Abstract
A Bristol study [1] established that screening and detection of psychological morbidity by general practitioners is distinctly influenced by patient's attribution style. In that study, consecutive patients were asked to interpret 13 symptoms, each with a ‘psychologising’, ‘somatising’ or ‘normalising’ optional response. Patients also completed the 12-item General Health Questionnaire (GHQ), with the general practitioner later judging whether the patient had an anxiety or depressive disorder. Using a GHQ cut-off score of 3 or more, the general practitioners demonstrated modest accuracy in detecting ‘cases’. However, scores on the attributional measure had a direct impact on detection rates, with ‘cases’ less likely to be detected when the patient had a high normalizing score and more likely when high psychologising scores, with effects persisting after controlling for age, sex and GHQ score. However, the degree of somatizing had no measurable effect on diagnostic rates.
As a consequence, we extend the Bristol research domain by examining the impact of attributional style on a number of parameters of depression (including lifetime depression and treatment recourse). Further, as attributional style may merely be a synonym for personality, we examine the impact of personality styles commonly observed in patients presenting with clinical depression to determine interdependence or independence of the personality and attributional constructs.
Method
Participants
Six general practices in Sydney agreed to data collection, with receptionists and research assistants distributing questionnaires, subject to the patient not being severely physically unwell, and being at least 16 years and competent in English. Patients were informed that the study was anonymous and that completed questionnaires should be ‘posted’ into a box near the receptionist.
Measures
An 18-item questionnaire assessed state depression in the medically ill [2] and a 70-item personality questionnaire [3] generated scores on six constructs (i.e. ‘anxious worrying’, ‘introversion’, ‘irritable’, ‘selfcentred’, ‘obsessional’ and ‘self-blame’) that we have argued as the most commonly observed personality styles in patients presenting with clinical depression. Sociodemographic details were sought.
Lifetime depression was assessed by asking subjects whether they had ‘been depressed (i.e. felt significantly depressed, hopeless and pessimistic about things, had a drop in self-esteem or self-worth, and not been able to cope as well as usual) for a period of at least two weeks?’ If affirmed, they were then asked to nominate their age at first episode and whether such episodes were best viewed as ‘normal blues’ or, whether ‘at times’ or ‘always’, a distinct disorder. Details were sought on duration of the longest episode, and if any episode (whether viewed as ‘normal blues’ or ‘distinct disorder’) had prevented them from working, required antidepressant medication or resulted in professional consultations.
As a measure of attribution style, we used an abbreviated version of the measure developed by Robbins & Kirmayer [4], with sample members provided only with three of the original 13 symptoms. They were asked, when experiencing each symptom (i.e. fatigue, insomnia, and appetite loss), for their commonest explanation, being invited to choose one of three (randomly ordered) options. Those who chose ‘I'm emotionally exhausted or discouraged’ (for fatigue), ‘I've been worrying too much or must be stressed about something’ (for insomnia) or ‘I am emotionally stressed’ (for appetite loss) scored one point (each) on our ‘psychological’ interpretation scale. Those choosing ‘There is a medical cause, such as anaemia’ (for fatigue), ‘There is likely to be a physical reason’ (for insomnia) or ‘I have a physical problem, such as a stomach ulcer’ (for appetite loss) scored one point (each) on our ‘somatic’ interpretation scale. Those choosing ‘I've been over-exerting myself or not exercising enough’ (for fatigue), ‘I'm just not tired’ (for insomnia) or ‘My body doesn't need as much food (or energy) at the moment’ (for appetite loss) scored one point (each) on our ‘normalising’ scale.
Results
Full completion of the questionnaire was commonly prevented by the patient being called through for assessment by the general practitioner. Thus, while more than 900 questionnaires were returned, only 638 were fully completed – and form the current sample. The mean age of sample members was 38.3 (SD = 16.5) years, with 55% female.
Mean attributional scale scores were: psychological = 1.4 (SD = 1.1), normalizing = 1.1 (SD = 0.9) and somatic = 0.36 (SD = 0.6). Psychological and normalizing scores were strongly (and inversely) associated (r = −0.73, p < 0.001). Somatic scores correlated negatively (p < 0.001) with both psychological scores (r = −0.40) and normalizing scores (r = −0.20).
Associations between depression variables and scores on each attributional measure were examined by use of t-tests. Table 1 data show a consistent phenomenon, whereby those rating as currently depressed, those reporting lifetime depression and it being more likely to be a disorder, those having depression prevent work and those receiving antidepressant medication returned differentially higher scores on the psychological scale and lower scores on the normalizing scale. Consulting a general practitioner or psychiatrist was associated with significantly lower normalizing scores. Somatic scores were of minimal influence, being significantly higher only for those reporting lifetime depression as always at the disorder level and in those consulting a general practitioner for depression.
