Attachment and coping with chronic disease

Journal of Psychosomatic Research 53 (2002) 763 – 773

Attachment and coping with chronic disease Silke Schmidta,*, Christof Nachtigallb, Olivia Wuethrich-Martoneb, Bernhard Straussc a

Department of Medical Psychology, University Hospital of Hamburg Eppendorf, Martinistrasse 52, Pav. 73, Hamburg D-20246, Germany b Department of Methodology, Institute of Psychology, University of Jena, Jena, Germany c Institute of Medical Psychology, University Hospital of Jena, Jena, Germany Received 4 September 2001; accepted 20 February 2002

Abstract Background: In this clinical study, attachment theory was applied to research in the field of coping with chronic disease. The approach was to integrate concepts of coping within a framework of attachment theory. It was hypothesised that attachment concepts have an influence on coping strategies, and that they may predict the subjective emotional and physical health status during the course of medical treatment. Method: One hundred fifty patients were investigated with an adult attachment interview (AAPR coding system) and a coping interview (Bernese Coping Modes). Self-reported coping modes, social support, the subjective health status and quality of life were also assessed by self-report measures at two or more sampling points of measurement. Three subsamples of patients, suffering from (a) breast cancer, (b) chronic leg ulcers and (c) alopecia, were studied in order to include a broad range of subjective impairment caused by a disease. Results: Findings indicate a moderate statistical effect of attachment patterns on coping strategies when controlling the influence of confounding

variables. Insecure attachment was related to less flexible coping. Coping strategies also differed between the different types of insecure attachment; however, there were differences depending on the perspective of the coping behaviour (self- vs. observer ratings) as well. From the observer perspective, ambivalently attached individuals showed more negative emotional coping while avoidantly attached individuals showed more diverting strategies. In the self-report, ambivalently attached patients revealed hyperactivating tendencies in their coping behaviour while avoidantly attached individuals revealed deactivating tendencies. Conclusion: As a conclusion, two levels of coping should be differentiated. One level strongly corresponds with affect regulation, in particular the regulation of attachment-related emotions and concerns, while the other level shows a stronger tendency to outwardly oriented coping. A more secure attachment might be considered to be an important inner resource in the emotional adaptation to chronic diseases. D 2002 Elsevier Science Inc. All rights reserved.

Keywords: Attachment styles; Coping; Chronic disease; Subjective health status; Adult Attachment Prototype Rating; Bernese Coping Modes

Introduction During the last three decades of the 20th century, there has been an extensive and continuing interest in conceptual and clinical issues of coping [1]. Coping may be considered to be one of the core concepts in medical and health psychology and has, therefore, generated a vast variety of empirical approaches. There has, however, been a lack of effort in integrating the interindividual and intraindividual approaches of coping [2]. Specifically, coping concepts in general, and in particular those concepts related to coping with chronic disease, have rarely been addressed from the perspective of

* Corresponding author. E-mail address: [email protected] (S. Schmidt).

developmental psychopathology. It is assumed that attachment theory [3] could be one of the most powerful theories to integrate the wide variety of coping concepts. Attachment theory has been developed by the British psychiatrist, John Bowlby [3], in order to describe the propensity of human beings to form close affectionate bonds. It is not only a normative theory, but also postulates qualitative interindividual differences in attachment security, which endure from early childhood to adulthood. Consequently, a variety of methods have been developed to assess attachment styles in children and adults. Along with the development of these methods, attachment theory has generated a vast amount of research in developmental, personality and clinical psychology; however, it has been rarely applied in the medical field. Yet, there are several reasons why attachment theory can be suitably applied as an umbrella theory for the study of coping

0022-3999/02/$ – see front matter D 2002 Elsevier Science Inc. All rights reserved. PII: S 0 0 2 2 - 3 9 9 9 ( 0 2 ) 0 0 3 3 5 - 5

764

S. Schmidt et al. / Journal of Psychosomatic Research 53 (2002) 763–773

concepts within a medical context. First, attachment theory is able to predict either the vulnerability or resilience to stressful life events to a significant degree [4]. In particular, attachment theory may predict why some persons are able to shift resourcefully from one strategy to another while others rely on very rigid ways of coping [5]. Second, Bowlby [4] postulated that the attachment behavioural system is activated in times of stress and disease. Thus, the attachment system may be considered to be an important motivational system involved in the dynamics of the coping process. Recent reconceptualizations of coping theories have focused on processes of emotion regulation [6]. In this respect, attachment theory is able to explain interindividual differences in the regulation of emotions or ‘‘affect regulation.’’ Fuendeling [7] has summarised the wide variety of findings related to different styles of affect regulation. He was able to show that in each phase of the coping process, attachment styles affect the way in which (a) attention to stressful events is managed [8], (b) information is processed [9] and (c) social coping resources are made accessible [10]. These stylistic differences in attachment styles can be attributed to the excessive processing of emotions of ambivalently attached individuals, or to the isolation or cutting off of emotions of avoidantly attached individuals [11]. Spangler and Zimmermann [12] concluded from several psychobiological studies on attachment and emotion regulation that the ‘‘rigid’’ styles of affect regulation cause larger discrepancies between different levels of coping in insecurely attached individuals than in securely attached individuals. It is striking, however, that the relationship among attachment, affect regulation and coping has rarely been investigated within the context of coping with chronic disease. There are, however, several studies that have focused on the relationship between attachment styles and subjective physical symptoms, illness behaviour and increased vulnerability to stress. Kotler et al. [13] have found in a student sample that avoidant attachment was significantly related to elevated physical and psychological symptoms. This relationship was mediated by a type of restrictive control of emotions, which was in turn related to avoidant coping, wishful thinking and self-blame. Feeney and Ryan [14] have found evidence that ambivalently attached individuals displayed a higher number of visits to health care professionals than securely attached individuals. The authors pointed out that this relation was mediated by negative affectivity. In a sample of patients with idiopathic spasmodic torticollis (IST), dismissing attachment was strongly overrepresented compared to a nonclinical control group [15]. The authors interpreted this finding in terms of an increased vulnerability to psychosocial stress in IST patients. These studies do not address the relationship between attachment and coping, but indicate that under certain conditions, attachment styles may be associated with maladaptive health behaviour and (subjective) symptoms. In this respect, Stuart and Noyes [16] have presented a conceptual framework linking physical com-

