Corresponding Author R.F. Baggaley, Department of Epidemiology and Population Health, London School of Hygiene & Tropical Medicine, Keppel Street, London WC1E 7HT, UK. Tel.: 02079272469; Fax: 02076374314; E-mail: email@example.com
Objective The K10 and K6 are short rating scales designed to detect individuals at risk for depressive disorder, with or without anxiety. Despite being widely used, they have not yet been validated for detecting postnatal depression. We describe the validity of these scales for the detection of postnatal depression in Burkina Faso.
Method The English language version of the K10 questionnaire was translated into West African French and local languages for use in Burkina Faso. Scores for 61 women were compared with the diagnostic interview made by a local psychiatrist within 3 days of administering the K10.
Results Clinical assessment found that 27 (44%) women were probable cases of depression. Internal consistency of K10 and K6 scores, defined by Cronbach’s alpha coefficient, was 0.87 and 0.78, respectively, indicating satisfactory reliability. The performance of the scores was not significantly different, with areas under the curve of 0.77 and 0.75 for the K10 and K6, respectively. To estimate prevalence of depression, we suggest cut-offs of ≥14 for the K10 and between ≥9 and ≥11 for the K6 for identifying women at high risk of depression. At ≥14, the K10 has 59% sensitivity, 91% specificity; at ≥10, the K6 has 59% sensitivity and 85% specificity.
Conclusion This study suggests that K10 and K6 are reasonably valid measures of depression among postpartum women in Burkina Faso and can be used as relatively cheap tools for estimating prevalence of postnatal depression in developing countries.
Objectif Les K10 et les K6 sont des échelles d’évaluation rapide conçues pour détecter les individus à risque de trouble dépressif avec ou sans anxiété. Malgré leur emploi étendu, elles n’ont pas été encore validées pour la détection de la dépression postnatale. Nous décrivons ici, la validité de ces échelles pour la détection de la dépression postnatale au Burkina-Faso.
Méthode La version anglaise du questionnaire K10 a été traduite en langues locales ouest africaines et française pour être utilisée au Burkina-Faso. Les scores pour 61 femmes ont été comparés à l’interview de diagnostique réalisée par un psychiatre local dans les trois jours suivant l’application du K10.
Résultats L’évaluation clinique a révélé que 27 femmes (44%) étaient des cas probables de dépression. La consistance interne des scores de K10 et K6 définie par le coefficient alpha de Cronbach’s était de 0,87 et 0,78 respectivement, indiquant une fiabilité satisfaisante. La performance des scores n’était pas significativement différente, avec des aires sous la courbe (AUC) de 0,77 et de 0,75 pour le K10 et le K6, respectivement. Pour estimer la prévalence de la dépression, nous suggérons des seuils limites ≥14 pour le K10 et comprises entre ≥9 et ≥11 pour le K6 pour identifier les femmes à risque élevé pour la dépression. Pour un score ≥14, le K10 a une sensibilité de 59% et une spécificité de 91%; à ≥ 10, le K6 a une sensibilité de 59% et une spécificité de 85%.
Conclusion Cette étude suggère que K10 et K6 sont des mesures raisonnablement valides pour la dépression chez les femmes qui ont accouché au Burkina-Faso et peuvent être utilisés comme outils relativement bon marché pour estimer la prévalence de la dépression postnatale dans les pays en voie de développement.
Objetivo K10 y K6 son escalas de clasificación corta diseñadas para detectar individuos a riesgo de desórdenes depresivos, con o sin ansiedad. A pesar de ser ampliamente utilizadas, no han sido aún validadas para detectar la depresión posparto. Aquí describimos la validez de estas escalas para la detección de la depresión posparto en Burkina Faso.
Método Se tradujo la versión inglesa del cuestionario del K10 al francés y a las lenguas locales de África Occidental, para su uso en Burkina Faso. Se compararon los puntajes de 61 mujeres con la entrevista de diagnóstico realizada por el psiquiatra local dentro de los 3 días de administración del K10.
Resultados La evaluación clínica encontró que 27 (44%) mujeres eran casos probables de depresión. La consistencia interna de los puntajes del K10 y el K6, definidos por el coeficiente alfa de Cronbach, era 0.87 y 0.78 respectivamente, indicando una fiabilidad satisfactoria. El desempeño de los puntajes no era significativamente diferente, con áreas bajo la curva de 0.77 y 0.75 para el K10 y K6, respectivamente. Para estimar la prevalencia de depresión, hemos sugerido puntos de corte de ≥14 para el K10 y entre ≥9 y ≥ 11 para el K6, con el fin de identificar mujeres con alto riesgo de depresión. A ≥ 14, el K10 tiene un 59% de sensibilidad, 91% de especificidad; a ≥ 10, el K6 tiene un 59% de sensibilidad y un 85% de especificidad.
