In the healthcare setting, the identification and classification of patients in relation to their degree of vulnerability (risk) is key to establishing nursing care times, adjusting them to adequate levels of quality and safety. In addition, it allows the institution to plan and schedule resources according to the dependency levels of each care area or unit.

To estimate the degree of vulnerability of hospitalized patients, scales are widely used to assess their state of health, functional level and risk, allowing the design of comprehensive treatments and care plans adapted to individual conditions. These scales assign a score to the patient that determines his or her degree of dependence or state in different areas, by means of the aggregation and weighting of the scores obtained in a set of test-type questions that make it possible to evaluate the patient. Each area of medical specialization uses a specific subset of these scales. Generally speaking, the scales are completed by the nurse responsible for the patient on admission and evaluated throughout the hospitalization process in the center’s Hospital Information System (HIS), and therefore their results are automatically recorded as part of the patient’s Electronic Health Record (EHR). However, the latter case is the least common and, even so, the results of the scales are interpreted manually by the nurse responsible for determining the risk of each patient in each area. Paradoxically, it has been shown that the main limitation and shortcoming of most of the HIS implemented in health centers is precisely the scarce development of clinical decision support modules, despite having access to a large volume of information. On the other hand, there is no global risk value for each patient resulting from the knowledge obtained with these scales, but rather the different risk values assigned to each patient are specialized by areas of operation, and depend entirely on the experience and expertise of the professionals responsible for their evaluation. This causes multiple problems (e.g. differences in the criteria of the healthcare personnel, loss of information, errors in the data, lack of general vision of the patient’s risk) which in the short-medium term produce important shortcomings in the alerts for detecting safety problems and in the quality of medical attention.

The riskML tool allows calculating the risk of a hospitalized patient from an AI model trained with information from hospitalized patient data and labeled by an expert nursing team. The tool avoids having to ask multiple questions corresponding to traditional nursing scales to, with very few questions, determine the patient’s overall risk.

Currently there is no nursing scale that assesses this type of risk. The information provided by this tool is very useful both for improving the quality and effectiveness of nursing care, as well as for hospital management, since it allows an appropriate allocation of resources according to the severity of the patient.