Short-Term Costs of Preeclampsia To The United States Health Care System

Report of Major Impact ajog.
org
Short-term costs of preeclampsia to the United States

health care system
Warren Stevens, PhD; Tiffany Shih, PhD; Devin Incerti, PhD; Thanh G. N. Ton, MPH, PhD; Henry C. Lee, MD; Desi Peneva, MS;
George A. Macones, MD; Baha M. Sibai, MD; Anupam B. Jena, MD, PhD
BACKGROUND: Preeclampsia is a leading cause of maternal RESULTS: Preeclampsia increased the probability of an adverse event
morbidity and mortality and adverse neonatal outcomes. Little is known from 4.6% to 10.1% for mothers and from 7.8% to 15.4% for infants while
about the extent of the health and cost burden of preeclampsia in the lowering gestational age by 1.7 weeks (P < .001). Overall, the total cost
United States. burden of preeclampsia during the first 12 months after birth was $1.03
OBJECTIVE: This study sought to quantify the annual epidemiological billion for mothers and $1.15 billion for infants. The cost burden per infant
and health care cost burden of preeclampsia to both mothers and infants in is dependent on gestational age, ranging from $150,000 at 26 weeks
the United States in 2012. gestational age to $1311 at 36 weeks gestational age.
STUDY DESIGN: We used epidemiological and econometric CONCLUSION: In 2012, the cost of preeclampsia within the first 12
methods to assess the annual cost of preeclampsia in the United months of delivery was $2.18 billion in the United States ($1.03 billion for
States using a combination of population-based and administrative mothers and $1.15 billion for infants), and was disproportionately borne by
data sets: the National Center for Health Statistics Vital Statistics on births of low gestational age.
Births, the California Perinatal Quality Care Collaborative Databases,
the US Health Care Cost and Utilization Project database, and a Key words: health care cost burden, hospital admission, maternal
commercial claims data set. morbidity, maternal mortality, perinatal morbidity, preeclampsia, preterm birth
P reeclampsia is among the top 6

causes of maternal mortality, severe
maternal morbidity, and adverse
(very early-onset preeclampsia), at
28e33 weeks (early-onset preeclampsia)
or after 34 weeks gestational age (GA). In
levels of these factors and the ratio be-
tween soluble fms-like tyrosine kinase-1
and placental growth factor correlate
neonatal outcomes in the United States addition, the clinical effects of pre- with the severity of preeclampsia as well
and globally.1-4 Within the United States, eclampsia may be limited to the mother as the latency period from diagnosis
the incidence of preeclampsia has risen (usually after 37 weeks of GA), the fetus, until delivery.11
steadily over the past 3 decades, from or both. Consequently, maternal and The clinical presentation of pre-
2.4% of pregnancies in 1980 to 3.8% of perinatal complications from preeclampsia is highly variable and may be
pregnancies in 2010.5,6 The incidence of eclampsia will depend on GA at onset, characterized by sudden progression
preeclampsia is rising, in part, because of severity of maternal manifestations, and requiring delivery within hours or days,
recent trends to delay pregnancy to a fetal condition.11 or it could remain stable and progress
later age and the increased rate of obesity The pathophysiological abnormalities very slowly over weeks. Nevertheless,
in pregnant women in the United in preeclampsia are characterized by pregnancies complicated by preeclamp-
States.6,7 endothelial dysfunction,12-15 altered sia are associated with substantial
The classic denition of preeclampsia immunological adaptation, exaggerated maternal and neonatal complications.
is characterized by hypertension plus inammatory response,16-19 increased These complications could happen
proteinuria; however, preeclampsia can coagulation with abnormal thrombin acutely with complete resolution, or the
occur in the absence of proteinuria.8,9 In production,20-22 increased oxidative mother and infant may be left with re-
this case, hypertension may be accom- stress,23-25 genetic predisposition,26 and sidual injury. In addition, neonates who
panied by systemic organ involvement reduced proangiogenic factors in asso- survive are at increased risk for long-
such as cerebral symptoms, abnormal ciation with increased antiangiogenic term decits related to residual injury
coagulation, and liver involvement.10 substances.12,27-30 Indeed, recent studies or as a result of fetal programming in
The clinical manifestations of pre- have found that the levels of placental utero. Moreover, some women who
eclampsia could develop at <28 weeks growth factor are reduced,31 whereas the survive may be left with residual decits
levels of soluble fms-like tyrosine kinase- as well as the effects of maternal vascular
132,33 and soluble endoglin34 are elevated programming (Table 1).
Cite this article as: Stevens W, Shih T, Incerti D, et al. at the time of diagnosis of preeclampsia. From an epidemiological perspec-
Short-term costs of preeclampsia to the United States
It has been suggested that the imbal- tive, preeclampsia is growing at a rate
health care system. Am J Obstet Gynecol 2017;xxx:xx-xx.
ance in proangiogenic and anti- more rapid than diabetes, ischemic
0002-9378/$36.00 angiogenic factors is responsible for heart failure, Alzheimers disease,
2017 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.ajog.2017.04.032 hypertension, proteinuria, and increased obesity, and chronic kidney disease,35-37
capillary permeability. As a result, the diseases for which substantial research
MONTH 2017 American Journal of Obstetrics & Gynecology 1

Report of Major Impact ajog.org
several low- and middle-income

TABLE 1
countries.38 Additionally, existing
Burden of early-onset preeclampsia in the United States
treatment options for preeclampsia are
Major maternal morbidity and Acute: death and intensive care unit complications limited, despite both the scale of the
mortality
e Eclampsia, stroke disease burden and its high growth
rate.
e Pulmonary edema
Several studies have estimated sepa-
e Myocardial ischemia rately the burden of preeclampsia on
e Admission to intensive care unit mothers and infants,39-43 while others
e Renal injury-failure with or without dialysis have examined overall trends in pre-
eclampsia by either maternal age at onset
e Abruption, disseminated intravascular coagulation
or by disease severity.44-47 No study has
e Liver dysfunction, hematoma simultaneously estimated the overall
Long-term: residual morbidity health and cost burden of preeclampsia
e Neurological deficit
by GA among both mothers and infants
in the United States within the rst 12
e Renal failure requiring dialysis months after delivery.48 This study esti-
e Cardiomyopathy mated the annual epidemiological and
Maternal programming caused by preeclampsia health care cost burden of preeclampsia
e Coronary artery disease
to both mothers and infants in the
United States in 2012.
e Chronic hypertension
e Metabolic syndrome Materials and Methods
e Renal insufficiency Study design and data sources
e Stroke
We conducted a retrospective cohort
study using secondary analysis of
e Retinal dysfunction multiple data sets to estimate the na-
e Premature death tional clinical and economic burden
Major perinatal morbidity and Acute: death and neonatal intensive care of preeclampsia in the United States.
mortality hospitalization complications We required data that were nationally
e Respiratory distress syndrome and representative, longitudinal, and
bronchopulmonary dysplasia linked mothers to their infants and
e Intraventricular hemorrhage, periventricular included International Classication of
leukomalacia Disease, ninth revision, codes to
e Necrotizing enterocolitis, late sepsis
ascertain both outcomes and costs
during the 6 months leading to the
e Retinopathy of prematurity birth event and the 12 months after
e Prolonged neonatal intensive care hospitalization delivery.
Long term: residual morbidity Because no single data source met all
of these requirements, we combined 5
e Cerebral palsy, neurological deficits, seizure
disorder of the data sets we reviewed (Appendix
1). To obtain mother-infant linkage and
e Learning disabilities
GA by week, we used data sources ob-
e Blindness and hearing deficits tained by California Perinatal Quality
e Chronic lung disease Care Collaborative: the California Ofce
e Chronic heart disease of Statewide Health Planning and
Development (OSHPD) data set linked
Fetal programming: fetal origin of adult disease
to vital statistics and birth certicate
e Metabolic syndrome data.49 The California OSHPD data
e Stroke, chronic heart disease, and others provide administrative and billing in-
Stevens. Short-term costs of preeclampsia in the US. Am J Obstet Gynecol 2017. formation on nearly all hospitalizations
in California and contains follow-up
data on infant and maternal outcomes.
and treatment funding have been preeclampsia in the United States has Only 2% of births are not linked to vital
allocated. This is notable because the recently exceeded the rate in other statistics in the cases of home births
rate of growth of early-onset high-income countries as well as without hospitalization, births in
2 American Journal of Obstetrics & Gynecology MONTH 2017

ajog.org Report of Major Impact
FIGURE 1
Methodology to estimate national cost burden of preeclampsia in 2012
Stevens. Short-term costs of preeclampsia in the US. Am J Obstet Gynecol 2017.
military hospitals, and freestanding Appendix 2. The study was approved adverse events, as listed in Appendix 3.
birthing centers. exempt by Western Institutional Re- Adverse outcomes are the main driver
Information on GA was obtained view Board since no patients were of cost differences between non-
from birth certicates. The linked data directly involved and all data were de- preeclamptic and preeclamptic preg-
constituted 2,021,013 linked maternal- identied. nancies and births. To contextualize
infant births during 2008e2011. We our cost estimates, we estimated the
used the National Center for Health Number of mothers with number of adverse outcomes in the
Statistics (NCHS) Natality File50 to esti- preeclampsia United States for preeclamptic mothers
mate national birth counts. Because the NCHS does not contain in- and their infants, the effect of pre-
For cost outcomes, we used a formation on preeclampsia, we conser- eclampsia on the probability of an
commercial claims data set that pro- vatively assumed that the national adverse outcome, and the association
vides detailed information on medical annual rate of preeclampsia was similar between types of adverse outcomes and
costs, health conditions, and de- to that observed in California OSHPD: costs.
mographic characteristics and allows 12.4% among mothers delivering <28
maternal-infant linkages in the data. weeks, 22.3% for 29e33 weeks, 13.0% National probabilities and counts
However, because claims data are not for 34e36 weeks, and 2.5% for 37 of outcomes
nationally representative, we incor- weeks.52 These estimates were multi- All outcomes were assessed during a 1
porated the Healthcare Cost and Uti- plied by the number of infants born to year time period after birth and consid-
lization Project51 data set, which does women in the United States in 2012 to ered an adverse maternal or infant
not allow for maternal-infant linkages, obtain counts. outcome. To predict the number of
but provides a nationally representa- adverse outcomes associated with pre-
tive sample of hospital costs by health Adverse outcomes eclampsia, we rst used the California
condition. These data sets are We identied a total of 8 types of OSHPD data to estimate logistic regres-
described in further detail in maternal and 10 types of infant sion models for each adverse maternal

TABLE 2
Characteristics of mothers delivering infants in California, 2008e2011a
Without preeclampsia With preeclampsia
(n 1,918,498) (n 69,193)
Characteristicb n % n %
Maternal age, y
<20 168,611 8.8 8089 11.7
20e24 408,610 21.3 14,736 21.3
25e29 516,508 26.9 16,290 23.5
30e34 482,990 25.2 15,338 22.2
35e39 271,336 14.1 10,583 15.3
40e44 66,094 3.4 3671 5.3
45e49 4031 0.2 431 0.6
50 or older 295 0.0 51 0.1
Infant male sex 982,604 51.2 36,014 52.0
Parity, live born
1 752,042 39.2 40,930 59.2
2e5 1,129,492 58.9 26,794 38.8
6e9 34,137 1.8 1327 1.9
10 or more 1553 0.1 86 0.1
Prenatal visits
0 8230 0.4 517 0.8
1e9 316,338 16.9 14,191 21.2
10e19 1,476,518 79.1 48,035 71.8
20e29 57,588 3.1 3429 5.1
30e39 7171 0.4 562 0.8
40e49 1420 0.1 142 0.2
50 or more 346 0.0 38 0.1
Multiples (yes) 27,146 1.4 4036 5.8
Body mass index
Underweight 72,568 4.1 1557 2.5
Normal weight 887,497 49.9 22,905 36.2
Overweight 458,589 25.8 17,400 27.5
Obese I 220,997 12.4 11,224 17.7
Obese II 88,773 5.0 5782 9.1
Obese III 50,763 2.9 4430 7.0
Maternal race
Non-Hispanic white 524,504 27.8 17,264 25.5
Non-Hispanic black 108,218 5.7 6241 9.2
Non-Hispanic Asian 187,933 10.0 3742 5.5
Non-Hispanic Pacific Islander 63,417 3.4 3231 4.8
Non-Hispanic American Indian/Alaska Native 8476 0.4 409 0.6
Non-Hispanic other 1545 0.1 63 0.1
Stevens. Short-term costs of preeclampsia in the US. Am J Obstet Gynecol 2017. (continued)

