CHANGE risk score identifies poststroke cognitive impairment
The CHANGE (chronic lacunes, hyperintensities, age, nonlacunar cortical infarcts, global atrophy and education) score is a newly developed tool that can reliably identify patients with risks of poststroke cognitive impairment, (PSCI) a new study has found.
“The performance of CHANGE has been demonstrated in both internal and external cohorts, and shown to be consistently comparable and stable with good utility,” said researchers.
Patients with PSCI or no cognitive impairment (NCI) were recruited from a tertiary stroke centre in Singapore. Variables that differentiated the two groups were included in binary logistic regression models for further testing. Those that were not statistically significant or were clinically irrelevant were dropped.
The resulting CHANGE risk score was internally validated in a separate cohort from the stroke clinic of the National Neuroscience Institute in Singapore. A third cohort of Chinese subjects from the Stroke Registry Investigating Cognitive Decline study in Hong Kong was used for external validation.
The risk score creation included data from 209 subjects (mean age 61.67±12.46 years; 32.1 percent female) of whom 37.3 percent (n=78) developed PSCI at 3 to 6 months. [Sci Rep 2017;7:12441]
The initial regression model included acute infarcts, intracranial stenosis, chronic lacunes, gender, education, age, white matter hyperintensity (WMH), hypertension, global cortical atrophy (GCA) and atrial fibrillation. Gender, atrial fibrillation and hypertension were eventually dropped, yielding the final 14-point risk score that could predict PSCI.
“The variables in CHANGE have been corroborated by the existing literature as being important risk factors for PSCI,” according to researchers who cited a 2005 review and a 2016 study that identified the same risk factors outlined in the study.
The risk score yielded an area under the receiver operating characteristic curve (AUROC) of 0.82 (95 percent CI, 0.76 to 0.88). It had an accuracy of 73.7 percent, sensitivity of 74.4 percent and specificity of 73.3 percent with an optimal cutoff score of ≥7.
Eight percent of subjects in the lower tertile of the risk score scale (0 to 4 points) had PSCI compared to 49 percent in the middle (5 to 9 points) and 75 percent in the upper (10 to 14 points) tertiles.
The CHANGE score was internally validated in 185 subjects, of whom 18.9 percent (n=35) developed PSCI. CHANGE remained significantly predictive of PSCI (p<0.001), with an AUROC of 0.78 (0.71 to 0.85), accuracy of 77.3 percent, sensitivity of 48.5 percent and specificity of 84.0 percent.
Particularly, the age (p=0.068), GCA (p<0.001) and chronic lacunes (p=0.075) components of the CHANGE score showed moderate statistical significance.
External validation was performed in 693 patients, of whom 50.8 percent (n=352) developed PSCI. Again, CHANGE was a significant predictor of PSCI (p<0.001), with an AUROC of 0.75 (0.71 to 0.79), accuracy of 67.6 percent, sensitivity of 71.5 percent and specificity of 63.6 percent.
The age (p<0.001), low education (p<0.001), WMH (p=0.023), acute infarct (p=0.022) and chronic lacunes (p=0.001) components were statistically significant in this cohort.
“CHANGE would be useful in identifying stroke inpatients at significant risk for delayed PSCI, especially those that show good functional recovery and would otherwise have been discharged,” said researchers.
“At-risk patients may then be prioritized for close clinical monitoring or prophylactic interventions with medication, rehabilitation or both,” they added.