A new study published in BMJ Open by medical schools and hospitals in Sweden and Boston explored the real-world value of new medical technologies. Return on healthcare spending suffers when patients do not live to benefit from the procedures. The study’s objective was to improve on the current ability to identify patients at the end of life.
The researchers trained machine learning models to predict patient mortality within 30 days of emergency department (ED) discharge. First, the team trained machine learning models with electronic health records and administrative data from 65,776 ED visits. In the second phase of the study, the team validated the AI models with 55,164 ED visits from a different hospital. The results showed that machine learning outperformed conventional indexes of short-term mortality.
According to the study authors, the use of new healthcare technologies contributes significantly to the rise in healthcare costs overall. Hospitals that make better predictions about end-of-life patients could help prevent over-treating patients. Weighing treatment decisions with better life expectancy data can increase treatment return on investment. Also, opting out of treatments that cause net harm during patients’ last days could improve the patients’ quality of life.