TABLE 2
Characteristics of mothers delivering infants in California, 2008e2011a (continued)
Without preeclampsia With preeclampsia
(n 1,918,498) (n 69,193)
Characteristicb n % n %
Hispanic US born 454,510 24.1 18,934 28.0
Hispanic foreign born 539,765 28.6 17,841 26.3
Insurance
Medi-Cal 917,590 47.9 33,802 49.0
Private 896,198 46.8 31,793 46.0
Uninsured 44,025 2.3 1538 2.2
Other 57,545 3.0 1909 2.8
Smoking during pregnancy 43,420 2.3 1505 2.2
Alcohol consumption 2125 0.1 128 0.2
Renal disease 349 0.0 260 0.4
Prior diabetes 14,771 0.8 2794 4.0
Gestational diabetes 138,057 7.2 8608 12.4
Cardiovascular disease 745 0.0 309 0.4
Chronic lung disease 621 0.0 266 0.4
a
A total of 33,322 linked records were missing on preeclampsia status; b Percentages are calculated among nonmissing values of each characteristic.
and infant outcome. However, because described below except GA. In other Estimating costs
women giving birth in California are words, we allowed the effect of pre- Predicting health care costs of
systematically different from those in eclampsia on outcomes to also include preeclampsia
the rest of the United States, we its effect on GA. We estimated the association between
derived sampling weights and esti- Generalized estimating equations preeclampsia and costs for mothers and
mated weighted logistic regression were used to account for the correlation infants in 2 stages, as displayed in
models using inverse probability between infants born to the same Figure 1. In the rst stage, we developed
weighting techniques (see Appendix 4 mother. For mothers with preeclampsia a population-representative data set that
for details). and their infants, we used the models to contained maternal covariates, health
The weighted models were used to predict the probability of an adverse outcomes, and costs. In the second stage,
predict the probability of each maternal outcome in both the status quo in which we used this data set to estimate the na-
and infant outcome for mothers with the mother has preeclampsia and a tional burden of preeclampsia. Here we
preeclampsia at the national level. The counterfactual scenario in which she describe the steps in more detail.
probabilities were then multiplied by does not. In the rst stage, we developed a
the total number of births to pre- population-representative data set that
eclamptic mothers in the United States Association of preeclampsia with contained maternal covariates, health
in 2012 to obtain absolute counts for gestational age outcomes, and costs by augmenting the
each adverse outcome. To assess the association between preexisting OSHPD data with nationally
eclampsia and GA, we used linear corrected predicted costs in 3 steps.
Association of preeclampsia with regression models with weekly GA as the First, we estimated the cost for
adverse maternal and infant continuous outcome and preeclampsia mothers and infants using regression
outcomes as the independent variable. We adjusted models in claims data (step 1,
To estimate the association between our models for maternal age, infant sex, Figure 1). Because preeclampsia often
preeclampsia and adverse outcomes, we nulliparity, prenatal care, multiple requires additional prenatal care, we
ran generalized estimating equations births, maternal body mass index, in- estimated maternal costs from 6
with logit links using all the variables surance status, cigarette smoking, months before to 12 months after
from the weighted logistic regressions alcohol consumption, and diabetes. birth. The infant model captured costs

In the second stage, we used linear

TABLE 3
regressions applied to the data set
Characteristics of mothers delivering infants in the United States, by
created in the rst stage to estimate the
geographya
relationship between preeclampsia and
National births nationally corrected imputed costs for
(excluding California) California births mothers and between GA and nationally
Characteristics (n 3,573,333) (n 545,758) corrected imputed costs for infants (step
Maternal age, y, mean (SD) 27.4 (6.2) 28.1 (6.3) 4, Figure 1). In these regressions, we used
Maternal education, n, % the same adjustments and sample
weights as those used to estimate adverse
Less than high school 725,819 20.5 148,680 28.1
outcomes, as described in the previous
High school 1,642,417 46.5 250,888 47.3 text.
College or higher 1,165,143 33.0 130,534 24.6 In the regression of nationally cor-
Infant male sex, n, % 1,829,339 51.2 278,858 51.1 rected imputed costs for infants, we
allowed costs by GA to differ by pre-
Maternal race, n, %
eclampsia status but found no evidence
Non-Hispanic white 2,564,272 72.3 402,652 74.9 that preeclampsia inuenced costs for
Non-Hispanic black 556,067 15.7 31,082 5.8 infants after conditioning on GA. We
Non-Hispanic Asian/Pacific 154,467 4.4 66,886 12.4 therefore calculated the average costs for
Islander mothers by GA and preeclampsia status
Non-Hispanic American 39,180 1.1 2752 0.5 and calculated the average costs for in-
Indian/Alaska Native/other fants only by GA (step 5, Figure 1).
Average predicted costs for mothers and
Hispanic US born 91,699 2.6 11,258 2.1
infants were multiplied by birth counts
Hispanic foreign born 142,661 4.0 23,282 4.3 to determine health care costs for pre-
Diabetes, n, % 133,133 3.7 13,360 2.5 eclamptic mothers and their infants
Chronic hypertension, n, % 37,738 1.1 1658 0.3 (step 6, Figure 1).
To estimate the marginal impact of
Gestational age, wks, mean (SD) 38.6 (2.6) 38.8 (2.4)
a
preeclampsia in step 6, we calculated
National Center for Health Statistics, 2004.
maternal costs caused by preeclampsia as
the net difference between pregnancies
with and without preeclampsia at each
for 12 months following birth. Second, multiplied imputed costs by correction GA, after adjusting for maternal cova-
we used the claims-based cost models factors from national Healthcare Cost riates.53,54 Net costs of preeclampsia for
to impute costs for each mother or and Utilization Project data to obtain infants were estimated by comparing
infant in the California OSHPD data as nationally representative costs (step 3, observed total costs of infants from
if they had been covered under private Figure 1). Appendix 5 details the eco- preeclamptic births to total costs in a
insurance (step 2, Figure 1). Finally, we nomic models and correction factors. counterfactual scenario in which we
TABLE 4
Number of adverse maternal outcomes in United States within 12 months of delivery among mothers with
preeclampsia by gestational age, using California Office of Statewide Health Planning and Development combined with
natality data from the National Center for Health Statistics
Mothers with CVA or Thrombo- Total
GA, wks preeclampsia ARF, % TIA, % MI, % Seizures, % cytopenia, % Hemorrhage, % DIC, % Death, % events
<28 3605 122 (3.4) 7 (0.2) 7 (0.2) 122 (3.4) 232 (6.4) 352 (9.8) 95 (2.6) 4 (0.1) 941
28e33 23,624 700 (3.0) 102 (0.4) 17 (0.1) 700 (3.0) 1231 (5.2) 1374 (5.8) 417 (1.8) 41 (0.2) 4581
34e36 41,856 576 (1.4) 91 (0.2) 11 (0.03) 576 (1.4) 1790 (4.3) 2475 (5.9) 474 (1.1) 14 (0.03) 6006
37 or more 87,595 389 (0.5) 131 (0.1) 14 (0.03) 398 (0.5) 3052 (3.5) 5288 (6.0) 637 (0.7) 13 (0.01) 9931
Total 156,681 1796 (1.1) 331 (0.2) 49 (0.03) 1795 (1.1) 6306 (4.0) 9489 (6.1) 1622 (1.0) 72 (0.1) 21,460
ARF, acute renal failure; CVA/TIA, cerebrovascular and transient ischemic accidents; DIC, disseminated intravascular coagulation; MI, myocardial infarction.

assumed the average cost of an infant, mothers. Across all GA, we estimated 72
had it not been born from preeclamptic maternal deaths among women with
8151
5268
5106
20,035
38,561
Total
pregnancy, was equal to the average cost preeclampsia during the 12 months after
of an infant born 2 weeks later. We did delivery.
Adverse outcomes in the United States within 12 months of delivery among infants born to mothers with preeclampsia in 2012, by gestational age
1413 (39.2)
642 (2.7) not include costs caused by the loss of life Table 5 describes adverse infant out-
179 (0.4)
173 (0.2)
2407 (1.8)
Death (%)
using estimates of the value per quality- comes born to mothers with pre-
adjusted life-year (QALY). eclampsia within the 12 months after
birth by GA. The predicted number of
Effect of outcomes on costs adverse infant outcomes was estimated
Seizures (%)
To provide further insight into how to be 38,561, with the most prevalent
42 (1.2)
213 (0.9)
171 (0.4)
4 (0.0)
430 (0.3)
adverse outcomes relate to costs, we adverse outcomes being neonatal respi-

estimated linear regression models of ratory distress syndrome (10.9%) and
BPD, bronchopulmonary dysplasia; IVH, intraventricular hemorrhage; NEC, necrotizing enterocolitis; PVL, cystic periventricular leukomalacia; RDS, respiratory distress syndrome; ROP, retinopathy of prematurity.
costs as a function of adverse outcomes, sepsis (7.4%).

1287 (35.7)
4783 (20.3)
Sepsis (%)
1949 (4.7)
1954 (2.2)
9974 (7.4)
separately for mothers and infants using In multivariable models estimating

claims data. Although claims data are the association between preeclampsia
not representative, we adjusted for a and GA, we found that mothers with
number of maternal covariates, so the preeclampsia in California had lower GA
1030 (28.6)
regression coefcients should provide by 1.6 weeks (P < .001). Nationally, we

981 (4.2)
79 (0.2)
99 (0.1)
2189 (1.6)
BPD (%)
reasonable estimates of the effects of the estimated that preeclampsia was associ-
adverse outcomes on cost.55,56 However, ated with a lower GA by 1.7 weeks
because costs tend to be higher for (P < .001).
10 (0.02)
commercially insured patients, we

79 (2.2)
127 (0.5)
4 (0.0)
220 (0.2)
PVL (%)
report both unadjusted mean differences Association of preeclampsia with

in costs between those with and without maternal and infant outcomes
preeclampsia and mean differences in Figure 2 examines the predicted proba-
693 (19.2)
27 (0.03)
1398 (5.9)
118 (0.3)
2236 (1.7)
costs adjusted by our estimated cost bility of maternal and infant adverse
IVH (%)
correction factor. outcomes for pregnancies with and

without preeclampsia. Overall, pre-
Results eclampsia increased the probability that
15 (0.02)
National probabilities and counts of a mother would have at least 1 adverse

304 (8.4)
582 (2.5)
75 (0.2)
976 (0.7)
NEC (%)
outcomes outcome from 4.6% to 10.1% and the

Table 2 presents characteristics of probability that an infant would have at
mothers who gave birth in California in least 1 outcome from 7.9% to 14.2%. For
965 (26.8)
18 (0.04)
14 (0.02)
920 (3.9)
1918 (1.4)
2008e2011 by preeclampsia status. mothers, preeclampsia had the largest

ROP (%)
Table 3 provides differences in de- inuence on the likelihood of hemor-

mographic and risk factor characteristics rhage (3.1% to 6.0%) and thrombocy-
used to calculate the probability of giving topenia (0.9% to 3.7%). For infants,
2256 (62.6)
9753 (41.3)
14,611 (10.9)
2225 (5.2)
376 (0.4)
birth in California vs elsewhere in the preeclampsia had the largest effect on the
RDS (%)
United States. In general, mothers giving probability of respiratory distress syn-

birth in California had lower educational drome (1.9% to 6.6%) and sepsis (3.0%
attainment, were less likely to be to 5.4%), although effects on predicted
non-Hispanic black, and were more probabilities were large for a number of
distress (%)
83 (2.3)
635 (2.7)
444 (1.1)
2439 (2.8)
3601 (2.6)
likely to be of Asian and Pacic Islander other adverse outcomes.

background. Diabetes and chronic hy-
Fetal
pertension were less common among Health care costs of preeclampsia

mothers in California. These differences in the United States
with preeclampsia
were weighted in national models. Table 6 presents national estimates of

