Our results with a yearly prevalence of 41% (95% confidence interval (CI) 37–46%) is in the upper range compared with studies worldwide . Our general paediatric subgroup used 37% (95%CI 28–47%), which is higher than the general paediatric outpatient UK study in 1996 . This may be a true increase in CAM use over that time, but confounders include; differences in the two outpatient populations, the definition of CAM used or the fact that our study was interview led, rather than a self-administered questionnaire, which may underestimate use. There was no statistical difference in the prevalence amongst the inpatient and outpatient groups with overlap of the 95% confidence intervals (Table 1). All of the four outpatients included patients with chronic illnesses and the high number of cystic fibrosis patients in the respiratory outpatients may partially explain their slightly higher prevalence. The prevalence in Cardiff is significantly lower the prevalence found in the Melbourne study (51%, 95% CI 47–56%) . The reasons for this difference have been explored in a separate comparative analysis between the two tertiary centres .
Socio-demographic factors associated with CAM use were parental tertiary education and high income. This supports the view that CAM is associated with higher socio-economic status [12, 14]. Higher levels of education may lead to exposure to these therapies, especially given their high profile in the media. The lack of other socio-demographic factors implicated in CAM use from our study supports many of the previous surveys [11, 13, 27].
Nearly two-thirds of medicinal CAM cost less than £5 per month, with about one-third spent nothing per month. This supports the assertion that CAM may not be as expensive as is commonly believed, as concluded by a large Canadian study which found that CAM was less than CAN$20 in 37% of cases and more than CAN$60 in only 6% . A BBC commissioned survey in 1999 of CAM use amongst adults in the United Kingdom approximated CAM expenditure of £14 per month with about one-third spending less than £5 per month . Non-medicinal CAM tended to be more costly as seen in this study, with about one-third spending more than £20 per month and 13% spending more than £40. This may limit their availability and therefore prevalence, which was significantly less than medicinal forms of CAM (12%. versus 41%). Of note some non-medicinal therapies were 'free' because treatments such as therapeutic massage or reiki were provided at no cost by friends or relatives.
Disclosure of CAM use to medical practitioners is important if the issue is to be discussed and any potential interactions or adverse effects detected. In our study 66% did not disclose CAM usage to their doctor, which is similar to previous studies [15, 18, 29]. Reasons for lack of disclosure may reflect a perceived lack of acceptance by the physician. It is important part of doctor-patient communication that issues such as CAM is discussed in a nonjudgmental and open manner. The "don't ask, don't tell" status quo,  is also reflected in the fact that there was no documentation of recent CAM use in the inpatient notes. Poor communication is supported by other studies, which found few doctors (16%) ask about CAM use .
In a study of paediatricians and nurses attitudes to CAM, many using CAM themselves, but only 40% actually asked about CAM . The main reason given for not discussing these therapies was a lack of confidence to do so because of minimal education about these therapies and the limited safety and efficacy data available.
Efficacy and perceptions
A review of 19 surveys between 1982 to 1995, looking into the referral patterns of physicians to CAM practitioners, highlighted the fact that decisions need to be based on efficacy studies, rather than "regional economics and cultural norms", and "somewhere between over enthusiastic belief and stubborn disbelief is a balanced perspective that will help patients and advance medical science" . The perception of parents and adolescents of doctors as not being interested was commonly found in this study, with quotes including: "they were not interested", "did not have the time" or "wanted to tell the doctor but feel that they don't want to hear".
Integrative medicine refers to "looking at a broad range of therapies and considering those that have the best evidence of safety and effectiveness in the context of holistic care".  It has been strongly supported by The Prince of Wales's Foundation for Integrated Health  and is becoming an important component of medical education. Cohen provides a practical example of integrative care, focusing on the chronically ill, "as many of these patients could benefit from the services of CAM practitioners... and as evidence emerges that some CAM are effective then it becomes ethically impossible for the medical profession to ignore them".  Studies have found that CAM practitioners are keen on integration. 
One of the major concerns for medical practitioners regarding CAM use is safety. Despite popular belief, CAM is not always "natural" and "safe" or free of side effects, and there are many case reports of adverse events in children. These include: death, neurological disability, organ failure and organ puncture. Possible interactions between medicinal CAM, particularly herbal remedies and prescription medications, is well documented . When questioned the majority (87%) of adolescents or parents in this study could not name a potential interaction. It was determined that there was a 3% risk of a participant in the study concurrently taking a herb and prescription medication. This is quite a small number in our hospital-based population and may mean that the interaction potential is less than previously believed. All ingested therapies, however, have the potential to interact so communication and reporting of interactions remains very important. Up-to-date information about CAM is becoming more available with pharmacists, local drug advisory services and some Internet resources useful [39, 40].
