Selective Opportunistic Screening for Hypercholesterolemia in Primary Care Practice

J Clin Epidemiol Vol. 51, No. 10, pp. 817–825, 1998 Copyright  1998 Elsevier Science Inc. All rights reserved.

0895-4356/98/$-see front matter PII S0895-4356(98)00068-7

Selective Opportunistic Screening for Hypercholesterolemia in Primary Care Practice Brian Hutchison,1,2,3,* Stephen Birch,2,3 C. Edward Evans,1 Laurie J. Goldsmith,2 Barbara A. Markham,3 John Frank,4 and Michael Paterson 5 1

Department of Family Medicine, 2 Department of Clinical Epidemiology and Biostatistics, 3 Centre for Health Economics and Policy Analysis, McMaster University, Hamilton, Ontario, 4 Department of Preventive Medicine and Biostatistics, University of Toronto, Toronto, Ontario, and 5 Institute for Clinical Evaluative Sciences, Toronto, Ontario, Canada ABSTRACT. Objectives: To assess the performance of selective opportunistic screening in a primary care group practice. Design: Cross-sectional survey of coronary heart disease risk factors and retrospective chart audit of cholesterol testing. Setting: Capitation-funded primary care group practice in Ontario, Canada. Subjects: 7785 enrolled patients between the ages of 20 and 69 years. Intervention: Protocol-based selective opportunistic screening program for hypercholesterolemia of 45 months duration. Main Outcome Measures: Targeting (proportion of screening tests that were appropriate), coverage (proportion of those meeting screening criteria who had a screening test performed), over-screening (proportion of those not meeting screening criteria who had a screening test performed), and screening ratio (likelihood that a screening test was performed on an individual who met screening criteria rather than one who failed to meet screening criteria). Results: 64.7% of patients tested met the practice criteria for screening. 37.7% of patients who met the practice screening criteria were tested and 24.9% of those not meeting practice screening criteria had a cholesterol test performed. The screening ratio was 1.52. Conclusion: Our findings bring into question the effectiveness of opportunistic approaches to preventive care. j clin epidemiol 51;10:817–825, 1998.  1998 Elsevier Science Inc. KEY WORDS. Screening, hypercholesterolemia, primary care, performance appraisal

INTRODUCTION Cardiovascular disease is the leading cause of death worldwide. In the late 1980s age-standardized mortality rates for cardiovascular disease among males ranged from 165 per 100,000 in Japan to 493 per 100,000 in Romania [1]. Among OECD countries, potential years of life lost (before age 75) per 100,000 population due to ischemic heart disease ranged from a low of 28 in females and 105 in males (Japan 1989) to a high of 204 in females (Ireland 1988) and 1019 in males (Finland 1988) [2]. Only relatively recently has it been shown, through randomized controlled primary prevention trials, that serum cholesterol levels and subsequent coronary heart disease (CHD) events can be reduced with diet and drug therapy [3–5]. Evidence to date does not indicate a significant reduction in overall mortality as a result of cholesterol lowering through diet or drug therapy in the absence of preexisting CHD [6], although, in a recently reported randomized controlled trial of pravastatin in middle-aged hypercho*

Address for correspondence: Dr. Hutchison, Centre for Health Economics and Policy Analysis, Faculty of Health Sciences 3H1E, McMaster University, Hamilton, Ontario, Canada L8N 3Z5. Accepted for publication on 28 April 1998.

lesterolemic men with no history of myocardial infarction, there was a 22% reduction in all cause mortality which fell just short of statistical significance (P ⫽ 0.051) [5]. A concern in the targeting of serum cholesterol as a risk factor is the considerable overlap in the serum cholesterol distribution between those who develop CHD and those who do not [7]. While elevations in cholesterol are associated with increases in the relative risk of later CHD events, the absolute event rates are low. In the Multiple Risk Factor Intervention Trial, 1.3% of the men in the highest decile of cholesterol (mean 7.5 mmol/L) died of CHD within 6 years [8]. In the Lipid Research Clinics Coronary Primary Prevention Trial, 2.3% of the men in this group died of CHD over a mean of 7.4 years of follow-up [3]. Yet screening and treatment programs for elevated serum lipids label at least a quarter of the population screened as ‘‘unhealthy’’ and give rise to life-long medical intervention with attendant risks and costs. Attempts to modify the cholesterol level of the population should, therefore, consider carefully inherent ‘‘trade-offs’’ in the targeting of the screening program and the setting of treatment cutoff levels. Practice guidelines for the detection and management of hypercholesterolemia have proliferated during recent years (e.g., Coronary Prevention Group [9], European Athero-

