ORLANDO – A new clinical prediction rule correlates well with performance status at 1 year after ICU hospitalization in patients over age 80.
Illness severity, comorbidities, baseline frailty, a primary diagnosis of stroke, and being male were all predictors of poor performance status at 1 year. A primary diagnosis of emergency coronary artery bypass grafting or valve replacement, a high baseline performance status, and being married were associated with good performance status at 1 year. The c-statistic for the model, a standard indicator of predictive power, was 0.811, a figure that indicates good predictive ability.
The findings from the REALISTIC 80 (Realities, Expectations ,and Attitudes to Life Support Technologies in Intensive Care for Octogenarians) study of 17 patient and illness characteristics allowed Dr. Daren Heyland, professor of medicine and epidemiology at Queen’s University, Kingston, Ont., and his coinvestigators in the Canadian Critical Care Trials Group (CCCTG), to conclude that eight factors were most predictive of performance status at 12 months for ICU patients aged 80 and over. REALISTIC 80 is a CCCTG project.
The values for the predictors are derived from responses to an online guided questionnaire called the ICU Workbook. The questionnaire is completed by patients’ family members or surrogates, and the responses are used to calculate the values that constitute the clinical prediction rule’s components.
Gathering this information may help health care providers and family members in end-of-life decision making, said Dr. Heyland. “For the very elderly, it is plausible that poor communication and decision making lead to overutilization of ICU resources and poor-quality end-of-life care,” he said. “Hopefully, with these strategies, we’ll improve clinical decision making and improve the quality of end-of-life care we provide for our older patients,” said Dr. Heyland, who is also director of CARENET, which hosts the online guided questionnaire and is an affiliation of Canadian researchers focused on end-of-life care. He spoke at the Society of Critical Care Medicine’s Critical Care Congress.
REALISTIC 80 enrolled 434 patients, aged 80-100 years (mean age, 84.6) who were admitted to ICUs at participating institutions.
Previous European and U.S. studies have shown an ICU mortality of 30%-35%, and an overall mortality rate of 60%-70% in the 12 months following ICU admission. In those studies, illness severity most strongly predicted short-term survival, and comorbidities best predicted long-term survival.
The primary outcome measure of REALISTIC 80 was the 12-month survival and health-related quality of life; “recovery from critical illness” was defined as a Palliative Performance Scale (PPS) score of greater than or equal to 60% at 12 months. Patients scoring at 60% on the 0%-100% scale of this functional status measure may have reduced ambulation, be unable to engage in housework or hobbies, have significant disease, need assistance, and be confused at times. An advantage of this scale, said Dr. Heyland, is that it eliminates survivorship bias in analyzing data, since a score of 0 is assigned to individuals who die.
About 50% of patients had died by 12 months; about 21% were alive, with a reduced health status below the threshold of 60 on the PPS; and about 29% were alive, with a PPS score above the predetermined quality of life threshold.
Dr. Heyland acknowledged that the study lost a significant number of participants – about 17% – to follow-up. The predictive model was derived from completed cases, and a sensitivity analysis using imputed data for missing patients showed that it retained its predictive value.
Dr. Heyland said the presence of advance directives didn’t appear to affect outcomes. “We subsequently in a different analysis showed that whether they had [a directive] or not, did not affect subsequent process or outcome of care in the ICU,” he said.
Dying in the ICU after days of mechanical ventilation or surviving with very low performance status “doesn’t sound like good quality of life to me, and it illustrates the challenge we have as clinicians in getting to what’s best for patients,” said Dr. Heyland. “Hopefully, prediction models will help, as well as better elicitation of authentic values and preferences from patients.”
The study was funded by the Canadian Institutes of Health Research and conducted under the auspices of the CCCTG and CARENET. The study investigators reported no other relevant financial disclosures.
On Twitter @karioakes