Born to mothers
Table 4 reports national predicted mean costs of preeclamptic pregnancies

counts of adverse maternal events within and related births in the United States in
23,624
41,856
87,595
156,681
3605
12 months of delivery in 2012 among 2012 by GA, separated into maternal and
women with preeclampsia, according to infant cost episodes. Lower GA births
GA. The total number of adverse were associated with higher costs
37 or more
TABLE 5
maternal events was predicted to be (Table 6 and Figure 3). For example, the
GA, wks
28e33
34e36
21,460, with hemorrhage (6.1%) and combined mean maternal and infant
Total
<28
thrombocytopenia (4.0%) constituting cost per pregnancy ranged from

the most common events among $311,700 for GA <28 weeks to $23,035

FIGURE 2
Predicted probability of adverse maternal and infant outcomes for pregnancies
These probabilities are determined in pregnancies with and without preeclampsia.

for GA 37 weeks. The share of total mothers, this implies $1.03 billion in showed that most adverse outcomes
health care costs accounted for by the increased total maternal health care were associated with high costs, espe-
infant was 26% for GA 37 weeks vs costs. cially for infants (see Table 2 in Appendix
91% for GA <28 weeks. Total health care The average cost of preeclampsia per 6 for detailed results). For instance,
costs for preeclamptic pregnancies, infant was calculated by GA based on the intraventricular hemorrhage, broncho-
including the normal costs associated costs associated with a 2 week reduction pulmonary dysplasia, periventricular
with birth, were estimated to be $6.4 in GA (as shown in Figure 3), multiplied leukomalacia, and infant death were all
billion, summed across mothers and by the infant prevalence by GA, and then predicted to increase costs by more than
infants for all GAs. summed across all GAs. We found that a $100,000 per patient.
Patients with preeclampsia had higher reduction in GA by 2 weeks was esti-
rates of underlying comorbidities that mated to increase costs for all infants by Comment
may be associated with higher infant and $1.15 billion (Appendix 6). Therefore, Principal findings
maternal costs independent of the effects preeclampsia was estimated to cost the We found that preeclampsia is associated
of preeclampsia. As such, it is important US health care system an additional with a greater cost of birth to mothers at
to understand what proportion of higher $2.18 billion above the usual maternal any given GA and that preeclampsia is
costs among preeclamptic pregnancies is and infant costs associated with birth. associated with a lower GA at birth by
attributable to preeclampsia vs other approximately 1.7 weeks. The total cost
confounding factors. In our maternal Relationship between preeclampsia burden of preeclampsia to the US health
cost model, we estimated that the exis- outcomes on costs system was $1.03 billion for mothers and
tence of preeclampsia increased costs by The results of multivariable linear $1.15 billion for infants in 2012. Costs
$6583 per birth; summing across our regressions of costs as a function of were considerably larger for infants born
estimated 156,680 births to preeclamptic maternal and infant adverse outcomes at lower GA: the cost burden of

TABLE 6
Estimated unit and total health care cost for preeclampsia patients in the United States, by gestational age at birth
(2012) using California Office of Statewide Health Planning and Development and commercial claims data
Costs <28 wks (3604) 28-33 wks (23,624) 34-36 wks (41,856) 37 wks or longer (87,596) All (156,680)
Maternal cost per birth $29,131 $24,063 $19,692 $17,021 $19,075
Infant cost per birth $282,570 $59,803 $11,112 $6013 $21,847
Combined cost per birth $311,701 $83,866 $30,804 $23,035 $40,922
Total health care cost $1.2 billion $2.0 billion $1.3 billion $2.0 billion $6.4 billion
Total cost because of infant cost, % 91% 71% 36% 26%
preeclampsia per birth ranged from as demonstrated by a validation study maternal preeclampsia. We assumed the
$150,000 for infants born at 26 weeks of of preeclampsia identied by Interna- national rate of preeclampsia was similar
GA to $1311 at 36 weeks of GA. tional Classication of Disease, ninth to that observed in California but expect
revision codes.58 As such, preeclampsia this assumption to be conservative, given
Strengths and limitations without severe features may be under- that prevalence of risk factors such as
The epidemiological component of this reported and more likely misclassied hypertension and diabetes is lower in
study has several limitations. First, as no preeclampsia than preeclampsia California than nationally.60,61
adverse outcomes are based on admin- with severe features. Similarly, more Third, because no population-based
istratively coded data, and under- severe and expensive conditions and data exist to describe both maternal
ascertainment is commonly reported in procedures are more likely to be coded and infant adverse outcomes within 12
perinatal research using administrative because of billing and reimbursement months of delivery, we relied on models
data.57 For instance, severe and more patterns.59 incorporating national sampling weights
expensive conditions are more likely to Second, birth certicate data, the to predict national outcomes. The last
be coded than milder forms of disease source of NCHS data, do not document NCHS data set to release geographical
information was from 2004, after which
state-specic information was withheld
for privacy purposes. To the extent that
FIGURE 3 geographic patterns have changed since
Mean decrease in costs for infants born 2 weeks later 2004, our sampling weights may not
precisely reect current geographic
distributions.
Finally, the epidemiological data do
not capture births at military hospitals.
This omission may lead to bias if the
relationship between preeclampsia and
health outcomes differs between military
hospitals and the hospitals in our
epidemiological data.
Our cost estimates are subject to a
number of limitations as well. First, the
California OSHPD did not contain cost
information, so cost was imputed using
insurance claims data. Second, we did
not incorporate the direct effect of pre-
eclampsia on infant costs, independent
of GA. Third, our cost correction factors
were estimated using birth event costs
only, which may not be representative of
The costs are by gestational age. subsequent costs following birth.
Stevens. Short-term costs of preeclampsia in the US. Am J Obstet Gynecol 2017. We took a number of measures to
assess the potential impact of these

limitations. First, we compared our suggested that preeclampsia increases have had preeclampsia compared with
epidemiological predictions to nation- infant costs by $14,144. Across all infants those with healthy pregnancies.72 Other
ally reported statistics. Our estimates are in the United States, this implies an in- studies have shown increased risk of
similar to 2010 estimates from the Cen- fant cost burden of approximately $2.2 stroke in adult offspring of pregnancies
ters for Disease Control and Prevention, billion, considerably higher than our with preeclampsia.73
which found that 64 deaths occurred baseline estimate. Future studies can improve upon this
among mothers with preeclampsia or Both our model checks and sensitivity research by using cost data that are
eclampsia.62 We estimated 72 maternal analyses suggest that our results are directly observable (rather than
deaths associated from preeclampsia. In robust. Our prediction models and sta- imputing costs). Furthermore, extend-
one study using mixed methods to tistical methodology generate pre- ing the time window beyond the rst 12
improve upon case identication and dictions that are similar to published months after birth will provide a better
cause of death, 17% of maternal deaths ndings, and our sensitivity analyses understanding of how preeclampsia has
were attributed to preeclampsia.63 The suggest that our burden estimates may, if an impact on women and children in the
number of predicted maternal deaths anything, be underestimated. The pri- long term. n
from preeclampsia from our model mary reason that our predictions are
represents 14.6% of all maternal deaths. similar to nationally reported statistics is References
The overall infant death rate from our that we used population-based data sets 1. Khan KSW, Wojdyla D, Say L,
model was 6.17 per 1000 births. In 2012, to ensure that our results were nationally Glmezoglu AM, Van Look PF. WHO analysis of
causes of maternal death: a systematic review.
the Centers for Disease Control and representative, which is a strength of our Lancet 2006;367:1066-74.
Prevention reported an infant mortality study. Overall, our analysis is the most 2. Knight M. Eclampsia in the United Kingdom,
of 5.98 per 1000 births.64 comprehensive effort to date to quantify 2005. BJOG 2007;114:1072-8.
Second, we examined the sensitivity of the national epidemiological and eco- 3. Kuklina EV, Ayala C, Callaghan WM. Hyper-
tensive disorders and severe obstetric morbidity
our results to certain assumptions. For nomic burden of preeclampsia to both
in the United States. Obstet Gynecol 2009;113:
example, we examined the sensitivity mothers and infants during the rst 12 1299-306.
of our maternal cost burden estimates months after delivery. 4. Liu S1, Joseph KS, Bartholomew S, et al.
by comparing our results using the Temporal trends and regional variations in se-
California OSHPD to estimates from Clinical and research implications vere maternal morbidity in Canada, 2003 to
insurance claims data. In the California Although our ndings suggest that pre- 2007. J Obstet Gynaecol Can 2010;32:847-55.
5. Ananth CV, Keyes KM, Wapner RJ. Pre-
OSHPD model, which is more eclampsia has a signicant burden on eclampsia rates in the United States,
population-based, preeclampsia was society, there are additional burdens that 1980e2010: age-period-cohort analysis. BMJ
estimated to increase cost per birth by we did not consider. For example, we did 2013;347:f656.
$6583. Using only insurance claims data, not measure health burden. However, 6. Wallis AB, Saftlas AF, Hsia J, Atrash HK.
this effect was $8361. If we apply the cost estimates of QALYs lost exist for several Secular trends in the rates of preeclampsia,
eclampsia, and gestational hypertension, United
correction factor to the estimate from infant adverse outcomes associated with States, 1987e2004. Am J Hypertens 2008;21:
the insurance data, the cost of pre- preeclampsia (see Appendix 7).65-67 521-6.
eclampsia for mothers is similar to our Given the incidence of these conditions 7. Jeyabalan A. Epidemiology of preeclampsia:
baseline estimate: $5468 per birth. in preeclamptic pregnancies and the net impact of obesity. Nutr Rev 2013;71(Suppl 1):
Applying the cost-corrected estimate loss of QALYs established in prior studies S18-25.
8. Hutcheon JA, Lisonkova S, Joseph KS.
from the claims data decreases the of these adverse outcomes, our cost Epidemiology of pre-eclampsia and the other
maternal cost burden of preeclampsia burden estimate is likely conservative, hypertensive disorders of pregnancy. Best Pract
from $1.03 billion to $813 million. One capturing only a part of the overall Res Clin Obstet Gynaecol 2011;25:391-403.
potential explanation is that the burden of preeclampsia to society. 9. Steegers EA, von Dadelszen P, Duvekot JJ,
commercially insured population is Moreover, our study estimates costs Pijnenborg R. Pre-eclampsia. Lancet 2010;376:
631-44.
healthier on average, although this is up to the 12 months after birth for 10. Sibai BM. Biomarker for hypertension-
unlikely to completely explain the dif- mothers and infants, but longer-term preeclampsia: are we close yet? Am J Obstet
ference between the coefcients. It is adverse outcomes associated with pre- Gynecol 2007;197:1-2.
further possible that in the overall eclampsia have been demon- 11. Barton JR, Sibai BM. Biomarkers for pre-
diction, risk stratication, and ruling out pre-
population, preeclampsia is more likely strated.40,41,68-71 Preeclampsia has been
eclampsia: what are the appropriate goals and
to be correlated with costly negative associated with an increased risk of car- objectives? Am J Perinatol 2017;34:415-8.
health outcomes than in the claims diovascular disease.40,71 Women with a 12. Roberts JM, Taylor RN, Musci TJ,
data. history of preeclampsia have twice the Rodgers GM, Hubel CA, McLaughlin MK. Pre-
We assessed the sensitivity of our in- relative risk of developing ischemic heart eclampsia: an endothelial cell disorder. Am J
fant results in a similar fashion. In disease or cerebrovascular disease and 3 Obstet Gynecol 1989;161:1200-4.
13. Redman CW, Sacks GP, Sargent IL. Pre-
particular, we regressed infant costs on times the risk of developing hyperten- eclampsia: an excessive maternal inammatory
maternal covariates and whether the sion.40,68 In addition, end-stage renal response to pregnancy. Am J Obstet Gynecol
infants mother had preeclampsia, which disease occurs more often in women who 1999;180:499-506.

