[End-of-life predictive models: risk of prognostic persistence?]

Machine learning techniques, applied in the palliative field, are able to define an increasingly accurate prognosis in patients with advanced neoplasms and to identify patients at greater risk of functional decline or short-term mortality. The improvement of predictive abilities can allow an enhancement of prognostic abilities and also a more accurate detection of the most complex needs of patients. Moreover, data, even scientific data, are not values, any intervention based on them must be endowed with meaning. Predictive models can therefore be useful but only as a complementary and above all optional tool for the doctor, one of the parameters to evaluate the usefulness in different specific situations. Otherwise, the risk is to add a new type of persistence, the prognostic one.