Zmudzki Fredrick, Smeets Rob J E M, Groenewegen Jan S, van der Graaff Erik
Department of Rehabilitation Medicine, Care and Public Health Research Institute, Faculty of Health, Life Sciences and Medicine, Maastricht University, Universiteitssingel 40, Room 3.544, P.O. Box 616, 6200 MD, Maastricht, 6229 ER, The Netherlands, 31 433882160.
Epoque Consulting, Sydney, Australia.
JMIR Rehabil Assist Technol. 2025 May 9;12:e65890. doi: 10.2196/65890.
Chronic musculoskeletal pain (CMP) impacts around 20% of people globally, resulting in patients living with pain, fatigue, restricted social and employment capacity, and reduced quality of life. Interdisciplinary multimodal pain treatment (IMPT) programs have been shown to provide positive and sustained outcomes where all other interventions have failed. IMPT programs combined with multidimensional machine learning predictive patient profiles aim to improve clinical decision support and personalized patient assessments, potentially leading to better treatment outcomes.
We aimed to investigate integrating machine learning with IMPT programs and its potential contribution to clinical decision support and treatment outcomes for patients with CMP.
This prospective pilot study used a machine learning prognostic patient profile of 7 outcome measures across 4 clinically relevant domains, including activity or disability, pain, fatigue, and quality of life. Prognostic profiles were created for new IMPT patients in the Netherlands in November 2023 (N=17). New summary indicators were developed, including defined categories for positive, negative, and mixed prognostic profiles; an accuracy indicator with high, medium, and low levels based on weighted true- or false-positive values; and an indicator for consistently positive or negative outcomes. The consolidated reporting guidelines checklist for prognostic machine learning modeling studies was completed to provide transparency of data quality, model development methodology, and validation.
The machine learning IMPT prognostic patient profiles demonstrated high accuracy and consistency in predicting patient outcomes. The profile, combined with extended new prognostic summary indicators, provided improved identification of patients with predicted positive, negative, and mixed outcomes, supporting more comprehensive assessment. Overall, 82.4% (14/17) of prognostic patient profiles were consistent with clinician assessments. Notably, clinician case notes indicated the stratified prognostic profiles were directly discussed with around half (8/17, 47.1%) of patients. Clinicians found the prognostic patient profiles helpful in 88.2% (15/17) of initial IMPT assessments to support shared clinician and patient decision-making and discussion of individualized treatment planning.
Machine learning prognostic patient profiles showed promising contributions for IMPT clinical decision support and improving treatment outcomes for patients with CMP. Further research is needed to validate these findings in larger, more diverse populations.
慢性肌肉骨骼疼痛（CMP）影响着全球约20%的人口，导致患者饱受疼痛、疲劳、社交和就业能力受限以及生活质量下降之苦。在所有其他干预措施均告失败的情况下，跨学科多模式疼痛治疗（IMPT）项目已被证明能带来积极且持续的效果。IMPT项目与多维机器学习预测患者档案相结合，旨在改善临床决策支持和个性化患者评估，有可能带来更好的治疗效果。
我们旨在研究将机器学习与IMPT项目相结合及其对CMP患者临床决策支持和治疗效果的潜在贡献。
这项前瞻性试点研究使用了一个机器学习预后患者档案，该档案涵盖4个临床相关领域的7项结果指标，包括活动或残疾、疼痛、疲劳和生活质量。为2023年11月在荷兰的新IMPT患者创建了预后档案（N = 17）。制定了新的汇总指标，包括阳性、阴性和混合预后档案的定义类别；一个基于加权真阳性或假阳性值的高、中、低水平准确性指标；以及一个持续阳性或阴性结果的指标。完成了预后机器学习建模研究的综合报告指南清单，以提高数据质量、模型开发方法和验证的透明度。
机器学习IMPT预后患者档案在预测患者结果方面显示出高准确性和一致性。该档案与扩展的新预后汇总指标相结合，能更好地识别出预测为阳性、阴性和混合结果的患者，支持更全面的评估。总体而言，82.4%（14/17）的预后患者档案与临床医生的评估一致。值得注意的是，临床医生的病例记录表明，约一半（8/17，47.1%）的患者直接讨论了分层预后档案。临床医生发现，在88.2%（15/17）的初始IMPT评估中，预后患者档案有助于支持临床医生和患者的共同决策以及个性化治疗计划的讨论。
机器学习预后患者档案对IMPT临床决策支持和改善CMP患者的治疗效果显示出有前景的贡献。需要进一步研究以在更大、更多样化的人群中验证这些发现。
