Data analysis/statistical methods

We entered the data in Microsoft Excel and analyzed the data with SPSS version 22. Analysis primarily involved descriptive statistics (proportion, mean, median, IQR, range). We categorized the patients using the GDS-15 scale into three groups; no depression, mild depression and severe depression with the cut off values of 4/5 and 8/9 respectively. Other demographic variables were dichotomized, as: educational status (no formal education vs any formal education), occupation (employed vs unemployed) and marital status (married vs single at present due to any cause). Categorical variables were compared with the chi square test. The Independent-samples Kruskal-Wallis test was used to compare the continuous variables with the three categorical variables. Multinomial regression analysis was applied to find specific factors associated with geriatric depression. The factors with P-value < 0.2 in former bivariate analysis were included for the analysis. Statistically, when P-value are < 0.05, the independent variables are likely to have impact on the dependent variable, depression.

A Kappa value between the research assistant and one of the authors (RS, APS) was performed to assess inter-rater reliability in 20 randomly selected cases. The Kappa result is interpreted as follows: values ≤0 as indicating no agreement and 0.01–0.20 as none to slight, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as substantial, and 0.81–1.00 as almost perfect agreement [36].

The validity of the five-question depression screen was compared to the standard 15 question depression screen by calculating sensitivity, specificity, positive predictive value and negative predictive values using a cutoff score of 2/3, and were reported as proportions using 95% confidence intervals (95% CIs) [37]. Sensitivity or true positive rate is referred to the proportion of patients who scored positive for depression on the GDS-5 and GDS-15. Specificity or true negative rate is the proportion who had scored negative for depression with the GDS-5 and GDS-15. Positive predictive value is the proportion of true positives among all positives and negative predictive value is proportion of true negatives among all negatives. The construct validity was measured by using the Spearman’s correlation test that compares GDS-5 with GDS-15. The Spearman correlation coefficient, r_s of + 1 indicates a perfect association of ranks, a r_s of zero indicates no association between ranks and a r_s of − 1 indicates a perfect negative association of ranks [38]. In order to study the discriminant validity, the receiving operating characteristic (ROC) curve was created, and the area under the curve (AUC) was calculated. The closer the AUC value is to 1.0, the greater the instrument’s ability to differentiate between depressed and nondepressed patients. An AUC higher than 0.75 confers to the tool a moderate discriminative validity; while an excellent one is demonstrated by a value ≥0.90.

The agreement between the GDS-15 and the GDS-5 was also calculated with a kappa coefficient. A reliability analysis was carried out on the GDS-15 scale items and the GD − 5 scale. The reliability of the GDS score was assessed in terms of internal consistency by calculating Cronbach’s alpha. Alpha values closest to 1 are considered best: excellent if > 0.9, good if > 0.8, and acceptable if > 0.7 [39].