Family Practice Vol. 20, No. 3, 340-346
© Oxford University Press 2003
Research Issues |
Developing clinically relevant and reproducible symptom-defined populations for cancer diagnostic research in general practice using a community survey
The Surgery, Manlake Avenue, Winterton, Scunthorpe DN15 9TA, UK.
Correspondence to Dr Nick Summerton; E-mail: N.Summerton{at}hull.ac.uk
Summerton N, Mann S, Sutton J, Rigby A, Theakston A, Clark J, Williams-Hardy H and Summerton A. Developing clinically relevant and reproducible symptom-defined populations for cancer diagnostic research in general practice using a community survey. Family Practice 2003; 20: 340–346.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodResultsDiscussionReferences |
|---|
Background. In developing cancer diagnostic studies, there is a particular need for the population of patients with symptoms of possible oncological significance who consult the GP in ‘real life’ to be comparable with the group of individuals with symptoms constructed as part of a research project.
Objectives. The objective of this study was to seek to assess whether a community-based symptom survey can be harnessed in order to produce clinically relevant and reproducible populations within which studies of more detailed indicants could be undertaken.
Method. A total of 3629 patients registered with a general practice at Winterton, UK, were sent a questionnaire enquiring about 10 symptoms of possible oncological significance together with their consultation intention in relation to these symptoms. Up to 1 month later, an identical questionnaire was applied to all patients reporting at least one symptom, and more detailed information was obtained by research nurses on each symptom.
Results. The overall response rate was 64.4%, and 850 patients reported one or more symptoms. For the majority of symptom reports, there was moderate to substantial agreement between the two applications of the questionnaire. The question on blood in the motions/toilet pan or on the toilet paper demonstrated almost perfect agreement. Slight agreement was found for abdominal pain for longer than 4 weeks and for black/tarry motions. In relation to the reliability of the patient consultation intention, there was substantial/moderate agreement for actions related to the majority of symptoms. For all symptoms, there was also a greater level of agreement for past activity than future intent.
Conclusion. The results of the study provide some support for a community survey as a mechanism to develop ‘clinically relevant’ populations for the iatrotropic symptoms rectal bleeding or indigestion/heartburn within which studies of more detailed indicants could be undertaken. There is also consistency with the work of others in relation to the numbers and characteristics of patients within the ‘clinically relevant’ population.
Keywords. Cancer, diagnosis, reproducibility, survey, symptoms.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionReferences |
|---|
GPs must have trust in the clinical indicants produced by diagnostic research if this information is going to be used within day-to-day clinical general practice. There is a particular requirement for the population of patients with, for example, rectal bleeding who consult the GP in ‘real life’ to be comparable with the group of individuals with rectal bleeding constructed as part of a research project (i.e. a ‘clinically relevant’ population). Furthermore, there must be confidence in the stability of the symptom that is used to characterize the study population; if the defining symptom exhibits significant unreliability, this would also tend to cast doubt on the status of the research population as a ‘clinically relevant’ population.
GPs have more reason for concern about selection bias in clinical research than their secondary care colleagues whose patients may be selected by the same processes as the individuals in most published studies. However, in developing a community-based approach in order to overcome such selective effects,1 it is essential to ensure that symptoms elicited from a general population survey can be linked to the iatrotropic symptoms that might bring a patient into the GP’s consulting room.
The primary purpose of the study described here was to seek to assess whether a community-based symptom survey can be harnessed in order to produce ‘clinically relevant’ and reproducible populations within which studies of more detailed indicants could be undertaken.
|
Method |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionReferences |
|---|
This study was based on the registered patient list of a single general practice at Winterton, North Lincolnshire, UK. The practice area covers the town of Winterton together with several surrounding villages: a community of 11 200 individuals within a semi-urban and rural setting. The majority (89%) of patients residing in the practice area are registered with the Winterton Medical Practice.
