Public Beliefs About the Helpfulness of Interventions for Depression: Effects on Actions Taken When Experiencing Anxiety and Depression Symptoms

Abstract

Objective: Previous research has shown that the public have different beliefs to mental health professionals about the helpfulness of interventions for mental disorders. However, it is not known whether the public's beliefs actually influence their behaviour when they develop psychiatric symptoms.

Method: A postal survey of 3109 Australian adults was used to assess beliefs about the helpfulness of a broad range of interventions for depression, as well as respondents' current level of anxiety and depression symptoms and any history of treated depression. A follow-up survey of 422 persons who had a high level of symptoms at baseline was conducted 6 months later. These people were asked which interventions they had used to reduce their symptoms. An analysis was carried out to see whether beliefs and other factors at baseline predicted subsequent use of interventions.

Results: There were some major discrepancies between the ranking of interventions as likely to be helpful and the ranking of how frequently they were actually used. Interventions involving mental health professionals were often rated as likely to be helpful, but were rarely used in practice. Other simple, cheap and readily available interventions were used the most frequently, but were not the most likely to be rated as helpful. The most consistent predictors across all interventions used were gender, history of treatment, current symptoms and belief in a particular intervention. Of particular interest was the finding that beliefs in the helpfulness of antidepressants predicted their use. However, beliefs were not predictors of use for all interventions.

Conclusions: Beliefs about the helpfulness of an intervention did not always predict actual use of that intervention, although beliefs did predict use of antidepressants. Therefore, campaigns that change public beliefs about effective treatments may also influence actual use of treatments. Interventions preferred by professionals are not frequently used at present. Most people with anxiety and depression symptoms rely primarily on simple self-help interventions, the effectiveness of which has been little researched.

The best treatment is not simply the most efficacious one, but the one that is actually put to use [1].

We have previously reported data from Australian national surveys of public and professional beliefs about the helpfulness of various interventions for mental disorders [2–4]. The surveys of professionals involved samples of psychiatrists, clinical psychologists and general practitioners. In all surveys, respondents were given a vignette describing a person with either major depression or schizophrenia. They were asked to rate a broad range of interventions as likely to be helpful, harmful or neither for the person described. The list of interventions included both professional and non-professional help and pharmacological and non-pharmacological treatments.

These surveys revealed some major discrepancies between public and professional beliefs. The public gave much higher ratings than the professionals to vitamins and minerals and to special diets for both depression and schizophrenia, and to reading self-help books for schizophrenia. On the other hand, the professionals gave much higher ratings for the helpfulness of antidepressants for depression, and of antipsychotics and admission to a psychiatric ward for schizophrenia. There were also numerous other less dramatic differences. For example, the public had a lower endorsement of the likely helpfulness of psychiatrists and clinical psychologists. Surveys in other countries of public attitudes towards treatment show similar findings [5–9].

The discrepancy between public and professional beliefs may have implications for help seeking and adherence to treatment when people develop a mental disorder. It is not enough to have evidence-based treatments available if the public do not believe in their efficacy and do not want them [9]. The findings also underscore the need for programs to improve the mental health literacy of the public, such as the Defeat Depression campaign in the UK which attempted to overcome misconceptions about anti-depressant treatment [8]. However, an assumption of such education programs is that beliefs about treatment actually influence behaviour. Here we report relevant evidence from a longitudinal study of a community sample in which beliefs about treatment were assessed at baseline. Those who had common psychiatric symptoms were followed up 6 months later and asked what actions they had taken to relieve their symptoms.

Andersen has proposed an influential ‘behavioural model’ of the predictors of health service use, which was used as a framework here [10]. According to this model, service use is dependent on a person's level of illness (illness level), the predisposition of the individual to use services (predisposing factors) and the individual's ability to secure services (enabling factors). Predisposing factors include beliefs about the helpfulness of services, as well as sociodemographic characteristics. Enabling factors would include the availability of services, being close to services and having health insurance. Although the Andersen behavioural model was developed to explain use of formal health services, we have extended it here to apply to self-help interventions as well. The hypothesis tested was that belief about the helpfulness of an intervention will be a predictor of actually using that intervention even when illness severity, enabling factors and other predisposing factors are statistically controlled.

