Gestational Diabetes Mellitus - Neonatal and Maternal Outcomes in Patients Treated with Insulin or Diet: A Propensity Matched Analysis
Description
Pregnant women worldwide face the risk of developing gestational diabetes mellitus (GDM), if left untreated, can cause complications. The study explores factors influencing the choice between diet control and insulin therapy for pregnant women with GDM. It aims to understand how these choices impact maternal and neonatal outcomes. In this quasi-experimental study, clinicians determined treatment (diet control or insulin) for 1030 patients with GDM at a private practice from 2010 to 2020 based on baseline characteristics. Propensity scores (PS), reflecting the probability of treatment allocation, were derived through multiple logistic regression.
Files
Steps to reproduce
The analysis data set involved only those subjects who continued with the same treatment modality throughout the gestational period. Nevertheless, there could be some systematic differences in the baseline characteristics of patients that could affect the treatment allocation by clinicians and, thereby, the eventual outcome. To account for this, propensity score analysis (PSA) was performed. A propensity score was obtained for each patient, which is the probability of receiving a specific treatment conditional to baseline covariates. PSA has the underlying assumption that two patients with similar scores have baseline covariates from a similar distribution. Multiple logistic regression was used to obtain the propensity score, which indicates the probability of specific treatment allocation to a patient for a given array of baseline covariate values. In the present study, the scores were obtained, indicating the likelihood of insulin treatment given the baseline parametric profile of the patient. The scores were used to match patients from diet and insulin-treated groups using the optimal algorithm to correct the allocation bias. Compared to alternatives like the nearest neighbor and genetic matching, the algorithm resulted in a maximum number of matched patients between two groups.