plaints and somatization to an anxious attachment history. The most prominent findings on this confounding effect were recently presented in a study of Taylor et al. [17] who identified, in a large sample of primary care attenders, a strong relationship between attachment styles and subjective physical symptoms. More specifically, patients with unexplained physical symptoms showed a higher incidence of insecure attachment and of psychiatric distress. The authors concluded that the attachment pattern exerted a mediating effect on the relationship between unexplained physical symptoms and psychiatric distress. The issue of this study was not to investigate the mediating influence of attachment patterns on subjective physical symptoms, but the statistical influence of attachment on coping strategies. However, even if one assumes that a causative relationship exists, current findings that point to a relationship between attachment patterns and maladaptive health behaviour also suggest that this link may be confounded by other variables. Consequently, it is not suitable to infer a causal effect of attachment history on coping strategies with chronic disease without also considering the potential impact of such confounders. In nonrandomized experiments or not fully controlled studies, conclusions drawn from the outcome of an experiment may not be regarded as a direct consequence of the treatment itself because third variables, or confounders, may have affected the results. Since a fully randomized experiment is often not feasible or even totally inappropriate because of ethical or cost reasons, an alternative is to outline statistical methods and procedures for controlling potential confounders in nonrandomized quasiexperimental studies. Though the term confounding is well known in epidemiology [18], it has not been mathematically and applicably formalized until the work of Steyer et al. [19,20]. In this theoretical framework, the relationship between a treatment and an outcome of this treatment is examined. Potential confounders are attributes of observational units, which might affect the treatment assignment and the relationship between treatment and outcome as well. On this basis, special tests for unconfoundedness have been developed and procedures for the computation of causal effects even in the case of confounding have been designed [21,22]. Applied to the attachment –coping model presented here, a third variable could be identified as a confounder if this variable has an effect not only on coping, but also on the attachment category. At the moment, there is no profound evidence in attachment-related coping research showing the existence of any confounders, which can be integrated into the design as covariates or factors. However, it may be hypothesised that there are specific types of diseases, or specific characteristics of the disease, which may have an effect on both the attachment styles and on coping strategies. It could, for instance, be assumed that in chronic diseases, there is an ‘‘unconfounded’’ relationship between attachment and coping. However, patients who suffer from diseases with less severe medical symptoms but who never-

S. Schmidt et al. / Journal of Psychosomatic Research 53 (2002) 763–773

theless seek professional advice over a long duration may differ in both their attachment patterns as well in their ways of coping. Coping strategies, for instance, have been found to vary substantially between acute and chronic conditions [23,24]. A variety of other variables related to the characteristics of a particular disease can also be assumed to influence both attachment styles and coping strategies, e.g., the severity of the disease. The primary goal of this study was to test the statistical effect of attachment patterns on coping strategies, which is assumed to exist irrespective of the disease and irrespective of other potential confounders such as the severity of the condition. Specifically, it was hypothesised: (a) that attachment styles are related to the flexibility/ rigidity of coping behavior; (b) that attachment styles indicate specific coping strategies (e.g., higher levels of ‘diverting’ in avoidantly attached, higher levels of ‘negative emotional coping’ in ambivalently attached and higher levels of ‘seeking attention and care’ in securely attached individuals); (c) that the discrepancy between self- and observer ratings is substantial and larger in patients with an insecure attachment than a secure attachment pattern. An additional question was to explore whether there is a contingency between psychosomatic conditions and the type of diseases and whether this contingency affects the relationship between attachment and coping. Considering this, confounding variables, which might bear influence on this causal relationship, must be analysed. If these confounders exist, they should be identified, and their effect on the relationship between attachment and coping must be formulated in statistical terms. It should be checked if they actually bias this influence and, in such a case, a corresponding adjustment for confounding should be performed. It should be noted, however, that such causal inference requires (a) that a temporal relationship exists (confounder ! treatment ! effect) and (b) that all confounders can be identified. Since neither the construct ‘‘attachment variable’’ nor the corresponding assessment procedure constitutes a treatment variable, this model cannot genuinely comply with the first prerequisite for causal inference on the basis of confounder analysis. Even if these requirements are not fully met, the adjustment for confounding variables is useful because it leads to an estimation of the effect of attachment on coping unbiased by the potential confounder, though in this case a causal interpretation should be made carefully.

765

and androgenetic alopecia (female type). Although the choice of these three particular diseases cannot be considered to be representative of all diseases, the selection was meant to cover a broad range of subjective impairments caused by different diseases. A number of eligibility criteria (e.g., no or low comorbidity) were defined in order to keep each disease category homogenous in terms of the main symptoms. Important inclusion criteria in patients with breast cancer were TNM stages < 3 and the primary diagnosis of breast cancer. With respect to patients with chronic leg ulcers, only conditions were included with a (mixed) venous aetiology; patients with leg ulcers of an arterial aetiology and patients suffering from diabetes mellitus Type 1a were excluded. Patients with hair loss were excluded if an alopecia areata, an alopecia totalis or an androgenetic alopecia (Ludwig III) was diagnosed. Because male and female hair loss are inherently different conditions not only in respect to the aetiology but also to treatment settings, only female patients with alopecia were included. The entire sample consisted of 150 patients with 54 patients suffering from primary breast cancer, 52 patients suffering from chronic leg ulcers and 44 patients suffering from female alopecia. The mean age was 52 ( ± 13.30) in patients with breast cancer, 65 ( ± 9.78) in patients with chronic leg ulcers and 45 ( ± 13.60) in patients with alopecia. A total of 131 patients were female and only 19 patients with chronic leg ulcers were male. Procedure The whole project was designed as a prospective investigation with prospective data being collected at two later points: (1) at discharge from the hospital or at the end of the medical treatment and (2) 6 months later. The study presented here was primarily cross-sectional. Research was conducted in the Departments of Dermatology and Gynaecology at the University Hospital of Jena. Patients were asked to participate at time of admission, except for alopecia patients who were asked at their first consultation appointment. All patients suffering from primary breast cancer had undergone mastectomy or breastconserving therapy. They were either interviewed after surgery or at the first follow-up examination. Patients with chronic leg ulcers received inpatient treatment at the Department of Dermatology. These patients received the standard conservative treatments including bandaging. All female patients with alopecia were seen in a consultation setting at the Department of Dermatology, specialised in the treatment of hair loss. The recruitment phase took place over a period of 12 months.