Method

Baseline survey

A postal survey on ‘Stress and Well-Being’ was carried out with persons aged 20–59 on the electoral roll for the electorate of Farrer, which includes the city of Albury and a surrounding rural area in New South Wales, Australia. Questionnaires for self-completion were posted to 8000 people. Those who did not respond within 1 month received a reminder and another copy of the questionnaire.

The questionnaire covered sociodemographic characteristics of the respondent, personality, social support, life events and psychiatric symptoms, beliefs about interventions for depression and attitudes to depression. The only questions of relevance here are those on beliefs about interventions, history of depression and treatment, current psychiatric symptoms and sociodemographic characteristics.

The section of the questionnaire on beliefs about interventions was based on a vignette of a person suffering from major depression which was used in previous surveys of the public and health professionals [2,3]. In the earlier surveys, the gender of the person described was randomly assigned to be either male (‘John’) or female (‘Mary’). However, since gender of the vignette made very little difference to the findings, we used only the ‘Mary’ vignette here. The vignette was: ‘Mary is 30-year-old. She has been feeling unusually sad and miserable for the last few weeks. Even though she is tired all the time, she has trouble sleeping nearly every night. Mary doesn't feel like eating and has lost weight. She can't keep her mind on her work and puts off making any decisions. Even day-to-day tasks seem too much for her. This has come to the attention of Mary's boss who is concerned about her lowered productivity.’

Respondents were asked ‘What would you say, if anything, is wrong with Mary?’ They were then asked to rate whether a number of people who could possibly help Mary were likely to be helpful, harmful or neither for her. These people were: a typical GP or family doctor; a typical chemist (pharmacist); a counsellor; a social worker; a telephone counselling service such as Lifeline; a psychiatrist; a psychologist; family; close friends; a naturopath or a herbalist; the clergy, a minister or a priest. Respondents were also asked, ‘Is it likely to be helpful, harmful or neither if Mary tries to deal with her problems on her own?’. Next, respondents were asked to rate some medicines as likely to be helpful, harmful or neither: vitamins and minerals, tonics or herbal medicines; pain relievers, such as aspirin, codeine or panadol; antidepressants; antibiotics; sleeping pills; anti-psychotics; tranquillisers such as Valium. Next were a series of activities: becoming more active physically, such as playing more sport, or doing a lot more walking or gardening; reading about people with similar problems and how they have dealt with them; getting out and about more; attending courses on relaxation, stress management, meditation or yoga; cutting out alcohol altogether; having an occasional alcoholic drink to relax; going on a special diet or avoiding certain foods; taking a holiday or taking a day off work now and then; cutting down on the number of commitments; having a massage or spinal manipulation; taking up new recreational activities. Last was the following list of treatments: counselling; psychotherapy; hypnosis; being admitted to a psychiatric ward of a hospital; undergoing electro-convulsive therapy (ECT). There was also an open-ended question: ‘Is there anything else that you think could help Mary that has not yet been mentioned?’.

To assess history of depression and treatment, respondents were asked: ‘Have you ever in your life been markedly depressed; that is, for several weeks or more, you felt sad, lost interest in things and felt lacking in energy?’. If the response was ‘yes’, they were asked: ‘Did you see a counsellor or a doctor for it at the time?’. The three symptoms mentioned in the first question are the core symptoms for Depressive Episode in ICD-10 [11]. The answers to these questions were used to determine whether respondents had a history of treated or untreated depression.

Other variables used here are age, gender, years of education, marital status, whether the person had private health insurance, the 12-item General Health Questionnaire (GHQ-12) [12] and the anxiety and depression scales of Goldberg et al. [13]. Rural residence was determined from the person's postcode and defined as living more than 100 km (i.e. a little over 1 hour's drive) from the city of Albury.

