Associations Between Hyperthyroidism and Adverse Obstetric and Neonatal Outcomes

A Study of a Population Database Including Almost 17,000 Women With Hyperthyroidism

Ranit Hizkiyahu; Ahmad Badeghiesh; Haitham Baghlaf; Michael H. Dahan


Clin Endocrinol. 2022;97(3):347-354. 

In This Article

Materials and Methods

We conducted a retrospective population-based study utilizing data from the Healthcare Cost and Utilization Project-Nationwide Inpatient Sample (HCUP-NIS) over 11 years from 2004 to 2014, inclusively. The HCUP-NIS is the largest inpatient sample database in the United States and comprised hospital inpatient stays submitted by hospitals in 48 states and the District of Columbia. This database is accessible online for the public (see "Data Availability Statement"). Each year, the database provides information relating to seven million inpatient stays, including patient characteristics, diagnoses, and procedures. The data are representative of ~20% of admissions to American hospitals and geographically represents over 96% of the population in the United States.

We evaluated deliveries by using the international classification of diseases, ninth edition, clinical modification codes for delivery-related discharge diagnoses (650.xx, 677.xx, 651.xx–676.xx, where the fifth digit is 0, 1, or 2) and birth-related procedural diagnosis (72.x, 73.x, 74.0–74.2). We limited our study group to the admissions that ended with delivery of fetuses at least 20 weeks of gestational age or maternal death, to guarantee that multiple admissions in the same pregnancy will be excluded. It should be noted that the database does not include any miscarriages or terminations of pregnancy. Within this group, all women with the diagnosis of hyperthyroidism were identified using ICD-9 diagnostic codes 242.0x, 242.1x, 242.2x, 242.3x, 242.4x, 242.8x, and 242.9x, and the remaining deliveries were categorized as nonhyperthyroidism births and comprised the reference group. The reference group may have suffered from other illnesses during pregnancy; therefore, not all the potential prepregnancy diseases with confounding effect have been controlled for in the groups.

Baseline clinical characteristics included the following: patient age, race, income, insurance type, previous cesarean section, multiple gestation, smoking history, obesity (body mass index ≥ 30 kg/m2), preexisting hypertension, and preexisting diabetes mellitus. Pregnancy outcomes include hypertensive disorders of pregnancy (gestational hypertension and preeclampsia/eclampsia), gestational diabetes mellitus, placenta previa, and venous thromboembolism (deep vein thrombosis and pulmonary embolism). Delivery outcomes include preterm premature rupture of the membranes, preterm delivery, placental abruption, cesarean section, maternal infection and chorioamnionitis, and hysterectomy. Postpartum hemorrhage, disseminated intravascular coagulation, and blood transfusions were also investigated. Neonatal outcomes included small for gestational age infants, Intrauterine fetal death, and congenital anomalies.

Statistical Analysis

An initial analysis was performed to identify the prevalence of pregnant women with hyperthyroidism over the entire duration of the study. We then compared baseline clinical and demographic characteristics between women with hyperthyroidism to those without hyperthyroidism, using χ 2 tests. If cases were five or less, Fisher's exact test was used and the low number of cases was made explicitly clear. Subsequently, logistic regression analyses were conducted to explore associations between hyperthyroidism and obstetrical and neonatal outcomes through the estimation of odds ratio (OR) and 95% confidence intervals (CIs). The regression models were adjusted for the potential confounding effects of maternal demographic, preexisting clinical characteristics, and concurrently occurring characteristics to generate adjusted CIs. A confounder was determined if it was part of the demographics or prepregnancy diseases and rates between those with hyperthyroidism and the controls gave a p < .05. All analyses were performed using SPSS 23.0 (IBM Corporation) software for all analyses.

This study used exclusively publicly accessible, anonymized data; therefore, according to Tri-Council Policy Statement,[17] institutional review board approval was not required.