Google's AI has already proven more accurate than even doctors at predicting the probability of a patient's death—the company published research in May that cited the case of an unidentified, late-stage lung cancer patient in the US.
Google’s AI has already proven more accurate than even doctors at predicting the probability of a patient’s death—the company published research in May that cited the case of an unidentified, late-stage lung cancer patient in the US. The hospital’s analysis of the patient’s health records and diagnostics gave her a 9.3% likelihood of dying during the course of her stay at the hospital. Google AI analysed over 175,000 data points on the patient, and pushed up the likelihood to 19.9%. She died within days of admission.
Of all the potential use of AI in healthcare, this is perhaps the one that is fraught with the most complex ethical, moral, and legal questions. There are many implications, including potential abuse by patients’ families, insurance providers, even hospitals. But, the need is also to look at where such use of AI scores—cold, hard data, could enable better decision-making. It could help free up limited hospital resources for giving timely treatment to those with a better shot at surviving. It could offer many a hard-pressed family relief from crushing medical-expense burdens—incurred, even when the outlook is “terminal”, out of misplaced hope or guilt. Most importantly, it could give the patient a dignified end.
That said, if they are to be used in the interest of rationality, death prediction tools must get quite accurate. For instance, just reading off a patient’s data-points can’t be judged enough. The neural networks Google’s death-prediction used must factor in all available data on progress in similar cases elsewhere, alternative treatment/therapies, emerging treatment methods, among others before they predict the chances of survival or otherwise. Policy will also need to respond to such technologies with appropriate regulation. A review of the prediction figure and any decision based on the same by a team of competent professionals should be made a must. Adequate checks against manipulation/breach of data must be provided for. Lastly, while Google AI may eventually be used to inform the decision on pulling the plug, the actual decision should be left to the discretion of the patient/her family and medical professionals.