Follow-up survey

The mean score on the GHQ-12 at baseline was 2.0 for males and 2.3 for females, suggesting the appropriateness of a 2–3 cutoff for screening for common mental disorders [14]. We chose to use a higher threshold of 3–4 to ensure a higher probability of including people with significant morbidity. Accordingly, those who had a score of 4 + on the GHQ-12 were sent a follow-up questionnaire 6 months later. This questionnaire repeated most of the questions asked in the baseline questionnaire. However, instead of assessing beliefs about the likely helpfulness of interventions, the follow-up questionnaire asked about ‘… things you might have done in the past 6 months to cope with stress, anxiety, depression or other emotional problems’. The questionnaire then listed the various actions rated at baseline and the respondent had to answer ‘yes’ or ‘no’ according to whether or not they had taken that action. A few additions were made to the list based on answers to the open-ended question at baseline. These were: ‘had sex to reduce stress’ and ‘meditated or prayed’. As well, the vitamins question was split into two components: ‘vitamins, minerals or tonics’ and ‘herbal medicines’, in view of the increasing popularity of St John's wort as a depression treatment. However, for the analyses reported here, the two were combined to allow comparison with the single question used at baseline.

Statistical analysis

We assessed the frequency with which symptomatic individuals rated each action as likely to be helpful at baseline, and the frequency with which they reported using each intervention at follow up.

To assess predictors of using each intervention, a series of multiple logistic regression analyses was carried out. In each analysis, the dependent variable was whether or not the person had used a specific intervention (such as see a GP, take vitamins, get out more etc.). The predictor variables were age, gender, education, being married, having private health insurance, living in a rural area, having a history of treated or untreated depression, the sum of anxiety and depression symptoms at baseline on the Goldberg et al. scales [13] and belief in the intervention (3 = helpful, 2 = neither, 1 = harmful). These predictors were chosen to cover the three components in the Andersen behavioural model of health service use: illness severity (history, symptoms), predisposing factors (sociodemographic characteristics, beliefs) and enabling factors (health insurance, rural residence) [10]. These analyses were carried out only with those interventions used by 10% or more of the respondents. Other interventions were used by fewer than 40 people and there was lower statistical power to evaluate predictors.

The measure of symptoms selected for the analysis (Goldberg anxiety and depression scales) was deliberately chosen to be different to that used for selecting symptomatic respondents (the GHQ-12), so as not to have an artificial minimum score imposed by the cut-off, and to eliminate the component of regression to the mean due to measurement error at baseline. Although the Mary vignette and the questions about history of treatment deal specifically with depression, the correlation between anxiety and depression symptoms was so high (0.73 in the whole sample and 0.60 in the symptomatic subsample) that we considered them together in the analysis rather than focusing only on depression.

To assess whether the interventions used had any effect on symptoms, a multiple linear regression analysis was carried out for each intervention. In these analyses the dependent variable was symptom change on the Goldberg scales (number of symptoms at baseline minus the number at follow up). The predictor variables were age, gender, education, being married, having private health insurance, living in a rural area, having a history of treated or untreated depression and whether or not the particular intervention was used. The aim of the analysis was to find out if using an intervention predicted symptom change controlling for the other predictors.

Results

Responses to the baseline questionnaire were received from 3109 persons (39% response rate). Compared with the 1996 census data for the electorate of Farrer, this sample had an overrepresentation of females and of people aged 50–59, and an under-representation of people aged 20–34. Of these 3109 respondents, 580 had a GHQ-12 score of 4 + (19%). Follow-up questionnaires were sent to these symptomatic persons 6 months later and responses were received from 422 (73%). There was a mean of 182 days (SD = 17) between completing the baseline and follow-up questionnaires. The non-responders to the follow-up questionnaire differed from the responders in that they tended to be younger and male, but they did not differ in education, marital status, anxiety and depression symptoms or residence (urban or rural).

At baseline, respondents to the follow-up questionnaire had a mean of 5.2 (SD = 2.3) depression symptoms and 6.3 (SD = 2.1) anxiety symptoms, indicating a high level of morbidity. According to Goldberg et al. [13], a score of 2 on depression or 5 on anxiety indicates a 50% chance of clinically important disturbance.