A questionnaire (A) was developed and piloted in order to enquire about 10 symptoms of possible oncological significance: constant cough, indigestion/heartburn, abdominal pain/discomfort, vomiting any blood during the last 6 months, coughing up any blood or pink-coloured sputum/ phlegm during last 6 months, blood in the motions/toilet pan or on the toilet paper during the last 6 months, passage of black/tarry stools during the last 6 months, change in the weekly number or consistency of bowel movements/ motions during the last 6 months, and involuntary weight loss during the last 6 months. In appreciation of the large number of patients presenting to GPs with transient symptoms,2 enquiries were also made as to whether the constant cough, the indigestion/heartburn or the abdominal pain/discomfort had lasted for at least 4 weeks. Participants were also asked to indicate whether they had seen a doctor or would consider visiting their GP about the symptom in order to define a ‘clinically relevant’ population.
An envelope containing questionnaire A, an information sheet, a consent form, a personally signed covering letter and a pre-paid return envelope was despatched to every patient in the practice aged between 45 and 75 years (n = 3629). This was achieved by dividing the Winterton practice area into six geographical sectors, with one new zone being targeted each month. As can be seen from Figure 1
, 19 people within the eligible age group were excluded from mailing, i.e. those with aphasia or dementia, or a history of lung, colorectal or upper gastrointestinal malignancy.
|
One month following dispatch of the original questionnaire A, a further questionnaire was sent out to non-respondents. Records of any remaining non-responders subsequently were tagged in order to seek their involvement in the study should they visit the Winterton Practice for any reason over the subsequent month. Any questionnaires received >3 months after the dispatch of the initial questionnaire A were excluded.
Clearly, it is possible that some people will consult their GP with symptoms but not complete a questionnaire A. In order to identify such false-negative responses, all the GPs working in the Winterton practice were asked to inform the research administrator of any patients presenting to them with one of the symptoms of possible oncological significance using notification cards. Secretarial and nursing staff were also requested to identify any patients they encountered who had been seen by the GP with one of the relevant symptoms. In addition, NS personally scrutinized all the radiology and endoscopy requests and reports, referral and consultant letters with a view to identifying any patients with one of the 10 symptoms of possible oncological significance.
NS checked all the positive responses on the returned questionnaire As in order to guarantee that every respondent satisfied the study inclusion criteria and also to ensure that no urgent intervention was required. Subsequently, and within 1 month of the receipt of the original questionnaire A at the Winterton Practice, all individuals reporting at least one symptom (n = 850) were contacted by a research nurse in order to complete a second questionnaire A. In reapplying the questionnaire A, the research nurses asked individuals about all 10 symptoms at the time when the original questionnaire A had been completed, and did not focus solely on those questions that had elicited a positive response. Furthermore, the research nurses also undertook a more detailed collection of information on the reported symptoms (using a questionnaire B). In the context of the study described here, this more detailed information served to assess the comparability of the ‘clinically relevant’ population in relation to some of the characteristics of patients with specific symptoms likely to consult GPs as identified by other researchers.
All the information obtained was entered onto an SPSS computerized research database at Winterton.3
Statistical methods
In order to assess the chance-corrected proportional agreement for the repeat applications of questionnaire A, weighted kappa (with 95% confidence intervals) was used.4 Landis and Koch5 have characterized different ranges of values for kappa with respect to the degree of agreement they suggest. Values >0.80 may be taken to represent almost perfect agreement, values between 0.61 and 0.80 substantial agreement, values between 0.41 and 0.60 moderate agreement, values between 0.21 and 0.40 slight agreement, and values of 
0.20 poor agreement beyond chance.
|
Results |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionReferences |
|---|
Sixty-four per cent of the study population returned a questionnaire A and, of these, 850 (36.4%) reported at least one symptom of possible oncological significance. A total of 814 patients completed a second questionnaire A, and 747 patients underwent more detailed questioning (using questionnaire B) by the research nurses.
Figure 1
illustrates the recruitment process and the age and sex characteristics of the respondents at each stage.
Amongst the 850 patients responding positively to at least one of the symptoms of possible oncological significance on the first questionnaire A, a total of 1943 symptoms were reported. A total of 549 patients reported one symptom only, 262 patients reported two symptoms, 125 patients reported three symptoms, 72 patients reported four symptoms, 26 patients reported five symptoms, eight patients reported six symptoms, three patients reported seven symptoms, and one patient reported eight symptoms. No patient reported nine or 10 symptoms of possible oncological significance.
