February 19, 2025
Measuring psychological well-being in a danish pregnancy cohort using the self-reported WHO-5 index | BMC Psychology

Questionnaires

The WHO-5 index

The WHO-5 index is derived from the WHO-10 index, which consists of the 10 most valid items from a questionnaire of 28 items used in a WHO multicentre study in eight European countries [17]. The index measures subjective psychological well-being using five items: Over the past two weeks; (1) “… I have felt positively charged”, (2) “… I have felt calm and relaxed”, (3) “… I have felt active and vigorous”, (4) “… I woke up feeling fresh and rested”, (5) “… my daily life has been filled with things that interest me”. When evaluating the WHO-5, the respondent is to score the five items using a scale that ranges from 0 (at no time) to 5 (all the time). This generates an overall sum score ranging from 0 to 25, that is to be multiplied by four to provide a standardized well-being percentage score from 0 to 100, with 0 as the lowest and 100 as the highest possible well-being percentage score [14]. A score of ≤ 50 is indicative of reduced well-being, and is suggestive of further testing of depressive symptoms, and a score of ≤ 28 is indicative of clinical depression [18].

The major depression inventory

The Major Depression Inventory (MDI) is a self-report tool developed to cover depressive symptoms in DSM-V major depression as well as in ICD-10 mild, moderate, or severe depression [15]. The MDI includes 10 items, of which three are classified as core items (1) feeling sad, (2) loss of interest, (3) lack of energy. The respondent is to score the items based on the past 14 days using a scale that ranges from 0 (at no time) to 5 (all the time), where a score of 4 or 5 indicates presence of a symptom [19]. In both DSM-IV and ICD-10 the items of depressed mood and loss of interests are considered as core symptoms of depression. MDI-2 consists of the first two core items of the MDI (feeling sad and loss of interest) and putatively captures core symptoms of depression according to DSM-V (depressed mood and markedly diminished interest or pleasure). A higher score indicates a higher severity of depression [19].

Population

The copenhagen pregnancy cohort (CPC)

This study population and its data were obtained from the Copenhagen Pregnancy Cohort (CPC) from September 2012 until December 2021. After booking a first trimester ultrasound scan, the pregnant women received a message in a digital mailbox providing an encrypted link to an online clinical questionnaire. The self-reported data is used for medical records and research purposes after permission from health authorities, and includes items about maternal and socio-demographic characteristics, obstetric history, lifestyle factors, mental health, etc. The CPC continuously collects patient-reported outcomes from pregnant women registered at the Department of Obstetrics at the Copenhagen University Hospital – Rigshospitalet. The department serves as a local birth facility in Copenhagen as well as a tertiary referral centre and has approximately 5,500 births annually.

From 2012 to 2019, only the WHO-5 index was used to screen for depression in respondents in the CPC. From November 2019 to August 2021, the MDI-2 was applied as a part of a two-step screening process, where women with a score of ≤ 50 on the WHO-5 index were additionally screened using the two core items from the MDI-2. Finally, from October 2021 to December 2021, both the WHO-5 index and two items from the MDI-2 were used independently to screen for depression in all women in the CPC.

Sociodemographic characteristics and lifestyle factors

The CPC questionnaire included covariates regarding socio-demographic characteristics and lifestyle factors. The covariates included were maternal age, pre-pregnancy Body Mass Index (BMI, kg/m2), parity, highest obtained educational level, occupational status, Danish language skills, cohabitation status, and lifestyle factors (alcohol consumption prior to pregnancy, smoking prior to pregnancy, and exercise prior to pregnancy). The covariates were categorized as shown in Fig. 1.

Statistical analysis

The distribution of selected covariates and outcomes was described as frequency (n) and proportion (%) for categorical data and as mean and standard deviation (SD) for continuous data.

Mean scores on the WHO-5 index and dichotomized scores on the WHO-5 index (using ≤ 28 and ≤ 50, respectively) were compared across the levels of each included variable. A list of the included variables (socio-demographic characteristics, obstetric history, lifestyle factors, health status and history of psychological health) and how they were categorized is described in Figs. 1 and 2. Many of these variables were described as risk factors of depression in other studies [20,21,22,23].

We evaluated the psychometric validity of the WHO-5 index in the population using confirmatory factor analysis (CFA) [24] and item response theory (IRT) [25, 26] models, which were used to test whether measures of a construct were consistent with a hypothesized measurement model. The first compared the empirical covariance matrix and the estimated covariance matrix of the best fitting model, while the latter estimated individual item locations and respondent locations on the well-being scale. Both provided ways of testing invariance.

CFA model fit were reported as a χ2 -test (insignificant tests indicating model fit) supplemented by the root mean square error of approximation (RMSEA; values below 0.05 indicating model fit), the Tucker-Lewis index (TLI; values above 0.95 indicating model fit) and the comparative fit index (CFI; values above 0.95 indicating model fit).

We used the most parsimonious IRT model, the Rasch model [27,28,29], assessing model fit using the Andersen conditional likelihood ratio test [30] and comparison of observed and expected item-rest-score associations [31]. Model fit was also evaluated graphically by dividing the sample into score groups and, for each item, plotting the item mean scores in each interval and comparing these to 95% confidence regions for the model expectations.

We used CFA for ordered categorical variables using the laavan [32] package in R [33] and used DIGRAM [34, 35] for Rasch analysis. In the CFA analysis correlated error terms are added based on modification indices (change in χ2 resulting from freeing fixed parameters) and invariance is tested using multiple groups CFA, while in the IRT analysis graphical Rasch models [36] were used. Correlated error terms indicate the presence of local response dependence, i.e., that the association between items is stronger than the latent variable can account for [37].

In the subsample of 1001 pregnant women where MDI-2 data were available, we considered respondents for whom one of the core symptoms were present most of the time or all the time during the past two weeks as cases. We evaluated different cut-offs on the WHO-5 score in terms of sensitivity (proportion of cases who are below the cut-off), specificity (proportion of non-cases who are above the cut-off), positive predictive value (PPV; proportion below the cut-off who are cases) and negative predictive value (NPV; proportion above the cut-off who are not cases).

Ethics

The study was approved by the National Data Protection Agency (file no: 2012-58-0004.RH- RH-2016-202), I-Suite nr.: 04778, 18. December 2017. The Medical Records Research, Health Research and Innovation Center for Regional Development, The Capital Region of Denmark, granted permission to disclose patient information from medical records for research (file no. R- -21043472, 5 January 2022).

link

Leave a Reply

Your email address will not be published. Required fields are marked *