14. Smrason AK, Sargent IL, Redman CW. 27. Levine RJ, Maynard SE, Qian C, et al. 42. MacKay AP, Berg CJ, Atrash HK. Preg-
Endothelial cell proliferation is suppressed by Circulating angiogenic factors and the risk of nancy-related mortality from preeclampsia and
plasma but not serum from women with pre- preeclampsia. N Engl J Med 2004;350:672-83. eclampsia. Obstet Gynecol 2001;97:533-8.
eclampsia. Am J Obstet Gynecol 1996;174: 28. Redman CW, Staff AC. Preeclampsia, bio- 43. Stoll BJ, Hansen NI, Bell EF, et al. Neonatal
787-93. markers, syncytiotrophoblast stress, and outcomes of extremely preterm infants from the
15. Endresen MJ, Morris JM, Nobrega AC, placental capacity. Am J Obstet Gynecol NICHD Neonatal Research Network. Pediatrics
Buckley D, Linton EA, Redman CW. Serum from 2015;213:S9.e1-4. 2010;126:443-56.
preeclamptic women induces vascular cell 29. Karumanchi SA, Granger JP. Preeclampsia 44. Ananth CV, Savitz DA, Luther ER,
adhesion molecule-1 expression on human and pregnancy-related hypertensive disorders. Bowes WA Jr. Preeclampsia and preterm birth
endothelial cells in vitro: a possible role of Hypertension 2016;67:238-42. subtypes in Nova Scotia, 1986 to 1992. Am J
increased circulating levels of free fatty acids. 30. Escudero CA, Herlitz K, Troncoso F, et al. Perinatol 1997;14:17-23.
Am J Obstet Gynecol 1998;179:665-70. Role of extracellular vesicles and microRNAs on 45. Haddad B, Deis S, Gofnet F, Paniel BJ,
16. Hartley JD, Ferguson BJ, Moffett A. The role dysfunctional angiogenesis during preeclamptic Cabrol D, Siba BM. Maternal and perinatal out-
of shed placental DNA in the systemic inam- pregnancies. Front Physiol 2016;7:98. comes during expectant management of 239
matory syndrome of preeclampsia. Am J Obstet 31. Ilekis JV, Tsilou E, Fisher S, et al. Placental severe preeclamptic women between 24 and 33
Gynecol 2015;213:268-77. origins of adverse pregnancy outcomes: po- weeks gestation. Am J Obstet Gynecol
17. Ferguson KK, Meeker JD, McElrath TF, tential molecular targets: an Executive Work- 2004;190:1590-5.
Mukherjee B, Cantonwine DE. Repeated mea- shop Summary of the Eunice Kennedy Shriver 46. Wu CS, Nohr EA, Bech BH, Vestergaard M,
sures of inammation and oxidative stress bio- National Institute of Child Health and Human Catov JM, Olsen J. Health of children born to
markers in preeclamptic and normotensive Development. Am J Obstet Gynecol 2016;215: mothers who had preeclampsia: a population-
pregnancies. Am J Obstet Gynecol 2017;216: S1-46. based cohort study. Am J Obstet Gynecol
527.e1-9. 32. Holme AM, Roland MC, Henriksen T, 2009;201:269.e1-10.
18. Sacks GP, Studena K, Sargent IL, Michelsen TM. In vivo uteroplacental release of 47. Auger N, Fraser WD, Paradis G, Healy-
Redman CW. Normal pregnancy and pre- placental growth factor and soluble Fms-like Prots J, Hsieh A, Rhaume M-A. Preeclamp-
eclampsia both produce inammatory tyrosine kinase-1 in normal and preeclamptic sia and long-term risk of maternal retinal
changes in peripheral blood leukocytes akin pregnancies. Am J Obstet Gynecol 2016;215: disorders. Obstet Gynecol 2017;129:42-9.
to those of sepsis. Am J Obstet Gynecol 782.e1-9. 48. Shih T, Peneva D, Xu X, et al. The rising
1998;179:80-6. 33. Maynard SE, Min J-Y, Merchan J, et al. burden of preeclampsia in the United States
19. Gervasi M-T, Chaiworapongsa T, Pacora P, Excess placental soluble fms-like tyrosine kinase impacts both maternal and child health. Am J
et al. Phenotypic and metabolic characteristics 1 (sFlt1) could contribute to endothelial Perinatol 2016;33:329-38.
of monocytes and granulocytes in preeclampsia. dysfunction, hypertension, and proteinuria in 49. Ofce of Statewide Health Planning and
Am J Obstet Gynecol 2001;185:792-7. preeclampsia. Obstet Gynecol Surv 2003;58: Development. Number of patients by type of
20. Tanjung MT, Siddik HD, Hariman H, Koh SC. 564-5. coverage: California primary care clinics,
Coagulation and brinolysis in preeclampsia and 34. Venkatesha S, Toporsian M, Lam C, et al. 2011e2115. Available at: https://www.oshpd.
neonates. Clin Appl Thromb Hemostas Soluble endoglin contributes to the patho- ca.gov/. Accessed Nov. 29, 2016.
2005;11:467-73. genesis of preeclampsia. Nat Med 2006;12: 50. Hetzel AM. History and organization of the
21. Lattov V, Prochzka M, Prochzkov J, 642-9. vital statistics system. Hyattsville, (MD): National
et al. Preeklampsie a trombin generac n test. 35. Moran AE, Forouzanfar MH, Roth GA, et al. Center for Health Statistics; 1997.
Ceska Gynekol 2013;78:466-72. The global burden of ischemic heart disease in 51. World Health Organization. Report of the
22. Erez O, Gotsch F, Mazaki-Tovi S, et al. Evi- 1990 and 2010: the Global Burden of Disease Commission on Macroeconomics and Health-
dence of maternal platelet activation, excessive 2010 study. Circulation 2014;129:1483-92. Macroeconomics and Health: investing in
thrombin generation, and high amniotic uid 36. Murray CJ, Abraham J, Ali MK, et al. The health for economic development. Geneva
tissue factor immunoreactivity and functional state of US health, 1990-2010: burden of dis- (Switzerland); 2001.
activity in patients with fetal death. J Matern Fetal eases, injuries, and risk factors. JAMA 52. Lee HCB, M.V.:Green, S.:Butwick, A.J.:
Neonat Med 2009;22:672-87. 2013;310:591-606. Druzin, M.:Melsop, K.:Ton, T.G.N. The burden of
23. Wikstrm A-K, Nash P, Eriksson UJ, 37. National Center for Health Statistics. Health, preeclampsia on preterm birth (abstract). Pre-
Olovsson MH. Evidence of increased oxidative United States, 2011: with special feature on sented at the American Academy of Pediatrics
stress and a change in the plasminogen acti- socioeconomic status and health. Hyattsville 2016 national conference and exhibition.
vator inhibitor (PAI)-1 to PAI-2 ratio in early-onset (MD): National Center for Health Statistics; 2012. 53. Rehm J, Mathers C, Popova S,
but not late-onset preeclampsia. Am J Obstet 38. Roberts CL, Ford JB, Algert CS, et al. Pop- Thavorncharoensap M, Teerawattananon Y,
Gynecol 2009;201. 597.e-8. ulation-based trends in pregnancy hypertension Patra J. Global burden of disease and injury and
24. Cohen JM, Kramer MS, Platt RW, Basso O, and pre-eclampsia: an international comparative economic cost attributable to alcohol use and
Evans RW, Kahn SR. The association between study. BMJ Open 2011;1:e000101. alcohol-use disorders. Lancet 2009;373:2223-33.
maternal antioxidant levels in midpregnancy and 39. Basso O, Rasmussen S, Weinberg CR, 54. Warren DK, Quadir WW, Hollenbeak CS,
preeclampsia. Am J Obstet Gynecol 2015;213: Wilcox AJ, Irgens LM, Skjaerven R. Trends in Elward AM, Cox MJ, Fraser VJ. Attributable cost
695.e1-13. fetal and infant survival following preeclampsia. of catheter-associated bloodstream infections
25. Zhou X, Zhang G-Y, Wang J, Lu S-L, Cao J, JAMA 2006;296:1357-62. among intensive care patients in a nonteaching
Sun L-Z. A novel bridge between oxidative 40. Brown MC, Best KE, Pearce MS, Waugh J, hospital. Crit Care Med 2006;34:2084-9.
stress and immunity: the interaction between Robson SC, Bell R. Cardiovascular disease risk in 55. Cameron AC, Trivedi PK. Micro-
hydrogen peroxide and human leukocyte anti- women with pre-eclampsia: systematic review and econometrics: methods and applications.
gen G in placental trophoblasts during pre- meta-analysis. Eur J Epidemiol 2013;28:1-19. Cambridge (United Kingdom): Cambridge Uni-
eclampsia. Am J Obstet Gynecol 2012;206. 41. Lykke JA, Langhoff-Roos J, Sibai BM, versity Press; 2005.
447.e7-16. Funai EF, Triche EW, Paidas MJ. Hyperten- 56. Solon G, Haider SJ, Wooldridge JM. What
26. Nakimuli A, Chazara O, Byamugisha J, et al. sive pregnancy disorders and subsequent are we weighting for? J Hum Resources
Pregnancy, parturition and preeclampsia in cardiovascular morbidity and type 2 diabetes 2015;50:301-16.
women of African ancestry. Am J Obstet mellitus in the mother. Hypertension 57. Ford JB, Roberts CL, Algert CS, Bowen JR,
Gynecol 2014:210-520.e1. 2009;53:944-51. Bajuk B, Henderson-Smart DJ. Using hospital

discharge data for determining neonatal Statistics, Division of Vital Statistics. Linked associated with increased risk of stroke in
morbidity and mortality: a validation study. BMC birth/infant death records, 2007e2013. Avail- the adult offspring the helsinki birth cohort
Health Serv Res 2007;7:1. able at: http://wonder.cdc.gov/lbd-current. study. Stroke 2009;40:1176-80.
58. Geller SE, Ahmed S, Brown ML, Cox SM, html. Accessed Oct. 13, 2016.
Rosenberg D, Kilpatrick SJ. International clas- 65. Feingold E, Sheir-Neiss G, Melnychuk J,
sication of diseasese9th revision coding for Bachrach S, Paul D. HRQL and severity of brain
preeclampsia: how accurate is it? Am J Obstet ultrasound ndings in a cohort of adolescents Author and article information
Gynecol 2004;190:1629-33. who were born preterm. J Adolesc Health From Precision Health Economics, Los Angeles, CA (Drs
59. Ferver K, Burton B, Jesilow P. The use of 2002;31:234-9. Stevens, Shih, Incerti, and Ton and Ms Peneva); Perinatal
claims data in healthcare research. Open Public 66. Gough A, Linden M, Spence D, Epidemiology and Health Outcomes Research Unit, Divi-
Health J 2009;2:11-24. Patterson CC, Halliday HL, McGarvey LP. sion of Neonatology, Department of Pediatrics, Stanford
60. Centers for Disease Control and Pre- Impaired lung function and health status in adult University School of Medicine and Lucile Packard Chil-
vention (CDC) Behavioral Risk Factor Sur- survivors of bronchopulmonary dysplasia. Eur drens Hospital, Palo Alto, CA and California Perinatal
veillance System (BRFSS). Prevalence of Resp J 2014;43:808-16. Quality Care Collaborative, Palo Alto, CA (Dr Lee); Division
Hypertension, 2011, US adults ages 20 and 67. Kamholz KL, Cole CH, Gray JE, of Maternal-Fetal Medicine, Department of Obstetrics and
older. Available at: http://www.cdc.gov/ Zupancic JA. Cost-effectiveness of early treat- Gynecology, Washington University School of Medicine,
dhdsp/data_statistics/fact_sheets/fs_bloodp ment for retinopathy of prematurity. Pediatrics Seattle, WA (Dr Macones); Department of Obstetrics,
ressure.htm. Accessed Jan. 31, 2016. 2009;123:262-9. Gynecology, and Reproductive Sciences, University of
61. Centers for Disease Control and Prevention 68. Bellamy L, Casas JP, Hingorani AD, Texas Medical School at Houston, Houston, TX (Dr Sibai);
(CDC) Behavioral Risk Factor Surveillance Sys- Williams DJ. Pre-eclampsia and risk of cardio- and Department of Health Care Policy, Harvard Medical
tem (BRFSS). Diabetes data and statistics. vascular disease and cancer in later life: sys- School, and Massachusetts General Hospital, Boston, MA
Available at: http://gis.cdc.gov/grasp/diabetes/ tematic review and meta-analysis. BMJ (Dr Jena).
DiabetesAtlas.html. Accessed Jan. 31, 2016. 2007;335:974. Received Sept. 16, 2016; revised April 15, 2017;
62. Centers for Disease Control and Prevention 69. Doyle LW, Peter J. Adult outcome of accepted April 18, 2017.
(CDC) National Center for Chronic Disease extremely preterm infants. Pediatrics 2010;126: The sponsor was not involved in the design of the
Prevention and Health Promotion. Safe moth- 342-51. study; the review and approval of the report; and the
erhood: advancing the health of mothers in the 70. Hennessy EM, Bracewell MA, Wood N, et al. decision to submit the report for publication. The funding
21st century, 2015. Available at: http://www.cdc. Respiratory health in pre-school and school age agreement ensured the authors independence in
gov/chronicdisease/resources/publications/ children following extremely preterm birth. Arch designing the study, interpreting the data, and writing and
aag/pdf/2015/safe-motherhood-aag-2015. Dis Childh 2008;93:1037-43. submitting the report for publication.
pdf. Accessed April 30, 2016. 71. Ray JG, Vermeulen MJ, Schull MJ, This study was supported by rEVO Biologics, Inc.
63. Mitchell C, Lawton E, Morton C, Redelmeier DA. Cardiovascular health after Drs Stevens, Shih, Incerti, and Ton and Ms Peneva are
McCain C, Holtby S, Main E. California maternal placental syndromes (CHAMPS): employees, and Dr Jena is a consultant at Precision
pregnancy-associated mortality review: mixed population-based retrospective cohort study. Health Economics, which provides consulting and other
methods approach for improved case identi- Lancet 2005;366:1797-803. research services to pharmaceutical, device, govern-
cation, cause of death analyses and trans- 72. Vikse BE, Irgens LM, Leivestad T, mental, and nongovernmental organizations. Dr Macones
lation of ndings. Matern Child Health J Skjaerven R, Iversen BM. Preeclampsia and the is a Steering Committee member for a randomized clinical
2014;18:518-26. risk of end-stage renal disease. N Engl J Med trial performed by rEVO Biologics, Inc. The other authors
64. United States Department of Health and 2008;359:800-9. report no conflict of interest.
Human Services, Centers for Disease Control 73. Kajantie E, Eriksson JG, Osmond C, Corresponding author: Anupam B. Jena, MD, PhD.
and Prevention, National Center for Health Thornburg K, Barker DJ. Pre-eclampsia is jena@hcp.med.harvard.edu