Table 1 provides more detail on the symptom reports together with an assessment of the level of agreement between the two applications of questionnaire A. Data on concordant and discordant responses are available from the first author on request, and as supplementary data on the Family Practice website (http://www.fampraoupjournals.org/content/vol20/issue3.).
|
Table 2
presents information on the symptoms of possible oncological significance in relation to the patient’s stated intention (i.e. have you seen a doctor or would you consider visiting your GP about it, defined as the ‘clinically relevant’ population) and also seeks to illustrate the degree of agreement in relation to the patient intentions. Data on concordant and discordant responses are available from the first author on request and on the journal’s website, as above.
|
Commentary on Tables 1 and 2
For the majority of symptom reports, there was moderate to substantial agreement between the two applications of questionnaire A. The question on blood in the motions/ toilet pan or on the toilet paper demonstrated almost perfect agreement. Slight agreement was found for abdominal pain >4 weeks and for black/tarry motions.
In relation to the reliability of the patient consultation intention, there was substantial/moderate agreement for actions related to the majority of symptoms. For all symptoms, there was also a greater level of agreement for past activity than future intent. Poor or slight agreement beyond chance was noted for patient consultation intentions when they had described vomiting blood, coughing up blood/pink phlegm or passing black/tarry motions.
Symptom notifications
Amongst the four full-time GPs, one notified 23 symptoms, two notified five symptoms and one notified four symptoms. For the three half-time partners, the notifications were 10, one and zero. A further 21 symptoms were reported by other staff (secretarial/nursing) following a GP consultation. Ninety-eight relevant symptoms were identified from the scrutiny of the written records by NS. As a result of this work, five false negatives were identified amongst the 20 reports of cough, three false negatives amongst the 15 reports of change in bowel habit and six false negatives amongst the 35 reports of abdominal pain. Only one false negative was identified for indigestion (out of 26 notifications) and one false negative for rectal bleeding (out of 17 notifications).
Detailed symptom characteristics
Details on some of the specific characteristics of the heartburn/indigestion and blood in the motions/toilet pan or on the toilet paper symptoms in relation to the patient’s stated consultation intention are provided in Table 3.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionReferences |
|---|
The overall response rate in this study was 64.4% which accords favourably with the response rates obtained by others in similar surveys (e.g. 60% by Crosland and Jones6). In terms of representativeness, the age and sex characteristics of all the respondents are similar to those of the study population. Furthermore, the overall population prevalence figures for some of the symptoms accords with work by others. On the basis of a systematic review of the available literature, Fijten7 concluded that the 6-month prevalence of rectal bleeding was 10%, and this is in keeping with the 11.4% found in this study. Comparable data on some other symptoms were obtained from a survey in Sweden by Agréus et al.,8 e.g. the community prevalence of heartburn was 16.7% (compared with 12.1%/20.5% in this study), weight loss was 2.3% (compared with 3.3% in this study) and black stools was 1.0% (compared with 1.6% in this study).
