30–39	No	878,158 (98.9)	9,709 (1.1)
	Yes	10,017 (1.1)	615 (6.1)	5.55 (5.12–6.02)	3.43 (3.15–3.78)
≥40	No	65,307 (98.6)	1,270 (1.9)
	Yes	941 (1.4)	78 (8.3)	4.26 (3.39–5.36)	3.00 (2.13–3.92)
Education
Less than high school	No	205,099 (98.6)	2,319 (1.1)
	Yes	2,837 (1.4)	152 (5.4)	4.74 (4.02–5.58)	3.08 (2.54–3.76)
High school graduate	No	513,135 (98.8)	5,651 (1.1)
	Yes	6,476 (1.2)	387 (6.0)	5.43 (4.90–6.02)	3.31 (2.94–3.79)
Some college	No	555,675 (98.8)	6,288 (1.1)
	Yes	6,738 (1.2)	391 (5.8)	5.13 (4.63–5.68)	3.06 (2.65–3.45)
Undergraduate degree	No	422,645 (98.9)	4,230 (1.0)
	Yes	4,591 (1.1)	286 (6.2)	6.22 (5.52–7.02)	3.99 (3.39–4.71)
Postgraduate degree	No	268,972 (98.9)	2,636 (1.0)
	Yes	2,936 (1.1)	178 (6.1)	6.19 (5.32–7.20)	3.88 (3.21–4.78)
Insurancec
Private	No	1,261,867 (98.9)	12,739 (1.0)
	Yes	13,817 (1.1)	787 (5.7)	5.64 (5.25–6.06)	3.51 (3.24–3.81)
Public	No	663,628 (98.6)	7,975 (1.2)
	Yes	9,346 (1.4)	584 (6.2)	5.20 (4.78–5.66)	3.27 (2.92–3.61)
Uninsured/Other	No	40,031 (99)	410 (1.0)
	Yes	415 (1)	23 (5.5)	5.41 (3.56–8.24)	3.63 (1.87–5.94)
IPIc
<6 months	No	97,576 (98.6)	949 (1.0)
	Yes	1,356 (1.4)	84 (6.2)	6.37 (5.10–7.96)	3.91 (2.65–5.26)
6–11 months	No	236,449 (98.8)	2,104 (0.9)
	Yes	2,884 (1.2)	175 (6.1)	6.82 (5.84–7.96)	4.13 (3.13–5.12)
12–17 months	No	307,955 (98.9)	2,745 (0.9)
	Yes	3,572 (1.1)	196 (5.5)	6.16 (5.33–7.12)	3.65 (3.05–4.45)
18–59 months	No	994,886 (98.8)	10,441 (1.0)
	Yes	11,959 (1.2)	694 (5.8)	5.53 (5.12–5.97)	3.37 (2.97–3.78)
60+ months	No	285,586 (98.8)	4,526 (1.6)
	Yes	3,433 (1.2)	217 (6.3)	3.99 (3.48–4.57)	2.82 (2.37–3.43)
Nativity
Foreign-born	No	514,511 (98.8)	5,513 (1.1)
	Yes	6,132 (1.2)	362 (5.9)	5.51 (4.95–6.13)	3.49 (2.95–4.06)
US-born	No	1,451,015 (98.8)	15,611 (1.1)
	Yes	17,446 (1.2)	1,032 (5.9)	5.50 (5.16–5.86)	3.38 (3.08–3.71)
Pluralityc
Singleton	No	1,936,079 (98.8)	19,858 (1.0)
	Yes	23,233 (1.2)	1,349 (5.8)	5.66 (5.36–5.98)	3.50 (3.28–3.73)
Multiple	No	29,447 (98.8)	1,266 (4.3)
	Yes	345 (1.2)	45 (13.0)	3.03 (2.25–4.08)	2.13 (1.44–2.83)
Race/ethnicityd
Asian American	No	282,710 (98.8)	3,222 (1.1)
	Yes	3,333 (1.2)	209 (6.3)	5.50 (4.78–6.33)	3.62 (3.12–4.30)
Hispanic	No	784,565 (98.7)	8,654 (1.1)
	Yes	10,470 (1.3)	558 (5.3)	4.83 (4.44–5.26)	3.03 (2.74–3.38)
Non-Hispanic White	No	734,578 (99)	6,517 (0.9)
	Yes	7,164 (1)	380 (5.3)	5.98 (5.39–6.63)	3.84 (3.34–4.32)
Non-Hispanic Black	No	106,676 (98.3)	1,997 (1.9)
	Yes	1,873 (1.7)	199 (10.6)	5.68 (4.91–6.57)	3.68 (2.98–4.40)
Mixed race	No	40,421 (98.7)	516 (1.3)
	Yes	531 (1.3)	37 (7.0)	5.46 (3.91–7.62)	3.50 (2.09–4.99)
Expanded Obstetric Comorbidity Index (first birth)
Low	No	1,457,422 (99.6)	12,775 (0.9)
	Yes	6,083 (0.4)	263 (4.3)	4.93 (4.37–5.57)	4.70 (3.88–5.57)
High	No	508,104 (96.7)	8,349 (1.6)
	Yes	17,495 (3.3)	1,131 (6.5)	3.93 (3.70–4.19)	3.19 (2.90–3.54)
Expanded Obstetric Comorbidity Index (second birth)
Low	No	1,500,305 (99.1)	5,650 (0.4)
	Yes	13,818 (0.9)	274 (2.0)	5.27 (4.66–5.94)	3.91 (3.27–4.93)
High	No	465,221 (97.9)	15,474 (3.3)
	Yes	9,760 (2.1)	1,120 (11.5)	3.45 (3.25–3.67)	2.95 (2.70–3.22)