B. Hutchison et al.

818

sclerosis Society [10,11], National Cholesterol Education Program Expert Panel [12,13], Canadian Consensus Conference on Cholesterol [14], Toronto Working Group on Cholesterol Policy [15,16], Canadian Task Force on the Periodic Health Examination [6], American College of Physicians [17], Task Force of the European Society of Cardiology, European Atherosclerosis Society, and European Society of Hypertension [18]). Although there is considerable disagreement among the creators of these guidelines as to which asymptomatic population groups should be targeted for screening, there appears to be a tendency to assume that once the appropriate target groups have been identified, cholesterol testing of persons included in those groups will occur in the normal course of events. Most assume (usually implicity) that the recommended screening or case-finding can and will occur during usual medical care. Many guidelines on cholesterol management implicitly [18] or explicitly [6,13,14,19] identify 100% of the population considered appropriate for screening as the target level of coverage. Among those that do not suggest full coverage as an objective [9,11,15–17], none suggests an alternative target level. The American College of Physicians [17], the Coronary Prevention Group [9], and the Toronto Working Group on Cholesterol Policy [15,16] all highlight the role of clinical judgment and shared decision making between clinicians and patients, without indicating what level of coverage might be anticipated using this approach. The European Atherosclerosis Society [11] advocates a combination of screening strategies that they suggest would lead over time to the assessment of CHD risk factors (including plasma cholesterol) ‘‘in the great majority of individuals in the physician’s practice.’’ A major goal of screening is to diagnose a disease (or a risk factor for a disease) in its earliest stages, before the individual is aware of any symptoms so that treatment can be initiated early in the disease process. Active screening involves recruitment of those targeted for screening, whereas opportunistic screening (also called case-finding) directs testing toward members of the target group as they present for health care. Selective screening aims to test people with particular characteristics related to health risks or responsiveness to intervention, while non-selective screening is directed toward an entire population. We have defined four basic screening strategies: active non-selective, opportunistic non-selective, active selective, and opportunistic selective (Table 1). Active non-selective screening (often denoted as universal screening) seeks to recruit and test an entire population or a demographically-defined subset of the population. Individuals are encouraged to present to their physician or a screening clinic for the purpose of detecting an asymptomatic disease. Active non-selective screening was favored in the recommendations for cholesterol screening developed by the National Cholesterol Education Program Expert Panel [12,13] in the United States.

TABLE 1. Types of screening

Screening

Non-selective

Active

Recruit and test a population

Opportunistic (case-finding)

Test all patients who present for care

Selective Recruit individuals who meet defined criteria for testing Test only patients who present for care and meet defined criteria for testing

A non-selective screening strategy can alternatively be opportunistic by design. That is, rather than recruiting patients to attend a clinic for the purpose of having screening tests performed, the strategy is to test all those who present for care. Opportunistic non-selective screening for hypercholesterolemia has been proposed by the European Atherosclerosis Society [10,11] and the Canadian Consensus Conference on Cholesterol [14]. The third screening strategy in our framework (active selective screening) has not yet been recommended for hypercholesterolemia detection. Active selective screening would recruit individuals who meet predefined criteria for testing (and are thus at high risk for the disease). In the context of cholesterol screening in a community family practice setting, such a strategy might involve sending a mailed questionnaire to all patients in the practice. Patients would be asked to identify their own risk factors. Those who met predefined levels of risk factors, would be encouraged to have their cholesterol measured. The fourth screening strategy, opportunistic selective screening, tests for the disease or risk factor in those patients presenting for care who meet predefined criteria for screening. Such a strategy has recently been recommended by the Toronto Working Group on Cholesterol Policy * [15,16], the Canadian Task Force on the Periodic Health Examination [6], the American College of Physicians [17], and, as an interim strategy, by the Canadian Consensus Conference on Cholesterol [14]. Although several expert guidelines for the detection and management of hypercholesterolemia propose opportunistic selective screening (in the case of the Canadian Consensus Conference as an interim strategy only), the effectiveness of this strategy has not been established. This article describes and reports the results of a study which retrospectively assessed the performance of opportunistic selective screening in a primary care group practice which had operated an organized program of hypercholesterolemia detection and management for 45 months prior to the beginning of the study. The study was conducted by investigators from *

The Toronto Working Group report was prepared for and endorsed by the Ontario Task Force on the Use and Provision on Medical Services, a cooperative venture between the Ontario Ministry of Health and the Ontario Medical Association.

