Fairness-Aware Graph Embeddings And Lightgbm-SHAP For Disease Prediction From Social Determinants Of Health
DOI:
https://doi.org/10.29284/ijasis.11.6s.2025.43-60Keywords:
Social Determinants of Health, LightGBM, SHAP, Fairness-Aware Machine Learning, Knowledge Graph, Disease Risk Prediction.Abstract
Social and Behavioral Determinants of Health (SBDH) affect long-term health risks but are rarely used as standalone predictors in disease modeling. This study investigates whether SBDH alone, without any clinical variables, can meaningfully predict disease vulnerability, using pressure ulcer (PU) development as a representative case. We constructed a fairness-aware prediction pipeline using data from the MIMIC-IV and MIMIC-SBDH datasets. We considered ten SBDH features and extracted them per patient, including housing stability, healthcare access, education, substance use, and social support etc. A knowledge graph (KG) was built to encode relationships among SBDH domains and between patients and their social profiles. Graph Convolutional Networks (GCNs) were trained with fairness constraints to produce de-biased patient embeddings. These embeddings, combined with the original SBDH features, were used to train a LightGBM classifier. Model interpretability was assessed using SHAP (SHapley Additive exPlanations), and fairness was evaluated via subgroup performance metrics. The model achieved strong discrimination (ROC AUC ≈ 0.72), good calibration, and equitable performance across subgroups. SHAP analysis identified housing instability, low social support, and poor healthcare access as the top contributors to risk. SBDH can serve as effective, fair, and interpretable predictors of disease risk. This approach enables upstream, community-level prevention, highlighting social context as a core component of equitable predictive healthcare.
References
[1] Busari, Muhammed, and Tunde Bolanle. "Integrating Social Determinants into Predictive Models for US Public Health Forecasting." ResearchGate, 2025. https://doi.org/10.13140/RG.2.2.23412.90244
[2] Adeniji, Olumide A., and Adeola Fagbemi. "Socio-Economic Inequalities in Stroke: A Global Perspective." IntechOpen, 2025. https://doi.org/10.5772/intechopen.1125587
[3] Nicholson, J., J. L. Paul, G. Piché, and A. Wittkowski. "Parents with Mental and/or Substance Use Disorders and Their Children, Vol III." Frontiers in Psychiatry, 2025. https://doi.org/10.3389/fpsyt.2025.1630833
[4] Xie, Z., G. Zhong, C. Xu, et al. "Trends and Cross-Country Inequalities of Alcohol Use Disorders: Findings from the Global Burden of Disease Study 2021." Globalization and Health, 2025. https://doi.org/10.1186/s12992-025-01124-5
[5] Zhou, Y., F. Chen, H. Li, and W. Wang. "Methodological Considerations in Assessing Modifiable Risk Factors Across Cancer Survivorship." Journal of Translational Medicine, 2025. https://doi.org/10.1186/s12967-025-06580-6
[6] Maganga, J. J., A. Katende, E. Luoga, et al. "Determinants of Physical Activity and Diet Consumption Among Adolescents Living with HIV in Southern Tanzania." HIV/AIDS—Research and Palliative Care, 2025. https://doi.org/10.2147/HIV.S519922
[7] Hamzens, M. F. "Culture and Non-Communicable Diseases: Determinants of Hypertension in Tenganan Pegringsingan Village, Karangasem, Bali." Society, 2025. https://doi.org/10.25077/society.11.1.1-10.2025
[8] Deku, J. G., F. A. Ofosua, G. E. Kpene, et al. "Prevalence and Risk Factors of Intestinal Parasitic Infections in Pregnancy: A Cross-Sectional Study in a Rural District in Ghana." BMC Pregnancy and Childbirth, 2025. https://doi.org/10.1186/s12884-025-07720-y
[9] Bergstrom, Nancy, Barbara J. Braden, Anne Laguzza, and Virginia Holman. The Braden Scale for Predicting Pressure Sore Risk. Nursing Research 36, no. 4 (1987): 205–210. https://doi.org/10.1097/00006199-198707000-00002
[10] Wright, Briana J., et al. "Health-related social needs and increased healthcare utilization among older adults." Journal of General Internal Medicine 38.4 (2023). https://doi.org/10.1007/s11606-022-07913-z
[11] Chen, X., Zhu, D., Li, C., Lin, Y., Lv, L., Ai, F., & Chen, Z. (2025). Modifiable factors affect cancer-specific survival: findings from a large population-based prospective cohort study. Journal of Translational Medicine, 23(1). https://doi.org/10.1186/s12967-025-06372-y
[12] Park, C., Kim, N., Won, C. W., & Kim, M. (2025). Predicting cognitive frailty in community-dwelling older adults: a machine learning approach based on multidomain risk factors. Scientific Reports, 15(1). https://doi.org/10.1038/s41598-025-00844-3
[13] Al-Zubaidy, Salih, et al. "The Impact of EHR-Incorporated Social Determinants on Risk Prediction Models." Frontiers in Digital Health 5 (2023). https://doi.org/10.3389/fdgth.2023.1135991
