This summary contains details. It was published in medRxiv.org as the form of a preprint.
Key Takeaways
-
An external validation study provided recommendations for necessary changes to an earlier model to predict a pregnant woman’s risk for developing gestational diabetes mellitus (GDM) at an early stage of pregnancy in a variety of settings, as well as updated ethnicity classifications.
-
The revision applied best practices using large datasets for validation and updating drawn from a diverse ethnic population with GDM diagnosis based on current criteria and a universal screening strategy and a GDM prevalence of 18.0 percent.
-
This report shows the advantages of starting with an validated model and then updating it to maintain predictive performance as time passes, rather than starting from scratch.
Why Matters
-
There is growing evidence for the integration of the GDM prediction model to routine practice in order to improve and speed up the treatment of women who are at high risk for GDM.
-
Both the original and revised models employ variables that are routinely collected in clinical practice. This eliminates the necessity to collect additional data and the associated costs.
Study Design
-
The researchers used routinely collected health information for 26,474 singleton pregnancies resulting in a birth between January 2016 to December 2018 at Monash Health, Australia’s largest health service that includes three maternity hospitals and caters to an ethnically diverse population.
-
They utilized the criteria of the International Association of Diabetes and Pregnancy study groups to establish the diagnosis of GDM.
-
The investigators modified the ethnicity classification system to reflect international ethnicity categories as well as self-reported ethnicity designations. The original model, however, relied on extrapolating ethnicity to a country of birth.
-
They selected the best model using the measures of predictive performance and a closed testing procedure. They used C-statistics (area under the curve of operating characteristic of the receiver) to assess and compare the models that they created.
Key Results
-
The model that was originally developed produced a C-statistic of 0.698 which indicates “reasonable” discrimination.
-
Model C2 was identified as the most popular model due to its similar calibration plot in the high prevalence region, a superior C-statistic of 0.732 and its use of more generalizable ethnicity categories, and because it showed an improved fit during closed testing.
-
The new model uses the variables of age and body mass index family history of diabetes, history of GDM and poor obstetric outcomes, and ethnicity.
Limitations
-
The revised model includes the continuous variables of body mass index as categorical variables. This method could be negative to predictive power and could be replaced by electronic risk calculators. Reestimating the relationship between body mass index and age as continuous variables and determining the presence of GDM would produce a completely new model for prediction, going beyond the scope of validation and updating.
Disclosures
-
The preprint does not include any information on disclosures by authors or sources of funding.
-
A previously published description of the design of the study revealed that it did not receive commercial funding and that none of the authors had any commercial disclosures.
This is a brief summary of a preprint research paper “External validation and updating of a prediction model for diagnosing gestational diabetes mellitus” Written by researchers who are mostly at Monash University, Clayton in Australia. The study was based on medRxiv, which was made available by Medscape. This study is not peer-reviewed. The full text of the study is available on medRxiv.org.
Mitchel L. Zoler is a reporter for Medscape and MDedge based in the Philadelphia area. @mitchelzoler
For more diabetes and endocrinology news Follow us onTwitter and Facebook.Facebook.
Follow Medscape on Medscape onFacebook Follow Medscape on Facebook,Twitter ,Instagram , andYouTube .
