Addressing e-health inequalities
Various barriers including lack of physical access, experience, attitudes, confidence or self-efficacy, knowledge, and social help may contribute to e-health inequalities [7, 29, 30]. Tackling e-health inequalities might address physical access through the provision of broadband (for example, the roll out of superfast broadband to rural Cornwall http://www.superfastcornwall.org/), or the provision and recommendation of e-health services, for example, in England there is marked variation of opportunities between different GPs  and different levels of referral by geographical area . Economic factors, particularly at a time of recession are also important but are mainly solved at a regional, national or global level. The focus in medical informatics literature over recent years has been on e-health literacy . But alongside e-health literacy we need to consider the support that is available from family, friends, or perhaps agencies such as Age UK, or in this case through University-run projects.
We had previously piloted email support for people recruited in outpatients, to use the Internet for health . The intervention seemed to offer some potential, but key issues included when in the 'patient's journey' to recruit, and how to recruit. Participants in our previous study had suggested that support in using the Internet would have been more useful earlier. In addition, we thought that many e-health opportunities for health promotion were more relevant to primary care. We therefore wanted to explore the feasibility and cost of recruiting those people who had access to the Internet, who saw the possibility of using it for a long term condition or lifestyle change, but who could benefit from email support.
Raising awareness in primary and secondary care for online recruitment
Online recruitment requires that potential participants are made aware of the study and can access the recruitment website. In our previous hospital study , participants suggested that having help to use the Internet would have been useful earlier in their condition. In this primary care study, we also recruited more people thinking of lifestyle changes or with mental health problems than in our hospital study. However, overall we recruited very few patients. The cost of recruitment was high, partly because considerable time was spent recruiting practices and agreeing what methods could be used. Although we recruited 18 practices, five 'enthusiastic' practices were responsible for 23/27 patients. These practices were easier and less costly to recruit than other practices.
In our previous hospital study, we recruited 29 participants over five weeks using leaflets in outpatient waiting rooms. We approached 864 people of which 29 (3.4%) consented via the study website. In this primary care study, 134 patients were approached by research assistants of which two (1%) consented. However, it may be easier to find and approach patients in outpatient waiting areas where there are often many patients with similar long term conditions waiting, compared to the few patients in different general practice waiting rooms. A detailed diary was not kept for our hospital study but, retrospectively, we estimated that a maximum of 70 hours were spent in recruitment and with other costs represent a total cost of £1500, i.e. about £38 per person recruited compared to, at best, £70-80 in this study.
Recruiting patients to studies in general practice
Recruitment in general practice has often been difficult [11, 12, 20, 33] and this study was no exception. In primary care, researchers have to negotiate access and to recruit 'at arm's length' in dispersed health centres. Practice managers act as gatekeepers to research and their views of particular research studies may be crucial. In this study, some practice manager comments showed that many were not convinced of the benefit of the proposed intervention. Unless practices are really receptive to the purpose of the research it is unlikely to 'work' . On the other hand, we recruited practices and patients in deprived areas, while other managers in more affluent areas were sceptical, suggesting practice manager views may not reflect the potential of patients to benefit. So, just working with more enthusiastic practices would not have resulted in a biased sample of patients.
Although in some countries identification of patients in terms of their eligibility for research is only permitted after explicit consent from patients is obtained, in the United Kingdom, provided the appropriate controls on who sees what data are in place, recruiting patients identified via GP computer systems is possible and has been used before [19, 20]. Some have reported problems in raising awareness 'offline' in general practice for online recruitment. Woodford et al. aimed to recruit patients with depression by identifying patients from GP practice registers, sending an invitation pack via post, and inviting expression of interest on a webpage . Although they only recruited seven people from eleven practices the reasons for failure included the study design (lack of equipoise) and poor coding of depression in GP records. On the other hand Kuyken et al. [22, 34, 35] were more successful using a similar method. They claimed  that getting the GPs to cooperate required them to work only in primary care settings eager to develop and support a research ethos, i.e. those practices who had a 'readiness to engage'. However, that they used 'assertive outreach' (in which a researcher contacted all of those sent the initial letter unless they opted out) may explain why more of the people initially identified agreed to participate. Such personal contact requires that researchers have full access to personal information on all eligible patients, something that not all UK ethical committees would agree to. Our study aimed to recruit anonymous participants.
The main practical and cost issue in deciding whether to recruit in primary or secondary care is the number and nature of professional 'gatekeepers'. In our hospital study we were able to contact patients with permission from one 'authority'.