Screening for Hypercholesterolemia

outside the practice. Features of the program included a nurse coordinator, agreed-upon criteria for cholesterol testing, a protocol for cholesterol testing and hypercholesterolemia management, and a group of care providers who were committed to the program. Performance appraisal of this program provides a measure of what can be expected of opportunistic screening under highly favorable circumstances. METHODS Setting The setting for the study was a capitation-funded primary care practice with an enrolled patient population of approximately 12,000, 7785 of whom were between the ages of 20 and 69 years. Enrolled patients agree to receive their health care from the practice unless they are referred to another physician or are travelling or temporarily living outside the practice area. The practice consisted of 7 family physicians (6.25 full-time equivalents), 3 nurse practitioners (2.0 fulltime equivalents), 10 nurses (4.7 full-time equivalents), clerical staff, and a part-time dietitian. Protocol for Cholesterol Testing For 45 months prior to the beginning of the study the practice had been performing protocol-based selective opportunistic screening for hypercholesterolemia among their adult patients. The protocol (based loosely on the report of the Toronto Working Group on Cholesterol Policy [15,16]) specified that total serum cholesterol testing be offered to persons with one or more of the following risk factors: (1) hypertension, (2) diabetes, (3) cigarette smoking, (4) severe obesity (⬎30% overweight), (5) family history of premature cardiovascular disease (myocardial infarction or sudden death of a parent or sibling before the age of 55), (6) previous myocardial infarction, (7) past or present angina, (8) previous cerebral vascular accident or transient ischemic attacks, (9) occlusive peripheral vascular disease. Risk factors 1 through 5 correspond to those identified by the Toronto Working Group (TWG) as most relevant for the detection and management of asymptomatic hypercholesterolemia. Risk factors 6 to 9 are designed to identify persons with symptomatic occlusive arterial disease. The protocol called for cholesterol testing to be offered under the following conditions: (1) during routine care for any of the above conditions, (2) during periodic health examinations whenever a risk factor is identified, and (3) at the initial visit of a new patient whenever a risk factor is identified. In addition, cholesterol testing is provided in the practice whenever a patient specifically requests it— regardless of risk factor status. Risk factor assessment forms, which both listed the practice’s criteria for cholesterol testing and served as a test requisition, were kept on physicians’ desks in ample supply through the duration of the program. Otherwise, no formal mechanisms to reinforce the protocol were used.

819

The protocol specifies that persons with a total serum cholesterol above 6.2 mmol/L are to be managed according to an algorithm which covers subsequent cholesterol testing, dietary intervention and, in the event of failure of dietary management, consideration of drug therapy. A designated nurse coordinates and documents patient follow-up and protocol adherence, and maintains detailed computerized records on patients who have their cholesterol tested. Three to four hours per week of nursing time was dedicated to implementing and administering the program during its initial year of operation; roughly 2 hours per week has been spent maintaining the program thereafter. Data Collection and Analysis To assess the effectiveness of this program we mailed a health questionnaire which included questions regarding coronary heart disease risk factors to all patients of the practice between the ages of 20 and 69 years, whether or not they had previously had cholesterol testing. We confined our study to patients 20 to 69 years of age based on the Toronto Working Group’s assessment that current evidence did not support screening beyond those age limits [15]. A second and third copy of the questionnaire were sent to nonrespondents and any remaining nonrespondents were surveyed by telephone. Performance of the screening program was assessed using coronary heart disease risk status information obtained from responses to the health questionnaire and cholesterol testing information from the practice records. The date on which individuals joined the practice was recorded in order to calculate the duration of potential exposure to the screening program. We assessed the validity of measuring CHD risk status using the health questionnaire by comparison with a clinical assessment of CHD risk status carried out by a research nurse who reviewed the patient’s clinical records and conducted a clinical interview in the patient’s home. The nurse measured height and weight. She did not measure blood pressure because a single reading is insufficient to establish the presence or absence of hypertension. The nurse assessed 100 randomly selected subjects, 40 of whom met Toronto Working Group criteria for cholesterol measurement (based on questionnaire responses) and 60 who did not. The program was evaluated in relation to the criteria for cholesterol testing contained in the practice’s screening protocol. In this article we focus on our evaluation of selective opportunistic screening for purposes of primary prevention. Subjects who reported a history of coronary heart disease (angina or myocardial infarction) were analyzed separately and are reported on and discussed only briefly. The practice criteria for cholesterol testing and our operational definitions of the criteria are shown in Table 2. Because the study was a prelude to a randomized controlled trial of active selective screening based on Toronto Work-