[14] Goldstein, Benjamin A., et al. "Opportunities and challenges in developing risk prediction models with electronic health records data." American Journal of Epidemiology 188.5 (2019): 870-875. https://doi.org/10.1093/aje/kwy292
[15] Cho, InSook, et al. "Comparison of prediction models for hospital readmission using structured data from EHR." International Journal of Medical Informatics 135 (2020): 104033. https://doi.org/10.1016/j.ijmedinf.2019.104033
[16] Obermeyer, Ziad, et al. "Dissecting racial bias in an algorithm used to manage the health of populations." Science 366.6464 (2019): 447–453. https://doi.org/10.1126/science.aax2342
[17] Mehrabi, Ninareh, et al. "A survey on bias and fairness in machine learning." ACM Computing Surveys 54.6 (2021): 1–35. https://doi.org/10.1145/3457607
[18] Shang, Jun, et al. "GAMENet: Graph augmented memory networks for recommending medication combination." AAAI (2019). https://doi.org/10.1609/aaai.v33i01.33012318
[19] Peng, Yuxin, et al. "Fairness-Aware Graph Neural Networks: A Survey." ACM Computing Surveys (2023). https://doi.org/10.1145/3589467
[20] Lyder, Courtney H., et al. "Hospital-acquired pressure ulcers: Results from the national Medicare Patient Safety Monitoring System study." Journal of the American Geriatrics Society 60.9 (2012): 1603–1608. https://doi.org/10.1111/j.1532-5415.2012.04106.x
[21] Cox, Jill. "Predictive power of the Braden Scale for pressure sore risk in adults: A meta-analysis." Decubitus 6.1 (1993): 24–30. https://doi.org/10.1097/01.NURSE.0000541266.72990.4e
[22] Sharp, Claire A., et al. "Social disparities in the burden of pressure ulcers: A systematic review." Journal of Tissue Viability 29.1 (2020): 10–20. https://doi.org/10.1016/j.jtv.2019.10.001
[23] Johnson, A., Bulgarelli, L., Pollard, T., Gow, B., Moody, B., Horng, S., Celi, L. A., & Mark, R. (2024). MIMIC-IV (version 3.1). PhysioNet. https://doi.org/10.13026/kpb9-mt58.
[24] Johnson, A.E.W., Bulgarelli, L., Shen, L. et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data 10, 1 (2023). https://doi.org/10.1038/s41597-022-01899-x
[25] Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., ... & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220.
[26] Ahsan H, Ohnuki E, Mitra A, Yu H. MIMIC-SBDH: A Dataset for Social and Behavioral Determinants of Health. Proc Mach Learn Res. 2021 Aug;149:391-413. PMID: 35005628; PMCID: PMC8734043.
[27] Gunningberg, Lena, et al. "Incidence and risk factors for pressure ulcers in ICU patients: A systematic review." Journal of Wound Care 29.3 (2020): 144–150. https://doi.org/10.12968/jowc.2020.29.3.144
[28] Weissman, Joel S., et al. "Social Determinants of Health in the ICU: An Underappreciated Risk Layer." Critical Care Medicine 48.5 (2020): 690–696. https://doi.org/10.1097/CCM.0000000000004302
[29] Agniel, Denis, et al. "Biases in electronic health record data due to processes within the healthcare system: retrospective observational study." BMJ 361 (2018): k1479. https://doi.org/10.1136/bmj.k1479
[30] Saxena, R. R. "Applications of Natural Language Processing in the Domain of Mental Health." Authorea Preprints, 2024. https://doi.org/10.36227/techrxiv.173014748.80471770
[31] Bonner, Sebastian, et al. "Graph representation of patient data: a systematic review." Briefings in Bioinformatics 22.1 (2021): 393–411. https://doi.org/10.1093/bib/bbz164
[32] Choi, Edward, et al. "Doctor AI: Predicting Clinical Events via Recurrent Neural Networks." Machine Learning for Healthcare Conference (2016): 301–318. https://doi.org/10.48550/arXiv.1511.05942
[33] Bazemore, Andrew W., et al. "Advancing population health improvement through community data partnerships." The Journal of the American Board of Family Medicine 29.6 (2016): 776–781. https://doi.org/10.3122/jabfm.2016.06.160086
[34] Himmelstein, Daniel S., et al. "Systematic integration of biomedical knowledge prioritizes drugs for repurposing." eLife 6 (2017): e26726. https://doi.org/10.7554/eLife.26726
[35] Kipf, Thomas N., and Max Welling. "Semi-supervised classification with graph convolutional networks." International Conference on Learning Representations (2017). https://doi.org/10.48550/arXiv.1609.02907
[36] Huang, Xin, et al. "Graph representation learning for healthcare: a survey." Briefings in Bioinformatics 23.2 (2022): bbab441. https://doi.org/10.1093/bib/bbab441
[37] Lundberg, Scott M., and Su-In Lee. "A Unified Approach to Interpreting Model Predictions." Advances in Neural Information Processing Systems 30 (2017): 4765–4774. https://doi.org/10.48550/arXiv.1705.07874
[38] Zhang, Yifan, et al. "Explaining Machine Learning Predictions in Healthcare with Tree SHAP: Applications and Optimization Techniques." Journal of Biomedical Informatics 144 (2024): 104461. https://doi.org/10.1016/j.jbi.2024.104461
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
This work is licensed under a Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