Method Measures Sample description This study comprised patients from three types of disease: (a) breast cancer, (b) chronic leg ulcers and (c) diffuse

Both an attachment and a coping interview were conducted with all patients individually. Self-report measures were used to assess sociodemographic data, coping strat-

766

S. Schmidt et al. / Journal of Psychosomatic Research 53 (2002) 763–773

egies, subjective health status, social support and quality of life (QoL) at several points of measurement. Medical details were obtained from the University Departments of Gynaecology and Dermatology; a special scale was designed for each disease category to measure diseasespecific characteristics. Attachment styles and ratings (observer report) Attachment styles were identified using the German version of the Adult Attachment Prototype Rating (AAPR [25]), which consists of an interview focusing on attachment behaviour in childhood and adulthood, a comprehensive observer rating scale and a corresponding self-report instrument. This method was originally developed by Pilkonis [26]. The modified German version has proved to be a reliable and valid method to assess attachment styles with an emphasis on attachment related interpersonal styles in adults. This perspective differs from the analytic strategy of the well-known AAI [27], which focuses almost exclusively on attachment representations including defensive strategies. The interviews are videotaped and observers must rate different aspects of attachment in subsequent analytic steps. First, clinical features of either secure, ambivalent or avoidant attachment patterns are identified. Second, ratings as well as rankings of attachment prototypes, which further differentiate primary attachment styles on a clinical level, are carried out. Each of these prototypes is defined by 10 five-point scale items as well as a comprehensive description of the prototype. The intraclass coefficient was, on average, 0.90, indicating a high interrater reliability [25,28]. Finally, by applying algorithms on the prototype rankings and ratings, a final decision on the attachment style is made. Cohen’s k related to the judgements of the primary attachment pattern were, on average, 0.68, thus revealing an acceptable interrater reliability of the primary attachment styles [25]. An additional criterion assessed with this method is the quantitative extent of attachment security. This criterion is partially based on the ratings, partially on the rankings of attachment prototypes. Subjects were categorised as ‘‘securely attached’’ if the prototype with secure features was ranked first and was given a value of 4 or more; ‘‘probably securely attached,’’ if rated 4 and ranked second, and ‘‘marginally securely attached’’ if the securely attached prototype was rated 3. Subjects not meeting any of the above criteria were classified as ‘‘insecurely attached.’’ The validity of the relatively new AAPR instrument has been tested in a series of clinical studies aimed at evaluating primarily the construct, clinical and prognostic validity of this instrument. During the last 10 years, the AAPR has been used in various studies in different clinical populations, in particular within psychotherapy research. These studies demonstrate both the construct and the prognostic validity [28]. The AAPR attachment categories have, for instance, been found to be related to interpersonal problems, to specific interpersonal orientations in the interpersonal cir-

cumflex or to different aspects of the psychotherapeutic alliance [25,28]. The attachment security scale has proved to be prognostically relevant in detecting changes in the psychotherapeutic process and to predict psychotherapeutic outcome after inpatient psychotherapy in a larger German multicentre study [28]. In an investigation of a nonclinical population, the convergent validity of the AAPR has been tested with the Close Relationship Questionnaire (CRQ [29]) and the Adult Attachment Scale (AAS [30]), indicating a substantially higher level of anxiety in the ambivalent attachment patterns [31]. So far, there has only been one study investigating the convergent validity with the AAI [25]. The findings of this study could not show a significant contingency between attachment categories identified with the AAI and the AAPR, but this may be either a result of the small sample size or of the AAI interview focus. At the moment, a powerful study is conducted by aiming at evaluating the convergence of the main self- and observer report measures in a large sample size. Coping strategies (observer and self-rating) The Bernese Coping Modes [32], an established self- and observer rating instrument, were used to determine the different ways of coping with a disease. This clinically oriented tool was originally developed with breast cancer patients in the 1980s but has meanwhile proved to be applicable for any diagnostic area or context [33]. The Bernese Coping Modes comprise 30 coping modes covering a broad spectrum of emotional and cognitive as well as behavioural coping strategies. In the observer rating version, independent and trained observers read transcripts or listen to audiotaped coping interviews carefully in order to assess the 30 coping forms, which are rated on both dichotomous and on ordinal rating scales. In the observer rating version, different factor solutions were identified, depending on the sample [33]. In this study, a five-factor solution suggested by Heim et al. [34] was used, which comprises the factors ‘‘denial,’’ ‘‘diverting,’’ ‘‘seeking attention and care,’’ ‘‘selfcontrol’’ and ‘‘negative emotional coping.’’ A factor analysis of the self-report version [35] has yielded three coping dimensions, called ‘‘diverting,’’ ‘‘negative emotional’’ and ‘‘seeking attention and care,’’ which have demonstrated a high internal and concurrent validity. The correlations among these three scales and the corresponding observer report scales ranged between 0.41 and 0.59 and were highly significant. An additional variable ‘‘flexibility of coping’’ was defined by the total number of coping modes used by an individual, independent of the intensity of each coping mode. Social support The short form of the social support questionnaire (F-SOZU-K22 [36]) was included to determine the perceived amount of social support (total score), the amount of emotional and practical support as well as social integration in the subscales of the instrument. This instrument is the

