SUBJECTS

A total of 1500 subjects was randomly selected from the Electorate of Lindsay list of voters.15 This area, which includes Penrith and surrounding suburbs of Sydney, has a population of 73 393 (3.6% of the Sydney population) and is sociodemographically very similar to the Australian population according to 1991 Census data, except that its inhabitants are slightly younger and it has a slightly higher socioeconomic status.15 ,16 The list of selected subjects was divided into three batches of 500 subjects, and mailings of the bowel symptom questionnaire (BSQ) were conducted at intervals approximately two months apart, for ease of administration.

Subjects were offered a lottery ticket in return for completing the questionnaire. Subjects were given the option of refusing. Reminder letters were sent at two week intervals. At week 6 this included another questionnaire. At week 10, we attempted to telephone subjects who had not yet responded. Finally, the survey was closed at 12 weeks.

Of the 1500 people in the general community to whom we sent the questionnaire, 323 (22%) had moved, and were excluded on that basis. Another 42 (3%) were excluded due to serious illness, death, or language difficulties, leaving us with a valid sample of 1135 subjects receiving the questionnaire. By the close of the survey at the beginning of November 1995, 360 (32%) subjects had explicitly refused to participate. A completed questionnaire was returned by 730 subjects, giving a response rate of 64%. Respondents were similar to the Australian population demographically but were slightly older (69% ⩾40 years compared with 61% in the Australian population).

QUESTIONNAIRE

The BSQ has been validated previously in Australia15and is a modified version of the previously validated and widely used Bowel Disease Questionnaire.17 ,18 It incorporates questions which identify the IBS.19 Visits to doctors and alternative therapists in the year before the survey and lifetime was also measured, as well as age, sex, marital status, and educational level.

We included in the BSQ two validated psychological questionnaires. The General Health Questionnaire (GHQ) was developed for population surveys.14 It was designed to detect non-psychotic psychiatric illness and is a state measure of psychiatric co-morbidity (assessing features in the previous few weeks); it does not provide a specific psychiatric diagnosis. The Eysenck Personality Questionnaire (10 item short neuroticism subscale) measures the personality trait neuroticism which we hypothesised would be important in promoting health care seeking.13

We included the sexual abuse questions validated by Drossman et al 8 ,20 that we have applied in a US study of abuse and IBS, and these were not altered.9 ,12 These questions were as follows:

• Has anyone ever exposed the sex organs of their body to you when you did not want it?

• Has anyone ever threatened to have sex with you when you did not want it?

• Has anyone ever touched the sex organs of your body when you did not want this?

• Has anyone ever forced you to have sex when you did not want this?

• Have you ever had any other unwanted sexual experiences not mentioned above?

 The Drossman physical abuse questions in addition were included, with minor modifications. Changes in grammar and format were made to ensure comprehension by the local population. The referent terms used to identify the perpetrator in the questions relating to childhood abuse were changed to eliminate positive responses to minor events such as fights at school or bullying. Although it is conceded that a child could experience some long term consequences from such incidents if they were repetitive, frequent or severe, these events are usually qualitatively distinct from child abuse as it is generally understood.20 Consistent with this definition, any violent behaviour by older siblings would be considered “abuse”, only if the sibling was in the position of being a care-giver for the child. This decision was adopted because in such a brief, self-report format it is not possible to discover the minority of cases where violence from older siblings or schoolmates occurred in circumstances sufficient to classify as abuse. Similarly, the change of wording for adult physical abuse was made to eliminate positive responses on the basis of one off events (e.g. mugging) or occasional violence from strangers or acquaintances (e.g. crime related or job related), as these events are outside the realm of abuse as conceptualised by our research definition.

 Additionally, on the basis of our previous population based study12 we incorporated four new questions to assess emotional and psychological abuse. These questions have not yet been formally validated. However, they have excellent face validity, and represent a dimension of abuse which has been inadequately studied in relation to gastrointestinal disorders. Response options for the emotional and psychological abuse questions were identical to those used in the Drossman physical abuse questions (i.e. never, seldom, occasionally, or often). To increase the specificity of the survey, subjects endorsing physical or emotional abuse occurring at least often were considered to have suffered abuse.

IRRITABLE BOWEL SYNDROME

The Manning (and for comparison Rome) criteria were applied to identify subjects with IBS. By the Manning criteria,21 ,22as used in other population based studies,23 ,24 IBS was defined as: abdominal pain that kept recurring over a period of more than three months in the previous year plus two or more of the following:

• looser stools at the onset of pain, often (more than 25% of the time);

• more frequent stools at the onset of pain, often;

• pain relief with defaecation, often;

• visible abdominal distension (defined as seeing the stomach or abdomen swell, or having to loosen a belt or clothes because of swelling), often;

• feelings of incomplete rectal evacuation, often;

• mucus per rectum.

IBS was also defined using the Rome criteria19 as more than three months of abdominal pain, in combination with one or more of the following symptoms:

• pain made better by bowel movement, often (more than 25% of the time);

• more and/or less bowel movements with pain, often;

• looser and/or harder stools with pain, often;

and  two or more of the following:

• more than three bowel movements daily, often and/or less than three bowel movements a week, often;

• stools very lumpy or hard, often and/or stools very loose or watery, often;

• incomplete rectal evacuation, often and/or urgency, often and/or straining, often;

• abdominal bloating or distension, often; and/or

• mucus in stools.

BOWEL HABIT SCORE

This was derived from counting the number of positive responses (0–7) to the following questions: >3 stools per day, often; <3 stools per week, often; lumpy or hard stools, often; straining, often; looser or watery stools, often; urgency, often; and mucus per rectum.

DISTENSION SCORE

This was categorised as nil, bloating only (without visible distension) and visible abdominal distension.

STATISTICAL ANALYSIS

Several demographic variables were grouped into coarser categories because of small frequencies in some categories. Education was grouped as tertiary, high school diploma or greater (but not tertiary) and less than high school diploma. Marital status was grouped as currently married/de facto or not. For education and marital status these groups were considered appropriate as they reflected either milestones in education or the existence of a supportive relationship, respectively. Pain severity and time since onset of symptoms were left ungrouped but were treated as ordered discrete categories in the logistic regression analyses.

Four health care seeking variables were used as dependent variables: (1) visits to a doctor for any reason in the past 12 months; (2) visits to an alternate care provider for any reason in the past 12 months; (3) visits to a doctor or alternate care provider for abdominal pain or discomfort in the past 12 months; and (4) visits to a doctor or alternate care provider for abdominal pain or discomfort ever in the time period more than 12 months previously.

All variables were dichotomised. Having visited a doctor in the past 12 months for any reason or having visited a doctor or alternate therapist for abdominal pain or discomfort ever were dichotomised as rarely (never or occasionally) versus more than rarely (sometimes or often). The other care seeking indexes were categorised as never versus ever. The different categories reflect different distributions of response. The original intention was to categorise all four care seeking indexes as never versus ever. Inspection of the distribution of outcome revealed that for (1) and (4) fewer than 25% of subjects fell into the never category (4% for (1) and 22% for (4)). We considered that breaking such small numbers down by independent variables could lead to unreliable findings.

Statistical analysis involved three steps: description of variation in care seeking behaviour with all independent variables, univariate logistic regression of probability of care seeking on single independent variables and multivariate analysis using stepwise logistic regression. The last of these analyses sought evidence of statistically independent correlates of care seeking. The pseudo-R²(deviance explained) is reported as a measure of the explanatory power of the multiple variable models.

All p values calculated were two tailed; the alpha level of significance was set at 0.05.