Objective: To optimise the accuracy of a clinical decision support system (CDSS) to support Self-Management, Remote monitoring and Timely review in Parkinson’s Disease (SMaRT-PD).

Background: Home Based Care (HBC) is a supported self-management pathway using remote monitoring data to inform person-centred specialist care at University Hospitals Plymouth, UK; scalability is limited by manual data entry and evaluation processes. A prototype CDSS comprising 58 rule-based decision trees (RBDTs) was created. RBDTs integrate multi-modal data from care records, a wearable sensor and patient reported outcomes (PROs) of motor and non-motor symptom burden and impact to generate personalised symptom-specific management recommendations for patients and clinicians (including alerts for clinical review). 48 RBDTs resulted in patient-facing recommendations: 31 for symptom management, 5 for wellbeing considerations and 12 for general management and pathway issues. 10 RBDTs resulted in outputs for the clinicians.

Method: Accuracy of the CDSS was tested by comparing CDSS and clinician-generated outputs. Retrospective off-line test-retest cycles were undertaken using representative historical HBC datasets. Prospective testing was performed using real-time data. CDSS outputs were classified as identical, different but appropriate or inappropriate. Inappropriate outputs triggered RBDT iteration prior to further cycles of testing. Test-iteration-test cycles were repeated until accuracy of each RBDT was 100%; testing of each RBDT was conducted proportionate to anticipated use (based on symptom frequency), with a minimum of 5 tests undertaken of the most recent version of each RBDT.

Results: 869 retrospective and 967 prospective tests resulted in 149 iterations across the 58 RBDTs. With iteration, the proportion of RBDTs with 100% accuracy increased from 20% to 100% for general management, 32% to 86% for symptom management (resulting in a mean accuracy of 97%), 60% to 100% for wellbeing and 20% to 100% for clinician report RBDTs. Iteration and prospective testing is ongoing.

Conclusion: We have demonstrated accuracy of the CDSS, SMaRT-PD, in generating patient and clinician-facing care recommendations. SMaRT-PD has the potential to support remote management of PD using data-driven decision making.

« Back to 2024 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/smart-pd-a-clinical-decision-support-system-for-the-management-of-parkinsons-disease/