We performed a retrospective analysis using insurance claims data for 2002-6. We identified claims for pre-eclampsia and eclampsia, chronic hypertension complicating pregnancy, pre-eclampsia or eclampsia superimposed on hypertension, and gestational hypertension diagnoses in a sample of women who had at least one delivery and had undergone bariatric surgery. We examined the proportion of hypertensive disorders in pregnancies after surgery compared with before surgery. We used women who underwent surgery after delivery as the comparison group because all patients undergoing bariatric surgery have met the criteria for either class II obesity with high risk comorbid conditions or class III obesity, and the dataset did not include data on BMI, which could have allowed us to identify a different comparison group of obese women.
We accessed claims data from seven BlueCross BlueShield health plans, providing coverage in seven states: Tennessee, Western Pennsylvania, Michigan, North Carolina, the city of Philadelphia in Pennsylvania, South Dakota and Iowa (included in same plan), and Hawaii. The dataset included insurance claims information, as well as variables for age, but did not consistently include height or weight to calculate BMI, or information on gravidity or parity. Inclusion in the dataset required at least one of the following criteria at any point during 2002-6: completed health risk assessment or other survey; claim for a diagnosis of obesity; claim for bariatric surgery (see appendix A on bmj.com); prescription claim for a weight reduction medication; or diagnosis code of hyperlipidaemia, type 2 diabetes, sleep apnoea, gallbladder disease, or metabolic syndrome. These diagnoses were identified in the claims by common procedural terminology codes, ICD-9-CM (international classification of disease, ninth revision, clinical modification codes), national drug codes, or diagnosis related group codes. We acquired data on enrolment files for administrative data; benefits information to determine medical coverage; and inpatient, outpatient, and pharmacy claims records containing ICD-9 diagnosis, common procedural terminology codes, prescription national drug codes, and costs and charges (submitted, allowed, and paid).
Selection of study sample
We selected women who had codes within the coverage period for both bariatric surgery (see appendix A on bmj.com) and a delivery, defined as one or more live births or a pregnancy loss after 25 weeks’ gestation (see appendix B on bmj.com), with an inpatient point of service code.
For women with more than one delivery date, we calculated the time between dates to try to determine if these were claims from the same pregnancy or if the woman had more than one pregnancy during 2002-6. For each possible pregnancy, we selected the actual delivery date as the date of service with the most delivery records/claims.
For each woman, the delivery date(s) was compared with the date of the claim for bariatric surgery. We excluded implausible delivery dates: either less than 280 days after bariatric surgery or less than 31 days before surgery. Women were classified into groups who delivered before or after surgery based on date of delivery in relation to surgery. If a woman had two or more deliveries, or had a delivery both before and after surgery, we selected the delivery closest to her bariatric surgery and classified the woman on the basis of this delivery.
We restricted our analyses to women aged 16-45 at the time of delivery who had continuous insurance coverage during pregnancy (calculated as delivery date minus 40 weeks) plus two weeks after delivery to capture the diagnosis codes for hypertensive disorders in pregnancy, the outcomes of interest. In a separate subgroup analysis of women who had deliveries both before and after surgery we did not apply the requirement of continuous insurance coverage but the rest of the inclusion criteria were the same.
Definition of outcomes of hypertensive disorders in pregnancy
Our outcomes of interest were diagnoses of hypertensive disorders in pregnancy during the pregnancy (40 weeks before delivery) plus the two weeks after delivery. We defined “pre-eclampsia” and as the presence of one or more ICD-9 codes for “mild” or “severe pre-eclampsia” (either 642.4x or 642.5x). “Eclampsia” was defined by its ICD-9 code (642.6x). We created mutually exclusive definitions for mild pre-eclampsia, severe pre-eclampsia, or eclampsia, which were classified by the presence of the more severe diagnosis code. We defined “chronic hypertension complicating pregnancy” as one or more ICD-9 codes for “benign essential hypertension complicating pregnancy” (642.0x) or “other pre-existing hypertension complicating pregnancy” (642.2x). We defined “gestational hypertension” as one or more ICD-9 codes for “transient hypertension in pregnancy” (642.3x) or “unspecified hypertension during pregnancy” (642.9x), as long as the woman did not also have chronic hypertension complicating pregnancy. The outcome of “pre-eclampsia or eclampsia superimposed on pre-existing hypertension” required the presence of its ICD-9 code (642.7x). In addition, women who met the definition for pre-eclampsia or eclampsia and then also had chronic hypertension complicating pregnancy or gestational hypertension were classified as “pre-eclampsia or eclampsia superimposed on pre-existing hypertension.” Women who met the definitions of any of the above hypertensive disorders of pregnancy were classified as “any hypertensive disorder in pregnancy.”
We used Flash Code software (version 2007 Q4, Medical Coding and Compliance Solutions (MCCS), LLC; the software division of PMIC, Turlock, CA) to search for and select the desired ICD-9 codes.
Covariates included a diagnosis of gestational diabetes (648.8x) or a diagnosis of pre-existing diabetes complicating a pregnancy (648.0x) during the 40 week pregnancy plus the two weeks after delivery. We identified multiple pregnancy using any ICD-9 code that referred to a pregnancy with more than one fetus. Type of bariatric surgery was assessed with common procedural terminology codes for surgery (see appendix A on bmj.com). We also included variables to represent the seven individual BlueCross BlueShield insurance plans (regions) as covariates.
We used descriptive statistics to compare the clinical characteristics of women who delivered before and after surgery. We used Student’s t tests for continuous variables and χ2 tests for categorical variables. We calculated the proportion of women with each of the hypertensive disorders in the two groups and tested for differences between groups using Fisher’s exact test. We used logistic regression to calculate the odds ratio and confidence intervals for each of the outcomes in unadjusted models and then in models adjusted for maternal age at delivery, multiple pregnancy, type of bariatric surgical procedure, pre-existing diabetes complicating pregnancy, and insurance plan.
We performed a subgroup analysis in 17 women who had deliveries both before and after surgery to compare the proportions of hypertensive disorders in pregnancy.
We conducted two additional analyses to evaluate for the possibility of selection bias. In addition to the covariates described above, we assessed the distribution, by group, of several conditions less likely to be affected by bariatric surgery or obesity. We also calculated a “risk of obesity score” in the year of bariatric surgery and the year of delivery to further evaluate selection bias at the time of surgery and delivery. These additional analyses are in appendix C on bmj.com.
P values less than 0.05 were considered significant. Statistical analyses were performed with SAS statistical software, version 9.1 (SAS, Cary, NC).