Survey method

A computer-generated random sample of individuals aged 18 years or over was selected from the database of Oxfordshire general practice patients. Ethical approval was obtained from the Oxford Psychiatric Research Ethics Committee and the host institution ethics committee, and patients' general practitioners were notified. A self-completion postal questionnaire was designed based on the results of a literature review and a qualitative interview study of 36 individuals concerning mental health information needs and use of the internet. Several questions investigated internet use, both in general and with respect to mental health issues. To investigate the relative frequency of use of the internet and its trustworthiness, respondents identified the three sources of mental health information they would be most likely to use, and the three sources that they regarded as the most accurate. This determined how many people listed the internet as one of their top three choices. The item responses for these lists were based on the interview results and literature review, and were refined during pre-testing. We avoided item response bias by reversing the item order in 50% of the surveys.

The survey also included demographic questions, a question on previous psychiatric history and the 12-item General Health Questionnaire (GHQ-12; Reference GoldbergGoldberg, 1972). This is a validated self-completion instrument to assess current mental health status, chosen for its brevity and validity. The layout and appearance of the questionnaire were informed by best practice in questionnaire design (Reference McColl, Jacoby and ThomasMcColl et al, 2001). It was brief, to encourage a high response rate (Reference Edwards, Roberts and ClarkeEdwards et al, 2002). Survey pretesting and piloting were undertaken with purposive samples of the general population.

To detect a 25% difference in use of the internet for health information between those who did and those who did not have current experience of mental health problems, with 80% power and 5% significance, the survey sample required 1800 individuals (assuming that there is a 50% response rate, 45% use the internet, 60% of these have used it for health information and 25% have some current experience of mental health problems). It was therefore mailed to 1800 potential respondents. Two duplicate mailings and one postcard reminder were sent following the initial mailing, and respondents could opt to be entered in a prize draw.

Analysis

Data were double-entered. Responses were confidential and data for analysis were anonymised. The Statistical Package for the Social Sciences version 13.0 for Windows and StatsDirect version 2.4.5 (StatsDirect Ltd, Sale, Cheshire, UK; http:.//www.statsdirect.com) were used for data analysis. To determine health status the GHQ-12 scores of respondents were calculated using the standard GHQ scoring method (Reference Goldberg and WilliamsGoldberg & Williams, 1988); a cut-off score of 2 or more indicated the presence of psychological distress (Reference GoldbergGoldberg, 1972). Univariate significance testing was carried out using chi-squared difference in proportions. Multivariate logistic regression explored the relationship between internet use, current mental health status (GHQ-12 score), past psychiatric history and socio-demographic variables. Except where stated, all P values refer to χ² comparisons of proportions.

Response rates

A total of 917 replies were received. After exclusion of patients who had died (n=13) and surveys returned unopened as ‘not known at this address’ (n=213), the adjusted response rate was 58.3% (917/1574). Respondents were significantly more likely to be female (P<0.001), older (P<0.001; unpaired t-test), and to come from less deprived areas (P<0.001; Mann-Whitney U-test) than non-respondents. The predominance of female respondents was more marked in younger age-groups.

Sample characteristics

The median age-group was 46-55 years for both men and women. There were 46.0% men and 54.0% women in the respondent sample. The mean GHQ-12 score was 1.8. Overall, 34.0% of respondents (n=312) had some evidence of current mental health disturbance (GHQ-12 scores of 2 or more) and 20.1% of respondents (n=184) had mental health disturbance rated as high or severe (GHQ-12 scores of 4 or greater). Of the sample, 18.2% had a self-reported history of significant mental health problems (166 out of 910 who answered this question), defined as a mental health issue or problem that had led to a consultation with a doctor or other health professional. There was a gender difference, with 15.7% of men (65/415) and 20.9% of women (101/484) reporting such history (P=0.039).

Internet use

Of the total sample, 58.8% reported ever having used the internet (539/917). There was no difference by gender, with 59.9% men (249/416) and 58.2% women (284/488) having used the internet (P=0.59). There was a large and significant difference by age, with 84.5% of respondents aged 45 years and under (299/354) reporting ever having used the internet v. 42.9% (240/560) of those aged 46 years and over (P<0.001). There was also a large and significant difference by level of educational attainment, with 37.9% (153/404) of those who did not have qualifications at A-level standard or above having used the internet, compared with 85.0% (335/394) of those who had this qualification or above (P<0.001). Of the whole sample, 37.4% (343/917) had used the internet for general health information ('finding out about any aspect of health or healthcare'). This represented 63.6% of those who had ever used the internet. There was no relationship between general internet use or health-related internet use and either current mental health status (GHQ-12 score) or having a previous episode of mental illness, once the effects of age and educational attainment were controlled for using logistic regression. For all regression analyses the effects of GHQ-12 caseness and of past psychiatric history were examined separately, owing to the lack of independence of these variables.

