Directly Measured Costs of Severe Maternal Morbidity Events During Delivery Admission Compared to Uncomplicated Deliveries

Abstract

Objective

To estimate the actual excess costs of care for delivery admissions complicated by severe maternal morbidity (SMM) compared with uncomplicated deliveries.

Study Design

This is a retrospective cohort study of all deliveries between 10/2015–9/2018 at a single tertiary academic center. Pregnant individuals ≥ 20 weeks’ gestation who delivered during a hospital admission (i.e., a “delivery admission”) were included. The primary exposure was SMM, as defined by CDC criteria, CDC criteria excluding blood transfusion, or by validated hospital-defined criteria (intensive care unit admission or ≥ 4 units of blood products). Potential SMM events identified via administrative and blood bank data were reviewed to confirm SMM events had occurred. Primary outcome was total actual costs of delivery admission derived from time-based accounting and acquisition costs in the institutional Value Driven Outcomes database. Cost of delivery admissions with SMM events was compared with the cost of uncomplicated delivery using adjusted generalized linear models, with separate models for each of the SMM definitions. Relative cost differences are reported due to data restrictions.

Results

Of 12,367 eligible individuals, 12,361 had complete cost data. 280 individuals (2.3%) had confirmed SMM events meeting CDC criteria. CDC criteria excluding transfusion alone occurred in 1.0% (n = 121), and hospital-defined SMM in 0.6% (n = 76). In adjusted models, SMM events by CDC criteria were associated with a relative cost increase of 2.45 times (95% CI 2.29 – 2.61) the cost of uncomplicated delivery. SMM by CDC criteria excluding transfusion alone was associated with a relative increase of 3.26 (95% CI 2.95 – 3.60) and hospital-defined SMM with a 4.19-fold (95% CI 3.64 – 4.83) increase. Each additional CDC subcategory of SMM diagnoses conferred a relative cost increase of 1.60 (95% CI 1.43 – 1.79).

Conclusion

SMM is associated with between 2.5 and 4-fold higher cost than uncomplicated deliveries.

U.S. rates of maternal mortality and morbidity exceed those of other high-income countries, and continue to climb.^1,2 For each maternal death, between 50–100 individuals experience a severe maternal morbidity (SMM) or “near-miss” event.^1,3–6 SMM complicates 1–3% of pregnancies, and 30–50% of these events may be preventable.^3,7,8

Numerous federal agencies, professional bodies, and maternal health advocacy groups recommend comprehensive reviews of SMM at the local, state, and regional levels as a means of providing instrumental data for both SMM and maternal mortality reduction.^9,10 However, implementation of SMM review has been limited for various reasons, including high case volume, complexity of review, and associated costs. Quantifying the costs of SMM may motivate policy changes to review and address SMM as well as provide the important inputs for evaluating cost-effectiveness of interventions.

A small number of existing studies on the cost of SMM show an increase in cost associated with SMM utilizing blunt methods of charge-to-cost ratios or claims data.^11–13 However, the methods for assigning value to the time and resources spent caring for these patients have important limitations. Costs derived from the ratio of costs to charges, for instance, often bear little resemblance to the actual cost of providing services.^14–16 Analyses using claims data estimate costs based on insurance payments, which vary widely across payors, increasing heterogeneity. Furthermore, claims data only reflect how much providers received as reimbursement for care, not the actual value of that care.¹⁵

The Value Driven Outcomes (VDO) tool, a unique hospital-system database of time-based and actual utilization costs, provides an opportunity to evaluate the true cost burden of SMM from the perspective of the hospital system.^17,18 The VDO offers valuable and unique insights compared with more coarse methods of estimating costs, particularly for health services such as delivery care, which are reimbursed using a global reimbursement in most cases. The objective of this study was to estimate the actual excess hospital costs of delivery admissions complicated by SMM compared with uncomplicated deliveries.

