Swapnajeet Sahoo1, Sutanaya Pal2, Avisha Mahla2, Satish Suhas3, Ekta Franscina Pinto4, Chittaranjan Andrade4. 1. Department of Psychiatry, Postgraduate Institute of Medical Education and Research, Chandigarh, India. E-mail: swapnajit.same@gmail.com. 2. Department of Psychiatry, Medical College Baroda, Vadodara, Gujarat, India. 3. Department of Psychiatry, National Institute of Mental Health and Neurosciences, Bengaluru, Karnataka, India. 4. Department of Psychopharmacology, National Institute of Mental Health and Neurosciences, Bengaluru, Karnataka, India.
Sir,In a post hoc analysis of data on the use of the Everyday Abilities Scale of India (EASI) and the Hindi/Bharmouri versions of the Mini-Mental State Examination (H/BMSE) as screening instruments to detect dementia in elderly, mostly illiterate persons in Himachal Pradesh, India, Raina et al.[1] found a significant negative correlation between the EASI and the H/BMSE. They concluded that the EASI may be used as an alternative to the H/BMSE as a dementia screen.We disagree with this conclusion. The value of the correlation was low (−0.252), indicating that the EASI explains <6.5% of the variance in the H/BMSE. Clearly, the EASI and the H/BMSE measure substantially different constructs. Furthermore, the method used to obtain this correlation was wrong; given a large number of outliers in the data shown in the figure in their paper, a Spearman's procedure should have been adopted, not a Pearson's procedure. In addition, the statistical analysis was inadequate. If the authors wished to evaluate the usefulness of the EASI in screening for dementia with the H/BMSE as the reference, they should have employed receiver operating characteristics (ROC) analysis to identify an EASI cutoff that identifies H/BMSE dementia with the best sensitivity and specificity; the correlation between EASI and H/BMSE scores is of little to no value. Most important of all, given that they clinically evaluated the subjects, the best case for the EASI would require a ROC analysis against clinically diagnosed dementia because the H/BMSE cutoff is not a gold standard.Also of concern, the authors provided no information whatsoever about the sample on which the presented data are based. Given that the entire paper served solely to present one correlation coefficient, there was no need for an entire table to present this number, let alone a figure to show the relationship that was captured by the number. In fact, the entire paper could have been condensed to a brief letter to the editor.As a final note: the correlation between EASI and H/BMSE was presented as −2.52 in both abstract and text. A correlation coefficient can never lie outside the boundaries of −1.00 to +1.00.