When asked what, if anything, was wrong with the person in the vignette, depression was mentioned by 59% of all participants at baseline and by 66% of the symptomatic subgroup that was followed up.

The symptomatic group used a mean of 6.9 (SD = 4.1) of the interventions listed at both baseline and follow-up. Including the two additional interventions only listed at follow-up (‘had sex to reduce stress’ and ‘meditated or prayed’) gave a mean of 7.4 (SD = 4.3) interventions used. Females used more interventions on average than males (means of 8.0 vs 6.4, p < 0.001). Only 3.3% of males and 1.9% of females used no interventions.

Table 1 shows the rank order of interventions rated as likely to be helpful at baseline, together with the rank order of actual use of interventions in the following 6 months. For example, counselling was rated as likely to be helpful by 93%, giving it the highest ranking, whereas only 15% actually used counselling in the following 6 months and it ranked as 17th most used intervention. Interventions which were ranked more highly for likely helpfulness than for actual use were having counselling, seeing a counsellor, learning relaxation, seeing a psychologist, having telephone counselling, seeing a psychiatrist and seeing a social worker. Interventions ranked more highly for actual use than for likely helpfulness were occasional drink, pain relievers, vitamins, special diet, consulting a chemist, sleeping pills and antibiotics. The ‘had sex to reduce stress’ and ‘meditated or prayed’ interventions are not shown in Table 1, but were used by 21% and 30%, respectively.

Table 1.

Predictors of using various interventions to ‘cope with stress, anxiety, depression or other emotional problems’: significant odds ratios from multiple logistic regression analysis

Table 2 shows the results of the multiple logistic regression analyses with interventions that were used by 10% or more of respondents. Gender, history of treated depression, number of anxiety and depression symptoms and belief in an intervention were the most consistent predictors of using an intervention. However, the pattern of predictors varied considerably. Not shown in Table 2 are the predictors of the ‘had sex to reduce stress’ and ‘meditated or prayed’ interventions for which beliefs at baseline were not assessed. Examining the remaining predictors, ‘have sex’ was used less by older people (OR = 0.97, p = 0.047) and more by males (OR = 0.47, p = 0.005), while meditation or prayer was used more by females (OR = 2.02, p = 0.005).

Table 2.

Rank ordering (and percentages) of interventions rated as likely to be helpful at baseline, and interventions actually used in following 6 months

The participants showed a tendency for their symptoms to improve from baseline to follow up, but there was considerable variability (mean improvement = 1.71, SD = 3.98). Linear regression analyses were carried out to see if use of any of the interventions predicted improvement in symptoms. The only significant effects were for use of pain relievers, which predicted a worsening of symptoms (standardised β = −0.17, p = 0.007) and physical activity, which predicted an improvement (β = 0.10, p = 0.04).

Discussion

The major aim of this paper was to assess whether beliefs about the helpfulness of an intervention predict its actual use. Relationships were found for some interventions, but not for others. It is reassuring that beliefs were a strong predictor of using anti-depressants, given that this was a major focus of the UK Defeat Depression campaign. One factor that might be relevant to whether or not beliefs predict use of an intervention is a person's prior history of using that intervention. It may be that some interventions are used on a regular basis and have been found to be helpful (e.g. long-term use of medication). In such cases, beliefs are likely to predict subsequent behaviour. On the other hand, some interventions may never have been tried, so that beliefs do not have much basis in experience. In such cases there may be weaker links to subsequent behaviour. Another relevant factor may be the strength of beliefs, which was not assessed in the present study. Beliefs about some interventions (e.g. seeing a natur-opath) may be more polarised than beliefs about others (e.g. having counselling).

Examining the other predictors of use of interventions, it is no surprise that history of treatment and number of current symptoms were often predictors. More interesting are the gender differences. Women were more likely to use a range of interventions, particularly those involving lifestyle changes (e.g. involving physical activity, close friends, massage, special diet, and a naturopath), while men had a more restricted repertoire. The only interventions used more by men were alcohol and sex. However, there were no gender differences on those interventions which we previously found to be favoured by professionals (seeing GPs, antidepressants, seeing a counsellor and having counselling) [3].