Glossary of terms
administrative claims data:
data collected by government departments and other organizations for the purposes of registration, transaction, record keeping, billing, and
payment, usually during the delivery of a service, which contain a comprehensive history of health care utilization.
cost episode:
the cost of all services provided to a patient with a medical problem within a specific period of time across a continuum of care in an integrated
system.
counterfactual scenario:
a hypothetical scenario that captures what might happen to the treated sample had it not been subject to the exposure, keeping other factors
constant.
generalized estimating equations:
the formula used in generalized linear models, which accounts for correlated data that can arise from longitudinal repeated measures or
clustering among observations that share similar characteristics.
interaction:
a statistical and epidemiological phenomenon in which the magnitude of the association between the outcome and exposure variable differs
according to a third variable.
linear regression:
a statistical approach for modeling relationships between variables, in which exposure variables and controls have an impact on outcomes in
a linear fashion.
logistic regression:
a statistical technique for estimating the relationship between a binary outcome and 1 or more independent variables, based on the logistic
function.
logit link:
link that specifies that the relationship between the linear predictor and the mean of the outcome variable follows a logit function, which is the
natural log of the odds that the outcome equals 1 of the possible 2 outcome categories.
retrospective cohort study:
a study design that involves identification and follow-up of subjects who are identified only after the follow-up period under study has ended.
quality-adjusted life-year (QALY):
a measure of disease burden, including both the quality and the quantity of life lived used in economic evaluation to assess the economic
value of medical interventions.
weighted multivariable linear regression:
an approach for estimating the relationship between an outcome variable and 1 or more explanatory variables, in which each data points is
assigned a weight and data points with higher weights contribute more to the estimated relationship.
MONTH 2017 American Journal of Obstetrics & Gynecology 12.e1

Appendix 21. HealthCore (WellPoint Anthem) data to identify all hospitalizations that
Appendices: short-term costs of pre- 22. Henry Ford Health Systems occur to either the mother or the infant
eclampsia to the United States health 23. Kaiser Permanente of Northern within 12 months of delivery.
care system California (KPNC) Natality data set from National Center for
24. Athena HealthCare Health Statistics
Appendix 1 25. Geisinger (MedMining) and The National Center for Health Sta-
Data sets reviewed for Quintiles tistics (NCHS) holds a collection of data
analysis 26. Pharmetrics IMS on births within the United States.2 The
Health outcomes only: These data sources 27. Pharmetrics Plus natality data set has information obtained
comprised of cohort or networks of research 28. Pregnancy to Early Life Longitudi- from all birth certicates in all states.
centers that focus primarily on health nal System (PELL) Birth certicates have been revised over
outcomes. 29. Premier Perinatal Safety Initiative the years, but the basic information in-
1. California Maternal Quality Care 30. Optum Stork cludes delivery events, selected maternal
Collaborative (CMQCC) 31. Optum Touchstone characteristics, and infant outcomes at
2. California Ofce of Statewide 32. Truven Marketscan delivery, with geographic information on
Health Planning and Development 33. Healthcare Cost and Utilization births by state. In 2005, the NCHS with-
(California OSHPD) Project (HCUP) held geographic information on the birth
3. California Perinatal Quality Care event from publicly available data sets.
Collaborative (CPQCC) Stevens. Short-term costs of pre- The most recent information with
4. Child Health Development Study eclampsia in the US. Am J Obstet Gynecol geographical data on the birth event
(CHDS): 1950e1970 2017. available to the public is 2004. We used
5. Danish Birth Cohort the 2004 data set to provide a distribution
6. Early Childhood Longitudinal Appendix 2 of mothers who gave birth within Cali-
Study Birth Cohort (National Cen- Data sources fornia or elsewhere in the country. To
ter for Education Statistics) California Ofce of Statewide Health extrapolate to national counts of adverse
7. European Birth Cohorts (many) Planning and Development (California outcomes, we used 3,960,796 total births
8. Extremely Low Gestational Age OSHPD) in the United States in 2012 as the
Newborns (ELGAN) We, the investigators of this study, denominator.
9. Pediatrix Center for Research, Ed- linked several California population- Commercial claims data
ucation, and Quality (CREQ) based data sets. The rst was a linked The claims database is a private-sector
10. Preeclampsia Foundation California vital statistics/California Of- health insurance claims-based data. A
11. Preeclampsia International ce of Statewide Health Planning and major advantage of the claims data set is
Network Development (California OSHPD) hos- that it combines information from a
12. UK Millennium Cohort Study (UK pital discharge administrative database plethora of different insurance pro-
MCS) for maternal hospitalizations.1 The Cal- grams, in contrast to most other claims
13. Vermont Oxford Network (VON) ifornia OSHPD data provide adminis- data sets that tend to be linked to only a
14. National Bureau of Economic trative and billing information on nearly single insurance program. As a result, the
Research / National Center for all hospitalizations in California. Only claims data set has greater representation
Health Statistics 2% of births are not linked to vital sta- than the data from a single insurance
15. National Perinatal Collaborative tistics in the cases of home births provider. The database captures all
Project (PCP): 1959e1974 without hospitalization, births in mili- health care claims, including prescrip-
16. National Survey of Family Growth tary hospitals, and freestanding birthing tion drugs, inpatient, emergency, and
(NSFG) centers. This maternal data set was ambulatory services for elderly and
17. Neonatal Research Network (NRN) linked to the corresponding OSHPD nonelderly individuals with employer-
18. NIAID Birth Cohorts (40 or more hospital discharge database for infants, provided insurance from more than 50
cohorts) with 96% of observations linked. Both large Fortune 500 rms. It contains a
19. Maternal-Fetal Medicine Units databases include diagnoses and pro- cumulative 82 million lives since 1993
Network (MFMU) cedures based on the International and 2 million to 3 million covered lives
20. Regenstrief Institute Classication of Disease, ninth revision per year for 1997e2008. Members reside
(ICD-9), codes. Vital statistics data pro- in the Northeast, Midwest, South, and
Health and cost outcomes: These vide further sociodemographic and West regions of the United States. We
data sources include health mainte- medical characteristics of the mother utilized these data to estimate a cost
nance organizations or claims-based and infant. We used linked data between model that predicts the medical costs
administrative data sets from which 2008 and 2011, constituting 2,021,013 from all health care claims by observed
both health and cost outcomes can be linked maternal-infant births during this health outcomes or adverse events. For
abstracted. period. We used the California OSHPD mothers, the model included the period
12.e2 American Journal of Obstetrics & Gynecology MONTH 2017

6 months prior to and 12 months after in HCUP.3 The NIS covers all patients, HCUP and the claims data to identify
delivery, while for infants, the model including individuals covered by Medi- systematic differences in costs between
included the period 12 months after care, Medicaid, or private insurance as the 2 data sets, controlling for health
delivery. well as those who are uninsured. HCUP outcomes. We then estimated this sys-
Health care Cost and Utilization Project also provides the Nationwide Emergency tematic difference in cost to correct costs
(HCUP) Department Sample. We used HCUP to predicted using our claims-based cost
The HCUP contains a database called obtain nationally representative esti- model.
the National Inpatient Sample (NIS), mates for cost to supplement our nd- Stevens. Short-term costs of pre-
which constitutes a 20% sample of dis- ings using the claims database. We eclampsia in the US. Am J Obstet Gynecol
charges from all hospitals participating combined data on birth events from 2017.
Appendix 3
TABLE 1
ICD-9 codes for preeclampsia and adverse maternal outcomes
Condition or outcome ICD-9 code Description
Preeclampsia
Preeclampsia without severe features 642.4x Mild or unspecified preeclampsia
Preeclampsia with severe features 642.5x Severe preeclampsia
642.6x Eclampsia, unspecified
642.7x Preeclampsia or eclampsia superimposed on
preexisting hypertension
Renal failure 584.x Acute renal failure
586 Renal failure, unspecified
593.9 Unspecified disorder of kidney and ureter
669.3x Acute renal failure following labor and delivery
Cerebrovascular accident 431 Intracerebral hemorrhage
432.x Other and unspecified intracranial hemorrhage
433.xx Occlusion and stenosis of precerebral arteries
434.xx Occlusion of cerebral arteries
436 Acute, but ill-defined, cerebrovascular disease
671.5x Other phlebitis and thrombosis
674.0x Cerebrovascular disorders in the puerperium
Transient ischemic attack 435.0x Unspecified transient cerebral ischemia
435.1x Vertebral artery syndrome
435.8x Other specified transient cerebral ischemias
435.9x Unspecified transient cerebral ischemia
Eclamptic seizure 780.3x Convulsions
Myocardial infarction 410.xx Myocardial infarction
Thrombocytopenia 287.3 Primary thrombocytopenia
287.33 Congenital and hereditary thrombocytopenic purpura
287.39 Other primary thrombocytopenia
287.4 Secondary thrombocytopenia
287.5 Thrombocytopenia, unspecified
289.84 Heparin-induced thrombocytopenia

TABLE 1
ICD-9 codes for preeclampsia and adverse maternal outcomes (continued)
Condition or outcome ICD-9 code Description
446.6 Thrombotic microangiopathy
776.1 Transient neonatal thrombocytopenia
Severe intra- and post-partum hemorrhage 666.0x Third-stage hemorrhage
666.1x Other immediate postpartum hemorrhage
666.2x Delayed and secondary postpartum hemorrhage
641.9 Unspecified antepartum hemorrhage
Disseminated intravascular coagulation 286.6 Defibrination syndrome
286.9 Other and unspecified coagulation defects
666.3x Postpartum coagulation defects
ICD-9, International Classification of Disease, ninth revision.
Appendix 3
TABLE 2
ICD-9 codes for adverse infant outcomes
Outcome ICD-9 code Description
Fetal distress 656.8x Other specified fetal and placental problems affecting management of mother
656.3x Fetal distress affecting management of mother
768.2 Fetal distress before onset of labor, in liveborn infant
768.3 Fetal distress first noted during labor and delivery, in liveborn infant
768.4 Fetal distress, unspecified as to time of onset, in liveborn infant
Respiratory distress syndrome 769 Respiratory distress syndrome in newborn
Bronchopulmonary dysplasia 770.7 Chronic respiratory disease arising in the perinatal period
516.34 Respiratory bronchiolitis interstitial lung disease
516.69 Other interstitial lung diseases of childhood
517.8 Lung involvement in other diseases classified elsewhere
518.89 Other diseases of lung not elsewhere classified
Retinopathy of prematurity, stage >3 362.25 Retinopathy of prematurity, stage 3
362.26 Retinopathy of prematurity, stage 4
362.27 Retinopathy of prematurity, stage 5
Necrotizing enterocolitis, Bells grade 2 777.53, 777.52 Stage 3 necrotizing enterocolitis in newborn
Intraventricular hemorrhage, stage 3 772.14, 772.13 Intraventricular hemorrhage, grade 4
Cystic periventricular leukomalacia 779.7 Periventricular leukomalacia
Sepsis 659.3x Generalized infection during labor
670.2x Puerperal sepsis
771.81 Septicemia [sepsis] of newborn
995.91 Sepsis
995.92 Severe sepsis