Use of the internet to find out about a mental health issue

Of the whole sample, 10.6% (97/917) had used the internet to find out about mental health, representing 18.0% (97/539) of all people who had ever used the internet. The equivalent figures were 15.1% for respondents with GHQ-12 scores of 2 or more (22.8% of those who had internet access) and 20.5% for respondents with a past history of mental health problems (31.5% of those who had internet access). Differences by age and educational level disappeared after allowing for differential access to the internet. Differences by past psychiatric history and GHQ-12 status remained statistically significant after allowing for internet access once the effects of age, gender and educational level were controlled for (Table 1). This analysis showed that internet users with current experience of mental health distress were more likely to have used the internet to find information about a mental health issue than those without current mental health distress (OR=1.82, 95% CI 1.14-2.89, P=0.012) and those with a past history of a serious mental health problem were more likely to have used the internet for this purpose than those with no such history (OR=2.83, 95% CI 1.71-4.68, P<0.001).

Table 1

Use of the internet for information on health issues and on mental health issues: logistic regression investigating explanatory variables among internet users (n=539)

	Use of internet for health information			Use of internet for mental health information
	Odds ratio (95% CI)	P	Odds ratio (95% CI)	P
Analysis with variables GHQ–12 score, age, gender and educational level
Age 45 years or under	1.11 (0.75–1.65)	0.59	0.94 (0.58–1.53)	0.81
Male	0.82 (0.56–1.20)	0.31	0.77 (0.48–1.24)	0.28
Educated to A-level or above	1.57 (1.05–2.34)	0.03	1.11 (0.67–1.85)	0.68
GHQ–12 score ≥ 2	1.03 (0.70–1.52)	0.87	1.82 (1.14–2.89)	0.01
Analysis with variables past psychiatric history, age, gender and educational level
Aged 45 years or under	1.14 (0.78–1.68)	0.52	1.05 (0.64–1.71)	0.86
Male	0.84 (0.57–1.24)	0.38	0.83 (0.51–1.36)	0.46
Educated to A-level or above	1.57 (1.05–2.34)	0.03	1.17 (0.70–1.94)	0.56
Previous consultation for mental health problems	1.32 (0.82–2.12)	0.25	2.83 (1.71–4.68)	<0.001

GHQ–12, 12-item General Health Questionnaire

Nine of the respondents had used the internet to chat (using live interaction in a chatroom or via instant messenger) with someone else about mental health issues (1.0% of the whole sample, 1.7% of internet users). These were four men and five women with a wide spread of age-group (18-75 years) and educational attainment (O-levels to degree). All nine had current mental health distress (GHQ-12 score of 2 or above) and five had experienced a serious mental health problem in the past.

Relative importance of the internet as a source of mental health information

Respondents were asked to select three sources they believed provided the most accurate information on mental health issues. Table 2 shows that 12.1% selected the internet as one of the top three most accurate sources of information on mental health issues. For the whole population and for those with and without current mental health problems, the results clearly demonstrate the trust put in health professionals, with mental health workers and general practitioners being rated the most accurate sources. There was no major difference by GHQ-12 caseness (χ²=11.74, d.f.=12, P=0.47). Third place was taken by leaflets produced by the National Health Service (NHS) or by voluntary organisations and charities. Fourth ranking was ‘someone else with the same mental health problems’, and this received relatively more votes from people with mental health problems. The internet was ranked eighth overall, and sixth by people with mental health problems, although the scoring between the fifth- and eighth-ranked sources was very close. There was a gender difference (χ²=31.76, d.f.=12, P<.01), explained by mental health professionals being ranked ahead of general practitioners by women, whereas men ranked general practitioners slightly ahead of mental health professionals.

Table 2

Proportion of respondents identifying sources of information providing the most accurate information on mental health issues

Rank	Source of information	All respondents (n=917) %	Respondents scoring 2 or more on GHQ–12 (n=312) %
1	Mental health professional	59.7	60.6
2	General practitioner	53.8	52.9
3	Leaflets from NHS or voluntary organisations or charities	29.8	29.5
4	Someone else with the same mental health problem	15.3	17.9
5	Television or radio programmes	15.0	15.1
6	Friend or family member	13.4	12.5
7	Newspaper or magazine articles	13.1	12.8
8	Internet	12.1	13.1
9	Charity or voluntary organisation telephone helpline	7.6	9.3
10	Home medical encyclopaedia or similar books	7.1	8.0
11	NHS Direct telephone helpline	6.8	6.1
12	Other	4.1	2.6
13	Alternative or complementary therapist	3.5	5.1

