Utilizing correspondence analysis to characterize the mental health of cardiac patients with diabetes

Abstract

The aim of this study was to examine the mental health of cardiac patients with diabetes and whether symptoms varied by gender and/or age. Screening for depression and posttraumatic stress symptoms was conducted on 1003 patients with cardiovascular disease. Correspondence analysis was utilized to identify clinical core profiles. Results suggested that cardiovascular disease patients with diabetes, particularly males, experience high rates of depression, suicide ideation, and posttraumatic stress symptoms. Clinical implications of these findings include targeted mental health screening options as well as offering a closer look at the specific concerns of cardiovascular disease patients with diabetes.

Keywords

coronary heart disease depression distress myocardial infarction quantitative methods

Introduction

Patients being treated for cardiovascular disease (CVD) report high rates of depression and posttraumatic stress disorder (PTSD) (Annunziato et al., 2008). Among CVD patients, mental illness can compromise overall well-being. For example, depression is associated with cardiac symptoms like angina, poorer physical functioning, and diminished quality of life (Spertus et al., 2000) as well as medication nonadherence (Gehi et al., 2005). Symptoms of PTSD are particularly common after a myocardial infarction (MI). In the first year after experiencing an MI, as many as 8–25 percent of CVD patients develop PTSD, which is also associated with poorer medical outcomes such as subsequent hospitalization (Bennett and Brooke, 1999; Shemesh al., 2004) and nonadherence to medications (Shemesh et al., 2001, 2004).

CVD is a broad term and within it there are many different types of patients. The mental health (MH) of some CVD patient types, like those who have had an MI, has been well studied. Because cardiac patients are at such high risk for depression and PTSD, our group implemented a MH screening program within an urban cardiology clinic. From a clinical perspective, we noticed during the program that many patients with diabetes were screening positive for PTSD, but they had not experienced a clear-cut life-threatening event like an MI. CVD and diabetes are frequent comorbidities that are strongly associated with depression (Clouse et al., 2003; Kinder et al., 2002). Yet, to our knowledge no studies have examined the experience of PTSD symptoms among CVD patients who are also diabetic. Therefore, the purpose of this study was to examine the MH of cardiac patients with diabetes more closely.

To achieve our study aim, we compared cardiac patients with diabetes to those with a previous MI as well as patients with both diabetes and a previous MI utilizing an innovative analytical method, correspondence analysis (CA; Benzécri, 1973, 1992; Greenacre, 1984; Murtagh, 2005; Nishisato, 1980). Patients with a prior MI were utilized as a comparison because of its established association with depression and PTSD (Annunziato et al., 2008). Gender and age were also included in the analyses given prior findings. Research has suggested that female CVD patients may experience a greater degree of depression, and differences in depression profiles between men and women have been observed, with women experiencing more somatic symptoms (Grace et al., 2013). Among older CVD patients, depression is associated with greater mortality (Ahto et al., 2007) and older patients with depression are less likely to experience posttraumatic growth (Leung et al., 2012). It is therefore important to look at the effects of gender and age in relationship to MH morbidity in patients with CVD and diabetes. The CA approach is able to identify core clinical profiles that take into account comorbidities such as gender and age.

Methods

Participants

Data have been evaluated for 1003 patients screened between June 2005 and November 2007 as part of a clinical program at the cardiology clinic of Elmhurst Hospital Center (EHC). EHC is located in the western portion of the borough of Queens in New York City; more than 40 nationalities are represented in this community. The primary cardiac diagnoses in the clinic are coronary artery disease and congestive heart failure. Since this was a clinical program, no specific inclusion/exclusion criteria were employed (all patients who presented to the clinic were offered the screen). Patients gave verbal consent for participating in the MH screening program and the questionnaires were stored in their charts. Institutional Review Board (IRB) approval was obtained to review screened patients’ medical charts.

Procedure

All patients who registered for an outpatient visit in the cardiology clinic received two questionnaires, available in English and Spanish, screening for depression and PTSD symptoms. The two measures together took about 5–10 minutes for completion. The objective of the screening program was to connect patients to appropriate services. A positive screening result generated immediate evaluation by the cardiology team (a nurse practitioner or a physician), who decided, with the patient, whether to suggest a referral for a psychiatric evaluation.