Methods

We conducted a retrospective cohort study of all deliveries at a single tertiary academic institution in the U.S. between 10/1/2015 and 9/30/2018. These dates coincide with the three years following nationwide transition to the International Classification of Diseases-2010 edition (ICD-10). Individuals who delivered at ≥ 20 0/7 weeks gestation were included. Data sources for the project included: (1) the institution’s enterprise data warehouse (EDW), which combines several administrative and clinical databases; (2) a database compiled by the institutional blood bank; and (3) the VDO database.

The primary exposure was SMM during a delivery admission. A delivery admission was defined as any admission during which delivery occurred, regardless of the indication for admission. SMM events were characterized using three standardized definitions for SMM, which are not mutually exclusive. The Centers for Disease Control and Prevention (CDC) define SMM as one of 21 subcategories of severe maternal health outcomes during pregnancy – ranging from blood transfusion to cardiac arrest – identified using administrative data.^19,20 Using CDC criteria, blood transfusion constitutes the majority of cases of SMM; of these, many involve 1–2 units of blood transfused non-emergently in the postpartum period, which some experts do not consider a “true” SMM event.^9,21 Therefore, in addition to reporting cases identified using the full criteria, the CDC also reports cases after excluding those in which transfusion is the only qualifying diagnosis. Another validated definition is a “hospital definition,” supported by the Society for Maternal-Fetal Medicine, that identifies cases by intensive care unit (ICU) admission or transfusion of ≥4 units of blood.^9,21–23 However, neither ICU admission nor quantity of units transfused can be easily ascertained via administrative data alone, limiting this SMM definition to studies with more granular data. Fig. 1 provides an overview of the definitions and terms used throughout this analysis. Cost was estimated separately for each definition of SMM.

Fig. 1.

Standardized definitions of severe maternal morbidity, and ICD-10 codes which define each of the subcategories of CDC criteria for severe maternal morbidity.

To identify and verify SMM events, EDW data were first utilized to identify individuals within the eligible cohort who had ICD-10 codes meeting CDC criteria for SMM or had an ICU admission during pregnancy. All delivery admissions were cross-referenced to institutional blood bank data, and any transfusion events not identified by ICD-10 codes were included. Trained research personnel then reviewed each record to ensure that: 1) a qualifying event had actually occurred during a delivery admission; and 2) that the correct codes were used for the event. Quantity of units transfused was obtained from the EMR. All data were entered into a secure Research Electronic Data Capture (REDCap) database.^24,25

Each verified SMM event was characterized according to the three, not mutually exclusive, definitions of SMM: full CDC criteria, CDC criteria excluding transfusion alone, and hospital-defined SMM criteria. We further characterized each SMM event by CDC diagnosis subcategory and by the count of the subcategories for which the event met criteria (e.g, if an individual experienced both renal failure and required ventilation, the CDC SMM subcategory count would be two). Due to low numbers of some subcategories, we combined three pairs of CDC subcategories (pre-specified by CDC guidelines) for a total of 18 subcategories (Fig. 1).²⁰

The primary outcome of interest was total cost of delivery admission determined using University of Utah’s VDO tool, which has been used in previous studies to evaluate cost.^18,26–29 The VDO utilizes time and activity-based accounting, unit utilization (supplies and human resources) and a ledger of the institution’s financial transactions to develop a robust calculation of the true cost of care.¹⁷ For this analysis, the total cost of the delivery admission comprises time-based costs for facility/hospital unit utilization (based on length of stay) and imaging (based on time in imaging suite), and actual acquisition costs for medications, supplies, and laboratory studies. Provider fees are excluded. Notably, the costs we report relate only to maternal care, and do not account for neonatal care nor costs associated with maternal transfer from outside facilities.