S. Schmidt et al. / Journal of Psychosomatic Research 53 (2002) 763–773

most common German social support self-report instrument. The internal consistency of this instrument has been found to be reasonable, with Cronbach’s a ranging between .79 and .94, and the validity has been shown in many clinical and nonclinical studies [37]. One primary objection to the validity of this instrument refers to the mediating relationship with the construct social desirability and its potential confoundedness with attachment styles [38]. Subjective health status To assess the subjective health status, the Nottingham Health Profile (NHP) was used since it has been extensively tested for its reliability and validity in Anglo-American and other countries [39,40]. It employs a total of 38 items to determine general aspects of health-related QoL in the following six scales: lack of energy, pain, emotional reactions, sleep disruptions, social isolation and physical mobility. Thus, it covers only adverse effects of the health status. It should be noted that the NHP has been, in the past, criticised for not discriminating between the severity of a disease affecting the patient, i.e., among patients with a less debilitating disease. However, Schmidt et al. [41] was able to show that even in patients with nonvisible symptoms of alopecia, the impairment to the health status was as strong as in patients with a severe and chronic disease, thus invalidating this argument. Internal consistencies of the scales in populations with chronic conditions have been found to vary between 0.65 (physical mobility) and 0.85 (pain) [39]. The summative score has been found to be useful as an outcome measure in clinical trials [41]. Statistical analysis The first analytic strategy was to test the causal effect of attachment styles on coping forms. Since the work of Rubin [42], the term average causal effect (ACE) has become the basis of causal inference and its application outside randomised experiments. It can be interpreted as the mean value of the causal effect that different treatments have on observational units [19,43,44]. As long as confounding variables exist, observed effects may differ from the ACE. The analysis of confounding developed by Steyer et al. [20] and Nachtigall et al. [22] allows to remove the biasing effect of a confounder and enables a test whether there is a causal effect or not, i.e., whether ACE = 0. This is analysed using a special software (CANOVA, causal analysis of variance [45]) that has been developed to test ACEs even outside randomised experiments. The first step of this statistical device is to identify potential confounders which might perturb the attachment and coping relationship. In case there is no confounding variable, conventional multivariate (MANOVA) and univariate analyses of variance (ANOVA) may be employed. If potential confounders are identified to be actual confounders, i.e., to induce a bias on the treatment means, a corresponding adjustment (for confounding) is performed.

767

Thus, the software combines the conventional statistical tests (ANOVA) and a method to adjust means in the subpopulations of the independent variable, which are the subpopulations evoked by the confounder.

Results The distribution of different attachment styles in each disease category is displayed in Table 1. Overall, the distribution of attachment styles was 36% secure, 10% avoidant, 24% ambivalent and 20% mixed ambivalent/avoidant. While patients with chronic leg ulcers and patients with breast cancer did not differ in terms of their attachment styles, patients with alopecia showed, in general, a higher proportion of the ambivalent and mixed ambivalent/avoidant attachment pattern. The difference of attachment styles between the types of disease was only significant on the dichotomous level, testing the amount of secure and insecure attachment patterns across the three types of disease. With regard to sociodemographic characteristics, it is important to note that all patients in the alopecia and breast cancer subgroups were female, while one third of the patients with chronic leg ulcers was male. Thus, the whole sample was, overall, more representative for female patients suffering from somatic diseases. Though some gender differences were found on the level of a few individual observer rating coping forms, these differences were not significant on the level of subordinate coping dimensions. With respect to attachment classifications, the shares of the avoidantly attached category were higher in males than in females; this effect was not significant. The second sociodemographic variable in which the subsamples showed substantial differences was the age, and consequently the state of employment. These variables, however, were not related to the attachment classification. Analysis of confounders Employing the CANOVA, the following two confounders were identified: (1) the type of disease (chronic diseases vs. alopecia) and (2) the acute impairment caused by the disease (more and less pronounced or debilitating symptoms). All other potential confounders analysed did not bear any influence on the relationship between attachment styles and coping modes.

Table 1 Distribution of attachment styles across the three diseases

Secure Ambivalent Avoidant Mixed

Breast cancer (%)

Leg ulcers (%)

Alopecia (%)

39 23 19 29

41 29 15 15

26 39 7 26

768

S. Schmidt et al. / Journal of Psychosomatic Research 53 (2002) 763–773

Table 2 Expected and adjusted means of ‘‘flexibility of coping’’ across the four attachment styles in both subpopulations of the confounders Subpopulation W Attachment style

Chronic disease (breast cancer/leg ulcers)

Alopecia

Ambivalent Avoidant Mixed ambivalent/avoidant Secure

15.11 12.44 13.94 15.60

9.83 11.00 8.50 13.91

Adjusted means across attachment patterns

M( Y | X = x)a

Madj( Y | X = x)b

Ambivalent Avoidant Mixed ambivalent/avoidant Secure

12.43 12.24 11.77 15.25

11.54 11.58 10.68 14.58

Analysis of statistical effect of attachment styles on copingc SSfactor

SSerror

dffactor

dferror

F

Significance

500.27

2000.60

3.00

142.00

11.84

.00

a b c

Expected means of ‘‘flexibility of coping’’ in the subpopulations of the confounder. Adjusted means of ‘‘flexibility of coping’’ in the subpopulations of the confounder. H0: all average causal effects (ace) are zero, i.e., Madj( Y | X = x) are equal for all x.

Importantly, both variables were only identified as confounders that influence the causative aspect of the dichotomous and categorical attachment classification on the two coping dimensions (i.e., ‘‘flexibility of coping’’ and ‘‘diverting’’). Neither the sociodemographic nor the disease-related characteristics were found to influence both the type of attachment and coping dimensions. Consequently, we only had to employ the CANOVA to test causal effects that were influenced by the two confounder variables. In regard to all other hypotheses, MANOVA and univariate ANOVAs were used. With respect to the analysis of the first confounder (= type of disease), Table 2 shows the results of the causal analysis of attachment styles on the flexibility of coping that takes these effects into account. The mean differences among the four attachment groups differ strongly when comparing alopecia patients with those suffering from chronic diseases. In particular, ambivalently attached patients have activated a wide variety of coping forms to deal with the chronic disease; however, this was not the case in the ambivalently attached patients with alopecia. It may be noticed that after the necessary adjustment performed automatically by CANOVA, the difference between attachment groups is not as big as before the adjustment; the adjustment has eliminated the bias due to this specific confounding variable. CANOVA adjusts the means according to the influence of the confounder so that the differences between attachment groups diminish. Nevertheless, the analysis showed that the causal effect of attachment style on the flexibility of coping was statistically significant. Since the statistical effect of attachment style on coping strategies remained uninfluenced by the confounder, it was intended to describe the nature of the confounder, i.e., the