Measures

Demographic and medical variables

Demographic (gender and age) and medical (cardiac diagnosis and comorbidities) variables were obtained from electronic charts. Diagnosis of diabetes or MI was only coded “yes” if the determination was made before the date the MH screening questionnaires were completed. If a patient presented with both a diabetes diagnosis and a previous MI, this was coded as a separate group.

Patient Health Questionnaire

This validated 9-item self-report depression measure (Spitzer et al., 1999) has been recommended for use in screening efforts in medical care settings, and its utility in identifying cases has been shown (Kroenke et al., 2001). Respondents are asked how often they have been bothered during the past 2 weeks by nine symptoms that correspond to Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM-IV) criteria for major depression. Items are scored from 0 (“Not at all”) to 3 (“Nearly every day”). The American College of Cardiology and American Heart Association recommend a threshold score of 10 for suggesting further MH evaluation (Lichtman et al., 2008), and we used both this threshold and total score in our analyses. Finally, we used the last item on the Patient Health Questionnaire–9 (PHQ-9) as an indicator of suicide ideation (SI) with any score above 0 on PHQ item 9 “thoughts that you would be better off dead or of hurting yourself in some way” as indicating the presence of SI.

Impact of Events Scale

The Impact of Events Scale (IES) is a 15-item self-report questionnaire that measures current subjective distress related to any stressful event or one that is specified (Horowitz et al., 1979). For this study, keying the items to “heart disease” was specified. The scale requires indicating how often one has experienced each symptom over the past 7 days on a 4-point Likert scale from 0 (“Not at all”) to 5 (“Often”). The IES also offers two subscales measuring experiences of avoidance (8 items) and intrusion (7 items) related to the traumatic event. It therefore serves as a useful screening tool for identifying possible cases of PTSD. The total score is calculated by summing responses to the 15 items while subscale scores on the avoidance and intrusion scales can be calculated as well. Good reliability and validity have been reported in multiple studies (Briere, 1997; Weiss and Marmar, 1997). A high score on the IES was shown to be associated with a higher cardiovascular risk profile (Shemesh et al., 2001, 2004) in patients with CVD. A screening cutoff of 19 is associated with excellent sensitivity and acceptable specificity (Wohlfarth et al., 2003). We used both this threshold and total score in our analyses. The IES was selected as the PTSD screen for several reasons including its strong psychometric properties, low patient burden for completion, its established use among CVD patients, and that it has also been validated among Spanish-speaking clients (Baguena et al., 2001).

Statistical analyses

Preliminary statistical analyses were performed with the SPSS 17.0 statistical package (SPSS Inc., Chicago, IL, USA). CA was conducted by using R-language (http://cran.r-project.org). A “p” value (given alpha level = 0.05) of 0.05 or less, two-tailed, was chosen as the level of statistical significance. Before CA was conducted, categorical variables were compared by chi-square analysis. Independent samples t-tests and analysis of variance (ANOVA) were used to compare outcomes on continuous measures. In the event of missing PHQ or IES data, patients were excluded from analyses using the respective measure(s).

This study utilizes the profile pattern approach (Davison et al., 2009; Frisby and Kim, 2008; Kim, 2010; Kim et al., 2007) in interpreting dimensions estimated by CA. The profile pattern approach (a) estimates dimensions from profile-type data where rows represent individual participants’ scores and columns represent measured variables, (b) views the participants’ scores and dimensions as their profile patterns, and (c) makes connections between these score profiles (or simply row profiles) and the dimension profiles. A similar paradigm is applied to the CA results. First, dimensions are estimated from a contingency table and then connections are made between row profiles and dimension profiles.

The ordinary chi-square statistic provides only the significance of the association between row and column categories. On the contrary, CA decomposes variation occurring from the row–column association and estimates dimensions for rows and columns separately. With the development of CA, row–column category contributions to the total variance occurring in the row–column associations are available. Moreover, since dimensional information from row–column categories is available like factor or principal component analysis, the category contributions for dimensions are also accessible. Therefore, utilizing the CA paradigm, researchers can obtain much richer information about row–column categories than the ordinary chi-square test.

