Jayant Mahadevan1, Bheemsain Tekkalaki2, Venkata Lakshmi Narasimha1, Ramdas Ransing3, Chittaranjan Andrade4. 1. Department of Psychiatry, National Institute of Mental Health and Neurosciences, Bengaluru, Karnataka, India. 2. Department of Psychiatry, Jawaharlal Nehru Medical College, Belagavi, Karnataka, India. 3. Department of Psychiatry, B. K. L. Walawalkar Rural Medical College, Ratnagiri, Maharashtra, India. 4. Department of Psychopharmacology, National Institute of Mental Health and Neurosciences, Bengaluru, Karnataka, India.
Sir,Johnson et al.[1] used the Clinical Interview Schedule-Revised (CIS-R) to screen 208 pregnant women for mental disorders; 12 women screened positive. We are concerned that the authors used a version of the CIS-R that had been translated without validation and without the establishment of the psychometric properties of the translation for the selected cut-off. Validation was also necessary for the population of interest, pregnancy, because nonspecific, pregnancy-related factors could have confounded the interpretation of the responses and hence the scores. The authors cited a Malaysian study in support of their cut-off when Indian studies are available;[2] there is no assurance that the psychometric properties of the Malaysian translation apply to pregnant Indian women studied using a new, unvalidated translation.Given that only 12 women screened positive, the authors could have attempted to establish the presence or absence of a definitive disorder through a formal psychiatric interview. This would have provided information about the Positive Predictive Value (PPV) of the translated instrument, in this special population, using the assigned cut-off.We suggest that, unless specifically necessary, continuous variables should not be converted into categorical variables [Table 1 in the authors’ paper]. This is because categories are often defined with arbitrary boundaries that do not exist in nature, because some values in adjacent categories may be closer to each other than some values within categories, and because categorisation of a continuous variable leads to loss of precision and hence statistical power.[345]Finally, we suggest that caution be exercised in the use and interpretation of statistical tests when the sample size is as disparate as n = 12 and n = 196, if only because a very small sample (n = 12) may not be representative of the population that it is expected to represent.