Setting

The eating disorders in-patient unit of the South London and Maudsley National Health Service (NHS) trust is a national and local referrals unit. All patients admitted to the unit and their families are encouraged to take part in a family-based intervention, part of the in-patient programme.

During the time of this study, patients (all of whom were female) and their families were randomly assigned to one of two family interventions. Families were either seen individually on a weekly basis for approximately 18 weeks (treatment as usual), or in conjunction with another family for an intensive 3-day workshop intervention. The aim of both interventions was to promote carer efficacy by providing psychoeducation and teaching motivational communication skills. The two interventions used similar techniques and were facilitated by the same mental health professionals.

At the end of the first day (or around the fifth session) carers were asked to complete a therapeutic writing task. Carers were instructed to write in their own time a letter describing what it is like to be the mother/father/sister/partner of someone with anorexia nervosa, and to bring it to the following session to read and discuss with the group. Carers were not given any time or length restraints for this exercise. All carers consented to their letters being used for teaching and research purposes. Ethical approval was obtained for this project and written consent was obtained from the participants.

Participants

Letters from 20 fathers and 20 mothers were collected and included in the analysis. For 13 patients, both parents’ letters were included; for the other 14 patients only one parent's letter was available. The mean age of carers was 52.7 years (range 40–66). The mean age of the mothers was 50.8 years (range 40–60) and that of the fathers was 54.4 years (range 45–66). The 27 patients had a mean age of 20.5 years (range 14–31) and a mean body mass index of 13.6 kg/m² (range 10.8–16.9) at admission. Approximately 45% of the patients had a duration of illness of greater than 5 years, and about 20% had a duration greater than 10 years. Approximately 65% of patients and their families participated in the family intervention offered within the in-patient service. This is a representative sample of families presenting to a local and national in-patient service, as only 10% of patients and their families refused to take part. Others could not participate for reasons such as lack of family and self-discharge.

Qualitative analysis

Letters were analysed using the ‘constant comparative’ method of the grounded theory approach (Reference Glaser and StraussGlaser & Strauss, 1967). Using this approach, no pre-existing category is imposed on the data but themes begin to emerge as the analysis takes place. The researcher generates as many themes as possible, while at the same time comparing them with previous incidents coded under the same theme. Similarities and differences between themes are examined. This allows the researcher to develop hierarchical coding systems and devise a theoretical framework (Reference Glaser and StraussGlaser & Strauss, 1967; Reference WilligWillig, 2001).

Two authors (J.W. and J.T.) initially examined the letters to identify emerging themes. Letters were separated into statements and a line-by-line open coding session was performed to ensure that the analysis was grounded to the data and comprehensive. The coding scheme generated by the lead investigator (J.W.) was used to code all the letters; this author continually checked and modified the coding framework to ensure that it incorporated all information within the transcripts. To ensure reliability of the findings, the devised coding scheme was given to an additional investigator (K.G.) to code a random set of 15 letters. A k coefficient of 0.61 was found, illustrating good interrater reliability of the coding scheme. The coding scheme was shown to be comprehensive and no new theme emerged. Once themes and categories were found, the authors examined models in the mental health and caregiving literature that corresponded with the findings.

Computerised text analysis

The narratives were also analysed using a computerised text analysis program, the Linguistic Inquiry and Word Count (LIWC; Reference Pennebaker and FrancisPennebaker & Francis, 1999). This program is designed to analyse text according to emotional, cognitive, structural and linguistic processes used within the writing. It conducts a word-by-word analysis and assigns words to over 70 categories, and then calculates the percentage of words in each category. Results from 43 individual studies were compiled by Pennebaker & Francis (Reference Pennebaker and Francis1999) to produce mean reference values for ‘emotional writing’ (when individuals are asked to write about deeply emotional topics) and ‘control writing’ (when asked to write about non-emotional topics). These published reference values were used for comparison. Analyses were conducted between genders using robust standard errors to control for the clustering of families.