This study analyzed a frequency table constituted by 1003 patients, where rows represent six age categories for female and male patients and columns represent binary (0, 1) clinical measure categories. The following abbreviations were used to denote these clinical measures categories: “mi” for MI, “dia” for diabetes, “both” in cases where the patient had a diabetes diagnosis and an MI, “ies” for a positive screen on the IES, “phq” for a positive screen on the PHQ, and “sui” for a positive screen on PHQ #9. The ending “0” refers to those who did not have the clinical presentation and the ending “1” refers to those patients who did. The 12 categories were therefore “mi0,” “mi1,” “dia0,” “dia1,” “both0,” “both1,” “ies0,” “ies1,” “phq0,” “phq1,” “sui0,” and “sui1.” The six age categories with male and females separately include the following: m17_35 and f17_35, m36_45 and f36_45, m46_55 and f46_55, m56_65 and f56_65, m66_75 and f66_75, and m76+ and f76+. The sample size for male patients was 625 and for female patients it was 378.

At the first stage, we estimated row dimensions that consist of column (clinical measures) category coordinates. These coordinates are used to depict row dimension profile patterns. At the second stage, we converted column dimensions coordinates into correlations of row (gender–age) categories. These correlations were used as matching statistics between the gender–age profiles and the row dimension profiles that are estimated in the first stage since our main interest was in examining relationships between row (gender–age) profiles and the row dimension profiles.

Results

Table 1 displays demographic characteristics of the sample. The mean age of the sample was 61.04 years (standard deviation (SD) = 13.96 years) with 62 percent males. Medical data were available for all patients. As expected for the EHC population, primary cardiac diagnoses of coronary artery disease (46.0%) and congestive heart failure (17.5%) were the most common. Among patients, 210 (21.2%) were diagnosed with diabetes, 181 (18.1%) survived an MI, 111 (11.2%) had both a diabetes diagnosis and MI history, while 501 (50.0%) patients were being treated for CVD but did not have diabetes or an MI. Within all three groups examined, a large percentage of patients screened positive for depression and PTSD. Table 2 depicts PHQ and IES scores and percentage of patients who scored above thresholds for cohorts of patients studied.

Table 1.

Demographic characteristics of the sample.

Age	M = 61.04, SD = 13.96
Gender
Male	62%
Female	38%
Ethnicity
Hispanic	50%
Southeast Asian	23%
Caucasian	11%
East Asian	9%
Black	5%
Unknown	<1%
Other	<1%
Study grouping
Diabetes	21%
MI	18%
Both	11%
None	50%

SD: standard deviation; MI: myocardial infarction.

Table 2.

PHQ and IES scores and percent above threshold by presentation.

	PHQ mean score (SD)	PHQ threshold	Suicide ideation	IES mean score (SD)	IES threshold
Diabetes	7.16 (6.63)	34.1	15.6	14.45 (15.09)	33.5
MI	5.62 (6.19)	21.3	10.7	15.04 (16.32)	31.9
Both	7.79 (6.76)	41.8	18.6	14.61 (16.46)	31.7
None	5.93 (5.92)	23.5	10.5	14.30 (14.66)	32.6

PHQ: Patient Health Questionnaire; IES: Impact of Events Scale; SD: standard deviation; MI: myocardial infarction.

Univariate analyses

Initially, univariate analyses were conducted to examine whether there were differences in PHQ and IES continuous and threshold scores between the different patient groups. For the PHQ, there was a significant difference between the group means (diabetes = 7.16, SD = 6.63; MI = 5.62, SD = 6.19; Both = 7.79, SD = 6.76; None = 5.93, SD = 5.92), F(3, 886) = 4.21, p = 0.01, $η_{p}^{2}$ = 0.01. Tukey’s post hoc tests revealed that patients with both diabetes and an MI had higher scores than patients who experienced an MI only or no diagnosis. There was no difference in IES score by group.