TABLE 2
ICD-9 codes for adverse infant outcomes (continued)
Meningitis 322.9 Meningitis, unspecified
320.xx Bacterial meningitis
321.xx Meningitis due to other organisms
322.xx Meningitis of unspecified cause
115.xx Histoplasmosis
114.2 Coccidioidal meningitis
112.83 Candidal meningitis
100.81 Leptospiral meningitis (aseptic)
098.82 Gonococcal meningitis
094.2 Syphilitic meningitis
091.81 Acute syphilitic meningitis (secondary)
090.42 Congenital syphilitic meningitis
072.1 Mumps meningitis
054.72 Herpes simplex meningitis
053.0 Herpes zoster with meningitis
049.1 Nonarthropod-borne meningitis due to adenovirus
047.9 Unspecified viral meningitis
047.8 Other specified viral meningitis
047.1 Meningitis due to echo virus
047.0 Meningitis due to coxsackie virus
036.0 Meningococcal meningitis
013.xx Tuberculosis of meninges and central nervous system
003.21 Salmonella meningitis
Seizures 780.31 Febrile convulsions (simple), unspecified
780.32 Complex febrile convulsions
780.33 Posttraumatic seizures
780.39 Other convulsions
345.40 Localization-related (focal) (partial) epilepsy and epileptic syndromes with complex
partial seizures, without mention of intractable epilepsy
345.41 Localization-related (focal) (partial) epilepsy and epileptic syndromes with complex
partial seizures, with intractable epilepsy
345.50 Localization-related (focal) (partial) epilepsy and epileptic syndromes with simple
partial seizures, without mention of intractable epilepsy
345.51 Localization-related (focal) (partial) epilepsy and epileptic syndromes with simple
partial seizures, with intractable epilepsy
345.80 Other forms of epilepsy and recurrent seizures, without mention of intractable
epilepsy
345.81 Other forms of epilepsy and recurrent seizures, with intractable epilepsy
Death Discharge status of death
798.0 Sudden infant death syndrome
798.1 Instantaneous death

TABLE 2
ICD-9 codes for adverse infant outcomes (continued)
798.2 Death occurring in less than 24 h from onset of symptoms not otherwise explained
798.9 Unattended death
348.82 Brain death
ICD-9, International Classification of Disease, ninth revision.
Appendix 3
TABLE 3
ICD-9 and CPT codes for birth events
Variable Code
ICD-9 640.81, 640.91,
diagnosis 641.01, 641.11, 641.21, 641.31, 641.81, 641.91,
642.41, 642.42, 642.51, 642.52, 642.6-642.62, 644.2, 642.71, 642.72,
643.01, 643.11, 643.21, 643.81, 643.91,
644.20, 644.21,
645.11, 645.21,
646.01, 646.11, 646.12, 646.21, 646.22, 646.31, 646.41, 646.42, 646.51, 646.52, 646.61, 646.62, 646.71, 646.81, 646.82,
646.91,
647.0-647.02, 647.1-647.12, 647.2-647.22, 647.3-647.32, 647.4-647.42, 647.5-647.52,
647.6-647.62, 647.8-647.81, 647.9-647.92,
648, 648.1-648.12, 648.2-648.22, 648.3-648.32, 648.4-648.42, 648.5-648.52, 648.6-648.62,
648.7-648.72, 648.9-648.91
650.xx
651.01, 651.11, 651.21, 651.31, 651.41, 651.51, 651.61, 651.71, 651.81, 651.91,
652.01, 652.11, 652.21, 652.31, 652.41, 652.51, 652.61, 652.71, 652.81, 652.91,
654.01, 654.02, 654.11, 654.12, 654.14, 654.21, 654.31, 654.32, 654.41, 654.42, 654.51, 654.52, 654.61, 654.62, 654.71,
654.72, 654.74, 654.81, 654.82, 654.91, 654.92,
655.01, 655.11, 655.21, 655.31, 655.41, 655.51, 655.61, 655.71, 655.81, 655.91,
656.01, 656.11, 656.21, 656.31, 656.41, 656.51, 656.61, 656.71, 656.81, 656.91,
657.01,
658.01, 658.11, 658.20, 658.21, 658.23, 658.30, 658.31, 658.33, 658.41, 658.81, 658.91
659.xx
660.00, 660.01, 660.03, 660.10, 660.11, 660.13, 660.20, 660.21, 660.23, 660.30, 660.31, 660.40, 660.41, 660.43, 660.50,
660.51, 660.53, 660.60, 660.61, 660.63, 660.70, 660.71, 660.73, 660.80, 660.81, 660.83, 660.90, 660.91, 660.93,
661.01, 661.11, 661.21, 661.31, 661.41, 661.91,
662.01, 662.11, 662.21, 662.31,
663.01, 663.11, 663.21, 663.31, 663.41, 663.51, 663.61, 663.81, 663.91,
664.01, 664.11, 664.21, 664.31, 664.41, 664.51, 664.61, 664.81, 664.91,
665.01, 665.11, 665.22, 665.31, 665.41, 665.51, 665.61, 665.71, 665.72, 665.81, 665.82, 665.91, 665.92,
667.02, 667.12,
668.01, 668.11, 668.21, 668.81, 668.91,
669.01, 669.11, 669.21, 669.41, 669.50, 669.51, 669.60, 669.61, 669.70, 669.71, 669.81, 669.91
670.02, 670.12, 670.22, 670.32, 670.82,
671.01, 671.11, 671.21, 671.81, 671.82,
672.02,
674.12, 674.22, 674.32, 674.42, 674.51, 674.52, 674.82, 674.92,
675.01, 675.02, 675.11, 675.12, 675.21, 675.22, 675.81, 675.82, 675.91, 675.92,

TABLE 3
ICD-9 and CPT codes for birth events (continued)
Variable Code
676.01, 676.02, 676.11, 676.12, 676.21, 676.22, 676.31, 676.32, 676.41, 676.42, 676.51, 676.52, 676.61, 676.62, 676.81,
676.82, 676.91, 676.92,
677
763.xx
V27.xx
V30.xx-V39.xx
ICD-9 72.xx-73.1; 73.22, 73.3, 73.4, 73.51, 73.59, 73.6, 73.8, 73.91, 73.92, 73.93, 73.94, 73.99, 74.0, 74.1, 74.2, 74.4, 74.99
procedure
CPT 59400-59410
procedure
59510-59515
59610, 59612, 59614, 59618, 59620, 59622
CPT, current procedural terminology; ICD-9, International Classification of Disease, ninth revision.
Appendix 4 models were conducted rst in Califor- Appendix 5

Epidemiological methods nia OSHPD data sets and adjusted for Sources and description of
Sampling weights maternal age, infant sex, nulliparity, cost estimates
To obtain nationally representative prenatal care, multiples, maternal body To predict the medical costs of health
estimates, we derived sampling weights mass index, maternal race, insurance care for mothers and infants in the
for the California data using the NCHS. status, cigarette smoking, alcohol con- California OSHPD data, we rst esti-
Geographic information available in sumption, and prior diabetes. An inter- mated models of maternal and infant
2004 in the NCHS was used to identify action term between gestational age and costs in the commercial claims data.
whether a mother delivered in the State preeclampsia was included in each We estimated these models using a
of California. Specically, we developed model. We used inverse probability sample of all women 18 years and older
statistical models to predict the proba- weighting techniques using sampling who had a delivery episode between
bility of giving birth in California (vs weights derived from the NCHS to 2004 and 2014 and were continuously
non-California) based on shared de- conduct weighted regressions in Cali- enrolled for 6 months before and 12
mographic and risk factor variables fornia data. All logistic regressions were months after birth. Birth events were
available between the NCHS and Cali- thus weighted, with weights equal to 1/P identied by an ICD-9 code or CPT
fornia OSHPD. These predictors for each observation in the sample,4,5 code for delivery event. Birth events
included maternal age, race, education, thereby reweighting the individuals in occurring within 30 days of each
diabetes, hypertension, renal disease, the OSHPD data to match the overall US other were considered to be the same
and number of prenatal visits. We esti- population, based on the demographic event.
mated model parameters using the and risk variables described previously. Total costs were dened as the sum of
NCHS data and then applied the esti- These models were then used to predict medical (including inpatient, emer-
mates to the California OSHPD to the probability of each outcome for those gency, and ambulatory services) and
obtain a predicted probability P of giving with preeclampsia in the United States. pharmacy claims from all payers. Medi-
birth in California for each mother and These national probabilities were then cal costs were adjusted for ination using
infant in our California data set. applied to the number of US mother- the annual CPI data published by the
Weighted logistic regression models infantelinked records in 2012 from the Bureau of Labor Statistics. Medical costs
We conducted analyses to assess the NCHS to obtain the absolute count of were adjusted using the following
association between preeclampsia and annual predicted outcomes nationally equation:
maternal and infant outcomes. Each during the 12 months after birth.
outcome was assessed in a separate NCHS, National Center for Health CPI Adjusted Cost
model. We used logistic models to assess Statistics; OSHPD, Ofce of Statewide CPIReference Year
the association between preeclampsia Health Planning and Development. Cost in Service Year
CPIService Year
and dichotomous outcomes (maternal Stevens. Short-term costs of pre-
and infant) that occurred within the 12 eclampsia in the US. Am J Obstet Gynecol For example, if the CPI in the refer-
month period following delivery. All 2017. ence year of 2014 is 435.292, and the CPI

in the service year 2007 is 351.054, then a iterative fashion to reduce potential er- dependent variable, cost. However, there
$100 medical cost incurred in 2007 will ror because of multicollinearity. did not appear to be asymmetry in the
be adjusted as follows: Although a variable for gestational age maternal data that would be corrected by a
was available in the claims data, it was log transformation. Furthermore, we
CPI Adjusted Cost 2014 dollars missing for a large majority of the data assessed the appropriateness of a general-
435:292 and was therefore excluded from the ized linear model with gamma errors and a
$100:00 $124:00 model. Because gestational age likely log link by tting these models to our
351:054
drives costs through the need for costly claims data and then comparing the root
health outcomes, the inclusion of the mean squared error between the predicted
We utilized the variables listed in health outcomes themselves in the and observed values. The results are pro-
Tables 2 and 3 (in the text) and Appendix 3 model minimizes any error introduced vided in Appendix 5 Table 1 below. The
as predictors in the cost model; how- by omitting gestational age as a variable. results of the regression models used to
ever, we restricted the covariates to For the infant model, maternal cova- predict maternal costs are provided in
include those measures available in both riates were linked to the infant data with Appendix 5 Table 2 below.
the claims data and the California the family identication variable. CPI, Consumer Price Index; CPT,
OSHPD data. Because our goal was to For both the maternal and infant current procedural terminology; ICD-9,
simply predict the actual cost of medical models, we assessed the appropriateness of International Classication of Disease,
care for each mother and infant, given a log transformation for the cost variable. ninth revision; OSHPD, Ofce of State-
the available data, we did not focus on Because there appeared to be skew both in wide Health Planning and Development.
methods that would allow for causal the raw cost data and in the residuals of the Stevens. Short-term costs of pre-
identication. We excluded variables untransformed infant cost model, we eclampsia in the US. Am J Obstet Gynecol
with P values such that P > .1 in an applied a log transformation to the 2017.
Appendix 5
TABLE 1
Comparison of predictive accuracy of OLS vs GLM in claims data
Variable RMSE Maternal RMSE Infant
Model
OLS (log transformation for infant costs) 16,583 150,997
GLM (gamma, with log link) 21,726 660,303
GLM, generalized linear model; OLS, ordinary least squares; RMSE, root-mean-square error.
Appendix 5
TABLE 2
Regression model of maternal costs
Total cost, $ b Robust SE P value
Preeclampsia 4188 487 .000
Age 190 12 .000
Multiparity 3384 1677 .044
Hypertension 5888 373 .000
Gestational diabetes 3749 264 .000
Diabetes 5529 538 .000
Cardiovascular disease 14,573 5468 .008