Conclusión Este estudio sugiere que el K10 y el K6 son medidas razonables para la depresión posparto en mujeres de Burkina Faso, y pueden utilizarse como herramientas relativamente baratas para estimar la prevalencia de esta condición en países en vías de desarrollo.
As awareness of depression in developing countries has increased, prevalence of depression and changes in prevalence over time need to be monitored. As gold standard tools for detecting depression, such as the Composite International Diagnostic Interview, are not feasible for large-scale epidemiological studies in developing countries due to the time and professional expertise required, easily administered and quick screening tools are required.
A number of screening questionnaires for depression exist, including the General Health Questionnaire, the Self Reporting Questionnaire, the Edinburgh Postnatal Depression Scale (EPDS) and the K10. The K10 is a short questionnaire developed to screen for depression, with or without anxiety, by determining a composite score based on participants’ responses (Kessler et al. 2002). It is ideal for monitoring prevalence of depressive disorders on a large scale. The K10 has been derived on the basis of a systematic process, extensively tested and validated in various regions (Andrews & Slade 2001; Kessler et al. 2002; Furukawa et al. 2003). It has strong psychometric properties and excellent ability to distinguish cases from non-cases. Its wording is simple, with short questions and clear response categories. Furthermore, it is the instrument being used in the World Mental Health Surveys, which include centres in West Africa, and it has recently been validated for use in Nigeria (Gureje et al. 2006). Despite this, relatively few studies have been conducted evaluating its use in Africa and there is currently no threshold score for determining detection of depression for use in the field in these settings. The K10 has not previously been validated as a tool for use in Francophone Africa or in predicting postnatal depression in any setting. The objective of this sub-study was to investigate the validity of the K10 and K6 in Burkina Faso for the detection of postnatal depression.
K10 and K6 scores
The K10 is a 10-item scale with five response categories ranked on a five-point scale, with the score being the sum of these responses (details in the Appendix). The K6 consists of a subset of six of these items. Scores are 0–40 for the K10 and 0–24 for the K6. The WHO-issued English language version of the K10 questionnaire was translated into West African French and the local languages of Mooré and Dioula using the standardised WHO translation and back-translation protocol (WHO) and pre-tested before being finalised (http://www.who.int/substance_abuse/research_tools/translation/en/index.html).
The K10 questionnaire was administered within a larger cohort study of postpartum women and women with early pregnancy loss (miscarriage, termination and ectopic pregnancy) by trained interviewers at 3, 6 and 12 months post-pregnancy (Filippi et al. in press). Interviewers took a 1-day training course with a local psychiatrist on the rationale and methods for the K10.
Of the cohort participants, 61 additionally completed a separate, diagnostic interview with a local psychiatrist, to provide the ‘gold standard’ for assessment of depression against which to compare the K10. The psychiatrist was blind to the subjects’ K10 scores. Women were selected in an attempt to over-sample from those with higher K10 scores in their most recent interview to gain a larger sample of probable cases of depression, but otherwise were chosen at random. Diagnoses were based on the ICD-10 criteria for Mental and Behavioural Disorders (http://www.who.int/classifications/icd/en/GRNBOOK.pdf), using a checklist of factors for consideration to facilitate standardisation of diagnoses. Women were classified as cases if diagnosed with depression, regardless of severity. Any woman diagnosed as depressed was given a follow-up appointment with the psychiatrist for further evaluation and treatment. The interview was taken within 3 days of a K10 assessment, at either the 3 month (n = 29, 48%) or 6 month (n = 32, 52%) postpartum interview.
Data were analysed using stata version 9. Cronbach’s alpha tests internal consistency by assessing how well a set of items measures a single, one-dimensional outcome by determining their correlation (Cronbach 1951). Alpha ≥0.70 was considered satisfactory (Nunnally & Bernstein 1994). Reliability of each item was also assessed by measuring the correlation between each item and the overall K10 and K6 scores using the Pearson product-moment correlation. To test whether the K10 and K6 are valid indices predicting risk of depression, the Kruskal–Wallis test was used to evaluate whether K10 and K6 scores were significantly different for cases and non-cases. Receiver Operating Characteristic (ROC) curves were constructed to investigate optimum cut-off scores for identifying cases.
Ethical approval was granted by the ethical committees of the London School of Hygiene and Tropical Medicine and Centre Muraz, Bobo-Dioulasso, Burkina Faso. Informed consent was obtained from all informants.
All 61 respondents completed all items of the K10 questionnaire. The age range was 17–46 years, mean 26 (SD = 7) years; 39 (64%) were married, and 28 (46%) had completed six or more years of education. The mean K10 and K6 scores were 10.7 and 7.0, respectively, while the clinical assessment found that 27 (44%) women were cases.