In general, the results were consistent with the Andersen behavioural model. However, the enabling factors of private health insurance and rural residence did not affect use of any of the interventions examined, indicating that such factors do not determine access. Rather, the rate of consulting mental health professionals was low for the whole population, perhaps due to low availability or reluctance to consult.

The present study also examined whether the interventions used predicted change in symptoms over 6 months. The only significant effects were for pain relievers, which were associated with a worse outcome, and physical activity, which was associated with a better outcome. The fact that other interventions were not associated with outcome should not be taken to mean they are ineffective. To examine these interventions properly requires a randomised, controlled trial. Unfortunately, randomised trials are either impossible or unlikely to be carried out for many of the commonly used interventions. Despite their limitations, naturalistic observation studies may be the only source of evidence available.

A striking finding was that the ranking of interventions as likely to be helpful was quite different to the ranking of how often they were actually used. The interventions that were ranked more highly for likely helpfulness than for actual use mainly involved the services of mental health professionals. There are several possible reasons for this discrepancy in rankings. First, mental health professional services are a scarce resource which is generally only available to people with the most severe disorders. However, scarcity of availability cannot be the only reason, because GPs are relatively inexpensive and readily available in Australia, yet were used by only 35% in the present study. Reluctance to consult for a mental disorder because of the stigma involved is another possible factor. However, in an analysis of other data from this study, we found that attitude measures about depression did not predict help-seeking [15]. Third, the respondents may have seen the person in the depression vignette as more impaired than themselves and so more in need of professional help. However, the vignette was written to barely satisfy the diagnostic criteria for major depression in DSM-IV and ICD-10. Although we did not have a diagnostic measure on our participants, the mean number of anxiety and depression symptoms would indicate that they were as impaired as the person in the vignette. Nevertheless, it is possible that the specific pattern of symptoms in the vignette differed from what some respondents themselves experienced, leading to different preferred options for intervention. For example, anxiety and/or depression as seen in general practice may often have a somatic presentation which is different to that described in the vignette. A related factor is that the ratings of helpfulness for a hypothetical other person may not be the same as the ratings that would be given if the respondents were asked what they themselves would do in the same situation.

As has been reported by others [16], persons with common mental disorders frequently used self-help interventions. Many of these interventions were cheap, readily accessible and easy to implement. Unfortunately, professionals believe that some of these self-help interventions are unlikely to be helpful [17] and one of the most commonly used interventions (pain relievers) was actually associated with less improvement in symptoms in the present study. However, it is reassuring that a few of the most commonly used interventions (physical activity and seeking support from family and friends) are backed by evidence as effective [18,19]. Indeed, physical activity was associated with greater symptom improvement in the present study. There is a need to ascertain what self-help interventions are most effective and to educate the public to use them. It is surprising that we have so little evidence on the effects of most commonly used interventions.

The major limitation of this study is the low baseline response rate. We cannot say the sample is representative. However, it does include individuals who were not receiving formal treatment, making it superior to a clinical sample. We decided not to carry out a household survey (with a likely higher response rate) because of greater cost and limited advantages in having a more representative sample to test our hypotheses. We were interested in whether certain associations exist, rather than in estimating parameters for a specific population. Another limitation is that use of treatments and other actions taken were assessed through self-report. We do not know how valid these are. One respect in which the sample differed from our representative national sample [2] is that 59% mentioned depression as what was wrong with the person in the vignette, compared with 39% previously. There are various reasons for this higher rate of recognition, including a higher response rate in persons with better mental health literacy, the opportunity to check other sources of information in a self-completed questionnaire, and the structure of the questionnaire which was preceded by questions about symptoms and history of depression. Another limitation is that the short time-frame did not allow examination of interventions which occur with low frequency, such as seeing psychiatrists, psychologists or counsellors.

Beliefs about the helpfulness of an intervention did not always predict use of that intervention, although beliefs did predict use of antidepressants and certain other treatments. Therefore, campaigns that change beliefs about effective treatments may also influence this use of treatments.