Improving the Algorithm: The Inclusion of a Socioeconomic Status Measure in Predicting Type 2 Diabetes Development in...

Area of Science:

• Endocrinology
• Public Health
• Data Science

Background:

• Prediabetes is a growing concern in youth, increasing their risk for type 2 diabetes.
• Socioeconomic factors, such as area deprivation, may influence diabetes development.
• Machine learning (ML) models offer potential for early risk prediction.

Purpose of the Study:

• To assess if the Area Deprivation Index (ADI) improves ML model predictions for type 2 diabetes development in adolescents with prediabetes.
• To compare the predictive performance of ML models with and without ADI data.

Main Methods:

• Supervised ML models were developed using 665 patient encounters from electronic health records.
• Models predicted type 2 diabetes development within one year of prediabetes diagnosis.
• The Area Deprivation Index (ADI) was calculated using census block data; logistic regression models with HbA1c alone and HbA1c + ADI were compared.

Main Results:

• 27.2% of patients (181 encounters) progressed to type 2 diabetes.
• The model incorporating HbA1c and ADI achieved a higher area under the receiver operating characteristic curve (0.73) compared to the HbA1c-only model (0.68).

Conclusions:

• The Area Deprivation Index (ADI) significantly enhanced the predictive performance of ML models for youth-onset type 2 diabetes.
• Integrating socioeconomic data like ADI into ML models is crucial for accurate risk stratification in prediabetic youth.