Cronbach’s alpha coefficient was 0.87 and 0.78 for K10 and K6 scores, respectively, indicating satisfactory reliability. Item-total correlation, assessed using the Pearson product-moment statistic, varied from 0.44 to 0.83 for the K10 scale and from 0.44 to 0.81 for the K6 scale, and all correlations were highly significant (P = 0.0005 or lower, data not shown), indicating satisfactory reliability of all items.
Cases had significantly higher K10 scores than non-cases [mean 14.3 (SD = 6.9) and 7.8 (SD = 6.2), respectively (P = 0.0003)]. Results were similar for the K6 score [mean 9.3 (SD = 4.4) and 5.3 (SD = 3.8), respectively (P = 0.0008)]. ROC curves plotting the relationship between each score and the clinical diagnosis of depression were not significantly different for the K10 and K6 (P = 0.239), with areas under the curve of 0.77 and 0.75, respectively (Figure 1). A ≥6 cut-off on the K10 score, the equivalent of which is used elsewhere to predict moderate risk of depression (Andrews & Slade 2001), gave sensitivity 85% and specificity 41%. The optimum cut-off for defining cases of depression for the K10 score was ≥12 according to the ROC curve, giving the optimum values of sensitivity (74%) and specificity (76%), but ≥14 gave the greatest percentage correctly classified (77% for ≥14 compared with 75% for ≥12). The ≥14 cut-off gave sensitivity 59% and specificity 91%. For the K6, cut-offs of ≥9 or ≥10 could have been used because no respondent had a K6 of 9 and so all outcomes were identical for the two cut-offs. Cut-offs of ≥9, ≥10 and ≥11 each give the highest percentage correctly classified (74%), but ≥9 and ≥10 perform better in the ROC analysis.
This small-scale validation study suggests that the West African French and local language versions of the K10 and K6 questionnaires are reasonably valid indicators of depression among postpartum women in Burkina Faso. It indicates higher K10 and K6 cut-off scores for depression for this setting compared to elsewhere (Andrews & Slade 2001). Such differences may be due to variation in the discriminatory power of different items in the K10 and K6 scores or differences in baseline prevalence of depression across settings, among other reasons (Goldberg et al. 1998).
While a cut-off of 12 for the K10 performed better in the ROC analysis, this was calculated for a group over-sampling for depression and therefore an artificially high prevalence of depression. A 14 cut-off may be preferable in this setting; for service provision where resources are scarce, positive predictive value is often a crucial factor. Similarly, a 10 cut-off for the K6 is recommended. The K10 measures risk of depression rather than attempting to diagnose it, and if it is to be used to identify women at risk for depression for referral for further evaluation, a lower cut-off would be recommended. In this particular context, the tool is used to monitor prevalence of depression and is thus used on a large sample; therefore the brevity of the questionnaire and achieving higher levels of sensitivity and specificity need to be weighed against each other.
These cut-offs provide categorical assessment of depression (dividing samples into those at high and low risk of depression), but dimensional assessments based on mean K10 and K6 scores are also recommended for epidemiological and sociological research, as they retain greater statistical power (Kessler 2002). The use of such an easily administered tool for mental health research will play a vital role in raising awareness of the extent and impact of depression within all resource-poor settings.
We thank the entire Immpact team in Burkina Faso, including all the interviewers, and particularly Thomas Ouédraogo and Nicolas Méda, as well as all the women who participated in the study. This work was undertaken as part of an international research programme – Immpact (http://www.abdn.ac.uk/immpact), funded by the Bill & Melinda Gates Foundation, the Department for International Development, the European Commission and USAID. The funders have no responsibility for the information provided or views expressed in this paper. The views expressed herein are solely those of the authors.
K10 and K6 score items
Questions from the English language version of the K10 and K6, upon which the West African French translation was based:
In the last 30 days, about how often did you feel…
(a) …tired out for no good reason?
(b) …nervous?1 (if never go to question d2)
(c) …so nervous that nothing could calm you down?
(e) …restless or fidgety?1 (if never go to question g2)
(f) …so restless that you could not sit still?
(g) …depressed? (if never go to question i2)
(h) …so depressed that nothing could cheer you up?1
(i) …that everything was an effort?1
1Denotes questions used for the K6 score.
2Replies of more than ‘never’ i.e. a score more than 0 for questions c, f or h would be dependent on a reply of more than ‘never’ for the preceding question in each case. Therefore if a respondent replied ‘never’ to question b (or question e or g) then they skipped question c (or question f or h), which was automatically coded as 0 i.e. never. This again saves time when administering the questionnaire.
Scoring of answers:
Never = 0
Occasionally = 1
Sometimes = 2
Most of the time = 3
All of the time = 4
The total K10 and K6 scores are computed using the equation:
with the result being rounded to the nearest whole number. If any required item has not been completed, it is excluded from the calculation and not counted as a valid item. Number of items is 10 for the K10 and six for the K6. If more than one of the items constituting the total score is not valid then the total score is set as missing.