Abstract

Background

The pathogenesis of preeclampsia remains poorly understood. Recent investigations have suggested that the incidence varies by season of conception and the season of delivery. A cross-sectional study was conducted to determine whether there was an association between the season of delivery and the prevalence of preeclampsia/eclampsia in Texas.

Methods

Retrospective analysis of hospital discharge records of 312,207 women who delivered in Texas in 2007 was performed. This statewide dataset was obtained from the Texas Department of State Health Services (Austin, TX). The season of admission for delivery was the independent variable: winter (December, January, February), spring (March, April, May), summer (June, July, August), and fall (September, October, November). The outcome was preeclampsia or eclampsia as defined by ICD-9-CM codes. Crude and adjusted prevalence odds ratios (OR) were calculated and reported with 95% confidence intervals (CI) and P values. The monthly prevalence of preeclampsia was also examined.

Results

Seasonal variation was minimal with the lowest prevalence detected in the fall (3.89%) and a peak of 4.1% in the winter. The highest monthly prevalence was found in January (4.4%). After adjusting for maternal age, race, and other potential confounders, women who were admitted in the fall for delivery were 6% less likely than women who were admitted in the winter to have preeclampsia: adjusted OR = 0.94, 95% CI: 0.89–0.99, P = 0.02).

Conclusions

A weak protective association between delivering in the fall (vs. winter) and the prevalence of preeclampsia was noted in this analysis of a statewide hospital database.

Preeclampsia is a multisystem disorder of pregnancy whose clinical signs include proteinuria and hypertension after 20-weeks gestation. Eclampsia is a severe complication in a woman carrying a diagnosis of preeclampsia and is manifest by new-onset seizures. Eclamptic seizures are rare, occurring in <1% of women who are diagnosed with preeclampsia.1 Approximately 5–7% of pregnancies are complicated by preeclampsia and complications of pregnancy-related hypertensive disease are the third leading cause of maternal deaths in the US.1 Early diagnosis and close observation of women with this condition are warranted as placental abruption, acute renal failure, cerebrovascular and cardiovascular complications, disseminated intravascular coagulation, and maternal death are all associated with preeclampsia.1,2

Some well established risk factors for preeclampsia include chronic hypertension, obesity, nulliparity, black race, age <20 years or >35 years, renal disease, and maternal infection.1,3 Smoking has been found to decrease the risk of preeclampsia among underweight and normal-weight women but not overweight/obese women.4

Previous studies have also indicated that there are seasonal trends in the incidence of preeclampsia/eclampsia. These fluctuations may be due in part to seasonal variation in ambient temperature and humidity. A retrospective study conducted by Immink, et al. analyzed over 11,000 deliveries in Tygerberg Hospital, South Africa and found that the occurrence of preeclampsia was highest in the winter (13.6%).5 In a separate study conducted in the tropical climate of Mumbai, India, Subramanian and colleagues analyzed 29,562 deliveries during a 36-month period from March 1993 to February 1996 and recorded the incidence of preeclampsia and eclampsia. They concluded that the incidence of eclampsia was significantly higher in the monsoon season when the weather is humid and cooler.6 Tam and colleagues also noted seasonal trends in the incidence of preeclampsia in singleton primiparae in Hong Kong between 1995 and 2002.7 After adjusting for maternal age, women who had conceived in June had the highest risk of developing preeclampsia compared to women who had conceived in October: odds ratio (OR) 2.8, 95% confidence interval (CI): 1.5–5.2). The authors concluded that further studies are needed to explore the role of ambient temperature and humidity in the development of preeclampsia.7

A study by Phillips et al. evaluated whether timing of delivery or timing of conception was more strongly associated with development of preeclampsia.8 Their sample size was comprised of 7,904 singleton primiparous gestations of which 142 were complicated by preeclampsia. These investigators found a seasonal variation in preeclampsia that appeared to be more strongly related to timing of conception than to the timing of delivery. Conception during the summer months compared with spring was associated with a 70% increase in the odds of preeclampsia (OR = 1.7, 95% CI 1.06–2.75).8

Using data from the Texas Department of State Health Services, we calculated the monthly prevalence of preeclampsia per 100 deliveries in Texas in calendar year 2007 and determined if the prevalence of preeclampsia varied by the season of delivery (winter, spring, summer, fall) in Texas in 2007 before and after adjusting for maternal age and other possible confounders. We also examined the association between season of delivery and preeclampsia after stratifying by maternal race/ethnicity.