Abbreviations: AIAN American Indian or Alaska Native, CI confidence interval, IPI interpregnancy interval, SMM severe maternal morbidity, Prev prevalence; RR risk ratio.

^aPrevalence per 1,000 births

^bAdjusted for age, education, insurance, nativity, race/ethnicity, and Expanded Obstetric Comorbidity Index from first birth

^cBootstrapping sample size of 80% used given the smaller sample size.

^dPrevalences and risk ratios for American Indian or Alaska Native, Pacific Islander, and other race/ethnicity are not presented due to <12 SMM events in second birth for individuals with SMM in first birth.

^e.Recurrence risk ratios refer to the recurrence of the same outcome in the second pregnancy that was experienced in the first pregnancy.

Discussion

Summary of Findings

This population-based study analyzed the recurrence of SMM in a cohort of 1,989,104 pairs of first and second births. Experiencing SMM or nt-SMM in first birth was associated with an approximately three-fold risk ratio for recurrence in second birth (3.4, 95% CI: 2.9–4.1 and 3.7, 95% CI: 2.6–5.3, respectively), even after accounting for other known risk factors for SMM. When considering specific SMM subtypes, recurrence risk ratios ranged from four-fold for transfusion SMM to 33-fold for cardiac SMM and over 100-fold for ‘Other medical’ SMM (comprised of puerperal cerebrovascular disorders and sickle cell disease with crisis); however, confidence intervals for these estimates were wide, with the exception of among other obstetric SMM and transfusion SMM. In stratified analyses, recurrence RRs tended to be highest among individuals with lower absolute prevalence of SMM (e.g., individuals with higher education, singleton pregnancies, and lower co-morbidity scores). Black individuals were the only group that had both higher absolute prevalence and higher recurrence RRs of nt-SMM, relative to other subgroups.

Findings in context of prior work

Recent studies have documented the elevated recurrence risk of SMM between pregnancies.^20–24,43 Similar to our study, data from Quebec (1989–2021, N=819,375 birth pairs) indicated a recurrence RR of 3.1 (95% CI: 3.0–3.3) between first and second recorded births.²³ Findings from smaller samples also report elevated recurrence risks: an odds ratio for SMM in subsequent birth of 8.2 (95% CI: 5.5–12.2, N=36,190), using birth certificate data from Iowa (2009–2014); RR of 3.7 (95% CI: 2.7–5.0, N=4,368) from a single US center from 2011–2020; and RR of 16.3 (95% CI: 10.8–24.6, N=852) from a single tertiary care center from 2015–2018.^20–22 Ukah et. al. did not report the recurrence risk of specific SMM subtype in both first and second pregnancy, but their findings for the risk of any SMM following a specific subtype of SMM, and the risk of a specific subtype of SMM following any SMM, align with our findings that the recurrence RRs was particularly elevated for cardiac SMM (32.6, 95% CI: 6.6–88.4) and other obstetric SMM (which includes eclampsia, 10.1, 95% CI: 6.3–13.5).²³