TABLE 2
Regression model of maternal costs (continued)
Total cost, $ b Robust SE P value
Acute chronic lung 14,148 3785 .000
Poor fetal growth 4377 234 .000
Excess fetal growth 2066 204 .000
Antepartum hemorrhage abruption 7730 493 .000
Cesarean delivery 9326 157 .000
Assisted vaginal delivery 3094 140 .000
Fetal death 3833 1354 .005
Meconium 6183 1676 .000
Premature rupture 6396 476 .000
Multigestational birth 15,774 795 .000
Eclamptic seizure 21,345 2754 .000
Cerebrovascular accident or transient ischemic attack 27,392 4728 .000
Thrombocytopenia 9921 2861 .001
Disseminated intravascular coagulation 12,273 3275 .000
Pulmonary edema 33,424 9690 .001
Intrapostpartum hemorrhage 6710 574 .000
Pulmonary embolism 22,852 5325 .000
Deep venous thrombosis 17,304 3406 .000
Renal failure 32,513 6705 .000
Myocardial infarction 41,076 16,442 .012
Shock 27,871 14,614 .056
Sepsis 24,930 5195 .000
Acute liver failure 7994 1107 .000
Acute fatty liver 7550 2102 .000
Blood transfusion 5176 295 .000
Hysterectomy 28,898 7399 .000
Death 2843 1556 .068
Constant 1954 363 .000
n 83,846
R2 0.2415

Appendix 5
TABLE 3
Regression model of infant costs
Log of total costs b Robust SE P value
Preeclampsia 0.176 0.017 .000
Maternal age at birth event 0.005 0.001 .000
Hypertension 0.025 0.012 .043
Acute chronic lung disease 0.264 0.083 .002
Poor fetal growth 0.026 0.009 .005
Excess fetal growth e0.041 0.008 .000
Antepartum hemorrhage abruption 0.070 0.015 .000
Vaginal delivery e0.148 0.022 .000
Cesarean delivery 0.147 0.006 .000
Assisted vaginal delivery e0.046 0.006 .000
Meconium e0.261 0.103 .011
Premature rupture 0.265 0.019 .000
Cord prolapse 0.156 0.078 .044
Multigestational birth 0.529 0.020 .000
Fetal distress 0.150 0.025 .000
Respiratory distress syndrome 1.516 0.026 .000
Bronchopulmonary dysplasia 1.224 0.046 .000
Necrotizing enterocolitis >grade 2 0.861 0.292 .003
Intraventricular hemorrhage >stage 3 0.791 0.120 .000
Cystic periventricular leukomalacia 1.061 0.176 .000
Sepsis 0.605 0.028 .000
Meningitis 0.515 0.060 .000
Seizure 0.801 0.035 .000
Postbirth death 0.882 0.159 .000
NICU/critical care admission 0.925 0.012 .000
Constant 8.141 0.019 .000
n 76,335
R2 0.4298
CPT, current procedural terminology; HCUP, Health care Cost and Utilization Project; ICD-9, International Classification of Disease, ninth revision; KID, Kids Inpatient Database; NICU, neonatal
intensive care unit; NIS, National Inpatient Sample; OSHPD, Office of Statewide Health Planning and Development.
The above models were applied to the costs were unrealistically high. This insured population may not match those
California OSHPD data to predict the change had an impact on only one of the overall US population. Therefore,
cost of each mother and infant observed observation in the data. we used cost correction factors to the
in these data. We used Duans smearing The predicted costs obtained through predicted costs to adjust for national
estimate to correct predicted infant costs the process previously mentioned as- representativeness.
for the log transformation of the cost sume all individuals in the data have To estimate cost correction factors, we
variable.6 For infant costs, following costs similar to a privately insured pop- appended maternal or infant birth event
expert input, we replaced the cost in ulation. However, these costs may not be commercial claims data with maternal or
observations for which the predicted nationally representative because the infant birth event data from the 2014
cost would exceed $12 million to be demographics, risk factors, and service HCUP NIS for mothers and KID for
truncated at $12 million because these utilization patterns of the privately infants. Birth events were again isolated

using ICD-9 and CPT codes in the NIS regression. Claims data were simply were calculated as the ratio of the mar-
and KID. In addition to the hospital cost given a sampling weight of 1. Using these gins for HCUP data vs the claims
data, we included all variables from regression estimates, we then estimated data.11,12 The predicted costs in the
Appendix 5 Tables 2 and 3 if they were the impact of the claims data dummy on California OSHPD data were multiplied
also available in the HCUP data. We costs by estimating the (weighted) pre- by this ratio to correct for national
merged the HCUP birth event data with dictive margin, or average predicted representativeness.
the yearly cost-to-charge les by hospital response, in a scenario in which the We then used the model to predict
identication number corrected hospital dummy was equal to 0 for everyone in costs for each mother and infant in the
charges using cost-to-charge ratios.7-10 the sample and a scenario in which California OSHPD data. Because of the
To calculate the cost correction factors the dummy was equal to 1 for everyone. small sample sizes, we assumed,
in our analysis, we performed a linear The resulting predicted costs are pro- conservatively, that costs for infants born
regression of costs on the demographics, vided in Appendix 5 Table 4. Costs less than 23 weeks were equivalent to
covariates, and outcomes, interacted for claims data correspond to predicted costs for infants born between 23 and 27
with a dummy for whether the obser- mean costs with the dummy equal to 1 weeks and that costs for infants born at
vation was from the claims data (vs the and costs for HCUP correspond to 40 weeks or later were equivalent to costs
HCUP data). National sampling weights predicted mean costs with the dummy for infants born between 37 and 40
from HCUP were also utilized in the equal to 0. Cost correction factors weeks.

Appendix 5
TABLE 4
Impact of claims data on predicted mean costs
Variable Predicted cost SE t statistic P value
Maternal
HCUP 4613.9 4.1 1123.1 .000
Claims data 7054.6 36.2 195.0 .000
Cost correction factor 0.654
Infant
HCUP 3612.1 18.2 198.2 .000
Claims data 5093.8 77.6 65.7 .000
Cost correction factor 0.709
HCUP, Health care Cost and Utilization Project; OSHPD, Office of Statewide Health Planning and Development.
Nationally corrected predicted costs in to estimate the impact of preeclampsia analysis estimating the impact of pre-
California OSHPD were then used in on nationally corrected imputed costs eclampsia on adverse outcomes.
linear regressions to estimate the rela- for mothers and between gestational Predictive margins were estimated to
tionship between preeclampsia and age and nationally corrected imputed measure the average cost per mother by
maternal and infant costs using Cali- costs for infants, adjusting the eco- gestational age and preeclampsia status.
fornia OSHPD mother-infantelinked nomic models and applying inverse Predictive margins were estimated for
data. We used linear regression models probability weights identically as in the infants only by gestational age.

Appendix 6
Impact of premature birth
due to preeclampsia on
health care costs
TABLE 1
Estimated impact of premature birth due to preeclampsia on total infant health care costs, by gestational age at birth
(2012) using California OSHPD and commercial claims data
Mean costs
Gestational age Observed Born 2 weeks later Cost burden due to premature birth Cost burden fraction Births Total cost burden
23 $243,016 $285,289 -$42,273 0.17 352 -$14,880,096
24 $365,604 $299,404 $66,200 0.18 477 $31,577,591
25 $285,289 $230,911 $54,378 0.19 547 $29,744,493
26 $299,404 $146,809 $152,594 0.51 663 $101,169,955
27 $230,911 $109,372 $121,539 0.53 742 $90,182,235
28 $146,809 $80,110 $66,699 0.45 1763 $117,590,989
29 $109,372 $63,323 $46,049 0.42 2125 $97,853,445
30 $80,110 $42,435 $37,675 0.47 2895 $109,068,054
31 $63,323 $29,244 $34,079 0.54 3728 $127,045,244
32 $42,435 $20,509 $21,926 0.52 5312 $116,472,771
33 $29,244 $12,067 $17,177 0.59 7801 $134,001,443
34 $20,509 $7153 $13,356 0.65 7891 $105,394,437
35 $12,067 $5841 $6226 0.52 12,273 $76,406,151
36 $7153 $5841 $1311 0.18 21,692 $28,443,071
37 $5841 $5841 $0 0.00 8486 $0
38 $5875 $5875 $0 0.00 16,233 $0
39 $6112 $6112 $0 0.00 29,482 $0
40 $6067 $6067 $0 0.00 19,368 $0
Total 141,830 $1,150,069,783

TABLE 2
Linear regression analysis of the association between adverse outcomes and maternal and infant costs
Claims data Cost-corrected
Variable Estimate Lower 95% CI Upper 95% CI Estimate
Maternal
ARF 37,479 24,211 50,747 24,511
CVA/TIA 32,202 22,353 42,052 21,060
Eclamptic seizures 25,588 20,139 31,036 16,734
Thrombocytopenia 12,888 7129 18,646 8429
Hemorrhage 8680 7509 9851 5677
DIC 21,795 15,061 28,529 14,254
Infant
Fetal distress 80 -3278 3439 57
RDS 89,478 82,540 96,417 63,440
IVH 376,579 201,149 552,009 266,994
PVL 263,327 37,851 488,802 186,699
BPD 187,786 164,301 211,271 133,140
Sepsis 29,141 20,986 37,296 20,661
Seizure 48,519 33,883 63,154 34,400
Death 133,950 43,094 224,807 94,971
Note: Coefficient estimates from multivariable linear regressions of cost on preeclampsia, the adverse outcomes reported in the table, and maternal covariates. Separate models were run for mothers
and infants. Maternal covariates include age at birth and indicators for obesity, smoking, alcohol use, renal disease, diabetes, gestational diabetes, hypertension, cardiovascular disease, and chronic
lung disease. Costs were not log transformed. Coefficients are only reported for adverse outcomes for which the number of observations was sufficiently large. ARF acute renal failure, CVA/TIA
cerebrovascular and transient ischemic accidents, DIC disseminated intravascular coagulation, RDS respiratory distress syndrome, IVH intraventricular hemorrhage, PVL cystic peri-
ventricular leukomalacia, BPD bronchopulmonary dysplasia.