Methods

This study was approved by the Texas Tech University Health Sciences Center Institutional Review Board, El Paso, Texas, and the Texas Department of State Health Services Institutional Review Board #1 (Austin, TX).

Source population and inclusion criteria.

Hospital inpatient discharge data (both public use and research data) leased from the Texas Department of State Health Services in Austin, Texas was used to perform a retrospective analysis. These data are from all state licensed hospitals except those that are exempt from reporting to the Texas Health Care Information Council. According to the data user manual, “Exempt hospitals include those located in a county with a population of less than 35,000, or those located in a county with a population more than 35,000 and with fewer than 100 licensed hospital beds and not located in an area that is delineated as an urbanized area by the United States Bureau of the Census... ” Hospitals that do not seek insurance payment or government reimbursement are also exempt from the reporting requirement.

This study evaluated the principal discharge diagnosis and up to 24 secondary diagnoses. Diagnoses found in this dataset had been coded using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). Patient records were included in this study if any of the 25 discharge diagnosis fields contained an ICD-9-CM code beginning with V27, indicating delivery during that hospital stay. Analyses were restricted to women 12–50 years of age. A unique patient identifier was used to ensure that only the first record of any woman who had multiple discharges during calendar year 2007 was included in this study as some women may have delivered more than one time or been admitted multiple times with a diagnosis of preeclampsia in 2007.

Definition of preeclampsia and eclampsia.

Five-digit ICD-9-CM codes beginning with 642.4, 642.5, 642.6, or 642.7 were used to identify women carrying the diagnosis of preeclampsia and eclampsia. If any of the 25 discharge diagnosis fields contained these codes then the patient was considered to be pre-eclamptic/eclamptic.

Statistical analysis.

SAS software version 9.2 (SAS Institute, Cary, NC) was used to manage and analyze the data. This study assumed that the month of admission equaled the month of delivery. The monthly prevalence of preeclampsia was calculated by dividing the monthly number of women discharged with preeclampsia by the monthly number of deliveries. The trend in the monthly prevalence of preeclampsia was assessed for statistical significance using a generalization of the Poisson model known as the negative binomial model. The latter model was preferred over the former in order to avoid the problem of overdispersion, which commonly occurs with Poisson regression models. The GENMOD procedure was used to perform the negative binomial regression.9 The month of delivery was treated as a continuous variable ranging from 1 for January to 12 for December. The natural logarithm of the total number of monthly deliveries was used as the offset variable.9,10 The change in the monthly prevalence of preeclampsia for each one month increase was calculated by subtracting the null value of 1 from the relative rate for the month term and multiplying the result by 100%. This figure was reported along with the P value associated with the month relative rate.

A cross-sectional prevalence study was conducted after the descriptive analysis with the exposure variable being season of delivery: winter (December, January, February), spring (March, April, May), summer (June, July, August), fall (September, October, November). The association between season and several potential confounders was examined using contingency tables and χ2P values. Logistic regression models were used to calculate unadjusted and adjusted prevalence ORs of preeclampsia/eclampsia. Season was entered in the models using three dummy (indicator) variables with winter as the referent group. ORs were adjusted for maternal age (eight age groups), maternal race-ethnicity, health insurance status, obesity, and diabetes.

Two variables found in the dataset (race and Hispanic ethnicity) were used to create the following maternal race-ethnicity variable: white Hispanic, black non-Hispanic, Other race-ethnicity (Native American, Asian, Black Hispanic, and Other race), and white non-Hispanic (referent). Obesity (unspecified or morbidly obese) was defined as an ICD-9-CM code of 278.00 or 278.01. Diabetes and gestational diabetes were included in the same category in this study, and their ICD-9-CM codes are 250.00 and 648.80, respectively. A variable that indicated the primary source of payment was used to create the health insurance variable. Over 20 levels were noted in the original payment variable. These were collapsed into a three-level variable and entered in the regression model using two dummy variables: Medicaid, Self-pay/indigent/charity/unknown, and other (includes Blue Cross/Blue Shield, Medicare, etc.). The referent group was the other category.

Finally, the association between the season of delivery and preeclampsia was quantified after stratifying by the maternal race-ethnicity variable. Within each race-ethnicity stratum, season ORs were adjusted for diabetes, age, obesity, and health insurance status.

