RESUMO
We studied methods to convert medical prescriptions in free text to a structured form for pharmacy insructions and planning nursing activities in hospitalized patients. We compared Natural Language Processing (NLP) with Parsing Process (PP), both for the Spanish language. We studied 87,750 and processed 65,000 prescriptions and recovered 62% and 65% with NLP and PP to a structured format respectively. The difference between the methods is significant (p < 0.001) and further work is needed to determine if combining them will have higher performance.
Assuntos
Processamento de Linguagem Natural , HumanosRESUMO
UNLABELLED: The objective of the study is to assess a nutritional risk alert using an Informatics System in hospitalized patients versus the conventional methodology. METHOD: We studied 400 medical patients at the Belgrano Hospital Critical Care Unit. We considered two groups of 200 patients. In the first the group (Control Group), nutritional risk was diagnosed in a traditional way. In the second group (Alarm Group), nutritional risk was diagnosed with an alert system. The alert was triggered when a patient showed low levels of, at least, two of the following variables: albumin, cholesterol, triglyceride, lymphocyte or low body mass index (BMI). RESULTS: Nutritional risk was detected in 20.3% of the Control Group patients while, in the Alert Group, nutritional risk was detected in 34.3% of the patients; the difference between the two groups was significant (p<0.001), with a sensitivity rate of 99% and a specificity rate of 98%. CONCLUSION: Malnutrition is more easily detected when using an alert system.