The article by Jaywant et al, Subgroups Defined by the Montreal Cognitive Assessment Differ in Functional Gain During Acute Inpatient Stroke Rehabilitation, published in Archives of Physical Medicine and Rehabilitation 2020;101:220-226 https://www.archives-pmr.org/article/S0003-9993(19)31059-7/fulltext
, contained errors. The article found that subgroups based on scores on the MoCA were associated with mean relative FIM gain (mRFG) and mean relative FIM efficiency (mRFE) in inpatient stroke rehabilitation. The analysis included prospectively collected and retrospectively collected data (via electronic medical record capture) from an inpatient rehabilitation unit. It recently came to the authors’ attention that some of the admission FIM scores captured retrospectively from the medical record were not accurate.
The authors corrected these scores and re-ran the analyses. All of the primary and secondary findings are consistent and none of the discussion points or conclusions have been affected. That is, MoCA subgroups continue to significantly predict mRFG and mRFE after accounting for age, sex, education, stroke number, and stroke severity. The authors continue to see a difference in the proportion of individuals who meet the minimal clinically important difference (MCID) on the FIM by MoCA subgroup (although the P-value is different). The regression models with MoCA subgroups continue to account for more variance than regression models using the Cognitive-FIM.
Below are updated findings. Given that some admission FIM scores have changed, the regression unstandardized β, standard errors, 95% confidence intervals, and P
-values have changed. The significance of the predictors remains largely unchanged. The exception is the NIHSS, which is now a predictor in the mRFG model. The updated regression models for MoCA subgroups now account for greater variance in predicting mRFG and mRFE (Adjusted R2
for mRFG is now .20; Adjusted R2
for mRFE is now .17).
1Update to Table 2: (admission FIM scores are the only changes)
1Update to Table 3 (mRFG):
1Update to Table 4 (mRFE):
Update to Comparison of FIM MCID by MoCA subgroup. The test statistic and the P-value of the MCID comparison by MoCA subgroup is changed (X2 = 3.68, P=.15) but the pattern of percentage difference per subgroup holds: 35% meet the MCID in the Moderately Impaired group, 42% meet the MCID in the Mildly Impaired group, and 49% meet the MCID in the Normal group.
Update to Exploratory comparison of MoCA subgroups and the cognitive-FIM. The updated regression models continue to show that the cognitive-FIM accounts for a smaller proportion of variance in predicting mRFG (F=8.64, P<.001, Adjusted R2 =.17) and mRFE (F=6.76, P<.001, Adjusted R2 =.14). Within each model, the cognitive-FIM continues to predict mRFG (unstandardized beta=.73, SE=.16, 95% confidence interval, 0.43 to 1.04; t=4.71, P<.001) and mRFE (unstandardized beta=.12, SE=.02, 95% confidence interval, 0.07 to 0.16; t=5.38, P<.001).
Published online: May 03, 2020
© 2019 by the American Congress of Rehabilitation Medicine