Machine Learning Uses Palliative Care Data to Identify Predictors of Illness Transition

Patient-reported palliative care data combined with machine learning could help signal when patients may transition between phases of illness, according to a study published in the Journal of Pain and Symptom Management.

“The present study demonstrated a novel application of network analysis to understand changes of phase in patient conditions. Previous research had focused on survival prediction and mortality rates, and have analyzed general patient health records rather than specific palliative care data,” researchers wrote. “As far as we are aware, this is the first time that palliative-specific data have been exclusively used in network analysis and machine learning.”

Researchers investigated whether machine learning could identify phases in patient palliative status—stable, unstable, deteriorating, or terminal—using information gathered on the Integrated Palliative Care Outcome Scale (IPOS). The 10-item IPOS measures symptom burden and is completed by patients, with the help of caregivers when necessary.

Palliative records from 804 adults in New Zealand receiving palliative care were analyzed for the study, which used a mix of statistical, machine learning, and network analysis techniques.

Six machine learning techniques revealed key variables that predicted transition between illness phases, according to the study. Poor appetite and loss of energy were crucial IPOS items, network analysis showed. Loss of energy was associated with drowsiness, shortness of breath, and lack of mobility. Meanwhile, poor appetite was associated with nausea, vomiting, constipation and sore and dry mouth.

“Future research using larger samples is needed to explore the phase changing by diagnostic groups and the application of wearable devices in the palliative care context,” researchers concluded. “Together with machine learning and network analysis, digital therapeutics has the potential to enhance clinical decision making to improve the quality of care across settings where palliative care is delivered.”

Reference:
Sandham MH, Hedgecock EA, Siegert RJ, Narayanan A, Hocaoglu MB, Higginson IJ. Intelligent palliative care based on patient-reported outcome measures. J Pain Symptom Manage. 2022 Jan 11;S0885-3924(21)00641-2. doi:10.1016/j.jpainsymman.2021.11.008