Additional independent variables were obtained via the EDW and derived either from ICD-10 diagnosis codes (noted) or from clinical data in the EMR. Maternal demographic data include age, race and ethnicity, and primary insurance payor at the time of admission. As a matter of institutional policy, hospital registrars ask patients to self-identify race and ethnicity, but self-identification cannot be confirmed. Maternal comorbidities include pre-existing diabetes (ICD-10), body mass index (BMI) at the time of admission, and pre-existing hypertension (ICD-10). Obstetric characteristics include gestational age at delivery, multiple gestation, nulliparity, cesarean birth, length of stay, and count of antepartum and postpartum admissions. These independent variables were chosen because each has a known association with SMM as well as the potential to independently increase the cost of delivery admission (e.g., via increased risk of cesarean birth, induction of labor, or prolonged hospital admission).^3,11–13,30

Statistical Analysis

We performed an a priori power calculation using historic data for the mean, range, and standard deviations of the cost of delivery at our institution between 2015–2018. We estimated a 2% rate of SMM as defined by CDC criteria out of the known total of 12,367 deliveries during the study period. We assumed that a 30% difference in the total costs of deliveries affected by SMM compared with the total cost of uncomplicated deliveries would represent a meaningful difference from a hospital system perspective. Under these assumptions, we calculated 100% power at an α of 0.05 to detect our pre-specified cost difference.

Bivariate analyses were conducted for all sociodemographic, obstetric, and outcome variables to assess differences between people who did and did not experience SMM. We utilized chi-square for categorical variables, and Kruskal-Wallis test or Mann-Whitney U test for continuous variables as appropriate.

Multivariable generalized linear regression (GLM) with a gamma distribution and log link function was used to assess the association between each definition of SMM and the cost of delivery admission. Though the definitions of SMM are not mutually exclusive, on the premise that narrower definitions of SMM (e.g., CDC definition excluding transfusion only and hospital-defined SMM) may also represent more serious SMM events, and to permit direct comparisons of costs of events meeting these different criteria, we also modeled costs using a single multi-category SMM variable. Those who experienced hospital-defined SMM remained in that category. Of the remaining individuals experiencing SMM but not meeting criteria for hospital-defined SMM, those experiencing CDC-defined SMM excluding transfusion alone were included in the next category. Finally, any remaining individuals not meeting either of the other criteria made up the third group, and uncomplicated deliveries were the referent. To evaluate the role of baseline demographic or clinical differences, we included any covariates that differed across SMM categories, or could be associated with increased costs. Variations of the models with non-significant covariates removed were compared using the Akaike Information Criteria (AIC), which estimates model fit.³¹ The final group of covariates included in the adjusted models had the lowest AIC, indicating the best model fit.

The presence of a dose–response relationship between CDC SMM diagnoses and cost was assessed using GLM models in which the count of CDC SMM subcategories (an interval variable) was included as the exposure of interest. Predicted margins were calculated from this model to provide estimates of the relative change in cost associated with one, two, three, and four or more SMM CDC subcategory diagnoses. In addition, we separately estimated the crude and adjusted costs of each CDC SMM subcategory diagnosis for which there were more than 10 cases using GLM models. Finally, most cost analyses of SMM do not include gestational age due to lack of data availability. However, given the relationship between SMM and preterm delivery in clinical studies, we modeled the relationship between preterm birth, SMM and cost using GLM models as described. An interaction term was included to assess whether relationships between SMM and cost vary between preterm and term birth.

Institutional policy precludes reporting absolute dollar costs. Therefore, all cost data are reported as relative change in costs compared with an uncomplicated delivery by exponentiating the coefficients of the GLM models. A p-value < 0.05 was used to define statistical significance, and all tests were two-tailed. No missing data were imputed. All analyses were performed using STATA version 14.2 (StataCorp, College Station, TX). This study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline.³² The study was approved by the University of Utah Institutional Review Board.