Initially, a sample of 313,224 records of women 12 through 50 years of age was identified. After deleting the records of patients who had missing or invalid values for the ethnicity, race, and health insurance variables, 312,207 records were available for analysis.

Results

The clinical and demographic features of the study sample are reported in Table 1 (N = 312,207). The highest number of deliveries (N = 80,180) occurred in the summer. Approximately 6% of the sample was diabetic. The prevalence of preeclampsia/eclampsia varied slightly by the season of delivery and was found to be highest in the winter months (4.1%) and lowest in the fall (3.89%). The monthly prevalence of preeclampsia per 100 deliveries is shown in Figure 1. The highest prevalence occurred in January (4.4%) with a nadir in December (3.5%). The downward trend was statistically significant (Figure 1): for each month that passed from January onwards, there was a 1.3% decrease in the monthly prevalence (P < 0.0001 from a negative binomial model).

Table 1

Clinical and demographic features of 312,207 women who delivered in Texas in a reporting hospital and were discharged in 2007

Monthly prevalence of preeclampsia/eclampsia per 100 deliveries in 312,207 women who were admitted for delivery and discharged in 2007 in Texas. Results are restricted to women 12–50 years of age. The downward trend was statistically significant (P < 0.0001 from a negative binomial model).

Unadjusted and adjusted prevalence ORs are presented in Table 2. Women who delivered in the fall were 6% less likely to have preeclampsia/eclampsia than those who delivered in the winter (P = 0.02) after adjusting for diabetes, maternal age, obesity, race, and health insurance status. Statistically significant seasonal trends were not observed after stratifying by maternal race-ethnicity (Table 3).

Table 3

Adjusted prevalence odds ratios (OR) and 95% confidence intervals (CI) for preeclampsia/eclampsia in 312,207 women who delivered in 2007 throughout Texas stratified by maternal race/ethnicity

Table 2

Unadjusted and adjusted prevalence odds ratios for preeclampsia/eclampsia in 312,207 women who delivered in 2007 throughout Texas

Discussion

Our analysis of a statewide hospital database found a weak protective association between delivering in the fall (versus winter) and the prevalence of preeclampsia. There is some difficulty in comparing the results of our study with those of previous investigators due to the varying climates of the study sites and also the choice of the exposure variable: season of conception or season of delivery. Nonetheless, a study of all deliveries that occurred in Norway in the years 1967–1998 (N = 1,869,388) found that the monthly prevalence of preeclampsia was lowest in women who delivered in August and highest in December.11 This relationship persisted after adjustment for parity, maternal age, and other factors (adjusted OR for December vs. August delivery = 1.26, 95% CI: 1.20–1.31). We also noted that the monthly prevalence of preeclampsia in Texas peaked in January. An analysis of data from the Collaborative Perinatal Project, a study of women seen at 12 US hospitals, reported that the prevalence of preeclampsia was highest in January deliveries and lowest in September deliveries.12

A variety of factors including ambient temperature, the number of daylight hours, dietary variations related to the seasonal food supply, infections, and changes in plasma volume due to changes in the weather have been proposed as mechanisms for explaining the seasonal variation in the frequency of preeclampsia/eclampsia.5,7–8,13 In a bovine model, heat shock was found to compromise the development of preimplantation embryos.14 It has been postulated that cold weather could lead to the type of vasospasm that constitutes a portion of the pathogenesis of preeclampsia.11

There is evidence that maternal race may modify the monthly variation in the frequency of preeclampsia. Bodar and colleagues conducted a cohort study of 20,794 white women and 18,916 black women.12 Among white women, a U-shaped pattern was noted with the highest incidence of preeclampsia found in both the winter months and the fall months with a nadir in mid-August (P < 0.05). In contrast, there was no association between the month of delivery and the risk of preeclampsia in black women (P = 0.81). We did not detect statistically-significant seasonal trends after stratifying by maternal race-ethnicity.