Carers’ illness perceptions

Impact of the illness

Emotional and cognitive processes

Maladaptive beliefs and responses

Detrimental appraisals were evident within carers’ narratives, highlighting misperceptions about their own or their daughter's role in causing the illness and their lack of efficacy, as illustrated in the following statements.

Gender differences

The main gender difference between the accounts was the tendency for mothers to express greater anxiety, whereas fathers produced more cognitive and detached accounts, with a greater use of cognitive and avoidant coping strategies.

Validation

The results of the LIWC analysis are presented in Table 1. The texts included a high proportion of emotional words, particularly negative emotions. The proportions of fear and anxiety words were double those reported in emotional writing norms. There was a significant gender difference in the use of emotional words. Mothers used a significantly higher proportion of emotion words (t=2.59, P=0.015), particularly negative emotions (t=2.21, P=0.036), with the greatest difference in the sadness and depressive domain (t=2.93, P=0.007).

Table 1

Text analysis using the Linguistic Inquiry and Word Count

LIWC word category	Reference values (emotional writing)1 %	Writing task			t	Robust standard error of test coefficient2	95% CI for difference between genders
		All carers, % Mean (s.d.)	Mothers only, % Mean (s.d.)	Fathers only, % Mean (s.d.)
Emotional processes (e.g. happy, bitter, ugly)	5.3	7.4 (3.1)	8.6 (3.4)	6.3 (2.4)	2.59	0.87	0.47 to 4.04
Negative emotions (e.g. hate, worthless)	2.6	4.5 (3.1)	5.4 (3.8)	3.5 (1.7)	2.21	0.88	0.14 to 3.73
Anxiety or fear (e.g. nervous, afraid)	0.6	1.3 (1.7)	1.6 (2.3)	0.9 (0.7)	1.37	0.52	-0.36 to 1.79
Anger (e.g. hate, kill)	0.7	1.0 (1.0)	1.1 (1.2)	0.9 (0.8)	0.68	0.29	-0.39 to 0.78
Sadness and depression (e.g. grief, cry, sad)	0.7	1.2 (0.9)	1.5 (1.0)	0.8 (0.6)	2.93	0.25	0.22 to 1.25
Positive emotions (e.g. happy, pretty, good)	2.7	3.0 (1.4)	3.1 (1.6)	2.8 (1.2)	0.82	0.47	-0.57 to 1.34
Positive feelings (e.g. happy, joy, love)	0.9	0.9 (0.8)	1.0 (0.9)	0.8 (0.7)	0.77	0.23	-0.29 to 0.64
Optimism and energy (e.g. certainty, pride, win)	0.5	0.8 (0.7)	0.8 (0.9)	0.8 (0.5)	0.32	0.22	-0.37 to 0.51
Cognitive processes (e.g. cause, know, ought)	7.8	8.3 (2.7)	8.9 (3.1)	7.8 (2.1)	1.30	0.85	-0.64 to 2.84
Causation (e.g. because, effect, hence)	1.1	1.1 (1.0)	1.2 (0.9)	1.1 (0.9)	0.20	0.33	-0.62 to 0.74
Insight (e.g. think, know, consider)	2.5	2.6 (1.3)	2.9 (1.4)	2.3 (1.1)	1.37	0.42	-0.28 to 1.43
Discrepancy (e.g. should, would, could)	2.7	2.6 (1.4)	2.9 (1.6)	2.3 (1.1)	1.47	0.45	-0.26 to 1.58
Inhibition (e.g. block, constrain)	0.3	0.5 (0.7)	0.6 (0.9)	0.4 (0.4)	0.77	0.27	-0.29 to 0.64
Tentative (e.g. maybe, perhaps, guess)	2.5	2.0 (1.2)	1.9 (1.2)	2.2 (1.2)	-0.66	0.35	-0.96 to 0.49
Certainty (e.g. always, never)	1.4	1.7 (1.2)	1.7 (1.6)	1.6 (0.7)	0.31	0.42	-0.72 to 0.98
Proportion of unique words	50.5	55.4 (11.8)	57.8 (14.8)	53.1 (7.3)	1.25	3.78	-3.03 to 12.48

