Machine learning methods for propensity and disease risk score estimation in high-dimensional data: a plasmode simulation and real-world data cohort analysis
Introduction: Machine learning (ML) methods are promising and scalable alternatives for propensity score (PS) estimation, but their comparative performance in disease risk score (DRS) estimation remains unexplored. Methods: We used real-world data comparing antihypertensive users to non-users with 6...
| Main Authors: | , , , , , |
|---|---|
| Format: | Journal article |
| Language: | English |
| Published: |
Frontiers Media
2024
|
| _version_ | 1826315375570059264 |
|---|---|
| author | Guo, Y Strauss, VY Català, M Jödicke, AM Khalid, S Prieto-Alhambra, D |
| author_facet | Guo, Y Strauss, VY Català, M Jödicke, AM Khalid, S Prieto-Alhambra, D |
| author_sort | Guo, Y |
| collection | OXFORD |
| description | Introduction: Machine learning (ML) methods are promising and scalable alternatives for propensity score (PS) estimation, but their comparative performance in disease risk score (DRS) estimation remains unexplored. Methods: We used real-world data comparing antihypertensive users to non-users with 69 negative control outcomes, and plasmode simulations to study the performance of ML methods in PS and DRS estimation. We conducted a cohort study using UK primary care records. Further, we conducted a plasmode simulation with synthetic treatment and outcome mimicking empirical data distributions. We compared four PS and DRS estimation methods: 1. Reference: Logistic regression including clinically chosen confounders. 2. Logistic regression with L1 regularisation (LASSO). 3. Multi-layer perceptron (MLP). 4. Extreme Gradient Boosting (XgBoost). Covariate balance, coverage of the null effect of negative control outcomes (real-world data) and bias based on the absolute difference between observed and true effects (for plasmode) were estimated. 632,201 antihypertensive users and nonusers were included. Results: ML methods outperformed the reference method for PS estimation in some scenarios, both in terms of covariate balance and coverage/bias. Specifically, XgBoost achieved the best performance. DRS-based methods performed worse than PS in all tested scenarios. Discussion: We found that ML methods could be reliable alternatives for PS estimation. ML-based DRS methods performed worse than PS ones, likely given the rarity of outcomes. |
| first_indexed | 2024-12-09T03:24:56Z |
| format | Journal article |
| id | oxford-uuid:c1311bbd-bd65-4249-9602-e2061d33e1de |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-12-09T03:24:56Z |
| publishDate | 2024 |
| publisher | Frontiers Media |
| record_format | dspace |
| spelling | oxford-uuid:c1311bbd-bd65-4249-9602-e2061d33e1de2024-11-27T20:03:55ZMachine learning methods for propensity and disease risk score estimation in high-dimensional data: a plasmode simulation and real-world data cohort analysisJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:c1311bbd-bd65-4249-9602-e2061d33e1deEnglishJisc Publications RouterFrontiers Media2024Guo, YStrauss, VYCatalà, MJödicke, AMKhalid, SPrieto-Alhambra, DIntroduction: Machine learning (ML) methods are promising and scalable alternatives for propensity score (PS) estimation, but their comparative performance in disease risk score (DRS) estimation remains unexplored. Methods: We used real-world data comparing antihypertensive users to non-users with 69 negative control outcomes, and plasmode simulations to study the performance of ML methods in PS and DRS estimation. We conducted a cohort study using UK primary care records. Further, we conducted a plasmode simulation with synthetic treatment and outcome mimicking empirical data distributions. We compared four PS and DRS estimation methods: 1. Reference: Logistic regression including clinically chosen confounders. 2. Logistic regression with L1 regularisation (LASSO). 3. Multi-layer perceptron (MLP). 4. Extreme Gradient Boosting (XgBoost). Covariate balance, coverage of the null effect of negative control outcomes (real-world data) and bias based on the absolute difference between observed and true effects (for plasmode) were estimated. 632,201 antihypertensive users and nonusers were included. Results: ML methods outperformed the reference method for PS estimation in some scenarios, both in terms of covariate balance and coverage/bias. Specifically, XgBoost achieved the best performance. DRS-based methods performed worse than PS in all tested scenarios. Discussion: We found that ML methods could be reliable alternatives for PS estimation. ML-based DRS methods performed worse than PS ones, likely given the rarity of outcomes. |
| spellingShingle | Guo, Y Strauss, VY Català, M Jödicke, AM Khalid, S Prieto-Alhambra, D Machine learning methods for propensity and disease risk score estimation in high-dimensional data: a plasmode simulation and real-world data cohort analysis |
| title | Machine learning methods for propensity and disease risk score estimation in high-dimensional data: a plasmode simulation and real-world data cohort analysis |
| title_full | Machine learning methods for propensity and disease risk score estimation in high-dimensional data: a plasmode simulation and real-world data cohort analysis |
| title_fullStr | Machine learning methods for propensity and disease risk score estimation in high-dimensional data: a plasmode simulation and real-world data cohort analysis |
| title_full_unstemmed | Machine learning methods for propensity and disease risk score estimation in high-dimensional data: a plasmode simulation and real-world data cohort analysis |
| title_short | Machine learning methods for propensity and disease risk score estimation in high-dimensional data: a plasmode simulation and real-world data cohort analysis |
| title_sort | machine learning methods for propensity and disease risk score estimation in high dimensional data a plasmode simulation and real world data cohort analysis |
| work_keys_str_mv | AT guoy machinelearningmethodsforpropensityanddiseaseriskscoreestimationinhighdimensionaldataaplasmodesimulationandrealworlddatacohortanalysis AT straussvy machinelearningmethodsforpropensityanddiseaseriskscoreestimationinhighdimensionaldataaplasmodesimulationandrealworlddatacohortanalysis AT catalam machinelearningmethodsforpropensityanddiseaseriskscoreestimationinhighdimensionaldataaplasmodesimulationandrealworlddatacohortanalysis AT jodickeam machinelearningmethodsforpropensityanddiseaseriskscoreestimationinhighdimensionaldataaplasmodesimulationandrealworlddatacohortanalysis AT khalids machinelearningmethodsforpropensityanddiseaseriskscoreestimationinhighdimensionaldataaplasmodesimulationandrealworlddatacohortanalysis AT prietoalhambrad machinelearningmethodsforpropensityanddiseaseriskscoreestimationinhighdimensionaldataaplasmodesimulationandrealworlddatacohortanalysis |