Screening for hypercholesterolaemia in primary care: randomised controlled trial of postal questionnaire appraising risk of coronary heart diseaseBMJ 1998; 316 doi: https://doi.org/10.1136/bmj.316.7139.1208 (Published 18 April 1998) Cite this as: BMJ 1998;316:1208
- Brian Hutchison (), associate professora,
- Stephen Birch, professorb,
- C Edward Evans, professorc,
- Laurie J Goldsmith, research coordinatord,
- Barbara A Markham, research coordinatore,
- John Frank, professorf,
- Michael Paterson, research coordinatorf
- a Department of Family Medicine, Centre for Health Economics and Policy Analysis, Department of Clinical Epidemiology and Biostatistics, McMaster University, Health Sciences Centre Room 3H1E, 1200 Main Street West, Hamilton, Ontario, Canada L8N 3Z5
- b Centre for Health Economics and Policy Analysis, Department of Clinical Epidemiology and Biostatistics, McMaster University
- c Department of Family Medicine, McMaster University
- d Department of Clinical Epidemiology and Biostatistics, McMaster University
- e Centre for Health Economics and Policy Analysis, McMaster University
- f Institute for Clinical Evaluative Sciences, Sunnybrook Health Science Centre, 2075 Bayview Avenue, Toronto, Ontario, Canada M4N 3M5
- Correspondence to: Dr Hutchison
- Accepted 27 November 1997
Objectives: To validate a self administered postal questionnaire appraising risk of coronary heart disease. To determine whether use of this questionnaire increased the percentage of people at high risk of coronary heart disease and decreased the percentage of people at low risk who had their cholesterol concentration measured.
Design: Validation was by review of medical records and clinical assessment. The questionnaire appraising risk of coronary heart disease encouraged those meeting criteria for cholesterol measurement to have a cholesterol test and was tested in a randomised controlled trial. The intervention group was sent the risk appraisal questionnaire with a health questionnaire that determined risk of coronary heart disease without identifying the risk factors as related to coronary heart disease; the control group was sent the health questionnaire alone.
Setting: One capitation funded primary care practice in Canada with an enrolled patient population of about 12 000.
Subjects: Random sample of 100 participants in the intervention and control groups were included in the validation exercise. 5686 contactable patients aged 20 to 69 years who on the basis of practice records had not had a cholesterol test performed during the preceding 5 years were included in the randomised controlled trial. 2837 were in the intervention group and 2849 were in the control group.
Main outcome measures: Sensitivity and specificity of assessment of risk of coronary heart disease with risk appraisal questionnaire. Rate of cholesterol testing during three months of follow up.
Results: Sensitivity of questionnaire appraising coronary risk was 87.5% (95% confidence interval 73.2% to 95.8%) and specificity 91.7% (81.6% to 97.2%). Of the patients without pre-existing coronary heart disease who met predefined screening criteria based on risk, 45 out of 421 in the intervention group (10.7%) and 9 out of 504 in the control group (1.8%) had a cholesterol test performed during follow up (P<0.0001). Of the patients without a history of coronary heart disease who did not meet criteria for cholesterol testing, 30 out of 1128 in the intervention group (2.7%) and 18 out of 1099 in the control group (1.6%) had a cholesterol test (P=0.175). Of the patients with pre-existing coronary heart disease, 1 out of 15 in the intervention group (6.7%) and 1 out of 23 in the control group (4.3%) were tested during follow up (P=0.851, one tailed Fisher's exact test).
Conclusions: Although the questionnaire appraising coronary risk increased the percentage of people at high risk who obtained cholesterol testing, the effect was small. Most patients at risk who received the questionnaire did not respond by having a test.
Of patients at high risk of coronary heart disease, 10.7% who received a risk appraisal questionnaire with a general health questionnaire and 1.8% of those who received the general health questionnaire alone had a cholesterol test within the following three months
Of patients at low risk, 2.7% of patients receiving the risk appraisal questionnaire and 1.6% of control subjects had a cholesterol test
Most patients at risk who received the risk appraisal questionnaire did not seek a test
Further research is needed to identify factors contributing to low uptake of cholesterol testing among people at high risk of coronary heart disease even when encouragement is given
Opportunistic approaches to screening for hypercholesterolaemia are widely advocated.1-8 However, studies of this approach for hypercholesterolaemia,9-11 cervical carcinoma,10-18 breast cancer, and hypertension have repeatedly shown that a substantial percentage of eligible patients, often most, are not screened, even when interventions designed to improve coverage are used. In only a few instances have rates of coverage been reported that might be considered satisfactory—for Papanicolaou smear testing,21-23 clinical breast examination,21 and blood pressure measurement.10 In our study of selective opportunistic screening for hypercholesterolaemia in a Canadian primary care group practice, 38% of patients who met the practice's criteria for screening were tested over 45 months.24
Among the factors that limit the effectiveness of opportunistic screening are non-attendance at the practice by healthy patients and the tendency for those who seek care to have immediate health problems that take precedence over preventive issues. These limitations might be overcome by active screening approaches that seek to recruit people who meet predetermined criteria for testing. To assess this strategy, we developed and evaluated an active screening intervention for hypercholesterolaemia using a postal self administered questionnaire appraising the risk of coronary heart disease.
