Predictor of Mortality in the Post-Cardiac Surgical ICU Found Accurate

The Cardiac Surgery Score (CASUS), which includes parameters for mechanical circulatory support, appears to be an accurate predictor of mortality in patients admitted to the ICU after cardiac surgery.

“CASUS is essentially a sequential organ failure score modified for unique pathophysiologies and therapies seen in patients admitted to the ICU after cardiac surgery,” said Brock Wilson, MD, FRCPC, critical care and cardiac anesthesiology fellow, and the lead author of the study, which was conducted at the University of Ottawa Heart Institute, in Ontario.

The researchers sought to validate the CASUS, which was first published in 2005 by Khosro Hekmat, MD, and his colleagues (Ann Thorac Surg 2005;79:1555-1562). Dr. Wilson presented the findings at the 2017 annual meeting of the Society of Critical Care Medicine (abstract 126).

“CASUS is a postoperative model, which includes six organ systems and 10 variables,” Dr. Wilson said. These include:

• partial oxygen pressure/fraction of inspired oxygen
• serum creatinine
• continuous venovenous hemofiltration/dialysis
• serum bilirubin
• platelet count
• neurologic status
• pressure-adjusted heart rate
• lactic acid
• intraaortic balloon pump
• ventricular assist device

Dr. Wilson explained the modification for mechanical circulatory support devices and the additional risk factor for renal replacement therapy in the CASUS model are not included in other ICU mortality risk prediction models.

Jason Adler, MD, MBA, FAAP, FCCP, medical director of the pediatric ICU at Joe DiMaggio Children’s Hospital, in Hollywood, Fla., and moderator of the abstract presentation session, cautioned: “Studies of modeling, such as Dr. Wilson’s, potentially help clinicians to gain a better understanding of the clinical factors associated with patient outcomes in cardiovascular surgical intensive care units. These scoring systems may augment clinical decision making, but they should not be used as a sole determinant in making decisions when it comes to care of the individual patient.”

Model Calibrates Well

The researchers performed a retrospective review of prospectively collected data from their institutional database looking at consecutive patients 18 years of age or older who underwent nonemergent major cardiac surgical procedures and required an ICU admission after cardiac surgery between July 1, 2012, and Sept. 30, 2015. The study excluded cases that required extracorporeal membrane oxygenation or minimal access aortic valve surgery. CASUS variables were collected on admission and postoperative days (PODs) 2, 4, 7 and 10 of the ICU stay.

There were 4,159 patients in the data set, with a mean age of 66.4±11.6 years; 72% were men, with a mean body mass index of 28.6±5.8 kg/m²; and 25.2% were New York Heart Association class III to IV. The rate of redo cardiac surgery was 7.5%, and 69.4% of cases were elective procedures. The predicted mortality as per EuroSCORE II was 3.1% (95% CI, 3.0%-3.3%).

Coronary artery bypass surgery (CABG) represented 49.7% of cases, with 8.9% being done off-pump. Isolated valve, combined CABG, and aortic valve, multivalve and thoracic aortic surgery represented 20.3%, 15.2%, 4% and 7.5%, respectively, of procedures. There were 19 ventricular assist devices (0.5%) and 49 intraaortic balloon pumps (1.2%) inserted. The rate of acute kidney injury leading to new dialysis was 3%. The median ICU stay was one day.

“The primary outcome of the study was ICU mortality,” said Dr. Wilson, who noted the overall ICU mortality rate was 1.5%.

Dr. Wilson said there was no significant difference between observed and expected ICU mortality when compared with a later publication by Dr. Hekmat’s group (Eur J Cardiothorac Surg 2010;38:104-109). “Hence, the model calibrated well.”

The model had good discrimination on admission and PODs 2 and 4, but on PODs 7 and 10 the model had slightly worse discrimination, although it was still considered fair. Dr. Wilson believed this was possibly due to patients becoming more homogeneous the longer they remain in the ICU.

Previous work by Dr. Hekmat and his group showed the additive CASUS model appeared to better predict ICU mortality compared with the EuroSCORE, which is based on preoperative variables, as well as models that use postoperative variables but were not developed specifically in the cardiac surgical population (Sepsis-related Organ Failure Assessment Score, Simplified Acute Physiology Score II and III, Acute Physiology and Chronic Health Evaluation II, and Multiple Organ Dysfunction Score).

Clinical prediction models do not replace clinical experience and judgment, and may not perform as well in all clinical situations, he warned. “Clinical prediction models can, however, help support clinical impressions as well as assist in family discussions by simplifying multisystemic variables into quantifiable prognostication,” he said.

Other potential benefits include identifying patients for enrollment in clinical trials or epidemiological analyses, and to assess the outcomes of new interventions. Moreover, he added, prediction models based on postoperative variables and ICU care could enhance reporting of provider performance and quality assurance.

He said further external validation of the CASUS model is required before its widespread adoption as a mortality risk prediction model in cardiac surgery ICUs.