We are unaware of prior studies examining recurrence risks of specific SMM subtypes or stratified by sociodemographic variables, which was possible due our study’s large study population of approximately 2 million birth pairs. No prior study adjusts for the Expanded Obstetric Comorbidity Index, which substantially attenuated recurrence risk ratios in our study. For example, when we previously published SMM recurrence risks in this study population (1997–2012, N=1,180,357 pairs), the adjusted recurrence RR was 6.5 (95% CI: 5.9–7.1; adjusted for education, insurance, race/ethnicity, age, and diabetes and hypertension before first pregnancy), which is comparable to the recurrence RR in this study from model 1 (5.3, 95% CI: 4.6–6.3; adjusted for age, education, insurance, nativity, race/ethnicity from first birth).²⁴ Upon the addition of the Expanded Obstetric Comorbidity Index to our models, we report an attenuation of the recurrence risk to 3.4 (95% CI: 2.9–4.1), suggesting that a substantial proportion of the risk is due to underlying maternal comorbidities (e.g., preeclampsia, chronic hypertension, placental abnormalities).⁴¹

We further considered how the recurrence risks differed among sociodemographic and clinical subgroups. For several variables, the subgroup that typically has higher absolute prevalence had lower recurrence risk ratios (e.g., lower education, public insurance, higher co-morbidity index, twin births). Thus, for individuals among whom there was a lower overall prevalence of SMM, the prognostic power of a first SMM event may be even stronger than for other groups. It is also possible that heightened surveillance during higher-risk pregnancies may lead to improved care and earlier detection of conditions that could develop into SMM in their second pregnancy. Our findings underscore the importance of monitoring subgroups for SMM recurrence risk that might otherwise be considered at a lower absolute risk for SMM.

Recurrence risk decreased with increasing IPI; this is partially consistent with studies showing that the risk of maternal adverse outcomes is highest among births with short IPI, however, we did not observe elevated recurrence risks among long IPIs, which are also shown to be associated with adverse outcomes.^22,44–46 Patterns of SMM and nt-SMM recurrence differed by race/ethnicity. Prior evidence from California supports that Hispanic, Asian and Black subgroups have a higher prevalence of SMM compared with White individuals⁸; in this study, White individuals experienced the greatest risk of SMM recurrence, followed by Black and Asian individuals, and with Hispanic individuals at the lowest risk of recurrence. Conversely, for risk of nt-SMM, Black individuals were at the highest risk of recurrence, followed by White individuals; Asian and Hispanic individuals had a lower risk of recurrence. Thus, in contrast to the pattern observed for other stratifying factors, Black individuals experience a at a higher absolute prevalence of nt-SMM in addition to their higher recurrence risks.^8,34,47

We conceptualize the presented RRs as descriptive and associational, i.e., we do not hypothesize that there is a direct causal association between SMM in first and second birth such that SMM in first birth causes SMM in second birth. Instead, we expect that SMM events in first and second births are associated due to shared upstream factors such as underlying maternal health and sociodemographic factors. The attenuation of the recurrence risk ratios with adjustment for sociodemographic and clinical factors support this. However, risk ratios remain elevated after adjustment, implying that other covariates likely contribute to the differences in risk – it is possible that the association could be further explained by overarching distal structural factors (e.g., neighborhood deprivation), physiology, and proximal clinical factors (e.g., hospital factors) not captured in these analyses.

Clinical and Research Implications

These findings highlight populations among whom there is opportunity to improve the quality of prenatal, peripartum, and postnatal care. Individuals who experience SMM in their first pregnancy and desire a subsequent pregnancy may require personalized counselling on reproductive planning as well as preventative care for health conditions between pregnancies, and prenatal care. Existing guidelines on postpartum care can be expanded to include timelines for follow-up and specific practice recommendations after an initial episode of SMM to provide counselling in subsequent risks and identify opportunities to prevent SMM recurrence.^48,49 Additionally, individuals who experience specific SMM subtypes such as cardiac SMM or other medical SMM indicators such as eclampsia should be monitored more closely during subsequent pregnancies given the pronounced recurrence risks. The high magnitude risk ratios presented in this work emphasize that prior SMM is a critical predictor of later SMM, particularly among groups that are generally experience lower absolute prevalence of SMM. Given the high preventability of SMM in clinical settings, this suggests that the prevention of SMM in early pregnancies could potentially reduce subsequent SMM.^3,50 Given these findings, next steps could include an investigation of risk differences for a complete understanding of the SMM recurrence risk; additionally, researchers could test methods for identifying higher risk subgroups for enhanced monitoring, such as a clinical risk score. Additionally, these results could inform counseling for specific indicators and more homogenous subtypes within their limitations, along with consideration of the known causes of SMM in first birth.

