Identificação de modelos de predição para risco de desenvolvimento de lesão por pressão em unidade de terapia intensiva.

Abstract

With a view to improving the pressure injury prognosis (PI) in intensive care, the assessment of the predictive capacity of the Braden scale and its components was carried out, the identification of potential predictors of development and potential models of prognosis of PI. The present study has a retrospective epidemiological character based on medical records of patients admitted to an Intensive Care Unit (ICU) of a public hospital in Ponta Grossa (Paraná - Brazil) with an observational and a prediction component. Data were collected from archived medical records of patients who were admitted to intensive care in a public hospital in the city of Ponta Grossa / PR, Brazil, and who were discharged from the unit between January 2017 and December 2018 (n = 764) . In total, 582 records met the eligibility criterion and were considered valid. Most patients were female (52.6%), with a mean age of 66.3 years and with at least one chronic condition (51.3% of hypertensive patients). Two hundred patients developed LPP, corresponding to an incidence rate of 34.3%. The scores on the Braden scale ranged between 08 and 20 points, with a mean of 12.57 and a median of 11. It was found that the group with the lowest scores on the Braden scale had a higher incidence of LPP (180 of 341) than the low risk group (20 out of 241). Observing the score of the components of the Braden scale, it was noticed that some as ‘Moisture’ and ‘Nutrition’ had a high concentration of notes, while others like ‘Sensory perception’ and ‘Activity’ showed a strong correlation with each other. ‘Sensory perception’ still showed a stronger correlation with LPP than the Braden scale itself. Braden showed moderate discriminative capacity with AUC equal to 0.751 and very similar to the component 'Sensory perception' which had AUC of 0.760. The factors that emerged as having the highest odds ratio for PI were ‘Sensory perception’, time on norepinephrine use, time on mechanical ventilation, length of stay in the ICU and age group. In all, fifteen models for predicting the risk of developing pressure injuries were identified in the present study, with five models considered more promising due to their robustness characteristics over the sampling period and consistency presented in the resampling procedure (bootstrapping), satisfying criteria that aim to eliminate spurious models. These logistic regression models with combinations of three variables resulted in models with good discriminative capacity with concordance statistics between 0.838 and 0.907 and proved robust to the sampling period (2017 and 2018 separate and total) and to the bootstrapping procedure. It is concluded that models with specific risk factors for ICU are more assertive and better describe the incidence of PI in ICU.