type of disease. It was evident that the confounding effect might have been related to the specific condition of alopecia, which had to be considered in more detail. One specific medical characteristic assessed in alopecia patients was the visibility of symptoms and this was the only feature that showed distinctive aspects differing from the other two populations. The visibility of symptoms was assessed by the dermatologist via three categories ranging from ‘‘not obviously visible,’’ to ‘‘marginally visible,’’ to ‘‘obviously visible’’. Table 3 shows a cross-tabulation of the categories related to the visibility of the symptoms and the attachment categorisation. The most striking finding was that all patients with nonvisible hair loss were insecurely attached. The chisquare test (Fisher’s test) for categorical variables was not significant when comparing all of the attachment patterns with the visibility categories [c(6)2 = 5.84, P >.05]. However, when all insecure attachment patterns were collapsed into one category, the chi-square test was significant [c(3)2 = 6.91, P < .05]. The second confounder identified in the study was associated with the (more or less) ‘‘objective’’ acute impair-

Table 3 Cross-tabulation of attachment styles and the visibility of symptoms in patients with alopecia

Ambivalent Avoidant Insecure mixed Secure

Nonvisible

Slightly visible

Obviously visible

7 1 4 0

6 2 6 8

5 1 4 3

Fischer’s test: c2 = 5.84, P > .05.

S. Schmidt et al. / Journal of Psychosomatic Research 53 (2002) 763–773

769

Table 4 ANOVA between attachment styles and coping dimensions Coping dimensions

Secure (n = 53)

Observer rating Seeking attention and care Denial Diverting Rigidity of coping

2.17 0.65 1.50 0.37

Self-report Diverting Seeking attention and care

n = 52 3.69 (0.74)a 2.62 (0.69)

(1.03)a (0.89)a (0.83) (0.26)a

Ambivalent (n = 45) 1.50 0.91 1.13 0.57

(1.00)b (0.83)a (0.82) (0.26)b

n = 38 3.59 (0.70)a 2.78 (0.69)a

F

P

h2*

(0.94)b (0.98) (0.84) (0.24)b

5.40 6.41 1.95 10.90

.002 .001 .124 .001

0.10 0.12 0.04 0.18

n = 25 3.19 (0.63)b 2.48 (0.61)

4.94 2.73

.003 .047

0.10 0.06

Avoidant (n = 21)

Mixed (n = 30)

1.50 1.56 1.50 0.62

1.40 1.37 1.18 0.65

(1.17) (0.99)b (0.96) (0.25)b

n = 18 3.35 (0.71) 2.25 (0.71)b

Means with different signs indicate significant differences in the Scheffe´ tests ( P < .05). Negative emotional in coping in both perspectives was not included because these scales were not normally distributed. All differences between the levels of each variable are significant at P < .05 except those differences with the same superscript. * h2 refers to the portion of variance explained by attachment patterns.

ment caused by the disease. This was a dichotomous variable related to disease characteristics as judged by the physician. This variable differentiated between patients who suffered from acute and strong impairment (i.e., treatment side effects, acute hair loss) and patients who suffered from less impairment. It should be noted that this confounder affected two dependent variables, i.e., diverting and the flexibility of coping, and only under the condition that all insecure attachment categories were collapsed into one category. The identification of this confounder revealed that in the case of acute distress, the statistical causal effect between attachment styles and the flexibility of coping was influenced ( P >.05). The means of securely attached patients in respect to the flexibility of coping behaviour did not diverge from insecurely attached patients in case of suffering from acute and severe conditions. Again the effect of attachment pattern on the flexibility of coping behaviour remained unaffected by the confounder ( F = 40.23; P < .05). The same was true for diverting coping strategies. Having identified the confounding variables, we were able to run conventional statistical tests on the ‘‘unconfounded’’ variables. ANOVA of attachment patterns and coping The remaining data were analysed by two-way MANOVAs and univariate ANOVAs for attachment styles and type of disease. All the tests were two-tailed. The variance of negative emotional coping was not homogeneous in the single cells according to Levene’s test. Therefore, nonparametric tests were applied to the analyses of this dependent variable. The MANOVA of all coping scales yielded a highly significant effect for Attachment Class [ F(27,330) = 3.0, P < .001] and for Type of Disease [ F(18,226) = 10.89, P < .001]. The interaction effect between attachment and type of disease was not significant [ F(54,580) = 1.28, P >.10]. The univariate ANOVAs indicated that the significant main effect was significant for all scales except for diverting in the observer report version. Scheffe´’s post hoc

tests (a =.05) yielded the following differences (Table 4): Securely attached patients relied to a greater extent on interpersonal resources (seeking attention and care) and were less rigid in their ways of coping than all other insecurely attached patients. In the self-report version, however, it was apparent that ambivalently attached individuals sought more attention and care than avoidantly attached individuals. Similar discrepancies between observer and self-report versions of coping occurred in the avoidant attachment style. Patients identified with this attachment pattern showed higher scores on denial as rated by observers; however, this finding was not consistent with findings in the self-reported diverting and with observerreported diverting. From the observer perspective, avoidantly and securely attached patients displayed more diverting strategies than ambivalently attached patients. From the self-report perspective, securely and ambivalently attached patients used more diverting strategies than the patients of the mixed subgroup. These discrepancies between the observer and selfreported version might be related to the attachment style. In Figs. 1 and 2, this association emerges more clearly (these figures represent z-standardised scores). Fig. 1 shows the attachment styles and coping dimensions in the observer report version. From this perspective, almost all scale scores or profiles seem to be congruent with theoretical assumptions. For instance, securely attached patients had high scores on seeking attention and care, avoidantly attached patients showed more diverting strategies while the coping strategies of ambivalently attached patients were characterised by negative emotions or a tendency for depressive coping; the mixed style displayed the highest amount of negative emotions. However, when comparing the profile of ambivalently and avoidantly attached patients, it is striking that in the self-report version, ambivalently attached patients showed high scores on all dimensions. The profile of avoidantly attached individuals indicating low scores on all dimensions reveals contrasting results. Thus, from this coping perspective, hyperactivating and deactivating strategies of these attachment groups are obviously pronounced

