AI has shown promise in saving patient’s lives by determining those most likely to develop rapid deterioration while in the hospital. Through a double case study in the peer-reviewed CMAJ, how the AI-based early warning system, CHARTWatch, effectivity contributed to minimizing the number of preventable mortalities in the patients admitted in the hospital was shown.
The system was designed and evaluated for three years and was implemented in the general internal medicine (GIM) ward of the st Michael’s Hospital in Toronto.
The health status of the admitted patients tends to worsen quickly resulting in transfer to the ICU this is an area that many previous technologies have attempted addressing with varying success. To determine whether CHARTWatch works as anticipated in identifying patients at high risk of deterioration, the research team which included members from Unity Health Toronto, University of Toronto and Institute for Clinical Evaluative Sciences conducted a study on the tool.
The research involved cross-sectional study of 13,649 patients aged 55 to 80 years by comparing the results between those patients who received treatment before the adoption of CHARTWatch AI system and those who received treatment after adoption of the said system.
About 4,023 cases from the post-intervention period were tagged by CHARTWatch while, 9,626 cases in the pre-intervention period were used for comparison. The findings evinced that a lower percentage of patients in the CHARTWatch group (1. 6 percent) died suddenly than the pre-intervention group 2. 1 percent).
The timeliness of the interaction with clinical teams was one of the main factors that have led to the achievement of the goals envisioned by CHARTWatch. Email alerts were produced twice a day for the nursing teams, and daily notifications were sent for the palliative care teams; high-risk patient obtained appropriate attention.
This includes increased coordination between nurses and physicians, constant evaluation of patients’ conditions, and better management of high-risk patients. Nurses were also encouraged more on the aspect of tracking the patients that their alarms alert them to get more proactive approaches to attending to.
‘The lead author of the study, Dr. Amol Verma is a clinician-scientist at Toronto’s St. Michael’s Hospital and a professor of education and research in the application of artificial intelligence at the University of Toronto noted the need to approach the use of AI tools in medicine cautiously.’ He said that such findings indicate that machine learning based early warning systems such as CHARTWatch are capable of lowering the rates of unfamiliar passing beyond general medical wards in hospitals.
Further contributing to the study, performance index co-author, Dr. Muhammad Mamdani, went further to point out that such a study was useful in understanding the practical application of AI in health care facilities, and how the same can be implemented in other institutions to enhance their performance on similar technologies.
Indeed, the study shows how the AI systems can be used as an effective tool in augmenting the work of the healthcare providers by providing better information for decision making, better communication to enhance the lives of patients.
The authors remain hopeful that with the increase in the use of such technologies such as CHARTWatch, there will be positive impact on patients leading to better outcomes, and less deaths that could have otherwise been prevented.
Although this study did not focus on evaluating the performance of S2. brain, a related article suggested in the study analyzed key implementation factors when it comes to using the AI scribe in clinical practice which include patients’ consent, the accuracy of notes generated by the AI scribe, and whether or not the notes generated are compliant with local privacy legislation.
Reference
Clinical evaluation of a machine learning–based early warning system for patient deterioration, Canadian Medical Association Journal (2024). DOI: 10.1503/cmaj.240132