Practical issues in recruiting patients in general practice
Most participating practices gave us access to use waiting room posters and leaflets as being non-intrusive on the work of the practice. However, our study has shown, that these 'passive' methods, posters and leaflets on their own, are not effective in recruitment. The most successful patient recruitment was in those practices which were easier to recruit for collaboration. There is no obvious reason why patients recruited from 'enthusiastic' practices for an e-health intervention should be different from other patients denied that possibility by practice staff. The 'enthusiastic' practices in this study came from across the city including both affluent and deprived areas. As suggested by White et al. , it would have been more cost effective to work only with those practices prepared to take a more pro-active approach to patient recruitment and it seems likely that this would not result in a biased sample of patients.
In our study the special mailshot was reasonably cost effective, recruiting 6 people (3% compared to less than half a percent recruited by mailshot in Woodford's study ). The integrated mailshot was not particularly effective, probably because of lack of incentive/enthusiasm from the practice, and the prominence of the message.
Although we suggested it to all 18 practices, only seven were able to put a website link to the study. Many of the other 11 saw it as being a difficult or time consuming task. Given that many of the practices were using the same software, or same web developer, this is probably associated with the IT literacy of the practice manager
Strategies for recruiting e-health novices
Recruitment of patients to studies in primary care is more difficult when the study is about communication or information seeking using the Internet, and as in this study, where recruitment of one specific group on the spectrum of 'e-health readiness' is required. Our target population for this type of e-health support was people who have physical access to the Internet and sufficient Internet skills that they could deal with a website registration and email, but were not confident users of the Internet. We do not know how many people are in that category. A quarter of people in Britain have not used the Internet and many of those are not interested in going online . This group was represented by some patients we contacted. But amongst those with Internet connections a proportion may benefit from support. Just over one third (36%) of Internet users say that they look up health information online, rising to 41% of Internet users aged 45-54. Overall levels of confidence among Internet users is high (87%) but drops to 73% for those aged 65 or over . The ultimate aim of this research was to offer email support in using the Internet for health. Raising awareness online to recruit online, shown to be effective in many situations, remains a possibility but may not attract Internet novices. Some form of direct face to face contact, or raising awareness via more traditional media may be more appropriate.
If National Health Service centred methods of raising awareness are to be used then the best combination of methods for this target population might be used. Given the density of people with long term conditions, patients with long term conditions might be recruited in hospital outpatient areas. People with mild to moderate mental health problems, or aiming to change to healthier lifestyles, might be recruited in primary care. However, it would be most cost effective just to use practices that are willing to implement a combination of methods including special mailshot and practice nurse recruitment.
Alternative ways of raising awareness for online recruitment
An alternative to raising awareness to patients via health services for online recruitment is to raise awareness directly to populations, either online or via the mass media. Knowing the cost-effectiveness of all methods is important to be able to decide on the best strategy. Researchers have used various media and methods to raise awareness for online recruitment, for example, Gordon et al. , in a study of a website supporting users of smokeless tobacco, used (a) thematic promotional releases to print and broadcast media, (b) Google ads, (c) placement of links on other Web sites, (d) limited purchase of paid advertising, (e) direct mailings to smokeless tobacco users, and (f) targeted mailings to health care and tobacco control professionals. Our own study of online advertising to recruit people to a website leading to online cognitive behavioural therapy , found costs per person clicking on the advert and following through to the onward link of about £1/person.
These studies show that online advertising can be an effective and inexpensive method of raising awareness of online interventions but the characteristics of those recruited need to be understood. Online recruitment may recruit those who are difficult to reach by traditional means, for example, Graham et al.  recruited a higher percentage of males, young adults, racial/ethnic minorities, those with a high school education or less, and dependent smokers compared to traditional methods. However, these may not be naïve users of the Internet who may benefit from e-health support. Online methods for e-health support may be worth exploring but cost-effective 'offline' methods are also needed.
This study was suggested by the findings of our previous outpatient based study. One possible confounder of our interpretation of our findings is that our previous hospital study was in a more affluent area . It may be that more patients had access to the Internet and were at a stage of Internet use where they were willing to be helped. However, there is no information about the e-health readiness of these two populations to know if this was the case. Furthermore this pilot study used a convenience sample of only one primary care trust, the practices and practice managers may not be typical of other parts of the United Kingdom. Given that the University research team was based in the same area it may be that our results were too optimistic. We estimated costs based on marginal time costs of researchers rather than include estate and other costs (mainly because this was a marginally costed project); this again means that our results were too optimistic. The results cannot easily be applied to other countries as they are fairly dependent on the way that primary care services are organised in the United Kingdom.