Colorectal cancer screening using the faecal occult blood test (FOBt): a survey of GP attitudes and practices in the UK

A random sample of GPs (n = 31,358) was selected using the MidReC database (derived from the Prescription Pricing Division Database for England). The sample was stratified by practice size, and, by linking practice postcode to the Index of Multiple Deprivation (IMD) 2004 [25], by deprivation quartile. The IMD 2004 is a weighted area level aggregation of a number of 'domains' of deprivation (income, employment, health inequality, disability, education, skills and training, barriers to housing services, crime and the living environment). Lower IMD scores indicate more deprived areas, whereas higher scores are associated with less deprived locations. A 'small' practice was defined as one having three GPs or fewer (n = 7,987; 25.9%), and 'multiple' as having four GPs or more (n = 22,874; 74.1%). 'Affluent' practices were defined as those with an IMD 2004 rank of 16242 or higher (n = 17,654; 57.2%), and 'deprived' practices as those with an IMD 2004 rank of 16241 or lower (n = 13,207; 42.8%). Those with no IMD rank were excluded (n = 497; 1.6%). Combining these factors, all practices in the sample were stratified according to practice size and IMD quartile, and each was allocated to one of four groups: small/affluent (n = 5,295; 16.9%), small/deprived (n = 2,692; 8.6%), multiple/affluent (n = 12,359; 39.4%), or multiple/deprived (n = 10,515; 33.5%). 800 GPs were randomly selected from each of these four groups.

After excluding nine GPs who had recently been contacted as part of a medical student project on attitudes to CRC screening, self-completion surveys were sent to a total of 3,191 GPs between August and October 2007. This mailing assumed a conservative 40% response rate, to yield a sample size sufficient to determine the overall proportion of respondents with negative attitudes towards, and resistance to, FOBt based screening with 5% precision (95% confidence), based on a worst case scenario of 50% reporting such attitudes. Recipients wishing to participate were able to return the survey directly to the research team using an enclosed Freepost envelope. Non-respondents received one reminder. No incentive was offered for survey completion. Mailing was undertaken in October and November 2007, during the period where the National Bowel Screening Programme was being rolled out.

The survey (Additional file 1) was a modified version of a postal questionnaire designed by the US National Cancer Institute (NCI) colorectal cancer screening team [26]. Prior to mailing, the modified survey was piloted using a sample of GPs within the Department of Primary Care, University of Birmingham to test ease of completion and to ensure comprehensibility. The survey included closed questions with categorical and Likert scale response options. Questions focused on the perceived effectiveness of cancer screening in general, and colorectal cancer screening in particular; factors influencing GP recommendation of CRC screening; perceived patient and system-related barriers to CRC screening, and current practice in relation to screening using FOBt for asymptomatic average-risk patients. Further questions gained information on socio-demographic characteristics of respondents (ethnicity, personal experience of CRC screening, affiliation with medical school) and GP practice attributes (practice setting, number of partners, list size, age profile of registered patients).

Analysis focused on the socio-demographic characteristics and general practice attributes of respondents, and their relationship with stated attitudes towards the recommendation or otherwise of FOBt. A score was calculated based on the responses to three survey questions to represent a proxy measure of GP attitudes to FOBt screening (Table 1). Possible attitude scores ranged from a minimum of three to a maximum of nine. Responses were dichotomised between those with an attitude score of seven and above (representing a positive attitude), and those scoring six or lower (negative attitude). Binary logistic regression was used to calculate bivariate Odds Ratios to evaluate the association of socio-demographic and practice attributes with attitudes towards FOBt. Data were analysed using parametric or non-parametric tests for comparison of means and proportions as appropriate. This included analysis of differences between survey responders and non-responders where data on the relevant characteristics for non-responders (practice size, deprivation quartile, and ethnicity) could be obtained from the MidReC database. All data were analysed using SPSS (version 14.0).

NHS ethical and Trust approvals were secured from Suffolk Local Research Ethics Committee (25^th May 2007; Ref: 07/Q0102/45), and Birmingham and Solihull PCT Consortium (Ref: 1087).