A strength of our study is its use of a state database. Rather than analyzing data from a single institution or a limited geographical area, we included deliveries to women who were discharged from reporting facilities located throughout Texas, a state that is large in both area (261,797 square miles) and population (estimated population in 2009: 24,782,302).15 Nonetheless, we only had access to one-year's worth of state hospital data containing the critical date of admission variable. Future similar studies should include multiple years of discharge data so as to distinguish a possible cyclical/seasonal trend from a long-term (secular) trend. For example, our analysis by month, rather than season, depicted in Figure 1 indicated that the highest prevalence of preeclampsia was in January while the lowest prevalence was in the month of December. This result could be best explored with several years' worth of state hospital data. If monthly data for subsequent years revealed a similar trend (peak frequency in January declining throughout the year with a sharp rise once again in January) then further investigations would be warranted to determine if this pattern was due to a reporting artifact.

Access to data covering several years would also allow one to stratify the analysis by region thereby separating, for example, the humid subtropical climate of the Houston area in east Texas from the dry, high-elevation desert of El Paso in far west Texas. Texas is found in the Temperate Zone of the Northern Hemisphere and contains both warm and cool regions spanning three major climatic types with no distinct boundaries dividing these climate types.16

Hospital discharge datasets may suffer from miscoding. For example, misclassification of the discharge diagnosis of preeclampsia is a possibility in our analysis. However, a study conducted by Yasmeen et al. found that the ICD-9-CM coding of preeclampsia (any type) in hospital discharge data was accurate: sensitivity = 88%, positive predictive value = 91%.17

A limitation of our analysis was its inability to control for parity. A limited number of ICD-9-CM V codes are available to identify primigravidas experiencing a high-risk pregnancy; however, these codes only identify a first pregnancy in women who are younger than 16 years of age or ≥35 years of age leaving a wide range of maternal age without codes. However, it is unlikely that parity affected our results. In order for a variable to be a confounder it must be both related to the exposure variable and independent of the exposure variable it must be a risk factor for the disease.18,19 While nulliparity is strongly correlated with our outcome of preeclampsia, it was most likely not related to the exposure variable which was season of birth. In other words, one would not expect to see a higher proportion of nulliparous women delivering in a particular season—this percentage should be constant throughout the year.

The low prevalence of obesity in our sample merits discussion. In each season, less than 3% of the patients had an ICD-9-CM code for obesity. In contrast, 28.5% of adult females in Texas in 2007 were classified as obese (defined as a body mass index of 30 or more) by the Centers for Disease Control and Prevention's Behavioral Risk Factor Surveillance System.20 In light of this information, one can conclude that either providers are failing to note a patient's obesity in her medical record and/or the ICD-9-CM codes for obesity are underutilized by hospital coders. However, residual confounding by this variable in our analyses is unlikely. As stated above, in order for a variable to be a confounder, it must be both related to the exposure variable, and independent of the exposure variable, it must be a risk factor for the disease. While obesity is linked to preeclampsia its frequency did not vary appreciably by the season of delivery and hence it is not associated with season of delivery and therefore cannot be a confounder.

In summary, this study found that delivering in the fall compared to delivering in the winter was associated with a small reduced odds of preeclampsia. This finding is consistent with the results of two previous investigations which reported a lower prevalence of preeclampsia in late summer/early fall and a peak prevalence in either December or January.11,12 Future epidemiologic studies of seasonal trends in preeclampsia should strive to capture detailed maternal environmental exposure data such as the time spent outdoors and related factors such as maternal serum vitamin D, low levels of which have been linked to severe preeclampsia.21,22 The possible role of vitamin D in the development of preeclampsia is an emerging area of investigation. The pathogenesis of preeclampsia involves a number of biological pathways that may be affected by vitamin D including immune dysfunction, placental implantation, and aberrant angiogenesis.21

Disclosure

The authors declared no conflict of interest.

References

1.
Wagner
LK
.
Diagnosis and management of pre-eclampsia
.
Am Fam Physician
 
2004
;
70
:
2317
2324
.
2.
ACOG Committee on Practice Bulletins--Obstetrics
.
ACOG practice bulletin. Diagnosis and management of preeclampsia and eclampsia. Number 33, January 2002
.
Obstet Gynecol
 
2002
;
99
:
159
167
.
3.
Rustveld
LO
Kelsey
SF
Sharma
R
.
Association between maternal infections and pre-eclampsia: a systematic review of epidemiologic studies
.
Matern Child Health J
 
2008
;
12
:
223
242
.
4.
Ness
RB
Zhang
J
Bass
D
,
Klebanoff
MA
.
Interactions between smoking and weight in pregnancies complicated by pre-eclampsia and small-for-gestational-age birth
.
Am J Epidemiol
 