Results

In total, 12,367 eligible people met inclusion criteria, of whom 6 were excluded due to missing cost data. Of the remaining 12,361 people, 422 met criteria for EMR review by SMM ICD-10 codes, and an additional 137 were identified as having transfusions using blood bank data. EMR review of these 559 pregnancies showed that 356 had confirmed SMM events, of which 280 (2.3%) occurred during a delivery admission (Fig. 2). Of these 280 delivery events meeting CDC criteria for SMM, 159 (1.3%) met criteria due to having a blood transfusion only, while 121 (1.0%) met CDC criteria after excluding transfusion alone. Sixty-six (0.5%) of these events met criteria by hospital-defined SMM. No individuals who met criteria for SMM did so by hospital-defined criteria alone; all SMM events by hospital-defined criteria also met criteria by at least one of the two CDC definitions.

Fig. 2.

Flow diagram for severe maternal morbidity case review and inclusion

Table 1 provides demographic, medical, and obstetric characteristics of the cohort, comparing people with an SMM event defined by CDC criteria to those with an uncomplicated delivery. Those experiencing SMM delivered at earlier gestational ages and were more likely to have pregnancies complicated by pre-existing hypertension. Cesarean delivery and multiple gestation were also more common among individuals with SMM. Maternal age and BMI were higher among people with SMM, who were also less likely to be non-Hispanic White. Both antenatal and postpartum admissions were more common, and delivery admission length of stay longer, among people with SMM.

Table 1.

Demographic, medical, and obstetric characteristics of pregnancies with severe maternal morbidity (as defined by CDC criteria) during delivery admission compared with uncomplicated deliveries^a

	SMM event by CDC criteria		Uncomplicated Delivery		p valueb
	(n = 280)		(n = 12081)
Demographic Factors
Maternal Age	30.8	(6.0)	29.6	(6.0)	p<0.001
Advanced maternal age (≥ 35 years)	73	26.2%	2362	19.6%	p = 0.01
Maternal Race and Ethnicity					p<0.001
Non-Hispanic White	133	48.4%	6305	53.1%
Non-Hispanic Black	6	2.2%	340	2.9%
Hispanic/Latina	88	32.0%	3897	32.8%
Asian	14	5.1%	645	5.4%
Native Hawaiian/Other Pacific Islander	14	5.1%	226	1.9%
American Indian/Alaska Native	3	1.1%	124	1.0%
Other or multiple races, non-Hispanic	17	6.2%	343	2.9%
Insurance status					p = 0.80
Private/Commercial	160	57.1%	7054	58.4%
Medicaid/Other government	71	25.4%	2857	23.7%
Self-Pay/Emergency Medicaid	49	17.5%	2169	18.0%
Medical Comorbidities
Body Mass Index	31.2	IQ (27.1, 36.0)	30.1	IQ (26.8, 34.4)	p = 0.01
Any diabetes mellitus	42	15.0%	1613	13.4%	p = 0.42
Pre-gestational hypertension	55	19.8%	1314	10.9%	p<0.001
Obstetric Factors
Gestational Age at Delivery	37.9	IQ (34.5, 39.6)	39.1	IQ (38, 39.9)	p<0.001
PTB <37 weeks	122	43.6%	1651	13.7%	p<0.001
PTB <34 weeks	54	19.3%	583	4.8%	p<0.001
PTB <28 weeks	17	6.1%	186	1.5%	p<0.001
Nulliparity	95	33.9%	4589	38.0%	p = 0.17
Multifetal gestation	20	7.2%	335	2.8%	p<0.001
Cesarean birth	186	66.4%	3069	25.4%	p<0.001
Antenatal admissionsc	31	11.1%	450	3.7%	p<0.001
Postpartum admissionsc	20	7.1%	157	1.3%	p<0.001
Delivery length of stay (days)	4.7	IQ (3.6, 6.3)	2.5	IQ (2.0, 3.3)	p<0.001

Abbreviations: CDC, Centers for Disease Control and Prevention; IQ, interquartile range; PTB, preterm birth; SMM, severe maternal morbidity.