B. Hutchison et al.

820

TABLE 2. Practice criteria for cholesterol testing

Practice criteria One or more of Hypertension Diabetes Cigarette smoking ⬎30% overweight Myocardial infarction or sudden death of a parent or sibling before age 55

TABLE 3. Performance measures

Study definitions One or more of Hypertension (controlled or uncontrolled) Diabetes Smoking Body mass index (BMI) ⬎27 Heart attack or death from coronary heart disease in a parent, brother, or sister before the age of 60 years

Previous cerebrovascular accident or transient ischaemic attack Occlusive peripheral vascular disease

ing Group criteria for cholesterol testing [20], we used the Toronto Working Group definitions of obesity (body mass index ⬎27) rather than the practice criterion (more than 30% overweight) and family history of coronary heart disease (event before age 60 rather than before age 55). To avoid undue complexity in the health questionnaire, we chose not to gather data regarding two criteria included in the practice protocol (previous cerebrovascular accident and occlusive peripheral vascular disease). This decision was based on the anticipated low frequency of these risk factors and the complexity they would add to our data gathering. We assessed screening performance using the measures shown in Table 3. Subjects were excluded from the analysis when information on relevant screening criteria was missing. We presented the results of the study at a meeting of practice physicians and nurses and then conducted semistructured individual interviews to explore their perceptions regarding our findings. We interviewed two nurses, including the nurse who served as coordinator for the practice’s cholesterol program, and all five physicians.* We employed the chi-square statistic to test for differences in sex distribution between respondents and nonrespondents to the health questionnaire and uncontactable subjects, and the Student’s t-test to test for differences in mean age. We calculated 95% confidence intervals (CI) on our measures of screening performance, using Confidence Interval Analysis [21].

Meets screening criteria Tested Yes No

Yes (a) Appropriately screened (c) Inappropriately not screened

No (b) Inappropriately screened (d) Appropriately not screened

Targeting: Proportion of screening tests that were appropriate ⫽ a/a ⫹ b. Coverage: Proportion of those meeting screening criteria who had a screening test performed ⫽ a/a ⫹ c. Over-screening: Proportion of those not meeting screening criteria who had a screening test performed ⫽ b/b ⫹ d. Screening ratio: Likelihood that a screening test was performed on an individual who met screening criteria rather than one who failed to meet screening criteria ⫽ a/a ⫹ c ⫼ b/b ⫹ d.

were not included in the study, 476 (6.1%) because they had left the practice and 624 (8.0%) because they could not be contacted by mail or telephone. Of the remainder, 5816 returned the questionnaire or were interviewed by telephone, representing 74.7% of subjects surveyed and 87.0% of contactable current patients. Respondents differed from uncontactable subjects and nonrespondents in age and sex composition. Fifty-three percent of respondents were female, compared to 47.9% of nonrespondents (P ⫽ 0.004) and 49.2% of uncontactable subjects (P ⫽ 0.02). The mean age of respondents was 42.9 years, compared to 37.2 years among nonrespondents (P ⬍ 0.0001) and 36.9 years among uncontactable subjects (P ⬍ 0.0001). The mean length of exposure to the opportunistic screening program was 39.1 months (SD 17.9 months). 5515 respondents reported that they had neither had a heart attack nor been told they had coronary heart disease or angina and were therefore assumed to be free of coronary heart disease. These subjects were considered to be candidates for primary prevention in the present study. Two hundred and three respondents reported a previous history of CHD. Clinical Validation of CHD Risk Status Measurement

RESULTS Study Sample

Among the 100 subjects assessed clinically by the research nurse, there were five questionnaire false-positives and five false-negatives. In three cases (one false-positive and two false-negatives), these misclassifications occurred because two subjects living in the same household completed each other’s questionnaire. The sensitivity of CHD risk status measurement by questionnaire was 0.875 (95% CI, 0.732– 0.958) and the specificity was 0.917 (0.816–0.972). With a prevalence of elevated CHD risk of 40% by clinical assessment, the positive predictive value of CHD risk status by questionnaire was 0.875 and the negative predictive value was 0.917.

