Predicting Readmissions Using Omics, Biostatistical Evaluate and Artificial Intelligence

Predicting Readmissions Using Omics, Biostatistical Evaluate and Artificial Intelligence

This study is a prospective registry that aims to predict readmissions in patients with heart failure, using -omics, machine learning, patient reported outcomes, clinical data and other high-dimensional data sources.

There is substantial need to better predict outcomes across the spectrum of heart failure (HF) phenotypes in order to provide more efficient care with greater precision. Specifically, no validated methods have been adopted to predict outcomes reflecting transitions in health status across the continuum of HF and changes in cardiac function. A key transition is hospitalization - either readmission or de novo cardiovascular hospital admission. This is a major unmet health care need, to be able to better predict who will require hospital admission.

Novel contributions of biomarkers, -omics, remote patient monitoring, and artificial intelligence (AI). It is anticipated that prediction of readmission and many other outcomes will be further improved by measurement of circulating biomarkers and by incorporating methods from AI including machine learning and probabilistic generative models that can incorporate the lens of how physicians and patients think. Machine learning that incorporates many different types of data, including physician interpretation and a broad array of biomarker/-omics molecular information can lead to significant improvements in predictive accuracy. Novel multimarker strategies coupled with machine learning may enable the ability of physicians to predict a range of outcomes (e.g., transitions in HF health status and LVEF) and refine clinical prediction models. Furthermore, the investigators will collect patient data, including patient reported outcome measures (PROMs), and physiological data (e.g. heart rate, blood pressure, and daily weights data) and integrate these data points into predictive models. The investigators will use the PROMs obtainable using Medly as a predictor of hospitalization, and as an outcome. In this proposal, the investigators will take advantage of recent advances in both deep and high throughput proteomics technologies to perform high-resolution analyses. These novel factors can be integrated into new electronic algorithms to improve HF care in the population.

Inclusion Criteria:

• Any patient aged 18 years or older admitted to hospital or seen in the emergency department with heart failure defined clinically
• The diagnosis will be guided by the Framingham criteria for HF and/or BNP. A BNP >400 will be defined as definite heart failure and BNP 100-400 classified as possible heart failure.
• Provides informed consent

Exclusion Criteria:

• Patients who cannot communicate due to dementia or severe cognitive deficits
• non-Ontario residents
• nursing home residents
• those who are not discharged home but are discharged to a skilled nursing facility (long-term care or chronic institution)
• those who are unable to communicate who do not have a proxy (e.g. spouse or close family member) to facilitate communication with the patient.