The initial pilot work confirmed a subjective impression of the face validity of questionnaire A in relation to the questions on health care-seeking behaviour, i.e. ‘have you seen a doctor’ or ‘would you consider visiting your GP’. However, although the comparability of the overall population prevalence results for rectal bleeding, heartburn, weight loss and black motions does provide some confidence in the mechanism used in order to identify individuals with symptoms in the community, many of the published surveys identified are of questionable applicability to the types of patients with symptoms likely to be encountered by UK GPs. A number of authors have highlighted the complex interactions between health status, social and psychological factors in determining why patients with dyspepsia,9,10 bowel symptoms11 or cough12 consult GPs. However, in a survey conducted within two general practices in Tyneside, Crosland and Jones6 noted that only 41% of all respondents with rectal bleeding had ever sought medical advice for the problem. Those aged over 60 years were more likely to have consulted (56%) in addition to those patients who reported blood mixed with the stools rather than simply on the toilet paper. In comparing our ‘clinically relevant’ population with the findings of Crosland and Jones,6 it is important to appreciate that the group defined by us as the ‘clinically relevant’ population was not restricted to those who had consulted their GP but it also included those with an intention to consult. However, amongst the respondents with rectal bleeding (i.e. blood in the motions/ toilet pan or on the toilet paper during the last 6 months), 145 (6.2%) indicated that they had seen a doctor about it, representing 54% of the total population of patients with rectal bleeding symptoms. Taking into account the slightly older age range of this population, the figure is comparable with that of Crosland and Jones.6 Furthermore, the more detailed information on the symptom characteristics of the patients within the ‘clinically relevant’ population demonstrates similar concordance. In particular, participants who noticed blood in the toilet pan are more likely to consult than those who see it only on the toilet paper. In Australia, Byles et al.13 examined a history of rectal bleeding for up to 5 years prior to a detailed patient interview. Of the 77 individuals who had noticed a first occurrence of rectal bleeding >3 months but <5 years prior to interview, 30% had either delayed or failed to seek medical advice about the symptom. This figure is remarkably similar to the proportion (32.6%) of patients in the total population of individuals describing blood in the motions/toilet pan or on the toilet paper who were not included within the questionnaire A defined ‘clinically relevant’ population.
In relation to indigestion/heartburn, Meineche-Schmidt and Krag14 undertook a comprehensive 1 year examination of consulters amongst a group of GPs in Denmark, and reported an annual incidence of patients with new onset dyspepsia consulting their GP of 4.6 per 100 patients (aged 55–74 years). Within our ‘clinically relevant’ population, it can be calculated that 97 patients developed new onset indigestion/heartburn (lasting for at least 4 weeks) in the year before the questionnaire B was completed, giving an incidence figure of 4.1 per 100 patients per year, comparable with the findings from Denmark. Furthermore, Holtmann et al.15 noted that patients who consult with heartburn have more severe and frequent symptoms. As can be seen from Table 3, patients within the ‘clinically relevant’ population tended to have more severe, frequent and longer bouts of indigestion/heartburn.
For both indigestion/heartburn and blood in the motions/toilet pan or on the toilet paper, it is reassuring that the weighted kappa values for both the symptoms (0.85 and 0.77) and the patient intentions (0.89/0.73 and 0.78/0.68) were substantial. One caveat relates to the lower level of agreement in relation to future consultation intentions. Validating what patients say they will do is difficult, but Ganther et al.16 recently have developed a scale to measure patient preferences for using medical care and, as a next step, it would be helpful to compare their scale against the simple question used in the current study.
Little information is available on the consultation prevalence of abdominal pain or cough, but other studies suggest an under-recording of these symptoms in the work described here. Agréus et al.8 noted a prevalence of 22.2% for central abdominal pain, 16.7% for upper abdominal pain and 14.5% for lower abdominal pain, whereas the figures obtained in the current study were 17.2% for all abdominal pain reducing to 13.4% for those lasting for at least 4 weeks. According to Patrick and Patrick,17 the prevalence of chronic cough is reported to range from 14 to 23% for non-smoking adults, but in this study only 7.4% of patients had a cough lasting for at least 4 weeks and, moreover, 16.9% of these stated that they were smokers. Furthermore, although few symptom notifications were received from the GPs at Winterton, it is of interest that the highest numbers of false negatives were amongst those with cough or abdominal pain.
Although the majority of the responses indicate a substantial/moderate level of agreement in relation to the test–re-test reliability of the questions on questionnaire A both for symptoms and for patient consultation intentions, a further development would be to assess the reliability of the questionnaire A using the same method of data extraction on both occasions, as the research nurses will presumably have had some effect in interpreting the questions and adjusting the patient’s understanding. There is also a possibility that an interviewer’s personal habits or attitudes influence respondents’ answers.18 One additional difficulty is that even if responses are in total agreement, there may still be important differences between, for example, doctors’/health care professionals and patients’ interpretations of the meaning of, for example, a change in bowel habit, and this also warrants further examination.