Questionnaires were mailed to all 7785 enrolled patients between the ages of 20 and 69 years. Of these, 1100 (14.1%)

Testing among Asymptomatic Patients

*

Two physicians (1.25 full-time equivalents) had left the practice during the course of the study.

Data on selective opportunistic screening in relation to the practice criteria are presented in Table 4. Nine hundred

Screening for Hypercholesterolemia

821

TABLE 4. Results of selective opportunistic screening (n ⴝ

4612)

or angina, 140 (69.0%) had a cholesterol test performed at least once during the 45-month study period.

Meets screening criteria

Tested Not tested Total

Yes (%)

No (%)

Total (%)

20.7 34.1 54.7

11.3 34.0 45.3

31.9 68.1 100

and three subjects (16.4%) were excluded from the analysis because of missing risk factor data, mainly family history of coronary heart disease. Based on practice protocol criteria for cholesterol screening, 34.1% of the adult practice population were inappropriately not screened and 11.3% were inappropriately screened. The main study results based on our principal measures of screening performance are presented in Table 5. The proportion of patients tested who met screening criteria (targeting) was 64.7%. The proportion of patients meeting screening criteria who had cholesterol testing performed (coverage) was 37.7%. The proportion of patients not meeting screening criteria who had a cholesterol test performed (over-screening) was 24.9%. The screening ratio was 1.52. To place this screening ratio in perspective, a hypothetical situation in which 80% of those meeting screening criteria were tested and only 5% of those not meeting criteria were tested, would yield a screening ratio of 16. The prevalence of CHD risk factors among tested and untested patients is shown in Figure 1. Except for cigarette smoking, risk factor prevalence was significantly higher among tested than untested patients. Tested patients were significantly less likely to be smokers than untested patients. The difference in risk factor prevalence between tested and untested patients was greatest for high blood pressure and family history of high cholesterol or coronary heart disease before 60 years of age. Figure 2 shows the proportion of patients tested by number of risk factors. The proportion tested increased from 20.4% of those with no risk factors to 50.5% of those with four or more risk factors (χ 2 for trend ⫽ 127.1, P ⬍ 0.00001). Testing among Patients with Symptomatic CHD Of the 203 patients who reported that they had either had a heart attack or been told they had coronary heart disease TABLE 5. Screening performance

Measure Targeting Coverage Over-screening Screening ratio

Patients

95% CI

64.7% 37.7% 24.9% 1.52

62.3–67.2 35.9–39.6 23.0–26.7 1.39–1.66

Interviews with Practice Personnel Practice physicians and nurses expressed disappointment and, in most cases, surprise when presented with the study results. All of the nurses and physicians interviewed felt that testing of low risk patients resulted mainly from patient requests for testing. Only one physician reported actively discouraging low risk patients from being tested. Most of the physicians and nurses felt that failure to test at risk patients occurred mainly because physicians and patients tend to give priority to immediate health care problems. Under these circumstances physicians might not have time to assess CHD risk and need for cholesterol testing, or might consider such an assessment inappropriate. Two physicians felt that they sometimes simply forgot to address the issue of CHD risk and cholesterol testing. Two physicians and both nurses cited patient refusal or noncompliance with testing as a source of failure to test at risk patients. However, only one of these thought that this would account for more than a small proportion of failures to test persons at increased CHD risk. DISCUSSION As is evident from Table 3, our performance measures of selective screening have parallels in the realm of diagnostic test evaluation. Our measure of targeting is analogous to positive predictive value in the diagnostic test lexicon, our measure of screening coverage is analogous to sensitivity and our measure of over-screening corresponds to the falsepositive rate (the complement of specificity). The screening ratio is analogous to the likelihood ratio for a positive test and, just as likelihood ratios are unaffected by prevalence, so the screening ratio is independent of the proportion of the population meeting screening criteria. It therefore provides a stable summary measure of the relative efficiency of screening performance, allowing valid comparisons of performance across different sets of screening criteria. Adequate levels of performance for a given screening program in terms of targeting, coverage, over-screening, and screening ratio, are ultimately a matter of judgment by care providers and health care policy makers. In retrospect, it would have been useful to encourage the members of the practice to identify what they considered acceptable levels of performance a priori and to compare those target levels with measured performance. At the conclusion of the study the care providers were uniformly dissatisfied with their screening performance. Given the level of commitment within the practice to selective opportunistic screening for hypercholesterolemia—as evidenced by the adoption of a protocol for the detection and management of hypercholesterolemia and the employment of a part-time nurse as program manager—