770

S. Schmidt et al. / Journal of Psychosomatic Research 53 (2002) 763–773

Fig. 1. Attachment styles and coping strategies in the self-report version (z scores).

in the self-report of coping modes. This effect appears only in the ambivalent and avoidant attachment styles. Both ambivalently attached patients as well as mixed ambivalent/avoidantly attached patients showed increased negative affect. The Kruskal – Wallis rank analyses showed significant differences between attachment patterns in negative emotional coping in the observer report scale [c(3)2 = 32.10, P < .001] as well as the self-report scale [c(3)2 = 12.44, P < .01]. Discrepancies between self- and observer reports of coping forms In order to test whether the discrepancy between self- and observer report rating scales is higher in insecurely attached patients than in securely attached patients, the differences of the z-standardised scores of each of the two corresponding dimensions were calculated. The mean differences, tested with univariate ANOVA, were significant with respect to Diverting Coping [ F(3,126) = 3.87, P < .01] and to Seeking Attention and Care [ F(3,126) = 2.83, P < .05]. In the post hoc analyses of seeking attention and care, the secure attachment pattern showed lower discrepancies between self- and observer reports than either the ambivalent pattern (in seeking attention and care) or the avoidant pattern (diverting). In respect to diverting, the ambivalent showed higher discrepancies than the avoidant attachment category.

interviews of patients at hospital and subsequent observer ratings. Although the AAI is the most common interview, we preferred to use a more behaviourally oriented interview measure, which focuses on adult attachment patterns. It is our assumption that early attachment history was, at least for this particular study, not to that extent clinically nor prognostically relevant as behavioural attachment patterns in adulthood. In fact, having this in mind, Strauss et al. [25] determined to modify and further develop a German adaptation of the AAPR rating. Existing studies [25,28] and especially the findings of the project associated with the presented study [39] support the usefulness of this measurement instrument in clinical investigations. There have been, however, some indications of possible limitations in the use of this measure. These limitations occurred when analysing possible confounders of the attachment– coping relationship. In case of acute distress, for instance, the numbers of insecurely attached patients were higher. This finding indicates some limitations of attachment research when applying the AAPR. Chronic diseases may have a traumatic effect on the patients or may cause a traumatic situation that disables the assessment of attachment patterns and coping strategies, or that leads to less reliable and valid classifications of these measures. The fact that this study was limited to a range of only three different diseases could also be considered to be a limitation; however, this is a common disadvantage when clinical studies must be conceptualised. Thus, a particular concern is to check application of our findings to other disease categories. Despite this, it must be considered that the gathering of multilevel data, as well as the range of clinical parameters, is advantageous. One aspect that needs further discussion is the theoretical, as well as methodological, overlapping of attachment and coping concepts. Theoretically, attachment theory is rooted in developmental psychology and coping theory is related to actual conditions. However, methodologically, both procedures were assessed by independent rater teams

Discussion Methodological considerations — validity of attachment measures A particular strength of this study is that it was clinically oriented, i.e., the measures used to assess attachment patterns and coping strategies have a clinical orientation. Both attachment and coping measures were collected by

Fig. 2. Attachment styles and coping dimensions in the observer-reported version (z scores).

S. Schmidt et al. / Journal of Psychosomatic Research 53 (2002) 763–773

in our study. The coping raters were not familiar with attachment theory and attachment raters were not trained in coping concepts. Thus, though construct overlapping can be considered as a critical point in our study and other studies investigating attachment and coping, the different context of both research traditions at least lends some support to assume the divergence of these concepts. From the perspective of coping research, one major limitation occurs in respect to the gender distribution, since in most studies gender was significantly related to coping [1,35]. Methodological consideration — interpretation of statistical causal effects The aim of this study was to infer an effect of attachment styles on coping strategies, although the inference of a statistical causal effect between attachment and coping is, in general, impossible due to the fact that we only assessed cross-sectional data. A variety of additional theoretical assumptions are required to ensure attachment style influencing coping strategies. In particular, the necessary temporal structure among confounder, treatment and effect was not met. Nevertheless, the study gives us an estimation of the potential statistical influence of attachment styles on coping strategies. The search for confounders was successful, and the CANOVA led to an adequate statistical adjustment with respect to those confounders. However, as in all nonexperimental studies, the existence of other, not observed confounders remains possible and puts limitations on the causal interpretability of the observed effects. Hypotheses In this study, we have taken a first step in the explanation of coping with disease as seen from a framework of attachment theory. The cross-sectional analysis has revealed a significant relationship between attachment style and coping strategies. Effect sizes were predominantly in the moderate range. In general, two levels of the association between attachment styles and coping emerged. These levels emerged since we assessed coping strategies by both a self-reported as well as an observer-reported method. With respect to the observer-reported versions of coping strategies, the empirical findings on the relationship between attachment and coping were partly consistent with theoretical assumptions. Securely attached individuals showed a strong tendency towards seeking social support as already indicated by findings of Feeney and Kirkpatrick [10]. Ambivalently attached individuals (as well as individuals assigned to the mixed pattern) were characterised by negative emotional coping, which has been shown in many studies [12], however, not by higher levels of seeking attention and care as indicated by Feeney and Kirkpatrick [10]. In respect to avoidant coping, avoidantly attached individuals as well as