Subjects and methods
The objectives of this research were (a) to validate assessment of the risk of coronary heart disease by self administered questionnaire and (b) to determine whether posting a questionnaire appraising the risk of coronary heart disease to patients in primary care increases the percentage of people at high risk of coronary heart disease who have serum cholesterol concentration measured and decreases the percentage of people at low risk who are tested.
Toronto Working Group's criteria for cholesterol measurement
For men aged 35-39, one or more, and for women aged 20-69 years and men aged 20-34 and 60-69, two or more of:
Hypertension (controlled or uncontrolled)
Smoking, especially when this is resistant to efforts to stop
Severe or abdominal obesity
Family history of hypercholesterolaemia or coronary heart disease, particularly myocardial infarction or death from coronary heart disease in a parent or sibling before the age of 60
The setting was a capitation funded primary care practice with an enrolled patient population of about 12 000, of whom 7785 were between the ages of 20 and 69 years. For the 45 months before the beginning of the research the practice had been performing protocol based selective opportunistic screening for hypercholesterolaemia among its adult patients. The research protocol received ethics approval from the Ethics Review Committee of the McMaster University Faculty of Health Sciences.
Validation of risk appraisal questionnaire
For a random subset of 100 subjects drawn from both intervention and control groups, risk of coronary heart disease on the basis of responses in the health questionnaire was validated by a face to face clinical assessment with a research nurse. Criteria for assigning risk were those of the Toronto Working Group on Cholesterol Policy (box). Three months after completing the questionnaire subjects were assessed in their homes by a nurse after she had reviewed their clinical records. The review of clinical records was intended to identify any history of myocardial infarction or angina, use of glyceryl trinitrate or other nitrates, smoking, hypertension, diabetes mellitus, and family history of hypercholesterolaemia or coronary heart disease. The nurse's clinical assessment included taking a history to identify coronary heart disease and risk factors for the disease, and any changes in these two since completion of the questionnaire; examination of current drug treatment; and measurement of height and weight. Blood pressure was not measured because increased blood pressure on a single occasion is insufficient to diagnose hypertension.
Randomised study of risk appraisal questionnaire
Figure 1 shows the overall design of the trial. All patients of the practice between the ages of 20 and 69 years who, according to the practice's computerised cholesterol programme database, had not already been tested were randomly allocated to receive either (a) a health questionnaire that determined whether they were at risk of coronary heart disease without identifying the risk factors as related to coronary heart disease (control group) or (b) the health questionnaire and a questionnaire appraising risk of coronary heart disease that encouraged those meeting criteria for cholesterol measurement to have a cholesterol test (intervention group).
The questionnaire appraising risk of coronary heart disease (fig 2) made operational the criteria for cholesterol measurement developed by the Toronto Working Group on Cholesterol Policy (box). To minimise labelling and arousing fear, people who met these criteria for cholesterol testing were not advised of their high risk of coronary heart disease but were encouraged to have a cholesterol test performed if they had not been tested in the previous five years. Although the Toronto Working Group is silent on the subject, the Canadian Task Force on the Periodic Health Examination recommends five years between cholesterol tests for people with initially normal results.1 Those not meeting the screening criteria of the Toronto Working Group were advised that they did not need cholesterol testing. A requisition for cholesterol testing that included the hours when the practice took blood samples was included in the intervention group's package. Covering letters were signed by the principal investigator and the practice doctors. The covering letters included a statement that a decision not to participate in the project would in no way jeopardise the patient's care.
Questionnaires were posted in 26 waves two weeks apart. This ensured that the work of drawing blood samples for cholesterol testing, processing test results, handling inquiries related to cholesterol, following up abnormal test results, and managing newly identified cases of hypercholesterolaemia would be spread over time so as not to overwhelm the practice resources. Patients in the intervention and control groups were followed up for three months after the questionnaire was initially sent to determine whether they had had a cholesterol test.
To minimise contamination, patients were allocated by household unit, rather than individually. A second and third copy of the questionnaire were sent to non-respondents, and any remaining non-respondents were surveyed by telephone. We conducted intensive follow up to obtain as complete information as possible about subjects' risk of coronary heart disease.
