IJCS | Volume 32, Nº5, September/October 2019

537 Table 1 - Scores for assessment of clinical deterioration Authors Number of patients Outcomes Diagnostic accuracy Year of publication Buist et al. 8 6,303 In-hospital mortality Positive predictive value: 16.2%. Four or more abnormal observations – 88.2% 2004 Goldhill et al. 18 63 ICU Admission Sensitivity – 97%; Specificity – 18% 1999 Goldhill & McNarry 19 548 Mortality within 30 days Sensitivity – 7.7%; Specificity – 99.8% 2004 Subbe et al. 20 709 ICU admission; Number of cardiac arrests; mortality within 60 days Endpoints ROC-curve analysis 2001 Hodgetts et al. 21 250 Cardiac arrests Sensitivity/specificity: 100/17%; 98/36%; 94/61%; 89/77%; 86/89%; 84/96% and 52/99% for scores 1,2,3,4,5,7, and 9, respectively. 2002 Kang et al. 5 3,889 Cardiac arrests and transfers to ICU Transfer to ICU – eCART > 54: sensitivity – 52.5% and specificity 88.5%. Cardiac arrests – eCART > 54: sensitivity – 80% e specificity -86%. 2016 Churpek et al. 15 56,649 Cardiac arrest risk and transfer to ICU Sensitivity – 65% and specificity – 93% 2014 Veiga & Rojas Performance of the rapid response systems Int J Cardiovasc Sci. 2019;32(5):536-539 Viewpoint failure in the rescue and prevention of adverse events, which results in inefficient systems. To minimize these issues, scores have been developed to improve detection of patients at risk. 10-11 These scores are easily executed and have a high reproducibility, and can predict elevated risks of cardiac arrest and need for ICU admission. 12 Nonetheless, scores that are based exclusively on vital signs have demonstrated limited accuracy, leading to lost opportunities to identify patients at risk of CA. Therefore, the use of electronic systems, as well as new models of stratification of patients at risk of clinical deterioration, has been gaining ground, with the aim of ensuring the early identification and appropriate treatment of these situations. 5,13-14 Churpek et al. 15 assessed a model of electronic data, which not only analyzes the patient’s vital signs, but also their laboratorial and demographic characteristics which, compared to models using vital signs alone, showed benefits in the early identification of patients at risk for CA, as well as their need for ICU admission. Similarly, Kang et al. 5 used an electronic score (eCART), in a study with 3,889 patients, which was able to identify risks at an earlier stage, compared to the usual RRT activation system. Combined outcomes of CA and transfers to ICU or death in hospitalization units were assessed by Churpek et al, in 2014, 15 in a study with data from five hospitals, which included 269,999 hospitalizations, and compared electronic data variables with the MEWS score. In all the outcomes assessed, the electronic scores were higher than MEWS score (p < 0.01) On the other hand, even though it increases the sensitivity of code activation, the structure proposed by the rapid response systems may be seen as a reactive response when the in-hospital patient is already at risk. There are some reports of proactive models, which are based on daily follow-up visits to patients considered at high risk, for example, those who have been recently transferred from intensive care units and surgery centers. 4,16 Other reports suggest the use of telemedicine units, whichwould provide support in patient care for the staff, while waiting for the rapid response team to arrive, leading to an earlier involvement of an intensivist in the management of the high-risk patient. 17 Table 1 describes the scores for clinical deterioration. 5,8,17,18-21 Results The practice of RRT is already well established when a decrease in the number of cardiac arrests is measured outside the ICU environment. Furthermore, there is also an influence between the time of implementation and the positive results, attributable to the organizational culture.