The Development and Validation of a Disability and Outcome Prediction Algorithm in Multiple Sclerosis Patients

Abstract

At the moment, multiple sclerosis (MS) is considered one of the major disability factors among the young population. Considering the high prevalence and severity of this chronic disease, the aim of this study was to develop a disability and outcome prediction algorithm in MS patients. The data from two MS patient groups was analysed-Group A (151 patients with the following drug therapies: interferon beta-1a, glatiramer acetate, teriflunomide, natalizumab) and Group B (58 patients treated with natalizumab). Considering certain demographical and clinical predictive variables, as well as different disability threshold values, several prediction models were developed and validated, which are able to estimate the disabilit y (Expanded Disability Status Scale, EDSS) and outcome probabilities. The prediction model validation on estimating the disability and outcome probabilities yielded a maximum Area Under the Receiver Operating Characteristic Curve (ROC AUC) Score of 80-82%. The overall results indicate that for Group A disability prediction based on EDSS is appropriate, while for Group B the outcome might be a better measure for probability estimation. Hence, the results suggest that the outcome is a more appropriate measure for monitoring natalizumab treated MS patients, since it depends on other clinical characteristics as opposed to the sole disability estimation through the EDSS score. A disability and outcome prediction model in MS patients was developed and validated. Despite the fact that the obtained results were satisfactory given the small dataset, by embedding more predictive demographical and clinical variables in the algorithm, as well as including more patients, the model could be used through an online platform in MS patients' monitoring and prioritization, in order to succeed in improving the quality of life and reducing disability progression. Rezumat La momentul actual, scleroza multiplă (SM) este considerată un factor major de dizabilitate în rândul populației tinere. Prin urmare, considerând prevalența și severitatea ridicată a acestei patologii, obiectivul prezentului studiu a fost reprezentat de dezvoltarea unui model de predicție a dizabilității și prognosticului pacienților cu scleroză multiplă. Au fost analizate datele provenite de la 2 grupuri de pacienți: grupul A (151 de pacienți cu următoarele terapii: interferon beta-1a, glatiramer acetat, teriflunomidă, natalizumab) și grupul B (58 de pacienți tratați cu natalizumab). Pe baza unor variabile predictive specifice (demografice și clinice), precum și prin considerarea mai multor valori prag de dizabilitate, au fost dezvoltați și validați mai mulți algoritmi de predicție, care au estimat dizabilitatea (scorul EDSS-Expanded Disability Status Scale) și prognosticul prin intermediul unui model probabilistic. Validarea modelelor de predicție din punctul de vedere al estimării dizabilității și prognosticului a condus la obținerea unui scor ROC AUC (Area Under the Receiver Operating Characteristic Curve) maxim de 80-82%. Rezultatele obținute au in...

AIGDR: An Artificial Intelligence Framework for Disability Research – A Multiple Sclerosis Case

“…Tommasin et al evaluated the accuracy of a data-driven approach, such as ML classi cation, in predicting disability progression in MS (49). Oprea et al developed a prediction model to estimate disability as measured by the EDSS and outcome probabilities (50). In the context of secondary progressive MS (SPMS), Law et al built algorithms based on DT, LR, and SVMs to predict SPMS disability progression using EDSS, MS Functional Composite component scores, T2 lesion volume, brain parenchymal fraction, disease duration, age, and gender (21).…”

Identification and Quantification of Risk Factors Related to the Quality of Life of Multiple Sclerosis Patients

“…The development of such machine learning algorithms is a growing need in many medical fields [24]. As for multiple sclerosis patients, the unpredictable nature of the disease, as well as the relapsing evolution of many individuals make such a purpose of an utmost importance, especially considering the fact that the majority of the machine learning algorithms focused on disease course prediction and diagnostic and prognostic modelling rather than on quality of life estimation [4,5,7,11,14,22,28].…”

“…Multiple sclerosis (MS) is a severe and progressive neurological disease, which affects both young and old adults. The high variability related to disease prognosis, as well as the consistent levels of disability can significantly influence the quality of life in MS patients [6,9,12,14]. Hence, relevant interventions must focus on limiting the disability, as well as on other factors which might influence the quality of life, from three important points of view: physical health, mental health and perceived quality of life [1].…”