^aData are displayed as mean (standard deviation), median (IQ range), or n (%) unless otherwise indicated

^bStatistical tests included Kruskal Wallis for medians, Mann Whitney U for means, and Chi-square for categorical outcomes

^cAntenatal and postpartum admissions are reported as the number (%) of participants with one or more admissions prior to admission for delivery (antepartum) or one or more readmissions after the delivery admission (postpartum)

Missingness: race and ethnicity – 207; insurance – 1; body mass index – 514; diabetes mellitus – 2; hypertension – 5; gestational age – 11; nulliparity – 6; multifetal gestation – 48; cesarean birth – 5

Table 2 provides crude and adjusted relative change in costs for models with each of the three definitions of SMM, count of CDC SMM subcategories, and each CDC SMM subcategory alone. In adjusted models, SMM by full CDC criteria was associated with a relative cost increase of 2.45 (95% CI 2.29 – 2.61). The adjusted relative cost associated with SMM by CDC criteria excluding transfusion alone was 3.26 (95% CI 2.95 – 3.60), and for hospital-defined SMM was 4.19 (95% CI: 3.64–4.83), compared with an uncomplicated delivery. Each additional SMM CDC subcategory diagnosed during a delivery admission was associated with an increase in costs by a factor of 1.60 (95% CI 1.43 – 1.79). In sensitivity analyses with adjusted models that directly compare the SMM definitions, the relative cost increases were overall similar (results not shown). Fig. 3 demonstrates the adjusted relative change in costs associated with each of the three definitions of SMM (modeled independently), as well as the dose response of increasing costs associated with increasing number of CDC SMM diagnoses. All adjusted models included maternal age, race and ethnicity, insurance status, pre-existing hypertension, nulliparity, cesarean delivery, and multiple gestation as covariates. BMI was excluded as an adjustment covariate as including it did not improve model fit and resulted in loss of data due to missingness.

Table 2.

Distribution of severe maternal morbidity with crude and adjusted^a relative changes in total cost of delivery hospitalization by SMM definition and CDC SMM subcategories

			Relative Change in Cost (95% CI)
			Unadjusted		Adjusteda
Delivery Categories	N	%	(n = 12361)		(n = 12092)
Uncomplicated delivery	12,081	97.7%	Referent	–	–	–
SMM: Full CDC criteria	280	2.3%	2.89	(2.67 – 3.14)	2.45	(2.29 – 2.61)
SMM: CDC criteria, excluding transfusion alone	121	1.0%	3.87	(3.40 – 4.40)	3.26	(2.95 – 3.60)
SMM: Hospital-defined	66	0.5%	4.83	(4.05 – 5.76)	4.19	(3.64 – 4.83)
Number of CDC diagnosis subcategoriesb			1.91	(1.81 – 2.01)	1.60	(1.43 – 1.79)
Deliveries with 1 CDC SMM category	218	1.7%
Deliveries with 2 CDC SMM category	43	0.3%
Deliveries with 3 CDC SMM category	12	0.1%
Deliveries with ≥4 CDC SMM category	7	0.0%
CDC SMM subcategories by frequency	N	% c
1) Blood transfusion	200	71.4%	2.42	(2.18 – 2.69)	1.98	(1.82 – 2.16)
2) Hysterectomy	51	18.2%	3.60	(2.89 – 4.48)	3.00	(2.51 – 3.58)
3) Shock	21	7.5%	4.21	(3.13 – 5.67)	4.03	(3.19 – 5.07)
4) Sepsis	19	6.8%	1.82	(1.44 – 2.30)	1.82	(1.52 – 2.19)
5) Acute renal failure	16	5.7%	2.60	(1.82 – 3.71)	1.76	(1.33 – 2.34)
6) DIC and coagulation defects	15	5.4%	6.23	(4.30 – 9.04)	5.54	(4.13 – 7.43)
6) Acute heart failure & pulmonary edema	15	5.4%	5.09	(3.59 – 7.23)	3.71	(2.80 – 4.91)
8) Adult respiratory distress syndrome	12	4.3%	5.17	(3.72 – 7.18)	4.27	(3.31 – 5.51)
9) Air and thrombotic embolism	5	1.8%	–	–	–	–
10) Ventilation or tracheostomyd	4	1.4%	–	–	–	–
10) Cerebrovascular disease	4	1.4%	–	–	–	–
12) Eclampsia	3	1.1%	–	–	–	–
13) Cardiac arrest/ventricular fibrillation or Conversion of cardiac rhythmd	2	0.7%	–	–	–	–
14) Acute MI or Aneurysmd	1	0.4%	–	–	–	–
14) Amniotic fluid embolism	1	0.4%	–	–	–	–
16) Heart failure/arrest during procedure	0	–	–	–	–	–
17) Severe anesthesia complications	0	–	–	–	–	–
18) Sickle cell disease with crisis	0	–	–	–	–	–

