Using Artificial Intelligence in Palliative Care
Updated: Jun 1
Vicent Blanes Selva
PhD Candidate at Biomedical Data Science Lab
Universitat Politecnica de Valencia
It has been demonstrated that palliative care offers different benefits for patients, such as better quality of life, improved mood or increased satisfaction with the treatment. However, determining which patients could benefit from this kind of program is not an easy task. Despite the virtues of earlier inclusion in palliative care programs, most referrals are made too late.
One of the most common inclusion criteria for palliative care is a short survival expectancy. However, scientific studies have shown that common tools, such as the surprise question (“Would I be surprised if this patient died in the next 12 months”), are not very accurate in identifying these patients. Other factors, such as the patient’s probability of becoming frailer during the following months, are also very hard to estimate.
However, technology could offer us a solution to improve the quality of the clinical prediction and therefore improve the detection of patients needing palliative interventions. In this case, the field of artificial intelligence, precisely what is known as Machine Learning, could help clinical staff predict a patient’s outcome.
Don’t be afraid! Machine Learning (or ML, for short) will not conquer the world in a Terminator-like fashion. The basic idea behind the technology is quite simple: an algorithm that ‘learns’ by examples. During the InAdvance project, one of the tasks assigned to our team was to create a predictive model for 1-year mortality. So, we used the available data from Hospital La Fe to train the algorithm so that it could predict the outcomes of new cases.
But ML models are far from perfect. Before their inclusion in clinical practice, they should pass a series of tests to ensure their quality, for example, checking their accuracy with data from other centres. Another crucial aspect of using ML in clinical practice is the software’s ability to adapt to clinical routines. We want to facilitate the job of clinical professionals and not overburden them with more tasks and difficult software!
In short, ML technology is an excellent opportunity to detect patients needing palliative care interventions. During the InAdvance project, our team created predictive models for mortality and frailty and prototyped them into a web application (https://thealeph.upv.es/) that we validated with clinical professionals. A lot of work is still needed to improve and gain the trust of clinical professionals. However, there is no doubt that ML algorithms will shape the future of personalized medicine.
Related scientific articles:
Blanes-Selva V, Ruiz-García V, Tortajada S, Benedí J-M, Valdivieso B, García-Gómez JM. Design of 1-year mortality forecast at hospital admission: A machine learning approach. Health Informatics Journal. 2021;27(1). doi:10.1177/1460458220987580
Blanes-Selva V, Doñate-Martínez A, Linklater G, García-Gómez JM. Complementary frailty and mortality prediction models on older patients as a tool for assessing palliative care needs. Health Informatics Journal. 2022;28(2). doi:10.1177/14604582221092592
Blanes-Selva, V.; Doñate-Martínez, A.; Linklater, G.; Garcés-Ferrer, J.; García-Gómez, J.M. Responsive and Minimalist App Based on Explainable AI to Assess Palliative Care Needs during Bedside Consultations on Older Patients. Sustainability 2021, 13, 9844. https://doi.org/10.3390/su13179844
Blanes-Selva, V., Asensio-Cuesta, S., Doñate-Martínez, A., Mesquita, F. P., & García-Gómez, J. M. (2022). Validating a Clinical Decision Support System for Palliative Care using healthcare professionals’ insights. MedRxiv. https://doi.org/10.1101/2022.06.03.22275904