Coping responses

Carers used a variety of coping responses to manage their anxiety and the demands placed upon them. Although some of the narratives illustrated adaptive coping responses such as focusing on their own interests, showing hope and optimism, seeking social support and externalising the illness, others produced themes suggesting maladaptive coping. For example, some carers, particularly mothers, expressed profound anxiety which in some cases appeared to reach clinical severity. Others expressed negative appraisals of the illness linked to hostility and criticism towards the patient and themself, and an overprotective response to the illness. The computerised text analysis of the narratives confirmed some of the findings from the thematic analysis. The level of emotion expressed within the narratives was higher than has been reported in the literature of other emotional writing settings (Reference Pennebaker and FrancisPennebaker & Francis, 1999), particularly for the mothers.

Previous research

These results are consistent with previous caregiving research which has found that women carers are more emotionally involved (Reference Stern, Doolan and StaplesStern et al, 1999), whereas men tend to distance themselves (Reference Morris, Woods and DaviesMorris et al, 1991) or use cognitive restructuring to manage the emotional impact (Reference Barusch and SpaidBarusch & Spaid, 1989). The heightened anxiety and overprotective response paired with the hostility and criticism correspond to the concept of ‘high expressed emotion’. This emotional response style in carers is widely recognised to be associated with a poor outcome in mental illness (Reference Wearden, Tarrier and BarrowcloughWearden et al, 2000) and has an important impact on the outcome (Reference Van Furth, van Strien and Martinavan Furth et al, 1996), withdrawal rates (Reference Szmukler, Eisler and RussellSzmukler et al, 1985) and response to treatment (Reference Eisler, Dare and HodesEisler et al, 2000) in anorexia nervosa.

Strengths and limitations

This is an exploratory study with a restricted methodology, which limits the conclusions that can be drawn. Parents’ accounts were produced as a therapeutic exercise, allowing open-ended and unguided narratives, although with the purpose of using the material in the process of treatment, which might have led to some bias in reporting. Specifically, there was a clear benefit of highlighting specific perceptions and responses that are viewed as unproductive. We used two approaches for the analysis of the texts. Although the thematic analysis of carers’ accounts allowed a comprehensive, unconfined exploration of the data, reliability and researcher bias are questionable within such approaches. Conversely, the computerised text analysis program LIWC is a widely used tool which provides a reliable, unbiased analysis of the data, but is limited by its rigid categories and purely literal analysis of the text, which restricts its use for meaning. These complementary approaches produced similar findings, suggesting a degree of validation in the results. The generalisability of the findings is uncertain, because the sample consisted of carers of patients who were receiving specialist in-patient care and thus were at the severe end of the illness spectrum. Many wrote about the illness in terms of a medical model (e.g. causes and chronicity of the illness). Their beliefs about the condition are likely to have been influenced by exposure to a psychiatric setting. It is possible that those caring for people with less severe illness, who have not received hospitalised care, would articulate more general or lay beliefs about the illness.

Clinical implications and recommendations

The overall aim of the study was to develop a conceptual framework for an intervention to optimise carers’ coping and reduce any interpersonal maintaining factors. The analysis of the parents’ accounts of caring for someone with anorexia nervosa has identified important clinical considerations for the development of new interventions.

CLINICAL IMPLICATIONS

1. ▪ Carers’ appraisals of the causes, consequences and outcomes of anorexia nervosa may be detrimental and contribute to an emotional and behavioural response that may be maladaptive.

2. ▪ Some carers, mothers in par particular, ticular, are highly distressed by the illness and might benefit from an assessment of their needs, as recommended by the National Service Framework for Mental Health.

3. ▪ Interventions that interrupt overprotective and highly emotional responses to the illness may improve the patients’ outcome.

LIMITATIONS

1. ▪ Narratives were collected with the purpose of using the material in the process of treatment. This might have led to some bias in reporting.

2. ▪ Although the design facilitated an unguided and open narrative, the material was limited and individual themes could not be examined in greater depth.

3. ▪ There is a question of generalisability of the findings, as the sample consisted of carers of patients receiving specialist in-patient care and thus at the severe end of the illness spectrum.