2008
;
168
:
427
433
.
5.
Immink
A
Scherjon
S
,
Wolterbeek
R
,
Steyn
DW
.
Seasonal influence on the admittance of pre-eclampsia patients in Tygerberg Hospital
.
Acta Obstet Gynecol Scand
 
2008
;
87
:
36
42
.
6.
Subramaniam
V
.
Seasonal variation in the incidence of pre-eclampsia and eclampsia in tropical climatic conditions
.
BMC Womens Health
 
2007
;
7
:
18
.
7.
Tam
WH
,
Sahota
DS
,
Lau
TK
,
Li
CY
,
Fung
TY
.
Seasonal variation in pre-eclamptic rate and its association with the ambient temperature and humidity in early pregnancy
.
Gynecol Obstet Invest
 
2008
;
66
:
22
26
.
8.
Phillips
JK
,
Bernstein
IM
,
Mongeon
JA
,
Badger
GJ
.
Seasonal variation in pre-eclampsia based on timing of conception
.
Obstet Gynecol
 
2004
;
104
:
1015
1020
.
9.
Allison
PD
.
Poisson regression
. In:
Logistic Regression Using the SAS system: Theory and Application
 .
SAS Institute
:
Cary, NC
,
1999
, pp
217
231
.
10.
Szklo
M
,
Nieto
FJ
.
Epidemiology Beyond the Basics
 
2000
.
Aspen: Gaithersburg
,
MD
, pp
310
311
.
11.
Magnus
P
,
Eskild
A
.
Seasonal variation in the occurrence of pre-eclampsia
.
BJOG
 
2001
;
108
:
1116
1119
.
12.
Bodnar
LM
,
Catov
JM
,
Roberts
JM
.
Racial/ethnic differences in the monthly variation of pre-eclampsia incidence
.
Am J Obstet Gynecol
 
2007
;
196
:
324.e1
324.e5
.
13.
Algert
CS
,
Roberts
CL
,
Shand
AW
,
Morris
JM
,
Ford
JB
.
Seasonal variation in pregnancy hypertension is correlated with sunlight intensity
.
Am J Obstet Gynecol
 
2010
;
203
:
215.e1
5
.
14.
Krininger
CE
3rd
,
Stephens
SH
,
Hansen
PJ
.
Developmental changes in inhibitory effects of arsenic and heat shock on growth of pre-implantation bovine embryos
.
Mol Reprod Dev
 
2002
;
63
:
335
340
.
15.
US Census Bureau
.
Texas QuickFacts from the US Census Bureau
. <http://quickfacts.census.gov/qfd/states/48000.html> Accessed 4 August 2010.
16.
Office of the Texas State Climatologist. The Climate of Texas
. <http://atmo.tamu.edu/osc/TXclimat.html>. Accessed 20 June 2011.
17.
Yasmeen
S
,
Romano
PS
,
Schembri
ME
,
Keyzer
JM
,
Gilbert
WM
.
Accuracy of obstetric diagnoses and procedures in hospital discharge data
.
Am J Obstet Gynecol
 
2006
;
194
:
992
1001
.
18.
Hennekens
CH
,
Buring
JE
.
Epidemiology in Medicine
 
1987
.
Little, Brown
,
Boston, MA
.
19.
Rothman
KJ
,
Greenland
S
,
Lash
TL
.
Modern Epidemiology
 
2008
,
3rd edn
.
Lippincott Williams & Wilkins
.
20.
Centers for Disease Control and Prevention
.
Behavioral Risk Factor Surveillance System. Prevalence and Trends Data Texas – 2007 Overweight and Obesity (BMI).
<http://apps.nccd.cdc.gov/BRFSS/sex.asp?cat=OB&yr=2007&qkey=4409&state=TX>. Accessed 16 June 2011.
21.
Bodnar
LM
,
Catov
JM
,
Simhan
HN
,
Holick
MF
,
Powers
RW
,
Roberts
JM
.
Maternal vitamin D deficiency increases the risk of preeclampsia
.
J Clin Endocrinol Metab
 
2007
;
92
:
3517
3522
.
22.
Baker
AM
,
Haeri
S
,
Camargo
CA
Jr
,
Espinola
JA
,
Stuebe
AM
.
A nested case-control study of midgestation vitamin D deficiency and risk of severe preeclampsia
.
J Clin Endocrinol Metab
 
2010
;
95
:
5105
5109
.