Giuseppina Affinito, Marcello Moccia, Roberta Lanzillo, Ruth Ann Marrie, Jeremy Chataway, Vincenzo Brescia Morra, Raffaele Palladino
{"title":"Neural networks to estimate multiple sclerosis disability and predict progression using routinely collected healthcare data.","authors":"Giuseppina Affinito, Marcello Moccia, Roberta Lanzillo, Ruth Ann Marrie, Jeremy Chataway, Vincenzo Brescia Morra, Raffaele Palladino","doi":"10.1177/13524585251347513","DOIUrl":null,"url":null,"abstract":"<p><strong>Background and objectives: </strong>Multiple sclerosis (MS)-related disability is conventionally measured using the Expanded Disability Status Scale (EDSS), which requires neurological examination and is generally embedded in clinical records, making it unavailable in administrative datasets. This limits its utility for population-level estimates and healthcare planning. This study aims to use routinely collected healthcare data to fill this gap.</p><p><strong>Methods: </strong>We conducted a population-based study using administrative data from Campania Region (Italy) to develop and validate neural network algorithms to estimate MS-related disability and predict its progression (2015-2021). We employed a deep learning approach to estimate the EDSS, and a hybrid model combining survival analysis with neural network predictions to forecast the risk of EDSS progression.</p><p><strong>Results: </strong>The model estimated EDSS with 0.68 accuracy, 0.68 precision, and 0.67 F1-score. The hybrid model had a predictive performance of 0.92. From 2016 to 2021, 9.01% of the population had EDSS ⩽ 3.0, 62.10% had EDSS between 3.5 and 5.5, and 28.89% had EDSS ⩾ 6.0. Looking at projections from 2021 to 2026, 67.68% people with EDSS ⩽ 3.0 are expected to progress to EDSS 3.5-5.5.</p><p><strong>Conclusion: </strong>These findings highlight the potential of advanced data analytics using administrative data to improve MS monitoring, healthcare planning, and decision-making.</p>","PeriodicalId":520714,"journal":{"name":"Multiple sclerosis (Houndmills, Basingstoke, England)","volume":" ","pages":"13524585251347513"},"PeriodicalIF":0.0000,"publicationDate":"2025-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Multiple sclerosis (Houndmills, Basingstoke, England)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1177/13524585251347513","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Background and objectives: Multiple sclerosis (MS)-related disability is conventionally measured using the Expanded Disability Status Scale (EDSS), which requires neurological examination and is generally embedded in clinical records, making it unavailable in administrative datasets. This limits its utility for population-level estimates and healthcare planning. This study aims to use routinely collected healthcare data to fill this gap.
Methods: We conducted a population-based study using administrative data from Campania Region (Italy) to develop and validate neural network algorithms to estimate MS-related disability and predict its progression (2015-2021). We employed a deep learning approach to estimate the EDSS, and a hybrid model combining survival analysis with neural network predictions to forecast the risk of EDSS progression.
Results: The model estimated EDSS with 0.68 accuracy, 0.68 precision, and 0.67 F1-score. The hybrid model had a predictive performance of 0.92. From 2016 to 2021, 9.01% of the population had EDSS ⩽ 3.0, 62.10% had EDSS between 3.5 and 5.5, and 28.89% had EDSS ⩾ 6.0. Looking at projections from 2021 to 2026, 67.68% people with EDSS ⩽ 3.0 are expected to progress to EDSS 3.5-5.5.
Conclusion: These findings highlight the potential of advanced data analytics using administrative data to improve MS monitoring, healthcare planning, and decision-making.