As a result of this study, we have reasonable confidence in the method described in this paper as a mechanism to develop ‘clinically relevant’ populations for the iatrotropic symptoms rectal bleeding or indigestion/heartburn within which studies of more detailed indicants could be undertaken within a semi-urban/rural practice. For example, for individuals reporting blood in the motions/ toilet pan or on the toilet paper, the symptom reported as well as the patient consultation intention appear to be reliable. There is also consistency with the work of others in relation to the numbers and characteristics of patients within the ‘clinically relevant’ population. Furthermore, only one false-negative notification (out of 17) was reported.
|
References |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionReferences |
|---|
1 Summerton N. Symptoms of possible oncological significance: separating the wheat from the chaff. Br Med J 2002; 325: 1254–1255.
2 Kroenke K, Jackson JL. Outcome in general medical patients presenting with common symptoms: a prospective study with a 2-week and a 3-month follow-up. Fam Pract 1998; 15: 398–403.
3 Norussis M. SPSS for Windows: Base System User’s Guide, Release 6.0. Chicago: SPSS Inc., 1993.
4 Cohen J. Weighted kappa: nominal scale agreement with provision for scaled disagreement or partial credit. Psychol Bull 1968; 70: 213–220.[CrossRef][Medline]
5 Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33: 671–679.[CrossRef][Web of Science]
6 Crossland A, Jones R. Rectal bleeding: prevalence and consultation behaviour. Br Med J 1995; 311: 486–488.
7 Fijten G. Rectal Bleeding, a Danger Signal? Amsterdam: Thesis publishers, 1995.
8 Agréus L, Svärdsudd K, Nyrén O, Tibblin G. Reproducibility and validity of a postal questionnaire. The abdominal symptom study. Scand J Prim Health Care 1993; 11: 252–262.[Medline]
9 Delaney BC. Why do dyspeptic patients over the age of 50 consult their general practitioner? A qualitative investigation of health beliefs relating to dyspepsia. Br J Gen Pract 1998; 48: 1481–1485.[Web of Science][Medline]
10 Lydeard S, Jones R. Factors affecting the decision to consult with dyspepsia: comparison of consulters and non-consulters. J R Coll Gen Pract 1989; 39: 495–498.[Web of Science][Medline]
11 Kettell J, Jones R, Lydeard S. Reasons for consultation in irritable bowel syndrome: symptoms and patient characteristics. Br J Gen Pract 1992; 42: 459–461.[Web of Science][Medline]
12 Cornford CS. Why patients consult when they cough: a comparison of consulting and non-consulting patients. Br J Gen Pract 1998; 48: 1751–1754.[Web of Science][Medline]
13 Byles JE, Redman S, Hennrikus D, Sanson-Fisher RW, Dickinson J. Delay in consulting a medical practitioner about rectal bleeding. J Epidemiol Community Health 1992; 46: 241–244.
14 Meineche-Schmidt V, Krag E. Dyspepsia in general practice in Denmark. A 1-year analysis of consulters in general practice. Scand J Prim Health Care 1998; 16: 216–221.[CrossRef][Web of Science][Medline]
15 Holtmann G, Goebell H, Talley NJ. Dyspepsia in consulters and non-consulters: prevalence, health-care seeking behaviour and risk factors. Eur J Gastroenterol Hepatol 1994; 6: 917–924.[Web of Science]
16 Ganther JM, Wiederholt JB, Kreling DH. Measuring patients’ medical care preferences: care seeking versus self-treating. Med Decis Making 2001; 21: 133–140.
17 Patrick H, Patrick F. Chronic cough. Med Clin North Am 1995; 79: 361–372.[Web of Science][Medline]
18 Johannes CB, Crawford SL, McKinlay JB. Interviewer effects in a cohort study. Results from the Massachusetts Women’s Health Study. Am J Epidemiol 1997; 146: 429–438.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  S. C Almond and N. Summerton Test of time BMJ, June 15, 2009; 338(jun15_1): b1878 - b1878. [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||