[MA 2023 25] Machine learning-based exploration of the effects of different anti-hypertensive medications on long-term blood pressure control in diverse ethnic populations
Department of Public and Occupational Health
Proposed by: Dr. Felix Chilunga [f.p.chilunga@amsterdamumc.n]
Introduction
Hypertension (HTN) is a growing public health issue among ethnic minority populations in Europe. (1) Despite progress made to prevent and control hypertension, it has been shown to disproportionately affect ethnic minority populations compared to host populations in several European countries. (2,3)
A recent publication by Van der Linden et al. highlights some of the determinants of a suboptimal blood pressure control in a multi-ethnic population. (4) However, this study did not look into the role which specific anti-hypertensives play a role in the blood pressure control. Due to differences in pathophysiological aspects of hypertension control such as salt sensitivity and chronic inflammation, it is possible that some anti-hypertensive medications control blood pressure better than others. We aim to investigate if there are specific anti-hypertensive medications that are better at regulating the blood pressure in participants of different ethnicities.
Description of the SRP Project/Problem
This study will utilize prospective data from the HELIUS cohort, a population-based, multi-ethnic study initiated in Amsterdam, the Netherlands. (5) The cohort, initiated in 2011, focuses on investigating the interplay between cardiovascular diseases, mental health, and infectious diseases in a diverse population. It includes 24,789 participants aged between 18 and 70 years with Dutch, South-Asian Surinamese, African Surinamese, Ghanaian, Moroccan, and Turkish origins, who were recruited using a stratified sampling strategy. Follow-up data was collected in between 2019 and 2021, with both baseline and follow-up data collected through self-administered questionnaires and physical examinations, including the collection of biological samples.
For this study, we will use a sub-sample of 1578 HELIUS participants who have been on anti-hypertensives from baseline to follow-up. Data on antihypertensive medications are provided in the cohort as generic names and Anatomical Therapeutic Chemical (ATC) codes. Only anti-hypertensive medications used by greater than or equal ten participants will be selected for further analyses (approximately 30 anti-hypertensives). Blood pressure control at follow-up will be categorized as "yes" or "no" according to WHO guidelines (less than 140/90 mmHg = good control, greater than or equal to 140/90 mmHg). Additionally, we have data on co-variates such as age, sex, education, occupation, physical activity, smoking, alcohol consumption, psychosocial stress, obesity, and comorbidities such as diabetes and chronic kidney disease.
To answer our question, we perform the main steps:
1. We are going to be using one or more subgroup discovery (SGD) techniques. (6,7) Specifically, we will examine if a relationship can be found between certain groups of antihypertensive medications and blood pressure control for with and without ethnic group as covariates.
2. In additional models, we will add other covariates such as (sex, being married, smoking, alcohol intake, obesity, cardiovascular disease (CVD) history, and family history of hypertension
3. We will evaluate the SGD techniques and the discovered subgroups using the following quality measures (or information-theoretic measures): coverage, support, rule length, significance, novelty (WRAcc), confidence and redundancy
4. We will evaluate the performance of our models using appropriate evaluation metrics such as the area under the receiver operating curve (AUROC), the area under the prediction/recall curve (AUPRC), accuracy, precision, recall, or F1 score.
Research questions
How well the SGD techniques and the discovered subgroups performs in terms of quality measures? How the subgroup changes when adding co-variates?
How well the predictive model predict blood pressure control? How the additional co-variates affect the performance when added to the model? How the discovered subgroups affect the performance when added to the model? Do they have predictive power?
Is there a differential effect on blood pressure control across ethnic groups based on treatment with specific (groups) of antihypertensive medications?
Expected results
The subgroup discovered and the performance of the SGD techniques and the discovered subgroups using the aforementioned quality measures.
The performance results of the prediction models in terms of appropriate evaluation metrics such as AUROC, AUPRC, accuracy, precision, recall, or F1 score.
Trained prediction models for predicting BP controls.
A master thesis written in a form of a scientific article.
Time periods
o November – June
o May - November
Contact
Dr. Felix Chilunga f.p.chilunga@amsterdamumc.nl
References
1. Modesti, P. A., Reboldi, G., Cappuccio, F. P., et al. (2016). Panethnic Differences in Blood Pressure in Europe: A Systematic Review and Meta-Analysis. PLoS One, 11(1), e0147601. https://doi.org/10.1371/journal.pone.0147601
2. Agyemang, C., Kieft, S., Snijder, M. B., et al. (2015). Hypertension control in a large multi-ethnic cohort in Amsterdam, The Netherlands: the HELIUS study. International Journal of Cardiology, 183, 180-189. https://doi.org/10.1016/j.ijcard.2015.01.007
3. Howard, G., Prineas, R., Moy, C., et al. (2006). Racial and geographic differences in awareness, treatment, and control of hypertension: the REasons for Geographic And Racial Differences in Stroke study. Stroke, 37(5), 1171-1178. https://doi.org/10.1161/01.STR.0000217230.20539.c2
4. van der Linden, E. L., Collard, D., Beune, E. J. A. J., Nieuwkerk, P. T., Galenkamp, H., Haafkens, J. A., Moll van Charante, E. P., van den Born, B. H., & Agyemang, C. (2021). Determinants of suboptimal blood pressure control in a multi-ethnic population: The Healthy Life in an Urban Setting (HELIUS) study. Journal of Clinical Hypertension, 23(5), 1068-1076. https://doi.org/10.1111/jch.14202
5. Snijder, M. B., Galenkamp, H., Prins, M., Derks, E. M., Peters, R. J. G., Zwinderman, A. H., & Stronks, K. (2017). Cohort profile: the Healthy Life in an Urban Setting (HELIUS) study in Amsterdam, The Netherlands. BMJ Open, 7(12), e017873. https://doi.org/10.1136/bmjopen-2017-017873
6. Helal, S. Subgroup Discovery Algorithms: A Survey and Empirical Evaluation. J. Comput. Sci. Technol. 31, 561–576 (2016). https://doi.org/10.1007/s11390-016-1647-1
7. Herrera, F., Carmona, C.J., González, P. et al. An overview on subgroup discovery: foundations and applications. Knowl Inf Syst 29, 495–525 (2011). https://doi.org/10.1007/s10115-010-0356-2