Mean attributional scores compared against study variables assessing depression and stress
Additional correlational analyses found no association between scale scores and age at first episode of any depression (r's of −0.05 to + 0.11). There was a weak association between higher normalizing scores and briefer episodes of depression (r = −0.18). Those scoring higher on the psychological scale returned higher state depression scores (r = 0.40, p < 0.001), while the converse held for normalizing scores (r = −0.36), as against no association (r = −0.06) for somatic scores.
We examined the influence of increasing normalizing scale scores, by use of χ 2 tests for categorical data and analysis of variance tests for dimensional data. Table 2 results show that as normalizing scores increase there was an increase in male representation, very sharp drops in state and lifetime depression rates, a decrease in judging any lifetime depression as a disorder and a decreased chance of seeking help from a professional or of receiving an antidepressant. Similar associations (in the converse direction) were demonstrated between psychological scores and study variables.
Associations between increasing normalizing scores and study variables
Attributional scores were intercorrelated with scores on the six personality scales. Higher psychological scores were associated (all p < 0.001) with higher anxious worrying (r = 0.41), irritability (r = 0.19) and self-blame (r = 0.17) scores. Conversely, higher normalizing scores were associated with lower anxious worrying (r = −0.37), self-blame (r = −0.16) and irritability (r = −0.15) scores. Such associations argued for determining whether personality variables dictated attribution scale influences on depression variables (i.e. were ‘attribution’ and ‘personality’ synonymous?).
Multiple regression analyses were undertaken to determine the independent contribution of age, sex and personality scores on each attributional scale. No predictor of somatic scores was identified, establishing that our personality scales were not linked with somatizing responses. Four variables were significant independent predictors of higher psychological scores (F = 37.2, p < 0.001): higher anxious worrying (Beta = 0.37, p < 0.001) and higher avoidant (Beta = 0.09, p < 0.05) personality scores, younger age (Beta = −0.14, p < 0.001) and being female (Beta = 0.13, p < 0.01). The same four variables were the only significant predictors of higher normalizing scores (F = 27.7, p < 0.001), albeit with the converse associational directions.
In refining our 70-item personality measure [3] we had imposed a 6-factor solution. We now repeated the factor analysis but added the three attribution scale scores to the personality items, again to determine interdependence or independence of personality and attributional scales. The somatic scale failed to load on any of the six factors, suggesting its independence from the personality dimensions. However, both the psychological and normalizing scales loaded highly (+ 0.66 and −0.60) on the first ‘anxious worrying’ factor, suggesting that the latter was a bipolar factor with ‘normalising’ at one pole and ‘psychologising’ at the other pole.
To further examine whether personality scores subsumed attribution scores or the converse, we undertook a series of logistic regression analyses. All three attribution scale scores, all six personality scores and both age and sex were entered to examine their independent prediction of depression variables, with analyses using the Wald statistic and with confidence intervals imposed in considering significance. Table 3 summarizes results, in listing the significant predictors identified in the seven separate analyses. The only consistent predictor was anxious worrying, the strongest independent predictor in six of the seven analyses. Higher self-blame and lower obsessional and selfcentred scores were the only additional personality scales to make a contribution. Higher psychological and lower normalizing attributional style scores were identified as independent predictors of state and lifetime depression and of receiving antidepressant medication, somatizing scores contributed to the prediction of attending a psychiatrist or a general practitioner for depression, while age, sex and other personality dimensions made additional contributions in several analyses.
Independent contribution of personality, attribution scale, age and sex on depression variables identified in logistic regression variables
Discussion
Our analyses indicated that respondents' choice of attributional responses to a set of somatic symptoms were powerful predictors of sample members' levels of reported state depression, lifetime depression variables, receipt of antidepressant medication and recourse to professional help.
In the Bristol study [1], the somatizing score did not influence the general practitioners' capacity to detect psychological caseness, which was interpreted as indicative of its irrelevance or due to a lack of statistical power. Here, somatizing scores made an independent contribution to the likelihood of subjects consulting both general practitioners and psychiatrists and to judging lifetime depression as being at the disorder level. Thus, such responses related more to seeking treatment than to depression prevalence. Normalizing and psychological scores had converse associations with depression variables, so that study results could be interpreted in respect to either one or other of those two converse styles.
Our Table 2 data are illuminating in considering the potential impact of attributional style on propensity to depression, its detection and thus, caseness estimates. Those with ‘extreme normalising’ (compared to those with ‘low normalising’) characteristics had one-twentieth of the chance of scoring as being currently depressed, one-quarter the likelihood of reporting lifetime depression and of having any such depression prevent them from working, and, in addition, not one viewed their depression as a disorder, and not one had received an antidepressant or had consulted a psychiatrist for depression.
