Value of APACHE II, SOFA and CardShock scoring as predictive tools for cardiogenic shock: A single-centre pilot study

Abstract

Aims: The aim of this study was to determine the value of the Acute Physiology and Chronic Health Evaluation II (APACHE II), Sequential Organ Failure Assessment (SOFA) and CardShock scoring systems in predicting the risk of in-hospital, 30 day and 3 year mortality in patients with cardiogenic shock (CS).

Methods: This was a single-centre observational study conducted between May 2016 and December 2017. Data from consecutive patients with CS admitted to the intensive cardiac care unit (ICCU) were included in the analysis.

Results: The study group comprised 63 patients with CS {median age 71.0 [interquartile range (IQR), 59-82]; 42 men}: 32 patients with ischaemic and 31 with non-ischaemic aetiology. The median APACHE II, SOFA and CardShock scores were 13 (IQR, 9.9-19.0) points, 8.0 (IQR, 6.0-10.0) points and 3.0 (IQR, 2.0-5.0) points, respectively. The in-hospital, 30 day and 3 year mortality rates were 39.7%, 41.3% and 77.8%, respectively. APACHE II and SOFA scores were significantly higher in the group of patients who died at 30 days (P = 0.043 and P = 0.045, respectively). The CardShock score was higher in patients with CS who died in hospital (P = 0.007) and within 30 days (P = 0.004). No score was statistically significant for 3 year mortality. Area under the curve (AUC) analysis showed that the CardShock score had the highest value in predicting in-hospital and 30 day mortality relative to APACHE II and SOFA, with a cut-off score of 5 points [AUC: 0.70; 95% confidence interval (CI): 0.59-0.81; P = 0.001] and 4 points (AUC: 0.71; 95% CI: 0.60-0.82; P < 0.001), respectively. The Bayesian Weibull model demonstrated the utility of all scales in estimating short-term risk in patients with CS, with the impact of APACHE II and SOFA on patient life expectancy decreasing to a non-significant level at approximately 32 days and CardShock at 33 days. The forest plots derived from the Bayesian logistic regression analysis show significant estimated coefficients with 94% highest density interval (HDI) for in-hospital and 30 day mortality. The use of invasive or non-invasive ventilation, a higher heart rate and a less negative fluid balance showed an unfavourable prognosis. Survival was associated with being in the pre-CS class, with a higher glomerular filtration rate and a higher platelet count.

Conclusions: APACHE II and SOFA could be used for the risk stratification of patients with CS admitted to the ICCU. CardShock proved to be a more appropriate tool for assessing short-term prognosis in patients with CS of all aetiologies, suggesting that there is potential for its promotion for use in daily clinical practice.

Keywords: APACHE II; Bayesian Weibull model; CardShock; SOFA; cardiac care facilities; shock, cardiogenic.

Figure 1

Flow diagram of study patients. ICCU, intensive cardiac care unit.

Figure 2

Receiver operating characteristic (ROC) curves of (A, D, G) Acute Physiology and Chronic Health Evaluation II (APACHE II) score for predicting in‐hospital, 30 day and 3 year mortality; (B, E, H) Sequential Organ Failure Assessment (SOFA) score for predicting in‐hospital, 30 day and 3 year mortality; and (C, F, I) CardShock score for predicting in‐hospital, 30 day and 3 year mortality.

Figure 3

Graphic representation of the disparity in hazard rates using the Bayesian Weibull model. Sub‐figure (A) presents the disparity for APACHE II, sub‐figure (B) displays the corresponding disparity for SOFA and sub‐figure (C) presents the disparity for CardShock. The dashed lines in the sub‐figures indicate the change points of a hazard rate difference function with 95% confidence intervals (CIs).

Figure 4

The Kaplan–Meier estimate (blue line) with its 95% confidence interval (light blue shaded area) against the Weibull model (red line) with its 94% highest density interval (light red shaded area). Sub‐figure (A) presents the estimation for APACHE II, sub‐figure (B) displays the estimation for SOFA and sub‐figure (C) presents the estimation for CardShock.

Figure 5

The forest plots derived from the Bayesian logistic regression analysis showing significant estimated coefficients with a 94% highest density interval (HDI) for in‐hospital (A) and 30 day (B) deaths. GFR, glomerular filtration rate; HR, heart rate; IV, invasive ventilation; NIV, non‐invasive ventilation; PLT, platelet; pre‐CS, pre‐cardiogenic shock stage.