With regard to thresholds, there was a higher frequency of patients with diabetes who screened positive on the PHQ than those with an MI, χ² = 6.93, p < 0.01, or no diagnosis, χ² = 7.33, p < 0.01. Similarly, patients with both diabetes and an MI also had higher frequencies of a positive PHQ screen than those with an MI, χ² = 12.49, p < 0.01, or no diagnosis, χ² = 13.70, p < 0.01. For SI, patients with both diabetes and an MI had higher rates than those with no diagnosis, χ² = 4.93, p = 0.03. There were no differences between the groups in frequencies above the IES threshold.

Profile analyses

Since there were 12 rows (gender–age categories) × 12 columns (clinical measure categories) of categorical data for analysis, the number of available dimensions was 11 utilizing the algorithm of min (rows, columns) − 1; for parsimony, the minimum number of meaningful dimensions should be identified. To determine the minimum number of meaningful dimensions represented by the data, like an ordinary factor analysis, we utilized percentages explained by eigenvalues. The maximum number of dimensions was 11 and the mean percentage for these 11 dimensions was 9 percent (=100%/11). To determine the number of dimensions for further analysis, we identified those with a percentage larger than or equal to the average of 9 percent. The percentages of the first three dimensions were accounted for by eigenvalues of 47 percent, 27 percent, and 12 percent, and in total, these three dimensions accounted for 86 percent of total variance. These dimensions elucidate different patterns of clinical symptoms.

Next, these three dimensions of clinical measure categories estimated by CA were plotted in Figure 1. Dimension 1 was labeled as “Typical” since there were peaks for mi1, dia1, ies1 and phq1. Dimension 2 was labeled as the “Diabetes” profile since there were peaks for dia1, both1, phq1, and sui1. Dimension 3 was labeled as the “Well-adjusted” profile since there were peaks for mi1 and both1 but no corresponding peaks on MH outcomes.

Figure 1.

Three patient profiles, (a) “Typical,” (b) “Diabetes,” (c) and “Well-adjusted,” identified by CA.

Then, we examined whether there were gender and age differences in the clinical measure categories. To assess how well the row (e.g. gender–age categories: m17_35-f76+) profiles were related to the row (clinical measure categories) dimension profiles, correlations were estimated. The correlations represent how closely the gender–age category profile patterns resemble the clinical measure profile patterns. Note that each gender–age category profile consists of frequencies in the 12 clinical measure categories as coordinates. Table 3 summarizes these results. In the table, r_D1, r_D2, and r_D3 refer to correlations between 12 gender–age category profiles and three clinical measure dimension profiles. R² values that were estimated by regressing the category profiles onto the dimension profiles stand for total variances accounted for by three dimensions and %_D1, %_D2, and %_D3 represent percentages of gender–age categories contributed to each dimension.

Table 3.

Correlations, R² values, and percentages accounted for by dimensions.

	r_D1	r_D2	r_D3	R ²	%_D1	%_D2	%_D3
m17_35	0.16	−0.96***	−0.03	0.95***	0.5	30.8	0.1
f17_35	−0.74***	−0.62**	−0.15	0.95***	12.5	15.2	2.0
m36_45	0.23	−0.90***	0.27*	0.94***	0.9	25.1	5.1
f36_45	−0.95***	−0.24	−0.01	0.95***	27.6	3.2	0.0
m46_55	−0.13	−0.23	0.85***	0.79**	0.3	1.4	42.4
f46_55	−0.93***	0.26*	−0.22*	0.98***	17.1	2.5	4.1
m56_65	0.21	0.75***	0.45*	0.80**	0.5	10.8	9.4
f56_65	−0.45**	0.63**	−0.52**	0.88**	3.1	10.4	16.7
m66_75	0.93***	−0.04	−0.20*	0.91***	28.0	0.1	4.9
f66_75	−0.16	0.04	−0.41*	0.20*	0.1	0.0	3.3
m76+	0.74***	−0.12	−0.42**	0.74**	8.9	0.4	11.7
f76+	−0.26	0.11	0.11	0.10	0.6	0.2	0.5

The R² values are estimated by regressing age/gender categories onto three dimensions. Explained percentages larger than the average 8.3% are given in bold.

***

p ≤ 0.001; **p ≤ 0.01; *p ≤ 0.05.