For half of the 26 waves, follow up was immediate. In the other half, follow up was delayed until the end of the three months' follow up for cholesterol testing to reflect how the risk appraisal questionnaire might ultimately be used in clinical practice—that is, with a single posting. Study results were analysed separately for the two follow up conditions to determine whether the follow up procedure influenced the likelihood of subjects in the intervention group having their serum cholesterol tested.
We estimated that the practice would provide a sample size of about 2400 subjects (1200 per group) who did not have pre-existing coronary heart disease and who met the Toronto Working Group's criteria for cholesterol screening. Setting α at 0.05 (two tailed) and β at 0.1, we assumed an 80% return rate on the health questionnaire and a 1.3% uptake of cholesterol screening in the control group over the three months of follow up (on the basis of screening uptake during the first 33 months of the opportunistic screening programme). This gave us greater than 99% power to detect an absolute difference between screening rates of 10% in the intervention and control groups, with a 95% confidence interval on the difference of 7.8% to 12.2%. We considered 10% to be the minimum clinically important difference.
A check on the completeness of the practice's cholesterol programme database found that a substantial number of cholesterol tests were recorded in patient charts but not in the database. As a result, the charts of all patients between the ages of 20 and 69 years for whom there was no record of cholesterol testing in the programme database were reviewed by trained chart abstractors. A search was conducted for all cholesterol tests recorded in the chart. Laboratory reports, hospital discharge summaries, reports of specialist consultations, and records obtained from previous family physicians were examined. Patients without a record of cholesterol testing in the five years before the questionnaire was sent were included in the analysis. Separate analyses were conducted for those with and without pre-existing coronary heart disease.
Because patients were randomised by household unit, rather than individually, we used the analytical procedure proposed by Donner et al to correct for the effect of cluster allocation in testing the statistical significance of and computing 95% confidence intervals on differences between the intervention and control groups in the percentage of patients who received a cholesterol test during follow up.25 We computed the required intracluster correlation (K) for cholesterol testing on the basis of all patients in the practice aged 20 to 69 years for whom data were available. For all other analyses that entailed comparisons of proportions we used χ2 tests or Fisher's exact test.
Validation of risk appraisal questionnaire
Table 1 shows the results of the clinical validation of measuring risk of coronary heart disease by questionnaire. Of 100 subjects assessed clinically by the research nurse, five had false positive and five false negative results with the questionnaire. The sensitivity of questionnaire measurement of risk of coronary heart disease was 87.5% (95% confidence interval 73.2% to 95.8%); specificity was 91.7% (81.6% to 97.2%).
Randomised study of risk appraisal questionnaire
Of the 7785 patients aged 20 to 69 years, 1063 had been previously tested according to the practice's computerised cholesterol programme database and were not included in the randomised study. Of the 6722 patients randomly allocated, 454 (6.8%) did not consider themselves to be part of the practice, 582 (8.7%) could not be contacted, and 872 (13.0%) did not respond. The response rate was 71.6% (4814/6722) among all randomised patients and 84.7% (4814/5686) when those who did not consider themselves part of the practice and those who could not be contacted were excluded from the denominator. Among contactable patients the response rate was 84.4% (2506/2970) for those receiving immediate follow up and 85.0% (2308/2716) for those who received delayed follow up.
The intervention and control groups were similar in age and sex distribution. The mean age of both intervention and control group subjects was 38.8 years (SD 12.5 years in both groups). Women comprised 53.6% of the intervention group and 53.7% of the control group. More subjects in the control group than in the intervention group (31.4% (504/1603) v 27.2% (421/1549); χ2=6.902, P=0.009) met the Toronto Working Group's risk criteria for screening. Although patients were allocated by household, most households included only one subject. The mean number of study subjects per household (cluster size) was 1.27 in the intervention group and 1.29 in the control group. The mean correlation corrected for chance for cholesterol testing within household clusters was 0.9754. This correlation was used in the analysis of results to adjust for the effects of cluster allocation according to the method developed by Donner et al.25
Table 2 shows the overall results. Of those without pre-existing coronary heart disease who met the Toronto Working Group's criteria for screening, 45 out of 421 subjects in the intervention group (10.7%) and 9 out of 504 subjects in the control group (1.8%) had a cholesterol test performed during the three months after the initial questionnaire posting (P<0.0001 after adjustment for cluster). In both the intervention and control groups the percentage of subjects tested was not significantly different in those receiving immediate or delayed follow up. In the intervention group 26 out of 228 (11.4%) who received immediate follow up and 19 out of 193 (9.8%) who received delayed follow up had a cholesterol test (χ2=0.266, P=0.606). In the control group 4 out of 275 (1.45%) who had immediate follow up and 5 out of 229 (2.18%) who had delayed follow up were tested (χ2=0.378, P=0.538). Of the patients without a history of coronary heart disease who did not meet the criteria of the Toronto Working Group for cholesterol testing, 30 out of 1128 subjects in the intervention group (2.7%) and 18 out of 1099 subjects in the control group (1.6%) had a cholesterol test during the three month follow up period (P=0.175 after adjustment for cluster). Of the 38 subjects with pre-existing coronary heart disease, 1 out of 15 subjects in the intervention group (6.7%) and 1 out of 23 in the control group (4.3%) had a cholesterol test during the three months of follow up (P=0.851 in one tailed Fisher's exact test).
