Integrating social and cardiometabolic risk factors in predicting cardiometabolic multimorbidity: a Bayesian model development and internal validation in Swedish middle-aged adults.
A Bayesian risk prediction model incorporating social determinants showed fair discrimination for any cardiometabolic disease (AUC 0.76) and good discrimination for cardiometabolic multimorbidity (AUC 0.89) in middle-aged Swedish adults, suggesting that social determinants and uncertainty estimates may provide useful information for cardiometabolic disease prediction.
Key Findings
Results
During follow-up, 6.0% of participants developed cardiometabolic disease(s), with 5.7% developing a single condition and 0.3% developing multimorbidity.
The study cohort included 11,964 adults without prior cardiometabolic disease from the Swedish CArdioPulmonary bioImage Study (SCAPIS)
Participants were enrolled between 2013 and 2016 and were middle-aged adults
Cardiometabolic disease was defined as incident type 2 diabetes (T2D), ischaemic heart disease/heart failure, or stroke
Follow-up period was 2-5 years
The low rate of multimorbidity (0.3%) reflects the relatively short follow-up window and middle-aged population
Results
The Bayesian prediction model showed fair discrimination for any cardiometabolic disease and good discrimination for cardiometabolic multimorbidity.
AUC for any cardiometabolic disease (≥1 condition) was 0.76 (95% CI 0.74-0.78)
AUC for cardiometabolic multimorbidity (≥2 conditions) was 0.89 (95% CI 0.86-0.93)
Regularised Bayesian logistic regression models were used for model development
Twenty-two social and cardiometabolic risk predictors were considered in model development
Models underwent internal validation
Results
The highest predicted risks were observed among older foreign-born males with hypertension, low HDL-C, high waist circumference, and current/ex-smoker status.
This highest-risk profile had a predicted risk of 34.7% (95% CrI 30.3%-39.5%) for any cardiometabolic disease
The same profile had a predicted risk of 4.4% (95% CrI 2.0%-7.9%) for cardiometabolic multimorbidity
The risk profile combined both social determinants (foreign-born status) and cardiometabolic factors (hypertension, low HDL-C, high waist circumference)
Smoking status (current/ex-smoker) was also included as a key risk factor in this highest-risk profile
Background
Social determinants were incorporated alongside traditional cardiometabolic risk factors in the prediction model, distinguishing it from existing single-disease prediction models.
Current risk prediction models typically focus on a single disease and rarely incorporate social determinants
Twenty-two predictors encompassing both social and cardiometabolic domains were considered
Foreign-born status was identified as a relevant social predictor, appearing in the highest-risk profile
The Bayesian framework provided uncertainty estimates (credible intervals) alongside point predictions, which is noted as a useful feature
Methods
The study used a prospective cohort design with a population free of prior cardiometabolic disease at baseline.
The cohort was drawn from the Swedish CArdioPulmonary bioImage Study (SCAPIS)
Enrolment occurred between 2013 and 2016
The sample included 11,964 adults without prior cardiometabolic disease
Participants were described as middle-aged Swedish adults
The model development included internal validation, with external validation noted as needed before clinical use
Conclusions
The authors concluded that further external validation is needed before the model can be used in clinical practice.
The paper states 'further validation is needed before use in clinical practice'
Only internal validation was conducted in this study
The model was developed and validated within the SCAPIS cohort only
The findings are described as suggesting that social determinants and uncertainty estimates 'may provide useful information' rather than confirming clinical utility
What This Means
This research suggests that it is possible to predict whether middle-aged adults will develop cardiometabolic diseases — such as type 2 diabetes, heart disease, heart failure, or stroke — using a statistical model that combines both traditional medical risk factors and social factors. The study followed nearly 12,000 Swedish adults who were free of these diseases at the start, and found that about 6% developed at least one cardiometabolic condition within 2 to 5 years, while a smaller group (0.3%) developed two or more conditions (called multimorbidity). The model performed reasonably well, with better accuracy for predicting multimorbidity (two or more conditions) than for predicting any single condition.
One distinctive feature of this study is that it included social factors — such as whether someone was born outside Sweden — alongside standard medical risk factors like high blood pressure, low good cholesterol (HDL-C), large waist circumference, and smoking. The combination of these factors identified a group at particularly high risk: older foreign-born men with several cardiometabolic risk factors, who had a predicted risk of about 35% for developing any cardiometabolic disease. The Bayesian statistical approach used also provided uncertainty ranges around predictions, giving a sense of how confident the estimates are.
This research suggests that including social determinants of health in risk prediction models may improve their usefulness, particularly for identifying people at risk of developing multiple conditions simultaneously. However, the model was only tested within the same Swedish dataset it was built from, so further testing in other populations is needed before it could be used in real clinical settings to guide healthcare decisions.
Anindya K, Merlo J, Lind L, Jernberg T, Weinehall L, Rosvall M, et al.. (2026). Integrating social and cardiometabolic risk factors in predicting cardiometabolic multimorbidity: a Bayesian model development and internal validation in Swedish middle-aged adults.. Scientific reports. https://doi.org/10.1038/s41598-026-67901-3