Abbreviations: CDC, Centers for Disease Control and Prevention; DIC, disseminated intravascular coagulation; IQ, interquartile range; MI, myocardial infarction; SMM, Severe maternal morbidity.

^aAdjusted for maternal age, race, insurance status, pre-existing hypertension, Cesarean delivery, multiple gestation

^bAdjusted and unadjusted relative cost increase associated with each additional CDC SMM diagnosis category

^cPercentages provided for each CDC subcategory are as a proportion of all SMM events (n = 280)

^dCategories were combined per CDC guidelines for low frequency subcategories

Fig. 3.

Relative change in costs of delivery admission associated with severe maternal morbidity compared with uncomplicated deliveries.^a

a) All estimated relative cost increases are statistically significant in comparison to uncomplicated deliveries. Relative costs of increasing counts of SMM subcategories are derived from predicted margins of GLM models. Whiskers at the top of each bar represent the upper bound of the 95% confidence interval. Abbreviations: CDC, Centers for Disease Control and Prevention; SMM, severe maternal morbidity.

The most common sub-categories of SMM included blood transfusion, hysterectomy, shock, sepsis, and acute renal failure; disseminated intravascular coagulation was associated with the largest relative increase in cost (5.54, 95% CI 4.13 – 7.43; Table 2). In adjusted models of the CDC definition of SMM including gestational age, each additional week of gestation was associated with an ∼5% decrease in costs (0.95, 95% CI 0.945 – 0.951) while SMM remained statistically associated with increased costs (2.07, 95% CI 1.24 – 3.43). The interaction term between gestation and SMM was not significant, indicating that the relative change in cost associated with SMM by CDC definition did not differ based on gestational age at delivery.

Discussion

SMM events, as defined by CDC criteria, are associated with an ∼2.5-fold increase in cost relative to uncomplicated deliveries. SMM events defined by CDC criteria excluding transfusion alone are associated with slightly more than 3-fold cost increases, while SMM defined by hospital criteria is associated with just over 4 times the cost of uncomplicated delivery. In addition, there is a dose–response between the number of SMM subcategory diagnoses and associated increased cost. Increased costs persist even when controlling for common procedures and diagnoses that have an impact on cost, including gestational age, cesarean birth, multiple gestation, and pre-existing hypertension.

These results provide additional context to previous data showing increased cost associated with SMM.^11–13 Howland et al found that, using charge-to-cost ratios, the excess cost associated with SMM (as defined by CDC criteria) in New York City delivery hospitalizations was approximately $6,000 after adjusting for covariates, representing a 70% increase in costs (∼1.7 times the cost of uncomplicated delivery). Vesco et al., utilizing MarketScan claims data, identified a cost increase of 37–47% associated with SMM compared with uncomplicated deliveries. It is worth noting, however, that due to the global insurance payment for deliveries, payor costs may underestimate true cost differences between deliveries complicated by SMM and uncomplicated deliveries. Indeed, we find that using actual hospital costs, the relative additional cost associated with SMM is much more pronounced than in either of these previous studies. This may be due to our ability to accurately identify cases via chart review in addition to the more granular cost accounting provided by the VDO Tool. We further expand upon the existing literature by introducing costs associated with different definitions of SMM and including an assessment of dose response.