The questionnaire appraising risk of coronary heart disease increased the percentage of people at high risk of disease who obtained cholesterol testing, but the effect was modest (and slightly less than the minimum clinically important difference we set before the study). Most of the patients at risk who received the risk appraisal questionnaire did not respond to its encouragement to obtain testing. Moreover, use of the questionnaire did not reduce the rate of cholesterol testing among people at low risk, even though it provided reassurance that cholesterol testing was unnecessary. Because few patients at low risk were tested during the three month follow up, our study had limited power to detect differences between intervention and control subjects in the proportion of low risk patients tested. Our results show, however, that the absolute impact on inappropriate testing in either direction is likely to be small.
The limited impact of the risk appraisal questionnaire may be partly because the practice had been conducting selective opportunistic screening for hypercholesterolaemia during the 45 months before the beginning of this trial. During that time 38% of patients who met the practice's criteria for screening and 42% of those who met the criteria of the Toronto Working Group were tested.24 The 60% of patients at high risk who had not been tested opportunistically would include those who were missed because they attended the practice infrequently and those who had been offered but refused opportunistic screening. During 1994, 75% of patients aged 20 to 69 (84% of women and 59% of men) were seen at least once. Because patients were exposed to the opportunistic screening programme for a mean of 39.1 months, comparatively few patients would not have been seen at all during the period of systematic opportunistic screening. In semistructured interviews conducted at the end of the evaluation of the opportunistic screening programme the doctors and nurses of the practice indicated that if patients at risk were not tested it was mainly because doctors and patients tended to give priority to immediate healthcare problems. All but one of them believed that patients' refusal of or non-compliance with testing would account for only a small proportion of failures to test people at increased risk of coronary heart disease.
Alternative screening strategy
An alternative active screening strategy would be to include cholesterol screening in a scheduled periodic health examination offered systematically to all middle aged patients. This strategy was used in the Oxcheck trial26 and in a Welsh general practice study.27 In the Oxcheck trial around 66% of registered patients aged 35 to 64 years attended for a health check after a two stage process which entailed a postal questionnaire and an invitation for respondents to receive a health check from a trained nurse. The health check included non-selective screening for hypercholesterolaemia.26 In a Welsh general practice serving 10 000 patients 62% of invited patients aged 25 to 55 years attended a nurse run lifestyle intervention clinic for the identification and treatment of risk factors for coronary heart disease.27 Implementation of screening strategies of this type, whether cholesterol testing was selective or non-selective, would require a substantial commitment of resources.
Further research to identify factors contributing to low uptake of cholesterol testing among people at high risk of coronary heart disease—even with encouragement to obtain testing—is clearly desirable. The failure of our questionnaire to have an important effect on cholesterol testing could be related to its low key content and advice giving (rather than information giving) nature. Although the questionnaire assessed risk, it did not explicitly say that a high score meant an increased risk of coronary heart disease. The questionnaire and covering letter provided no information about coronary heart disease or its risk factors but advised those with high scores to obtain cholesterol testing and reassured those with low scores that cholesterol testing was not required. Perhaps an instrument that was more explicit about the risk of coronary heart disease and the potential benefits of lowering cholesterol concentration might have had more impact. Patients' perceptions about the risk of coronary heart disease and about lowering cholesterol concentration may underlie non-response to this intervention, which means that patients may be resistant to information as well as to advice.
This work is dedicated to the memory of C Edward (Ted) Evans, who contributed enormously to this project and whose death was a great loss to all of us who were privileged to work with him.
Contributors: BH initiated the study with SB, assembled the research team, participated in the conceptualisation of the project, drafted the methods section of the research protocol, supervised research staff in the collection and analysis of data, and drafted the paper. He is guarantor of the study. SB and BAM participated in all phases of the research from conceptualisation to writing of the paper. CEE (deceased) participated in the development of the protocol, designed the coronary heart disease risk appraisal questionnaire, and participated in the implementation of the study. LJG played a major role in data collection and analysis and participated in the writing of the paper. JF participated in the conceptualisation of the project, and the development of the research protocol, and the writing of the paper. MP participated in the data analysis and writing of the paper.
Funding: BH and SB are supported as national health research scholars by Health Canada. This study was supported by a research grant from the Ontario Ministry of Health.
Conflict of interest: None.
C Edward Evans has since died