Appendix 7 blindnessis another serious condition are positively associated with functional
Sources and description of often found in premature babies. Studies and cognitive abnormalities in child-
social value estimates have found that infants born with ROP hood as well as poor elementary school
Although preeclampsia has immediate have signicant visual impairment later outcomes.21-23 Fiengold et al.24 show
effects on child health after birth, a large in childhood.17,18 Other research has that these impairments lead to signi-
part of its burden occurs over longer examined the impact of loss of visual cant losses in quality-of-life in early
time periods. This is especially true acuity on quality-of-life and shown that adulthood. They use a measurethe
because preterm infants are likely to infants with ROP accrue, on average, CDC HRQOL -14 Healthy Days Mea-
develop a number of health conditions 21.3 QALYs over a lifetime.19 surethat is not designed for creating a
with long-lasting effects. Table 1 below Premature birth also increases the preference-based measure of quality-of-
produces rough estimates of lifetime risk of respiratory distress syndrome life needed to calculate QALYs, but their
QALYs (with a 3% discount rate) for (RDS), a condition that usually de- results are comparable to those QALY
infants with some of these conditions velops shortly after birth and makes it losses seen in BPD studies. Lifetime
and compares them to expected QALYs hard for infants to breath. Some in- QALY losses due to premature birth are
for a healthy baby.a fants fully recover from RDS while substantial and could total as much as 9
One of the most devastating condi- others develop chronic lung disease QALYs for infants with NEC or ROP.
tions caused by preterm birth is necro- (CLD). As a result, adult CLD survi- If we were to assume the value of a
tizing enterocolitis (NEC), which has a vors have worse quality-of-life in QALY of $150,000 in the US, this would
case fatality rate of 15-30%.13 NEC sur- adulthood: mean utility based quality- mean a health burden of $20 billion per
vivors are also at risk for a number of of-life scores for adults with CLD as year, on top of the health care costs of $7
long-term complications including infants in a recent study were 0.84 billion. Our use of $150,000 is based on a
short-bowel syndrome (SBS), growth while scores for full-term infants were recent series of studies and reviews on
abnormalities, and neurodevelopmental 0.93.20 If we extrapolate the utility the value of a life-year, or health-
disabilities.14,15 SBS is a signicant score of 0.84 over a lifespan of 78 adjusted life-year, in a US setting.25-27
morbidity that occurs in one-fourth of years, then total discounted QALYs are Looking more widely, we also accoun-
NEC survivors.16 We used results from 26.0. By contrast, discounted QALYs ted for the guidelines suggested by the
these studies to provide back-of-the- are 28.7 for a full-term infant with a World Health Organizations Macro-
envelope calculation of QALYs lost from utility score of 0.93. There are conse- economic Commission on Health,
the disease. In particular, if we assume quently fairly large QALY losses from which was set up over a decade ago to
(1) that 25% of infants with NEC die bronchopulmonary dysplasia (BPD) evaluate the cost-effectives of global
from the disease and the life expectancy even if we assume that BPD has no health interventions. This report sug-
for these infants is 1 month, and (2) a effect on mortality. gested a health care intervention was cost
quality-of-life weight of 0.8 for the 25% Grade III or IV intraventricular hem- effective if it saved a health-adjusted
of NEC survivors with SBS, then infants orrhage (IVH) and periventricular leu- life-year at a cost of under three times a
with NEC can expect to obtain 21.9 komalacia (PVL) are brain injuries that nations GDP per capita.28 For the US,
QALYs over a lifetime.b are more common in preterm newborns. this is currently $160,000, so our esti-
Retinopathy of prematurity (ROP) It is common for babies with IVH to also mate may be considered slightly
the second leading cause of childhood have areas with PVL. Both IVH and PVL conservative.
a
A healthy full-term infant is assumed to have a life ex-
pectancy of 78 years and to live all years in full health.
Although it is not realistic to live all years in perfect health,
future health is discounted so most QALYs are accrued
early in life. A second case in which all life years are
weighted at 0.93based on evidence from Gough et al.
(2014)is also shown in the table.
b
We could not find a quality-of-life weight for SBS in the
literature. The score of 0.8 was chosen based on pref-
erence scores for disorders of the digestive system re-
ported in Bell et al. (1997). All calculations use a 3%
discount rate.

TABLE 1
Estimates of lifetime QALYs lost by outcome
Outcome Lifetime discounted QALYs Net QALYs lost Cases in preeclamptic pregnancies QALYs lost Social burden
Healthy infant 30.9
NEC 22 8.9 976 8686 $1303 m
ROP 21.3 9.6 1918 18,413 $2762 m
BPD 26 4.9 2189 10,726 $1609 m
IVH / PVL 27 3.9 2456 9578 $1437 m
Death 0 30.9 2407 74,376 $11,156 m
Total 121,780 $18,267 m
Notes: Social burden estimate values a QALY at $150,000. NEC necrotizing enterocolitis, ROP retinopathy of prematurity, IVH intraventricular hemorrhage, PVL cystic periventricular
leukomalacia, BPD bronchopulmonary dysplasia.
11. Graubard BI, Korn EL. Predictive margins 21. Elbourne D, Ayers S, Dellagrammaticas H,
References
with survey data. Biometrics 1999;55(2):652-9. Johnson A, Leloup M, Lenoir-Piat S. Rando-
1. Ofce of Statewide Health Planning and 12. Williams R. Using the margins command to mised controlled trial of prophylactic etamsylate:
Development. https://www.oshpd.ca.gov/. estimate and interpret adjusted predictions and follow up at 2 years of age. Archives of Disease in
Accessed November 29, 2016. marginal effects. Stata Journal 2012;12(2):308. Childhood-Fetal and Neonatal Edition
2. Hetzel AM. History and organization of the 13. Lin PW, Stoll BJ. Necrotising enterocolitis. 2001;84(3):F183-7.
vital statistics system. Hyattsville, Maryland: The Lancet 2006;368(9543):1271-83. 22. Gross SJ, Mettelman BB, Dye TD, Slagle TA.
National Center for Health Statistics.; 1997. 14. Hintz SR, Kendrick DE, Stoll BJ, et al. Neu- Impact of family structure and stability on aca-
3. Knight M. Eclampsia in the United Kingdom rodevelopmental and growth outcomes of demic outcome in preterm children at 10 years
2005. BJOG: Int J Obstet Gynaecol extremely low birth weight infants after necro- of age. The Journal of pediatrics 2001;138(2):
2007;114(9):1072-8. tizing enterocolitis. Pediatrics 2005;115(3): 169-75.
4. Mansournia MA, Altman DG. Inverse proba- 696-703. 23. Piecuch RE, Leonard CH, Cooper BA,
bility weighting. BMJ 2016;352:i189. 15. Stanford A, Upperman JS, Boyle P, Schall L, Sehring SA. Outcome of extremely low birth
5. Solon G, Haider SJ, Wooldridge JM. What are Ojimba JI, Ford HR. Long-term follow-up of weight infants (500 to 999 grams) over a 12-year
we weighting for? Journal of Human resources patients with necrotizing enterocolitis. Journal of period. Pediatrics 1997;100(4):633-9.
2015;50(2):301-16. pediatric surgery 2002;37(7):1048-50. 24. Feingold E, Sheir-Neiss G, Melnychuk J,
6. Duan N. Smearing estimate: a nonparametric 16. Berman L, Moss RL. Necrotizing enteroco- Bachrach S, Paul D. HRQL and severity of brain
retransformation method. Journal of the Amer- litis: an update. Paper presented at: Seminars in ultrasound ndings in a cohort of adolescents
ican Statistical Association 1983;78(383):605-10. fetal and neonatal medicine2011. who were born preterm. Journal of adolescent
7. Afana M, Brinjikji W, Cloft H, Salka S. Hospi- 17. OConnor AR, Stephenson T, Johnson A, health 2002;31(3):234-9.
talization costs for acute myocardial infarction et al. Long-term ophthalmic outcome of low birth 25. Braithwaite RS, Meltzer DO, King JT Jr,
patients treated with percutaneous coronary weight children with and without retinopathy of Leslie D, Roberts MS. What does the value of
intervention in the United States are substantially prematurity. Pediatrics 2002;109(1):12-8. modern medicine say about the $50,000 per
higher than medicare payments. Clinical cardi- 18. Palmer E, Hardy R, Dobson V, et al. 15-year quality-adjusted life-year decision rule? Medical
ology 2015;38(1):13-9. outcomes following threshold retinopathy of care 2008;46(4):349-56.
8. Agarwal S, Menon V, Jaber WA. Outcomes prematurity: nal results from the multicenter trial 26. Neumann PJ, Cohen JT, Weinstein MC.
after acute ischemic stroke in the United States: of cryotherapy for retinopathy of prematurity. Updating cost-effectivenessthe curious resil-
Does residential ZIP code matter? Journal of the Archives of ophthalmology 2005;123(3):311-8. ience of the $50,000-per-QALY threshold. New
American Heart Association 2015;4(3): 19. Kamholz KL, Cole CH, Gray JE, England Journal of Medicine 2014;371(9):
e001629. Zupancic JA. Cost-effectiveness of early treat- 796-7.
9. Chan T, Kim J, Minich LL, Pinto NM, ment for retinopathy of prematurity. Pediatrics 27. Weinstein MC. How much are Americans
Waitzman NJ. Surgical volume, hospital quality, 2009;123(1):262-9. willing to pay for a quality-adjusted life year?
and hospitalization cost in congenital heart sur- 20. Gough A, Linden M, Spence D, Medical care 2008;46(4):343-5.
gery in the United States. Pediatric cardiology Patterson CC, Halliday HL, McGarvey LP. 28. World Health Organization. Report of the
2015;36(1):205-13. Impaired lung function and health status in Commission on Macroeconomics and Health-
10. Finkler SA. The distinction between cost and adult survivors of bronchopulmonary dysplasia. Macroeconomics and Health: Investing in
charges. Annals of internal medicine 1982;96(1): European Respiratory Journal 2014;43(3): Health for Economic Development. Geneva.
102-9. 808-16. 2001.

Short-Term Costs of Preeclampsia To The United States Health Care System

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Short-Term Costs of Preeclampsia To The United States Health Care System

Uploaded by

Copyright:

Available Formats

Report of Major Impact ajog.

Short-term costs of preeclampsia to the United States

P reeclampsia is among the top 6

MONTH 2017 American Journal of Obstetrics & Gynecology 1

several low- and middle-income

2 American Journal of Obstetrics & Gynecology MONTH 2017

Stevens. Short-term costs of preeclampsia in the US. Am J Obstet Gynecol 2017.

MONTH 2017 American Journal of Obstetrics & Gynecology 3

4 American Journal of Obstetrics & Gynecology MONTH 2017

MONTH 2017 American Journal of Obstetrics & Gynecology 5

In the second stage, we used linear

6 American Journal of Obstetrics & Gynecology MONTH 2017

adverse outcomes relate to costs, we adverse outcomes being neonatal respi-

costs as a function of adverse outcomes, sepsis (7.4%).

separately for mothers and infants using In multivariable models estimating

regression coefcients should provide by 1.6 weeks (P < .001). Nationally, we

commercially insured patients, we

report both unadjusted mean differences Association of preeclampsia with

correction factor. outcomes for pregnancies with and

National probabilities and counts of a mother would have at least 1 adverse

outcomes outcome from 4.6% to 10.1% and the

2008e2011 by preeclampsia status. mothers, preeclampsia had the largest

Table 3 provides differences in de- inuence on the likelihood of hemor-

Stevens. Short-term costs of preeclampsia in the US. Am J Obstet Gynecol 2017.

United States. In general, mothers giving probability of respiratory distress syn-

likely to be of Asian and Pacic Islander other adverse outcomes.

pertension were less common among Health care costs of preeclampsia

were weighted in national models. Table 6 presents national estimates of

Table 4 reports national predicted mean costs of preeclamptic pregnancies

thrombocytopenia (4.0%) constituting cost per pregnancy ranged from

MONTH 2017 American Journal of Obstetrics & Gynecology 7

These probabilities are determined in pregnancies with and without preeclampsia.

8 American Journal of Obstetrics & Gynecology MONTH 2017

MONTH 2017 American Journal of Obstetrics & Gynecology 9

10 American Journal of Obstetrics & Gynecology MONTH 2017

MONTH 2017 American Journal of Obstetrics & Gynecology 11

12 American Journal of Obstetrics & Gynecology MONTH 2017

MONTH 2017 American Journal of Obstetrics & Gynecology 12.e1

12.e2 American Journal of Obstetrics & Gynecology MONTH 2017

MONTH 2017 American Journal of Obstetrics & Gynecology 12.e3

12.e4 American Journal of Obstetrics & Gynecology MONTH 2017

MONTH 2017 American Journal of Obstetrics & Gynecology 12.e5

12.e6 American Journal of Obstetrics & Gynecology MONTH 2017

Appendix 4 models were conducted rst in Califor- Appendix 5

MONTH 2017 American Journal of Obstetrics & Gynecology 12.e7

12.e8 American Journal of Obstetrics & Gynecology MONTH 2017

MONTH 2017 American Journal of Obstetrics & Gynecology 12.e9

12.e10 American Journal of Obstetrics & Gynecology MONTH 2017

MONTH 2017 American Journal of Obstetrics & Gynecology 12.e11

12.e12 American Journal of Obstetrics & Gynecology MONTH 2017

MONTH 2017 American Journal of Obstetrics & Gynecology 12.e13

12.e14 American Journal of Obstetrics & Gynecology MONTH 2017

MONTH 2017 American Journal of Obstetrics & Gynecology 12.e15

12.e16 American Journal of Obstetrics & Gynecology MONTH 2017

You might also like