Funding is a major issue for producing quality evidence-based research into CAM as resources are still allocated predominantly through traditional biomedical channels and in 1996 only 0.08% of funding in medical sciences was going into CAM . As stated by Ernst; "The scientific establishment criticises the paucity of data and weakness of CAM hypotheses with one breath and with their next breath they withhold the money that would be essential for changing this situation" .
Following the House of Lords Science and Technology Committee report on CAM in 2000, there was a push for some CAM practitioners such as Osteopaths and Chiropractors to follow the medical model of statutory regulation as seen with the General Medical Council . Alternatives to statutory regulation are self regulation, with development of national standards regarding training and competence. Welsh et al. looked in depth at the tensions and difficulties in CAM practitioners seeking self-regulation . The Prince of Wales's Foundation for Integrated Health has also played an important role in this area .
All professions are prone to rogue practitioners, such as those outlined in the study of Traditional Chinese therapists treating atopic eczema with topical remedies that were shown on analysis to contain high levels of potentially hazardous steroids . In these situations it is important that the public sees effective and visible disciplinary procedures.
Limitations of this study
There were a number of potential areas of bias in this study, namely: 1. Questionnaire 2. Interviewer, 3. Interviewee, 4. Sampling, 5. Language bias. The questionnaire itself may have produce bias towards the listed agents as it is not possible to include all CAM therapies and there was a free text area. This was minimised by extensive review of the literature and including 7 medicinal and 13 non-medicinal therapies in the interview led questionnaire. Of note 5 of the 20 therapies accounted for the majority of CAM use, with some CAM only being used by less than 2% of the study population.
Despite calibrating the interviewers with a view to obtaining consistent responses, it remains possible that inconsistencies in asking questions and recording answers occurred The interviewers [NC, DC, AH, DR] had variable amounts of CAM disclosed (Table 4) and in particular CAM use was much higher (63%) in the surveys conducted by DW. He was a pharmacy student whilst the other three interviewers were paediatric trainees. All interviewers were similarly attired, with no identifying features. DW interviewed only a small proportion of participants (13%).and none of the inpatient subgroup. DW may have had higher CAM disclosure because as the only pharmacist amongst the interviewers he may have been better at gaining information about medication use and CAM. Alternatively there could have been some potential bias in the parents/patients perceptions of the different interviewers, thus affecting disclosure. In particular, the role of the three doctors as interviewers may have led participants to feel that they would not approve of CAM use. This would produce an under-estimate of CAM use in our population. The fact that the interviews were conducted in a busy outpatient clinic and on the inpatient wards may have inhibited disclosure of CAM use in some cases.
Different responses may also be obtained when interviewing the patient or either parent about CAM. The knowledge and opinions regarding CAM may differ within the family. The mother (75%) was the person most commonly interviewed in the survey. The prevalence may be under-estimated in cases where a guardian with less knowledge about the child completed the questionnaire. This study asked about CAM use in the past year, which is prone to recall bias, as participants may find it difficult to remember exactly what therapies were used. Repeating the questionnaire over the phone showed that it had stability with only a 3.3% difference in responses. There was, however, no differentiation made between positive or negative differences and individuals may have started or stopped taking CAM in the two months between the two surveys. It is also possible that respondents were more educated about CAM after becoming involved in the study, affecting the responses in the repeat questionnaire.
A representative sample is difficult to obtain, but by approaching consecutive inpatient admissions and outpatient presentations, the potential for sampling bias was minimised. Only 80 cases were excluded and 10 of those were because they had been interviewed previously. Information was not obtained to directly compare the demographics or other details of the 80 excluded cases from the 500 respondents.
This study excluded a small number of subjects (5 in total) who had insufficient knowledge of English, leading to language bias. Whilst this is a small number overall, it may be expected from the literature and on general principles that this group is more likely to use CAM. Cost was a factor as this study was not funded and there were insufficient resources to provide questionnaires in different languages or provide independent interpreters. This study was also designed with the aim of doing a direct comparison of CAM use between Cardiff and Melbourne, so their methodology was replicated, including exclusion of those with insufficient English language skills. Whilst this strengthens the comparability of data it highlights the need for further research into specific ethnic groups and CAM subtypes utilised by them.
Applicability of the study
As outlined previously methodologies used in CAM surveys vary, reducing the comparability of studies and hence the applicability of our study elsewhere. The study population was a tertiary paediatric group based at one location, comprising children with acute and chronic illnesses. The results whilst not applicable to the general population, may apply to other tertiary paediatric centres with similar demographics and case-mix. Well-conducted population based studies are require to ascertain the overall use of these therapies in children.