Augmented intelligence to predict 30-day mortality in patients with cancer

Area of Science:

• Oncology
• Medical Informatics
• Biostatistics

Background:

• Identifying cancer patients at high risk for short-term mortality is crucial for timely interventions.
• Existing methods may not fully capture the dynamic risk profiles of cancer patients.

Purpose of the Study:

• To develop and validate an augmented intelligence tool for predicting 30-day mortality risk in cancer patients.
• To facilitate early identification of patients requiring palliative care or other interventions.

Main Methods:

• An algorithm was developed using socioeconomic and clinical data from a large community hematology/oncology practice.
• Patients' 30-day mortality risk was scored weekly.
• Algorithm performance was validated against actual patient death dates in electronic health records.

Main Results:

• The highest-risk group exhibited a 4.9% 30-day mortality rate.
• The low-risk group had a 0.7% 30-day mortality rate.
• The tool identified a 7.4-fold increased mortality risk in the highest-risk patient group.

Conclusions:

• The development of a decision support tool to accurately predict short-term mortality risk in cancer patients is feasible.
• This tool can aid clinicians in identifying at-risk patients for targeted support and care planning.