GHQ–12, 12=item General Health Questionnaire; NHS, National Health Service

Respondents were also asked to indicate which three sources of information they would be most likely to use if they were seeking information on a personal mental health issue. Table 3 shows that 24% of both the general population and those with current mental health distress indicated that the internet was one of the three sources they would use, suggesting the internet is used more than it is trusted as an accurate medium. Table 3 also shows that general practitioners and mental health workers not only were considered the most accurate sources, but also were the most likely to be used. The internet was rated fourth overall and third equal by those with mental health problems: there was no major difference by GHQ-12 caseness (χ²=9.72, d.f.=13, P=0.72).

Table 3

Proportion of respondents identifying sources of information on mental health issues that they would be most likely to use

Rank	Source of information	All respondents (n=917) %	Respondents scoring 2 or more on GHQ–12 (n=312) %
1	General practitioner	74.2	73.1
2	Mental health professional	57.1	57.1
3	Leaflets from NHS or voluntary organisations or charities	27.7	24.0
4	Internet	23.6	24.0
5	Friend or family member	18.0	18.3
6	Someone else with the same mental health problem	13.0	13.8
7	Home medical encyclopaedia or similar books	11.7	9.9
8	Charity or voluntary organisation telephone helpline	8.8	9.3
9	NHS Direct telephone helpline	7.0	8.0
10	Newspaper or magazine articles	6.3	8.0
11	Alternative or complementary therapist	5.5	6.4
12	Television or radio programmes	3.5	3.5
13	Other	1.3	0.6

GHQ–12, 12-item General Health Questionnaire; NHS, National Health Service

Limitations of the study

Cross-sectional sampling can only identify the views and reported behaviour of respondents at one point in time. Further work observing actual behaviour or following individuals prospectively might be helpful. The response rate was moderate despite the use of duplicate mailings, postcard reminders and entry into a prize draw. Non-respondents to the GHQ-12 are known to have a higher prevalence of psychiatric morbidity than respondents (Reference Williams and MacdonaldWilliams & Macdonald, 1986). Because of the nature of the NHS, general practice registers in the UK are generally considered to provide adequate population samples; however, they are known to suffer from ‘list inflation’, whereby people who have been registered with a general practice can remain registered after dying or moving away (Reference Carr-Hill and RobertsCarr-Hill & Roberts, 1999). On average non-responders were 5.5 years younger, from areas with slightly higher deprivation scores, and were more likely to be male. These are known associations with population survey non-respondents (Reference Purdon and NicolaasPurdon & Nicolaas, 2003). However, it is difficult to judge what effects non-response bias might have had on the findings: for example, internet use is more common in younger age-groups (over-represented among non-respondents) but also in less deprived groups (under-represented among non-respondents). For generalisability, it is reassuring that the prevalence of general internet use is in line with the findings of other UK population surveys (Reference Dutton, Di Gennaro and Millwood HargraveDutton et al, 2005).

Implications

Most people with minor mental health problems seek help from family and friends rather than professionals (Reference Oliver, Pearson and CoeOliver et al, 2005). Practitioners and policy makers must also take note of the role of the internet in help-seeking behaviour. Almost a third of internet users with a history of psychiatric disorder had used the internet to seek mental health information. A few respondents had used the internet for live chat with others about mental health issues and this is an area likely to see future expansion. As an information source the internet has advantages of privacy, anonymity and widespread accessibility at low or no cost (Reference Cline and HaynesCline & Haynes, 2001). However, it appears that the public also recognises the frequently expressed concern of professionals regarding the accuracy of online information (Reference Christensen and GriffithsChristensen & Griffiths, 2000). The internet is mostly unregulated, but there are voluntary initiatives to encourage quality assessment of health information sites (Reference Risk and DzenowagisRisk & Dzenowagis, 2001; Reference Griffiths and ChristensenGriffiths & Christensen, 2005). However, poor-quality information has always existed in various forms, and there have been only a few isolated case reports of individuals coming to harm from online information (Reference Crocco, Villasis-Keever and JadadCrocco et al, 2002). The need is for a better understanding of how individuals actually use the internet, what they do with the information they find and how internet help-seeking relates to other help-seeking behaviour. By examining the role of the internet in meeting information needs, psychiatric services and practitioners could harness the internet as a tool to educate and support patients. This is particularly important in mental health, where the internet may have a role in supporting those for whom stigma inhibits help-seeking through more traditional routes (Reference Berger, Wagner and BakerBerger et al, 2005).