The study cohort was identified using a dataset (the Cesarean registry) sourced from a previous multicenter study by the National Institute of Child Health and Human Development Maternal-Fetal Medicine Units (MFMU) Network.
9 (link) Briefly, between 1999 and 2002, data were collected in women who underwent primary cesarean delivery, repeat cesarean delivery, or trial of labor after cesarean, and who delivered infants ≥ 20 weeks or ≥ 500 g at 19 academic centers. The final 2 years of the study included only women who underwent repeat cesarean delivery or vaginal birth after cesarean. Data were collected through detailed chart review at delivery, and information regarding perioperative morbidity was collected from discharge summaries. Patients and hospitals were deidentified by the MFMU.
For the current study, we included only women from the Cesarean registry who had undergone a primary or repeat cesarean delivery. Based on a definition described by Kuklina et al,
2 (link) we defined a prolonged postpartum LOS as a postpartum hospitalization (number of hospital days between cesarean delivery to hospital discharge) with a postpartum LOS ≥ 90th centile. In the Cesarean registry dataset, all dates were expressed as day numbers. Maternal demographic, antepartum, perioperative, and neonatal variables were compared between women with and without prolonged LOS. Maternal characteristics included: age, race/ethnicity, body mass index (BMI) at delivery, diabetes, chronic hypertension, and number of prior cesarean deliveries. Obstetric variables included: gestational age at delivery, pregnancy-associated hypertension, labor, or induction before the cesarean delivery, and chorioamnionitis. Perioperative variables included: type of uterine incision, mode of anesthesia, intraoperative red blood cell transfusion, hysterectomy, and postpartum complications, such as endometritis and wound complications. We also compared birth weights of neonates of women with versus without prolonged LOS.
We performed bivariate analyses to compare maternal, obstetric, and neonatal characteristics between women with and without prolonged LOS after cesarean delivery. Categorical variables were compared using the chi-square test; bivariate analyses did not account for missing data. Based on bivariate analyses, unadjusted odds ratios (ORs) and accompanying 95% confidence intervals (CIs) were calculated. Candidate variables that were associated with prolonged LOS on bivariate analyses (
p ≤ 0.1) were included as covariates in an unconditional multivariable logistic regression model. The final model was determined using a traditional backward elimination, with all variables initially included and then selectively removed if not significant (
p < 0.05). To determine the presence of collinearity between independent variables, variance inflation factor (VIF) testing was performed. Collinearity was determined to be insignificant as VIF scores ranged from 1.01 to 1.49 with a mean VIF score of 1.18. Population attributable fractions (PAFs) were used to calculate the proportional reduction in risk of prolonged LOS that would occur by eliminating the exposure of interest from the population while the distribution of other risk factors remained unchanged. PAFs were calculated for selected risk factors that were considered modifiable by using adjusted ORs (aORs) from the final multivariate model.
10 (link) We calculated the area under the receiver-operating characteristic curve (AUROC) using standard methodology to assess the predictive performance of the final model.
Based on data for the date of hospital admission and discharge, we performed a secondary analysis to assess risk factors for prolonged total length of hospital stay, defined as the interval from admission to discharge. For the total period of hospital stay we defined a prolonged delivery hospitalization as a total hospital LOS ≥ 90th centile. We did not count the day of admission in the calculation for the total hospital LOS.
Data analyses were performed using SAS 9.2 (SAS Inc., Cary, NC) and STATA version 12 (Statacorp, College Station, TX). As the Cesarean registry contains deidentified data, our study was deemed institutional review board exempt by the Stanford Institutional Review Board.
Blumenfeld Y.J., El-Sayed Y.Y., Lyell D.J., Nelson L.M, & Butwick A.J. (2015). Risk Factors for Prolonged Postpartum Length of Stay Following Cesarean Delivery. American journal of perinatology, 32(9), 825-832.