Strengths and Limitations

The large, diverse, population-based dataset, including virtually all births in California, is a major strength of this study. To the best of our knowledge, this study is the largest sample of linked birth pairs in which SMM recurrence has been investigated.

Our study has limitations. First, despite our large sample size, we had insufficient outcome events among certain racial/ethnic subgroups (e.g., Pacific Islander) and SMM subtypes (e.g., sepsis); we were unable to present recurrence risk estimates due to data privacy policies or issues with model convergence. Second, SMM is a composite outcome and as such, can be challenging use to inform exact clinical care. To address this, while maintaining adequate statistical power for these rare outcomes, we employed subtypes of SMM. However, these subtypes themselves also have limitations, such as the heterogeneity of outcomes grouped under ‘Other obstetric SMM’ or ‘Other medical SMM’. We tried to balance homogeneity of grouped conditions, sample size limitations, and completeness in conversation with clinical experts. Third, while we adjusted for a robust range of theoretically informed risk factors, it is important that future studies assess factors that we were unable to study, that may help us understand what explains, and thus could prevent, recurrence (e.g., hospital factors, quality of clinical care, multi-level social determinants of health). Fourth, given that the majority of the existing literature on this topic considers relative risk, we opted to present risk ratios that summarize the risk of SMM recurrence in the exposed and unexposed groups; however, risk ratios can overestimate results when not interpreted in context of the lower baseline prevalence of an outcome. In order to provide a complete picture of recurrence risk and clinical comparison of risks, future work should consider examining risk differences as well. Fifth, our study covered a long time period, 23 years, which was necessary for adequate sample size for SMM subtypes and indicators. It is possible that the consistency in the coding of SMM has changed over the years, however, prior work has shown that SMM prevalence was consistent over the transition from ICD-9 to ICD-10. ³¹Lastly, experiencing SMM can delay subsequent births⁵¹ – thus our sample could be subject to selection bias from the absence of individuals who are censored because experiencing SMM in first birth led to them (1) opting not to have a subsequent birth, or (2) having a subsequent birth outside our study time window. If so, this would lead to our sample comprising of a lower-risk cohort, and hence potentially underestimate the recurrence risk of SMM.

Conclusions

In a large, population-based cohort of births from California (1997–2020), we observed an approximately 3-fold adjusted risk ratio for recurrence of SMM and nt-SMM, between first and second birth. Groups that typically are associated with a lower absolute prevalence of SMM, such as individuals with singleton births or higher educational attainment, had higher recurrence risks relative to their counterparts. As an exception, Black individuals, who experience a high absolute prevalence of nt-SMM, also experience a high recurrence risk. This suggests the importance of close monitoring for the potential recurrence of SMM, regardless of baseline risk based on known risk factors.

AJOG at a glance:

1. Why was this study conducted?

   • Prior research on SMM shows an elevated risk of recurrence but has not investigated if it varies by sociodemographic subgroups or by subtype of SMM.
2. What are the key findings?

   • People who experienced SMM in first births had a three-fold higher prevalence of SMM in second births (5.9%), compared to people who did not have SMM in first births (1.1%).
   • Recurrence risk ratios for specific SMM subtypes ranged from four-fold for transfusion SMM to 100-fold for ‘Other medical’ SMM.
   • Subgroups with lower absolute prevalence of SMM (such as individuals with higher educational attainment, 20 to 39 years of age, lower comorbidity index scores, and singleton pregnancies) tended to have higher recurrence risk.
3. What does this study add to what is already known?

   • Results highlight the importance of appropriate counseling among patients with SMM in their first birth, regardless of baseline prevalence.