771

securely attached patients revealed higher levels of diverting. In coping research, diverting has been interpreted as an adaptive strategy [33] while in attachment research, diverting sometimes has the connotation of a maladaptive strategy. In the self-reported version, some contrasting findings occurred. From this stance, ambivalently attached individuals showed high elevations on all of the three coping dimensions. This might reflect a hyperactivating strategy, which is consistent with the assumed strategy of regulating attachment emotions in this attachment style [46]. It could be assumed that an ambivalently attached individual aims at seeking any kind of guarantee in his coping efforts while observers do not assume that this individual really seeks attention and care. In contrast, avoidantly attached individuals showed low scores on all coping dimensions reflecting deactivating strategies of coping. One might conclude that deactivating and hyperactivating strategies of regulating attachment experiences and emotions are revealed in coping strategies. This effect strongly underlines the significant advantages to be gained by the application of an attachment perspective within the context of coping behaviour in patients with chronic disease. Adaptive or maladaptive coping may thus not be interpreted in terms of a single dimension, but by a whole pattern of strategies. The fact that discrepancies between the observer-reported and self-reported versions of seeking attention and care were smaller in the secure attachment style than in all of the insecure attachment styles supports the findings of Spangler and Zimmermann [12]. These researchers concluded that the coherence between different response levels of coping, i.e., the biological, emotional, cognitive and behavioural response levels, differed between secure and insecure attachment styles. Thus, the discrepancy between how a patient perceives him/herself and how that person is perceived by another might also be interpreted as a further signal of the incoherence of response patterns in insecurely attached individuals. In this study, however, the smaller discrepancy was only identified in one strategy and not in the a whole range of response patterns. Both attachment and coping theories have many principles in common that predispose an advantageous conjunction of both approaches. For instance, both theories are interactional models that postulate a continuous directional process of person-by-situation interaction [1,4]. From a perspective of contemporary coping research, tribute has been paid to both models. However, the study and application of psychological aspects in the course of medical treatment of patients have been noticeably neglected in the field of research on attachment behaviour. In our data, we found some hints for the meaning of situational factors in identifying attachment styles. Whether the patient was under severe stress caused by a disease or not affected the attachment categorisation and the flexibility of coping. Though the relationship between attachment styles and the flexibility of coping was still significant, the impact of such a variable might be greater in other samples or smaller

772

S. Schmidt et al. / Journal of Psychosomatic Research 53 (2002) 763–773

sample sizes. Another confounding variable was the type of disease. The association between attachment and coping was much stronger (and different) in the patients with chronic diseases than in those patients with alopecia. The main reason for this confounding influence was related to the fact that a special subgroup of alopecia patients that displayed some dysmorphophobic features was identified. Most of these patients did not show obviously visible signs of alopecia [41]. Consequently, the existence of psychosomatic features might hamper the attachment coping relationship, which corresponds to a specific vulnerability in terms of attachment in psychosomatic diseases [15]. This specific finding strongly supports results presented by Taylor et al. [17] on the mediating effect of attachment patterns on unexplained or subjective physical symptoms. It might be concluded that it is not possible to study the attachment coping relationship without considering the potential influence of this confounder. Though this was not one of the central questions of our investigation, the findings referring to the identification of these confounders thus raise some important issues on the limitations of attachment– coping research. First, the relationship between attachment and coping still has to be replicated in psychosomatic conditions. Second, under specific disease conditions, i.e., acute distress or even traumatic experiences, other factors may be more important in the coping process of the patients. It is particularly important to apply these findings when attachment concepts are used in intervention programs. If, for instance, a patient is under severe distress caused by the treatment, it might not be appropriate to activate ‘‘attachment resources.’’ An important question is not only whether integrating an attachment perspective into coping research would be a helpful approach, but also which particular patients would derive benefit from this. There seem to be some limitations to this approach, as in the application to traumatised patients or in specific treatments; however, in general, from our studies, it can be concluded that an implementation of attachment theory might lead, for patients — as well as their physicians — to a better understanding of the roots of one’s coping processes in the presence of chronic disease.

[6] [7] [8] [9] [10]

[11] [12]

[13] [14]

[15]

[16] [17]

[18] [19]

[20]

[21]

[22]

[23]

References [24] [1] Zeidner M, Endler NS. Handbook of coping: theory, research, applications. New York: Wiley, 1996. [2] Parker JDA, Endler NS. Coping and defense: a historical overview. In: Zeidner M, Endler NS, editors. Handbook of coping: theory, research, applications. New York: Wiley, 1996. pp. 3 – 23. [3] Bowlby J. Attachment and loss. Attachment, vol. 1. New York: Basic Books, 1969. [4] Bowlby J. A secure base. Clinical applications of attachment theory. London: Routledge, 1988. [5] Grossmann KE, August P, Fremmer-Bombik E, Friedl E, Grossmann A, Scheuerer-Englisch H, Spangler G, Stephan C, Suess G. Die Bindungstheorie: modell und entwicklungspsychologische Forschung (Attachment theory-model for research in developmental psychology),

[25]

[26] [27]

In: Keller H, editor. Handbuch der Kleinkindforschung. Berlin: Springer, 1989. pp. 51 – 97. Lazarus RS. From psychological stress to the emotions: a history of changing outlooks. Annu Rev Psychol 1993;44:1 – 21. Fuendeling J. Affect regulation as a stylistic process within adult attachment. J Soc Pers Relat 1998;15:291 – 322. Mikulincer M, Nachshon O. Attachment styles and patterns of selfdisclosure. J Pers Soc Psychol 1991;61:321 – 31. Zuroff DC, Fitzpatrick DK. Depressive personality styles: implications for adult attachment. Pers Individ Differ 1995;18:253 – 365. Feeney BC, Kirkpatrick LA. Effects of adult attachment and presence of romantic partners on physiological responses to stress. J Pers Soc Psychol 1996;70:255 – 70. Kumin I. Pre-object relatedness. New York: Guilford, 1996. Spangler G, Zimmermann P. Attachment representation and emotion regulation in adolescence. A psycho-biological perspective of the internal working model. Attachment Hum Dev 1999;1:270 – 90. Kotler T, Buzwell S, Yolanda R, Bowland J. Avoidant attachment as a risk factor for health. Br J Med Psychol 1994;67:237 – 45. Feeney JA, Ryan SM. Attachment style and affect regulation: relationships with health behavior and family experiences of illness in a student sample. Health Psychol 1994;13:334 – 45. Scheidt CE, Waller E, Schnock C, Becker-Stoll F, Zimmermann P, Lucking CH, Wirsching M. Alexithymia and attachment representation in idiopathic spasmodic torticollis. J Nerv Ment Dis 1999;187: 47 – 52. Stuart S, Noyes R. Attachment and interpersonal communication in somatization. Psychosomatics 1999;40:34 – 43. Taylor RE, Mann AH, White NJ, Goldberg DP. Attachment style in patients with unexplained physical complaints. Psychol Med 2000;30: 931 – 41. Mc Neal D. Epidemiological research methods. New York: Wiley, 1996. Steyer R, Gabler S, Rucai AA. Individual causal effects, average causal effects, and unconfoundedness in regression models. In: Faulbaum F, Bandilla W, editors. SoftStat ’95. Advances in statistical software 5. Stuttgart: Lucius and Lucius, 1996. pp. 203 – 10. Steyer R, Gabler S, von Davier AA, Nachtigall C. Causal regression models: II. Unconfoundedness and causal unbiasedness. Methods Psychol Res Online 2000;5(3):55 – 86. Wuethrich-Martone O, Mueller M, Steyer R. Controlling qualitative confounders in nonrandomised experiments: a method and its implementation in SPSS. In: Nunez-Anton V, Ferreira E, editors. Proceedings of the 15th International Workshop on Statistical Modelling, Bilbao, Spain 2000;364 – 7. Nachtigall C, Steyer R, Wu¨thrich-Martone O. Causal effects in empirical research. In: May M, Ostermeier U, editors. Interdisciplinary perspectives on causality. Bern studies in the history and philosophy of science. Norderstedt: Libri Books on Demand, 2001. pp. 81 – 100. Endler NS, Kocovski NL, Macrodimitris SD. Coping, efficacy, and perceived control in acute vs. chronic illnesses. Pers Individ Differ 2001;30:617 – 25. Endler NS, Parker JDA, Summerfeldt LJ. Coping with health problems: developing a reliable and valid multidimensional measure. Psychol Assess 1998;10:195 – 205. Strauss B, Lobo-Drost A, Pilkonis PA. Einschaetzung von Bindungsstilen bei Erwachsenen-erste Erfahrungen mit der deutschen Version einer Prototypenbeurteilung (Assessment of attachment styles in adults — first results of the German version of the Adult Attachment Prototype Rating), Z Klin Psychol, Psychiatr Psychother 1999;47: 347 – 64. Pilkonis PA. Personality prototypes among depressives: themes of dependency and autonomy. J Pers Disord 1988;2:144 – 52. George C, Main M, Kaplan N. The Berkeley Adult Attachment Interview. Unpublished manuscript. Berkeley: University of California, 1985.