To highlight the impact of including these variables, the gender–age categories that occupied more than the average percentage, 8.3 percent, of variance in a dimension are in bold type. Females between the ages of 17 and 55 years and males between the ages of 66 and 76+ years were dominant in Dimension 1, “Typical.” The majority of male patients of m17_35, m36_45, and m56_65 and female patients of f17_35 and f56_65 were dominant in Dimension 2, “Diabetes.” However, in Dimension 3, most of age 46 years or above groups, f46_55, m56_65, f56_65, and m76+, were dominant, but no gender effect appeared in the dimension.

Table 3 also summarizes regression results. Most of the gender–age categories were significantly related to the dimensions except f76+ (which was not significantly related to any of the dimensions). To test the significance of these correlations, each gender–age category was regressed onto the clinical measure dimensions and we obtained the statistical significance of the correlations and of R² values. The positive correlations in the table show that the gender–age category profiles are similar to the pattern of the clinical measure dimensions. The negative correlations indicate that these category profiles have mirror images of the dimension profiles, and peaks in the dimension profiles appear as valleys in the mirror image profiles. Taken together, these results illustrate three common sets of symptom presentation (dimensions) in our symptom and their association with gender and age.

Discussion

We aimed to learn more about the specific MH concerns of CVD patients with diabetes using an innovative statistical approach in conjunction with traditional statistics. Our findings suggest that for CVD patients, diabetes is associated with a substantial risk for depression, suicidality, and also symptoms of PTSD. In fact, it appears that having diabetes and CVD may be more concerning from a MH perspective than an MI.

Across the board, above-threshold symptoms of depression were identified in 20 percent to almost 40 percent of the specific patient types included, while about 30 percent of all patients reported above-threshold levels of PTSD symptoms. Perhaps, the most consistent finding of all was the strong relationship between diabetes and depression. The incremental increase in depression risk among CVD patients with diabetes has been established (Egede, 2005). Similarly, our findings showed that patients with diabetes and an MI were most likely to screen positive for depression. Additionally, symptoms of depression were higher in the combined group as well as for patients with diabetes alone relative to those who had an MI. Both the univariate analyses and CA showed that patients with diabetes appear to be suffering more than those who have had an MI.

PTSD and diabetes

Although the risk of developing PTSD in the wake of an MI has been widely documented, in this study, rates of PTSD positive screens were high, but uniform among the patient types included. A novel finding is that CVD patients with diabetes confer great risk for symptoms of PTSD as well as depression. With regard to PTSD symptom presentation, diabetic patients with CVD were also no different from MI patients; this is surprising given that an MI is a clear-cut life-threatening event, while diabetes is not. It may be that symptoms associated with PTSD are common generally in cardiac settings. Indeed, research has shown that benign chest pain is linked to depression and anxiety among patients presenting in cardiology settings (Rosenbaum et al., 2012). Such findings suggest that PTSD-like symptoms may be more widespread among cardiac patients and not specific to those with history of MI.

Utilization of CA

More precise results were only revealed when using a novel technique, profile pattern approach with CA which revealed three symptom dimensions. The focused findings that this technique helped uncover, embracing key demographic variables, greatly augmented the results derived from univariate techniques. By using this analytical strategy, a clearer picture of those patients most in need of MH services was obtained. The first dimension, “Typical,” primarily captured the established relationships between diabetes and MI individually with depression as well as PTSD symptoms. The second dimension, “Diabetes,” bolstered the univariate analyses by elucidating the grouping of patients with diabetes only, patients with diabetes and an MI, and positive screens for depression and SI. The particularly salient MH concerns of CVD patients with diabetes, no matter whether an MI was experienced, were accentuated here. This dimension also captured the established connection between diabetes and depression, with the addition of SI. The use of CA isolated the association with SI that was not shown in univariate analyses but is a critically important clinical finding. Finally, the third dimension, “Well-adjusted,” unearthed a subset of patients who despite their significant medical needs (diabetes, MI, and/or both) did not display elevated levels of MH symptoms.

