Integrating Machine Learning and Operation Research for Optimized Care Pathways in Neurological Disorders

Neurological disorders pose an increasing challenge in low- and middle-income countries (LMICs), where healthcare systems are hindered by delays, inefficiencies, and restricted access. This research analyzed the combination of machine learning (ML) predictions and operations research (OR) optimization to improve neurological care pathways in Oyo State, Nigeria. Two hundred patients participated in the study (average age 56.8 ± 20.5 years; 56% female, 44% male), with diagnoses such as stroke (33.5%), Alzheimer’s (25.5%), epilepsy (19.5%), Parkinson’s (14%), and multiple sclerosis (7.5%). Diagnostic methods included CT (55%), MRI (45%), EEG (35%), PET (20%), and biomarker analysis (40%), with wait times varying from 2.5 days (MRI) to 4.2 days (PET). ML models exhibited impressive predictive ability: Logistic Regression (accuracy 0.79, AUC 0.83), Random Forest (0.86, AUC 0.91), XGBoost (0.89, AUC 0.94), and Neural Networks (0.91, AUC 0.96). OR-driven interventions enhanced system efficiency, cutting rehabilitation wait times from 4.5 to 2.8 days (− 38%), boosting low rehabilitation usage from 65% to 88% (+ 23%), elevating medication compliance from 42% to 70% (+ 28%), and reducing surgical delays from 2.2 to 1.4 weeks (− 36%). Clinical results were enhanced in 57% of stroke cases, 56% of epilepsy cases, 50% of Parkinson’s cases, 40% of multiple sclerosis cases, and 24% of Alzheimer’s cases, with follow-up durations ranging from 8.7 to 12.4 months. The economic assessment indicated that standard care expenses were $2,450 per patient (0.58 QALYs; cost/QALY = $4,224), whereas optimized pathways amounted to $2,780 with 0.89 QALYs (cost/QALY = $3,124), achieving a 26% improvement in efficiency. Combining ML prediction with OR prescription shows promise for delivering real-time, economical, and resource-efficient neurological care in LMIC environments.