Unrecognized geriatric depression in the emergency Department of a Teaching Hospital in Nepal: prevalence, contributing factors, and metric properties of 5 item geriatric depression scale in this population

BACKGROUND

Depression is prevalent but poorly recognized in the Emergency Department (ED). We aimed to identify the frequency of unrecognized geriatric depression and its possible determinants in the ED using the 15-item geriatric depression scale (GDS). We also aimed to explore the performance of the shorter, five-item GDS as compared to the 15-item GDS.

METHODS

This was a cross-sectional study of the ED patients ≥ 60 years. The previously validated Nepali version of GDS-15 screened the sample into "no", "mild-moderate" and "severe" depression using cutoff values of 4/5, and 8/9 respectively. Demographic and socioeconomic factors, comorbidities and health seeking behaviors were studied and the relation was assessed with binary (Chi-square and Kruskal-Wallis test) and multinomial regression analysis. The performance of GDS-5 was compared with the GDS-15 as standard. Cronbach's alpha was calculated to investigate reliability. Validity was assessed by calculating sensitivity, specificity, positive predictive value, negative predictive value, Spearman's correlation, receiver operating characteristic curve, and kappa coefficient.

RESULTS

Two hundred eighty patients were enrolled with an overall prevalence of unrecognized depression of 45.7% [104 (81.3%) mild-moderate depression, and 24 (18.8%) severe depression]. The mean age of the sample was 71.36 with female predominance (61%), and 82.5% were illiterate. In the binary analysis, those who had more pain, visited the ED more often, had musculoskeletal diseases and sleep problems, mobility problems, visited local healers previously, and who reported self-perceived "heart and mind" disease showed statistically significant differences among the three categories. In multinomial regression analysis, visits to local healers, sleep problems and frequency of pain were significantly related to depression. The sensitivity, specificity, area under curve and Spearman's correlation of GDS-5 were 75.8%, 96%, 0.919, and 0.827 respectively. Cronbach's alpha for GDS-5 was low (0.416), therefore a new version was proposed which improved the sensitivity to 90.6% and Cronbach's alpha to 0.623.

CONCLUSIONS

Unrecognized geriatric depression was highly prevalent which urges the need for ED-based interventions for screening and referral. The proposed brief GDS-5 correlated well with the GDS-15 with better validity and internal reliability and offers a more expeditious form of screening for geriatric depression in emergency settings in Nepal.

Psychometric features of an assessment instrument with likert and dichotomous response formats

OBJECTIVE

To assess the psychometric properties of a Likert-formatted assessment instrument after altering the responses to a dichotomous format.

DESIGN AND SAMPLE

This methodological study used a 15-item instrument to obtain data from 183 participants who responded in both Likert and dichotomous formats. Response sets from each format were compared.

MEASURES

Each response set underwent factor analysis, Kuder-Richardson 20, Cronbach's α coefficient, item-total correlation, and parallel form equivalence tests.

RESULTS

Factor loads of the instrument varied between .362 and .754 when responses were Likert-formatted and between .370 and .713 when responses were dichotomous. The Cronbach's α coefficient with Likert-formatted responses was .858; the Kuder-Richardson 20 coefficient of the dichotomous responses was .827. Parallel form equivalences were significant at the level of r = .753.

CONCLUSIONS

The instrument had valid results when either Likert or dichotomous responses were obtained.