The number of nursing home residents who need help with their functional health, i.e., ability to perform the Activities of Daily Living, ADLs, is constantly increasing. Clinicians, patients, and their caregivers would benefit from information on likely next ADL events (both progression and recovery). While much of the focus in long-term care and the nursing home is on preservation and restoration of these activities, very little information is available to guide patients, providers, families, and policymakers regarding expectations for the sequence, likelihood, and timing of functional loss and recovery. 

Predictive modeling, particularly Machine Learning, and availability of large data enables clinicians and patients to better understand what is likely to occur next. These forecasts can also set priorities for clinical interventions. Predictive modeling could also be used for planning purposes as it provides quantitative information on the timing and severity of the next disability event. Caregivers would like to know how long they need to provide care. Families would like to understand the likely quality of life of their loved one. 

The objective of this study is to provide a data-driven analytical model, based on Machine Learning techniques, for predicting functional status change for residents in long term care facilities. Particularly, we predict:

1.     the sequence in which ADLs were lost and recovered using large longitudinal data,

2.     the days between ADL loss and recovery from these losses,

3.     the likelihood of loss and improvement. 

This information can help in planning for end-of-life disabilities. It can also be used to set “pay for improvement” incentives for providing long term care to older adults.

The sequence of functional loss and recovery among long term care residents is accurately predictable. This study provides benchmarks for the sequence, likelihood, transitions, and timing between various combinations of ADL deficits. While the majority of patients were able to recover from ADL deficits, the mean recovery time from a single ADL deficit suggests that recovery occurs over a long period and care planning needs predicting such timelines. This information can also be used to pay more for long term care facilities that perform better than average. In these payment systems, the percentage of patients that improved their condition is compared to the average percentage and additional payments are provided for the facilities that report better than average improvement rates.