To what extent are our findings suggestive of a ‘real’ attributional effect or, alternatively, more likely to be ‘artifactual’? The former explanation supposes that there are individuals who truly normalize bodily cues and are perhaps indifferent to stressful events themselves. Those individuals may then be more indifferent to bodily cues, reflecting intrinsic resilience (being phlegmatic, cool and unconcerned). Such processes would lead to a higher tolerance both for distressing circumstances and for recognizing personal distress, greater coping capacities, a decreased likelihood to develop depression or, if it developed, to be less aware of its impact and any disabling components.
Artifactual explanations would argue that such individuals are just as likely to be aware of psychological dysfunction and to develop depression, but that they tend to deny its presence. Intermediate explanations allow that normalizing individuals may be aware of psychological problems but that they possess a degree of protective fatalism, quiescently viewing bodily cues as part and parcel of life and therefore not worthy of pathologizing to disorder status, while less normalizing individuals have a lower threshold to such status and a more ‘emotion focused’ coping style.
Such attributional style influences are of key importance due to their capacity to impact on case definition and detection. Those who were GHQ positive ‘cases’ in the Bristol study [1] were more likely to be so diagnosed by their general practitioner if they showed a psychologising style compared with those who showed a normalizing (i.e. 62% vs 15%) style. In an accompanying commentary, Heath [5] suggested that such results provide us ‘with a superlative example of the folly of medicalization’. She was critical of the view that those who normalize (or are normalizing) are truly depressed and escape diagnosis by the doctor colluding with their minimizing approach. However, the Bristol study authors suggested that those with normalizing attributions and ‘commonsensical’ overtones risk influencing ‘the doctor to join with the patient in minimizing and even dismissing the symptoms’. Heath argued an alternate view – that such individuals ‘may have already begun a process of finding meaning, making sense and learning to cope’, and queried whether there were any data to suggest that diagnosing and treating such patients as if ‘depressed’ would improve their outcome. Which view is most likely to be correct? This, and related issues, are substantive and require resolution.
Our study further advanced beyond the Bristol one by examining the possible impact of depressive personality style on attributional scores. In another report [3] we noted that those who present with a non-melancholic depression are highly likely to have anxious personality styles (expressed most commonly as internalized ‘anxious worrying’ or, alternately, as an externalizing ‘irritable’ style). Therefore, we established that the anxious worrying style was consistently and distinctly associated with all depression study variables, while less consistent and weaker associations were observed between depression variables and the irritable, avoidant and self-blaming styles. In the current study component, we undertook several analyses considering whether personality ‘style’ might determine attributional scores and therefore subsume their independent contribution. While we have examined both attributional style and personality contribution in a relatively limited way, our study has the advantage of building on relevant earlier studies.
In the logistic regression analyses (examining all personality and attributional scores, as well as age and sex), a higher anxious worrying score was consistently identified in association with every depression variable and made the strongest contribution to the prediction apart from lifetime depression (where lower normalizing scores dominated in this analysis). However, in all analyses apart from examining predictors of ‘work impairment’, attributional scores made an additional significant contribution, favouring some independence of constructs. We judge then that anxious worrying contributes to offering psychological attributions and against offering normalizing attributions, but that such attributional responses are not simply or entirely driven by such a personality or temperament style. Multiple other factors must be assumed, including sociocultural and family modelling influences.
Findings suggest an extraordinary influence of attributional style and an anxious worrying personality style on depression reporting, and argue for more fine-focused studies, with personality and attributional style examined in greater detail. Even if our attributional cues operated largely to detect anxious worrying, this should not reject the potential utility of an attributional-based approach to depression risk and detection. Our measure was simple and brief, and established quite striking associations. The impact of a general attributional style in the Bristol study [1] was distinct. Findings in the current study also raise substantive questions about valid definition of ‘caseness’, and the recognition and detection of depression, whether in general practice or community settings, but also have relevance to an equally important topic – the extent to which human distress risks being ‘medicalized’ as a disorder more on the basis of attributional style and reporting rather than by the actual severity of the disorder.
Findings allow that a normalizing style may protect against depression onset and persistence and/or influence reporting and thus risk invalidating self-report measures as a valid strategy for assessing depression. Such findings thus impact on many current strategic approaches to measuring depression in community and general practice settings.
Footnotes
Acknowledgements
We thank the many contributing general practitioners, Dusan Hadzi-Pavlovic, Kerrie Eyers, Christine Boyd, Therese Hilton and other Mood Disorders Unit staff.
Supported by NH & MRC Grant (222708) and an Institute Infrastructure grant from the NSW Centre of Mental Health.