The gender–age findings revealed by CA were in some ways consistent with clinical expectations but also helped to home in on those who may be more likely to experience compromised MH. Dimension 1, “Typical,” largely comprised women from a wide age range (17–55 years). However, males of all ages dominated the “Diabetes” dimension. It appears that men who have diabetes experience a considerable impact on their MH. It is unclear to what extent this finding is specific to our sample. At EHC, the majority of patients are from other countries with more traditional gender roles (e.g. males are the primary breadwinners). Male CVD patients often express distress over loss of or threat to this role wrought by their medical limitations and this may be aggravated by diabetes. Broadly, this finding suggests that diabetic males have a substantial burden. In addition, some age groups of female patients were also included in this dimension accentuating the importance of attending to the MH generally of patients with diabetes.

Finally, the third dimension, “Well-adjusted,” was interestingly mainly composed of middle-age to older patients with no gender effects. Anecdotally, EHC does tend to serve patients who are from more collectivistic cultures and it may be that there is inherently more community and family support provided to those with significant medical needs (Goodwin and Plaza, 2000). Similarly, a recent study examined “posttraumatic growth” among a large sample of cardiac patients finding several correlates of resiliency including lower levels of depression and social support (Leung et al., 2012). The emergence of a “positive” dimension in concert with other studies of such trajectories among cardiac patients certainly suggests that further research is necessary to better understand what factors differentiate these patients from others.

Limitations

A major limitation of the study, because it is cross-sectional, is that the pathway of our findings is unknown. For example, it is unclear whether MH symptoms emerged because of poor physical health. Because CVD and diabetes are chronic conditions, it may be that MH symptoms fluctuate over time and this can only be captured in a longitudinal design. The measures selected for screening were given only in their English and Spanish versions, which were not the primary languages of all patients. Ideally, patients would be able to complete these measures in their primary language, but in the event that patients felt uncomfortable in English or Spanish, they were not screened. And finally, although the IES specified that the questions were with regard to their cardiac illness, it is possible, given the uniformly high rates, that participants based their responses on other traumatic events as well.

Conclusion

Our results bolster the importance of MH screening among CVD patients (Lichtman et al., 2008) and offer possible options for more targeted screening which given scarce resources may be critical to more successful programming. Patients with diabetes may benefit greatly from MH screening that encompasses assessment of SI. And in general, CVD patients with diabetes appear prone to PTSD symptoms irrespective of MI status. And finally, the profile pattern approach with CA demonstrated here offers a promising future in health psychology by better elucidating patterns drawn from dichotomous or any categorical source data.

Footnotes

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

References

Ahto

Isoaho

Puolijoki

. (2007) Stronger symptoms of depression predict high coronary heart disease mortality in older men and women. International Journal of Geriatric Psychiatry 22: 757–763.

Annunziato

Rubinstein

Sheikh

. (2008) Site matters: Winning the hearts and minds of patients in a cardiology clinic. Psychosomatics 49(5): 386–391.

Baguena

Villarroya

Belena

. (2001) Propiedades psicometricas de la version Espanola de la Escala Revisada de Impacto del Estressor (EIE-R) [Psychometric properties of the Spanish version of the Impact of Event Scale—Revised (IES-R)]. Analisis y Modificacion de Conducta 27: 581–604.

Bennett

Brooke

(1999) Intrusive memories, post-traumatic stress disorder and myocardial infarction. British Journal of Clinical Psychology 38: 411–416.

Benzécri

(1973) L’Analyse des Données, Tome 1: La Taxinomie, Tome 2: L’Analyse des Correspondances. Paris: Dunod.

Benzécri

(1992) Correspondence Analysis Handbook. New York: Marcel Dekker.

Briere

(1997) Psychological Assessment of Adult Posttraumatic States. Washington, DC: American Psychological Association.

Clouse

Lustman

Freedland

. (2003) Depression and coronary heart disease in women with diabetes. Psychosomatic Medicine 65: 376–383.

Davison

Kim

S-K

(2009) Factor analytic modeling of within person variation in score profiles. Multivariate Behavioral Research 44(5): 668–687.

10.

Egede

(2005) Effect of comorbid chronic diseases on prevalence and odds of depression in adults with diabetes. Psychosomatic Medicine 67: 46–51.

11.