The rate of SMM and the distribution of SMM subcategories at our institution is comparable overall to that reported by the CDC, as blood transfusion, hysterectomy, and disseminated intravascular coagulation were the top three CDC categories in 2014.³³ Of note, the rate of hysterectomy at our institution is influenced by the fact that it serves as a regional referral center for placenta accreta spectrum disorders.

Strengths of this study include the use of multiple data sources to ensure complete case ascertainment, as well as record review of SMM events to assure validity. Using administrative data alone may lead to inclusion of incorrectly coded events, which would necessarily skew any difference in costs toward the null. Additionally, VDO data provide estimates of actual hospital costs, in contrast to less accurate methods used in other studies. The ability to reliably identify ICU admissions and the volume of transfusion increases the utility of the study by providing a cost assessment of multiple definitions of SMM. While CDC criteria for SMM have the best receiver-operating characteristics for identifying true SMM events on a state-wide level, they have diminished utility at the institutional level, where hospital-defined SMM is frequently used.²¹ Therefore, our results provide data that both institutions and states can use to motivate SMM review and identify opportunities for prevention and cost reduction.

For instance, an institution with 2,500 deliveries per year, an average cost of delivery of $3,500, and a 2% rate of SMM (as defined by CDC criteria) could estimate an excess cost of $428,750 per year for SMM during delivery admissions. One-third to one-half of SMM events may be preventable,³⁴ and therefore a significant fraction of these excess costs could potentially be recovered with dedicated case review and intervention development. More impressively, a state with ∼50,000 births per year and a 1–2% rate of SMM could estimate an excess cost of nearly $5 million per year, with one-third to one-half related to preventable events. Systematic, state-wide maternal morbidity review may not only improve quality of maternal care, reduce maternal deaths, and potentially correct inequities in care, but may also result in significant savings to the health care system. Recovered funds could more than offset the costs of SMM review and implementing interventions.

Our study has several limitations. For rare outcomes (e.g., ≥ 3 CDC diagnosis subcategories), cost data are derived from relatively few cases, which may bias cost estimates in an unclear direction. Point estimates for costs of these rare outcomes should be interpreted cautiously. Additionally, a substantial proportion of the identified cases ultimately did not meet criteria after EMR review. As demonstrated in Fig. 2, the CDC criteria had a 65% positive predictive value. However, this is similar to other reported data. Many of these cases screened positive due to ICD-10 codes that represented historical rather than current diagnoses or were the result of inappropriate “upcoding” (e.g., pre-eclampsia coded as eclampsia). These discrepancies speak to the importance of state and local maternal morbidity reviews.

The VDO excludes provider fees, which may cause an underestimation of cost; however, even with this omission, we find substantially higher additional cost associated with SMM compared with other studies. The proprietary nature of absolute cost in the VDO prevents us from sharing actual dollar amounts, which may limit the interpretation of our data. As a single institution retrospective review, the generalizability of these results to other institutions is also limited. However, hospitals and states can utilize the provided relative costs to estimate the range of true absolute costs associated with SMM within their own institution or region. Finally, as an observational study, we cannot exclude the possibility of unmeasured confounding.

Ultimately, our findings suggest that, when actual costs are measured at the hospital level, SMM (depending on definition) is associated with an increase in costs between 2.5 and 4-fold higher than an uncomplicated delivery. The accuracy of the VDO too for identifying actual costs confers confidence in these results. Such confidence may assist hospitals and states in placing appropriate value on the important undertaking of local and state-wide maternal morbidity review, which may ultimately provide the data needed for evidence-based intervention.

Key Points.

• Severe maternal morbidity, as defined by CDC criteria, confers a 2.5-fold increase in delivery hospitalization costs

• ICU admission or ≥ 4 units of blood products confer a 4-fold increase in cost

• Costs of maternal morbidity may motivate SMM review