The majority of survey respondents were White British (n = 719; 74.9%), with those considering themselves to be Asian/Asian British constituting 15.6% of responses received (n = 150). Most respondents had no personal experience of CRC screening, but 98 individuals (10.2%) had been screened in the past for CRC with one or more of the four available CRC screening modalities (FOBt, flexible sigmoidoscopy, colonoscopy or double contrast barium enema). Just over a quarter of respondents had some form of medical school affiliation (n = 259; 27.0%). In terms of GP practice attributes, GPs from mid-sized practices constituted the majority of respondents: 51.6% (n = 495) worked in practices with between two and four GPs, and a further 249 respondents (25.9%) were from practices with between five and seven GPs. Most respondents were partners or principals in their practice (87.1%; n = 836). Those in single handed practices comprised 7.1% of responses received (n = 68). Responses were split fairly evenly on the basis of deprivation quartile and practice list size (Table 3).

Established national cancer screening programmes were considered by the majority of respondents to be effective in reducing mortality in average-risk patients, with 95.1% of GPs (n = 893) perceiving cervical cancer screening to be either 'very' or 'somewhat' effective, and 96.7% (n = 908) believing the same of mammography. This compared with 77.7% of respondents (n = 729) who considered FOBt as similarly effective. FOBt was recommended to patients by GPs far less frequently than either cervical smears or mammography: only 11% of respondents (n = 105) stated that they recommended FOBt 'almost always' when appropriate for asymptomatic average-risk patients, compared with 96.5% (n = 926) recommending cervical screening and 87.2% (n = 837) who routinely endorse mammography.

Respondents were asked whether they considered each of the four principal CRC screening modalities to be 'very', 'somewhat', or 'not at all' appropriate for population based CRC screening. Despite colorectal cancer screening typically being recommended less frequently than other cancer screening tests, FOBt was considered by the majority of survey respondents to be an appropriate test. FOBt was considered 'very appropriate' by 50.1% of GPs (n = 481), compared with 11.8% (n = 113) believing the same of flexible sigmoidoscopy; 12.8% (n = 123) endorsing colonoscopy, and only 3.1% (n = 27) perceiving the double contrast barium enema to be a 'very appropriate' screening test (Table 4).

FOBt was considered an inappropriate screening test by 8.2% of GPs (n = 79), compared with 42.2% (n = 405) agreeing that flexible sigmoidoscopy was inappropriate; 56.9% (n = 546) believing the same of colonoscopy, and 76.2% (n = 665) finding double contrast barium enema 'not at all appropriate'.

Of those who responded to the question, 31.4% of GPs (n = 72) agreed with the current recommended starting age for FOBt screening (60 years old). 148 GPs (64.6%) recommended beginning screening at a younger age, with only 3.9% of those responding (n = 9) believing that screening should begin at a later age than 60. The majority of respondents advocated less frequent testing than is currently recommended by the NHSBCSP. Of those who responded, 22.5% (n = 43) agreed with the current recommendation of biennial screening. 13.1% (n = 25) believed that people should be screened more frequently, and 64.4% (n = 123) advocated less frequent screening. The preferred option was screening every three years (37.2%; n = 71). There was little consistency in the reporting of a recommended stopping age for CRC screening in the asymptomatic average-risk population.

Respondent attitudes towards FOBt were assessed through the calculation of an attitude score. The mean score for all respondents (n = 960) was 5.9. Only 43 respondents (4.5%) scored the maximum of nine (indicating a very positive attitude), and 6.8% (n = 65) scored the minimum of three (very negative attitude). Mean scores were calculated according to both practice attributes and socio-demographic characteristics of respondents, and remained very similar across all groups and sub-groups (Table 3). Binary logistic regression was used to calculate bivariate Odds Ratios evaluating the association of socio-demographic and practice attributes with FOBt attitude scores. No significant associations were found between FOBt attitude and practice list size; percentage of registered patients aged over 50; the number of GPs working in a practice, or whether or not a respondent indicated an affiliation with a medical school, although those with such an affiliation (n = 259; 27.0%) were more likely to have a positive attitude towards FOBt than those without (bivariate OR: 1.3; CI 1.0 to 1.8).

It might be expected that survey respondents from non-White British ethnic groups may be disproportionately concentrated in socio-economically deprived areas. However, even after controlling for the possible confounding influence of practice deprivation quartile on FOBt attitude by ethnic group, the significance of the observed associations remained, although they were less pronounced (multivariate OR for ethnic group: 1.8; CI 1.2 to 2.6; p = 0.002; multivariate OR for deprivation quartile two: 0.7; CI 0.5 to 0.9; p = 0.039).