Frisby

Kim

S-K

(2008) Using Profile Analysis via Multidimensional Scaling (PAMS) to identify core profiles from the WMS-III. Psychological Assessment 20(1): 1–9.

12.

Gehi

Haas

Pipkin

. (2005) Depression and medication adherence in outpatients with coronary heart disease: Findings from the Heart and Soul Study. Archives of Internal Medicine 165: 2508–2513.

13.

Goodwin

Plaza

(2000) Perceived and received social support in two cultures: Collectivism and support among British and Spanish students. Journal of Social and Personal Relationships 17(2): 282–291.

14.

Grace

Yee

Reid

. (2013) Measurement of depressive symptoms among cardiac patients: Should sex differences be considered. Journal of Health Psychology. Epub ahead of print 30 April. DOI: 10.1177/1359105313482165.

15.

Greenacre

(1984) Theory and Applications of Correspondence Analysis. London: Academic Press.

16.

Horowitz

Wilner

Alvarez

(1979) Impact of Event Scale: A measure of subjective stress. Psychosomatic Medicine 41: 209–218.

17.

Kim

S-K

(2010) Evaluating the invariance of cognitive profile patterns derived from Profile Analysis via Multidimensional Scaling (PAMS): A bootstrapping approach. Journal of International Testing 10: 33–46.

18.

Kim

S-K

Davison

Frisby

(2007) Confirmatory factor analysis and profile analysis via multidimensional scaling. Multivariate Behavioral Research 42(1): 1–32.

19.

Kinder

Kamarck

Baum

. (2002) Depressive symptomatology and coronary heart disease in Type I diabetes mellitus: A study of possible mechanisms. Health Psychology 21(6): 542–552.

20.

Kroenke

Spitzer

Williams

(2001) The PHQ-9: Validity of a brief depression severity measure. Journal of General Internal Medicine 16: 606–613.

21.

Leung

Gravely-Witte

Macpherson

. (2012) Post-traumatic growth among cardiac outpatients: Degree comparison with other chronic illness samples and correlates. Journal of Health Psychology 15(7): 1049–1063.

22.

Lichtman

Bigger

Jr Blumenthal

. (2008) Depression and coronary heart disease: Recommendations for screening, referral, and treatment: A science advisory from the American Heart Association Prevention Committee of the Council on Cardiovascular Nursing, Council on Clinical Cardiology, Council on Epidemiology and Prevention, and Interdisciplinary Council on Quality of Care and Outcomes Research: Endorsed by the American Psychiatric Association. Circulation 118(17): 1768–1775.

23.

Murtagh

(2005) Correspondence Analysis and Data Coding with Java and R. Boca Raton, FL: Chapman & Hall/CRC.

24.

Nishisato

(1980) Analysis of Categorical Data: Dual Scaling and Its Applications. Toronto, ON, Canada: University of Toronto Press.

25.

Rosenbaum

White

Gervino

(2012) The impact of perceived stress and perceived control on anxiety and mood disorders in noncardiac chest pain. Journal of Health Psychology 17(8): 1183–1192.

26.

Shemesh

Rudnick

Kaluski

. (2001) A prospective study of posttraumatic stress symptoms and nonadherence in survivors of a myocardial infarction (MI). General Hospital Psychiatry 23(4): 215–222.

27.

Shemesh

Yehuda

Milo

. (2004) Posttraumatic stress, nonadherence, and adverse outcome in survivors of a myocardial infarction. Psychosomatic Medicine 66: 521–526.

28.

Spertus

McDonell

Woodman

. (2000) Association between depression and worse disease-specific functional status in outpatients with coronary disease. American Heart Journal 140(1): 105–109.

29.

Spitzer

Kroenke

Williams

(1999) Validation and utility of a self-report version of PRIME-MD: The PHQ primary care study. JAMA 282: 1737–1744.

30.

Weiss

Marmar

(1997) The Impact of Event Scale—Revised. In: Wilson

Keane

(eds) Assessing Psychological Trauma and PTSD. New York: The Guilford Press, 399–411.

31.

Wohlfarth

van der Brink

Winkel

. (2003) Screening for post-traumatic stress disorder: An evaluation of two self-report methods among crime victims. Psychological Assessment 15(1): 101–109.