S. Schmidt et al. / Journal of Psychosomatic Research 53 (2002) 763–773 [28] Strauss B, Buchheim A. Interview methods in adult attachment research. In: Strauss B, Buchheim A, Raechele H, editors. Handbook of clinical attachment research. Goettingen: Hogrefe, 2002. pp. 22 – 53. [29] Fraley RC, Waller NG, Brennan KA. An item – response theory analysis of self-report measures of adult attachment. J Pers Soc Psychol 2000;78:350 – 65. [30] Collins NL, Read SJ. Adult attachment, working models, and relationship quality in dating couples. J Pers Soc Psychol 1999;58:644 – 63. [31] Heinrichs M, Ditzen B, Fink N, Zeugin A, Schmidt S, Strauss B, Ehlert U. Effects of attachment style and social support on endocrine and subjective stress responses in human. 60th Annual Scientific Meeting of the Society of Psychosomatic Medicine, Barcelona, March 13 – 16, 2002. [32] Heim E, Augustiny K, Blaser A, Schaffner L. Berner Bewaeltigungsformen BEFO (The Bernese Coping Modes). Bern: Huber, 1991. [33] Heim E, Valach L. Berner Bewaeltigungsformen BEFO-ein Instrument zur Selbst- und Fremdbewertung der Krankheitsverarbeitung (The Bernese Coping Modes — self and an observer rating scales to assess strategies of coping with chronic disease). Rehabilitation 1996; 35:31 – 42. [34] Heim E, Valach L, Schaffner L. Coping and psychosocial adaptation: longitudinal effects over time and stages in breast cancer. Psychosom Med 1997;59:408 – 18. [35] Hessel A, Heim E, Geyer M, Braehler E. Normierung und Skalierung der Berner Bewaeltigungsformen in der Selbstbeurteilung und Bestimmung personaler und situativer Einflussfaktoren. Psychother, Psychosom Med Psychol 2000;50:309 – 42. [36] Sommer G, Fydrich T. Entwicklung und Ueberpruefung eines Fragebogens zur Sozialen Unterstuetzung (F-SOZU) (Development and evaluation of the social support questionnaire). Diagnostica 1991;37: 160 – 78.

773

¨ berpru¨fung des Fragebogens zur Sozia[37] Franke GH. Testtheoretische U len Unterstu¨tzung (Psychometric criteria of the social support questionnaire), Z Med Psychol 1994;3:168 – 77. [38] Schmidt S. Bindung und Coping — Eine empirische Untersuchung zu Krankheitsverarbeitungsprozessen aus einer entwicklungspsychologischen Perspektive [Attachment and coping — a study on coping strategies from a developmental perspective]. Unpublished dissertation, University of Jena, 2000. [39] Hunt SM, McKenna SP, McEwen J, Williams J, Papp E. The Nottingham Health Profile: subjective health status and medical consultations. Soc Sci Med 1981;15:221. [40] Kohlmann T, Bullinger M, Kirchberger-Blumstein I. German version of the Nottingham Health Profile (NHP): translation and psychometric validation. Soz Praeventivmed 1997;42:175 – 85. [41] Schmidt S, Fischer T, Elsner P, Strauss B. Quality of life and strategies of coping in female patients with alopecia. Br J Dermatol 2001;144: 1038 – 43. [42] Rubin DB. Estimating causal effects of treatments in randomized and nonrandomized studies. J Educ Psychol 1974;66:688 – 701. [43] Rosenbaum PR. From association to causation in observational studies. The test of strong ignorable treatment assignment. J Am Stat Assoc 1986;79:41 – 8. [44] Steyer R, Gabler S, von Davier AA, Nachtigall C, Buhl T. Causal regression models: I. Individual and average causal effects. Methods Psychol Res Online 2000;5(2):39 – 71. [45] Wuethrich-Martone O, Nachtigall C, Mueller M, Steyer R. CANOVA: a Program for testing causal effects. Metheval Report, 2002, 1 (Internet: www.uni-jena.de/svw/metheval/). [46] Kobak RR, Sceery A. The transition to college: working models of attachment, affect regulation, and perceptions of self and others. Child Dev 1988;88:135 – 46.