B. Hutchison et al.

822

FIGURE 1. Proportion of patients with coronary heart disease risk factors

and the fact that the program had been in operation for 45 months at the point of evaluation, the accomplishments of the practice seem modest in terms of efficient use of screening resources. Only 37.7% of asymptomatic patients who met practice criteria for cholesterol testing actually had a cholesterol test performed. Even among asymptomatic patients reporting four or more (of six) CHD risk factors, only 50% were tested. Almost 30% of patients with established CHD remained untested. On the other hand, 24.9% of those who did not meet the practice risk factor criteria for screening were tested and only 64.7% of patients tested met risk factor criteria. Not only did the practice fail to reach most of those targeted for screening, but much of their

FIGURE 2. Proportion tested by number of risk factors*

screening activity was directed toward patients who had a very low probability of benefiting from intervention because of their low CHD risk. It follows that much subsequent clinical activity related to the monitoring and treatment of hypercholesterolemia would be similarly misdirected. This is particularly worrisome in light of evidence from a metaanalysis of randomized controlled trials of cholesterol lowering treatments that total mortality is increased with treatment among patients at low CHD risk [22]. Although these findings have been challenged by other meta-analysts using different analytic approaches and selection criteria for primary studies [23,24], they suggest a need for careful reflection on aggressive treatment of hypercholesterolemia

Risk Factors ⴝ Diabetes, high blood pressure, obesity, smoking, family history of CHD before age 60 or high cholesterol, male aged 35–59 years. Chi-square test for linearity ⴝ 127.1, P ⬍ 0.00001; No significant deviations from the straight line (Departure from linearity chi-square ⴝ 4.4, P ⴝ 0.22). *

Screening for Hypercholesterolemia

among low risk patients. Had all screening activity been directed toward individuals who met screening criteria, 58% of those meeting practice criteria could have been tested. A number of factors contributed to these results. Failure to provide testing to at risk individuals occurred because some at risk patients did not attend the practice in the 45 months during which the program operated prior to evaluation. However, this factor is unlikely to explain more than a small proportion of failures to test at risk patients. Based on data from the practice’s computerized appointment and encounter data systems, 75% of patients 20 to 69 years of age (84% of females and 59% of males) were seen at least once during 1994. In light of the fact that patients included in this study were exposed to the opportunistic screening program for a mean duration of 39.1 months, relatively few patients would not have been seen at all during the 45month study period. Some at risk patients attended the practice but had health problems which preoccupied both the patient and the health care provider so that preventive issues were not addressed. Others were offered cholesterol testing but declined or failed to comply with it. The relative contributions of these potential influences are unknown. However, the consensus among health care providers in the practice is that the most important factor was a tendency for patients and providers to focus on immediate health problems. Low risk individuals who did not meet risk criteria for screening were sometimes tested at their request and, occasionally, on the initiative of a health care provider. Physicians rarely actively discouraged such individuals from being tested. Physicians’ willingness to order cholesterol testing for low risk patients at their request, is consistent with usual Canadian practice. It would be unusual for a physician to refuse to order a requested test that was an insured benefit under the universal health insurance plan when the test offered a potential clinical benefit and carried a minimal risk of harm. This would also be true in the case of a non-insured service for which the patient was willing to pay out-ofpocket or through private insurance. Some testing of low risk patients and failure to test those at risk may be apparent rather than real because of the differences between the screening criteria included in the practice protocol and our operational definition of those criteria for study purposes. For example, the practice protocol defined family history of coronary heart disease as a myocardial infarction (heart attack) or sudden death of a parent or sibling before the age of 55, whereas we used a cutoff of age 60. This difference would result in some patients who were low risk by practice criteria being identified as at risk by our criteria. Occlusive peripheral vascular disease and a previous history of cerebral vascular accident (stroke) or transient ischemic attack (TIA) were criteria for screening in the practice protocol but were not included in our operational definition of those criteria. This difference would lead to patients who met the practice protocol criteria for in-

