Cardiac Depression Scale: Mokken scaling in heart failure patients

Heart failure (HF) is a leading cause of morbidity and mortality worldwide that imposes a considerable human and economic burden[1‐4]. Depression and depressive symptoms are common in patients with HF though estimates of prevalence vary depending on method and timing of assessment[5, 6]. A meta-analytic review on depression in HF reported clinically significant depression in 21.5% of patients, but varied by use of questionnaires versus diagnostic interview and HF severity[6]. Accurate identification of depression in patients with HF is critical because of significant associations with increased hospitalisations[5, 7, 8], poorer functional limitations[9, 10], lower survival rates[8, 11, 12], and reduced quality of life[13]. Despite an American Heart Association (AHA) advisory recommending depression screening for all patients with coronary heart disease[14], depression remains under recognised and under treated in this population[15]. Given the potential health consequences of untreated comorbid HF and depression, the value of using reliable and valid instruments to measure depression cannot be underestimated. This is paramount when one considers the numerous indirect costs of depression such as absenteeism, decreased productivity, increased risk for secondary complications, and malignant effects on dependent family members[16, 17].

Recognition of depression in patients with HF is complicated by an overlap in risk factors such as smoking, excessive alcohol consumption, obesity, and lower physical activity, and symptoms such as fatigue, lack of interest in activities, appetite gain or loss, psychomotor impairment, poor concentration, and depressed mood[1, 18]. These similarities have the potential to result in contrasting outcomes such as screening for depression generating a number of false positive results, and depression remaining undiagnosed for a substantial period of time. Hence, accurate assessment of depression in patients with HF is necessary to aid early intervention.

The AHA advisory endorsing depression screening, referral, and treatment of all individuals with coronary heart disease[14] has provoked considerable debate[19‐23], with much of the contention surrounding the lack of evidence regarding the choice of instruments to measure depression, specifically their psychometric properties. Of note, is that depression screening will only identify depressive symptoms, i.e. those at risk of depression, as a diagnosis of depressive illness can only be confirmed when a person has a number of depressive symptoms consistently over a couple of weeks or more.

The Cardiac Depression Scale (CDS) was developed for the specific purpose of providing a valid and reliable instrument to measure depression in cardiac patients[24]. The CDS is the only measure for depression derived from responses of cardiac patients. It offers a more responsive and sensitive alternative to other measures that are not always suitable for cardiac patients because they have been developed and validated in non-cardiac populations[25]. It is an easily administered (taking five minutes to complete and one minute to score) disease-specific, self-rating scale that assesses the full range of depression seen in cardiac patients. Whilst having 97% sensitivity and 85% specificity for diagnosing major depression, it also assesses the range of less severe ‘reactive’ depression[26]. Validation studies have applied the CDS to a general cardiac population and encouraged its continued use and evaluation in other cardiac patient groups[27‐29]. Thus, we aimed to validate the CDS in patients with HF using Mokken scaling analysis (MSA), a method related to item response theory that analyses multivariate databases for unidimensional hierarchies of items[30].

A unidimensional hierarchy of items can be identified using Loevinger’s coefficient (H) along with other indicators of reliability, probability and monotone homogeneity of items[31]. Recent versions of software for MSA are capable of analysing for invariant item ordering (IIO)[32], an important property of questionnaires whereby, for a set of items displaying IIO, every respondent responds to all of those items in exactly the same order in the hierarchy[33]. IIO is not necessary for the utility of a questionnaire showing a Mokken scale of items: unidimensionality and monotone homogeneity are sufficient as such a scale can be used to order individuals. However, IIO is desirable. Mokken scaling has recently been applied to several scales designed to measure psychological morbidity and other phenomena[34‐36]. We investigated if a hierarchy of items, according to Mokken scaling criteria, existed in the CDS when applied to patients with HF. This paper contributes to a series of validation papers[34‐38] that have implemented MSA to determine scale item unidimensional hierarchy.

We administered the CDS in face-to-face interviews to 603 HF patients attending the out-patient heart failure clinic of a major metropolitan hospital. Gender (male 68% and female 32%) and age distributions (x = 70, SD = 14) were typical for this patient group. As part of the usual clinic assessment protocol developed by one of the authors, all patients had the CDS routinely administered by a cardiac nurse prior to seeing the heart failure cardiologist. Clinical characteristics of study participants are outlined in Table 1.

The study was approved by the institutional Human Research Ethics Committee (Approval No. H2006/02657). Written informed consent was obtained from the participants.

The results of the MSA are shown in Table 2; 22 items formed a unidmensional Mokken scale of low strength (H<0.40) but high reliability (Rho>0.8) which was statistically significant. Fifteen items showed IIO but at low30 accuracy (H^T<0.). The hierarchical pattern of responses to items in the Mokken scale, ordered by their mean scores in terms of difficulty from low difficulty (high mean score) to high difficulty (low mean score) is interpretable in terms of respondents more easily endorsing items related to general concern about their condition (for example: ‘My problems are not over yet’; ‘I am not the person I used to be’; ‘I may not recover completely’) through a range of items to serious concerns about their condition with a sense of hopelessness (for example: ‘Things which I regret about my life are bothering me’; ‘There is only misery in the future for me’) and even expressions of wishing for death (‘Dying is the best solution for me’).

Depression in patients with HF is associated with increased hospitalisations, decreased medication adherence, poorer health outcomes, increased mortality, and significant economic costs[5‐13]. The high prevalence of depression in patients with HF[6] and the AHA advisory to screen for depression[14] indicates the need for a valid and reliable instrument. We provide evidence that the CDS is an appropriate prognostic indicator for identifying depression in patients with HF.

Although most parameters in biological systems are continuous variables, for pragmatic purposes we also heuristically organise information into identifiable groups. Thus, whilst symptoms of depressed mood and the components of diagnosed “depression’ are actually continuous variables, the diagnosis of “depression” has a dichotomous meaning that is useful for both management and determining prognosis. Whilst the CDS was originally developed to measure the full range of depressive symptoms in cardiac patients, it also has excellent accuracy for the “diagnosis” of major depression as a clinical entity (26). Nevertheless this current study was designed to examine the utility of the CDS over the full range of depressive symptoms, with the finding of a gradation from easily endorsed items through to those items representing more severe depression.

The application of Mokken scaling to the CDS has demonstrated the existence of a hierarchy of items which can be interpreted in terms of the increasingly serious effects of depression occurring as a result of a HF. Unresolved problems, frustration and irritability are at one end of the hierarchy, whilst guilt, misery and wanting to die (measures of severe depression) are at the other. The demonstration of this hierarchy provides new information (a non-parametric measure of IRT) about the structure of the CDS and adds utility to the scale in clinical practice when managing patients with HF.

Invariant item ordering was not demonstrated for the items retained in the Mokken scale of the CDS; this means that there is no guarantee that all (HF) respondents to the CDS will respond to the hierarchy of items in precisely the same order. This could be partly explained as, within each patient cohort, there are patients who have pre-existing chronic depression as well as those with a more acute 'adjustment disorder" who might respond more strongly to a slightly different range of items. Also the lack of IIO indicates that the conceptual distance between the items is likely to be small and that there is some overlap of the item response functions because of cohesion in the scale items. Nevertheless, the Mokken scale is still useful for ordering respondents on the basis of their mean scores.

In conclusion, we found the CDS, developed as a valid and reliable cardiac-specific measure for depression, to contain a set of items that when applied to patients with HF are hierarchical. The CDS is an appropriate depression screening instrument for patients with HF that allows for early identification and better management, especially of those at increased risk of functional limitations, mortality, and impaired quality of life.