Prediction Modelling and Data Quality Assessment for Nursing Scale in a Big Hospital: A Proposal to Save Resources and Improve Data Quality

Nursing scales play an important role in the evaluation of patients’ clinical and social frailty. Filling out correctly the scales, allows the early identification of patients at risk of prolonged hospitalization or difficult discharge, and enables an estimation of care complexity during hospitalization. Given the high predictive value of these scales, it is important that the measurements are reported precisely. In this paper, we provide a general methodology to estimate the quality of data related to nursing scales and we introduce an approach to infer a certain scale based on other ones that share common fields in their calculation. The former is for measuring the reliability of the scoring values, and the latter is to reduce data entry time and costs. Our experimental setting focuses on two scales: Blaylock Risk Assessment Screening Score Index (Brass scale), that evaluates the risk of difficult discharge, and Care Dependency Index (IDA scale), that evaluates the degree of care complexity. These two scales have several fields which often provide similar or correlated information about the patient’s clinical condition. Preliminary results show the possibility to reduce oversights in filling out the scales use an automatic evaluation metric, which allows reproducing better the clinical condition of the patient. Moreover, the opportunity to predict value of one scale using the data of the other one allows the nurse to reduce the time for completing and to focus more on patient care.