823

creased CHD risk being identified as low risk by our definition. Because the number of patients with peripheral vascular disease, previous TIA or stroke, or a parent or sibling experiencing a heart attack or sudden death between age 55 and 60 is likely to be small relative to the number of patients with other risk factors, these discrepancies would not account for more than a small proportion of testing of low risk patients or failures to test patients at risk. Some further misclassification may have occurred because risk status was measured at the end of the study period. For some patients risk status may have changed during the study period because of either the emergence of new risk factors (e.g., hypertension, diabetes, or obesity) or the disappearance of old ones (e.g., smoking habit, obesity). These patient’s risk status classification would have been incorrect for some part of the study period. Our results are broadly consistent with the only other studies of selective opportunistic screening for hypercholesterolemia that we could identify. In their recent study of cholesterol screening in 20 Dutch general practices, van der Weijden et al. assessed screening performance in relation to cholesterol guidelines of the Dutch College of General Practitioners [25]. Those guidelines recommend cholesterol testing of patients aged 18 to 65 years with one or more of the six CHD risk factors. In their study, CHD risk factors were identified by chart review rather than, as in our study, by patient questionnaire. This may have accounted for the low prevalence of risk factors in the study practices (mean proportion of patients aged 18–65 years with one or more risk factors was only 12.5%). The overall level of cholesterol testing over a 2-year period was 9.6%. The mean level of coverage (proportion of patients with risk factors who had cholesterol testing) was 31% (compared to 37.7% in our study). Over-screening (proportion of those not meeting screening criteria who had a screening test performed) was much lower than in our study (mean 8% per practice versus 24.9%). Ovhed and colleagues assessed the performance of selective opportunistic cholesterol screening in a Swedish primary health care center staffed by six general practitioners [26]. Criteria for screening were left to the judgment of the individual practitioners. Over a 3-year period 27% of the residents of the health center’s catchment area aged 25–59 years were screened. This compares with a screening rate of 31.9% of patients aged 20–69 years in our study. We believe that our performance appraisal of this program of selective opportunistic screening provides a measure of what can be expected of such screening under highly favorable circumstances. We think that performance is likely to be substantially worse under usual practice circumstances in which screening protocols have not been established and no special resources are allocated to preventive activities. Our findings regarding the prevalence of CHD risk factors among tested and untested patients are largely in the expected direction of a higher prevalence among tested pa-

B. Hutchison et al.

824

tients. However, tested patients were less likely than untested patients to be smokers. Several explanations are possible. Smokers may tend to deny health risks. Providers may give priority to smoking cessation rather than cholesterol screening among smokers. Finally, smokers who were tested and found to have increased cholesterol may have discontinued smoking as part of a risk reduction effort. How does the performance of the selective opportunistic screening program we evaluated compare with other approaches to cholesterol screening in primary care? In addition to the two studies of selective opportunistic screening cited above [25,26], we have identified one study of opportunistic non-selective screening [27] and five studies of active non-selective screening [28–32] for hypercholesterolemia. Hahn reported on the implementation of a health maintenance protocol for patients 18 years of age or older by a single U.S. physician in private practice [27]. Evaluation of the implementation of the protocol, which included cholesterol testing, was based on a random sample of records for an 18-month period. Of 192 eligible patients whose records were reviewed, 186 (97%) were offered cholesterol testing and 181 (95%) had cholesterol testing performed. The findings from this study of opportunistic, non-selective screening by a single physician may not be generalizable. Of the five studies of active non-selective screening for hypercholesterolemia we identified, four were conducted in U.K. general practices [29–32] and one in a New Zealand general practice [28]. In each case patients were extended an invitation to attend for a ‘‘health check’’ performed by a nurse. The health checks included cholesterol testing and, in most cases [28–30,32], focussed on cardiovascular risk factors. The invitations were extended either at the time of office visits or by mail. The target populations were young and middle-aged adults. Coverage reported as a proportion of the target population that received health checks was lower (24.9% [31] and 62% [32]) than coverage reported as a proportion of those invited that received health checks (79% [28], 82% [29], 73% [30], and 94% [32]). Although these studies of active non-selective screening indicate that substantial proportions of those invited to attend for cardiovascular risk assessment will respond, the resource requirements of such programs are substantial. Our findings (and those of the studies discussed above) bring into question the wisdom of relying exclusively on opportunistic approaches to preventive care. Perhaps the difficulties of integrating illness care and prevention are such that opportunistic approaches are unlikely to have more than modest success. Organizations such as the Canadian Task Force on the Periodic Health Examination may need to review their firm and consistent commitment to providing preventive care in the context of care for other reasons. An examination of what we call active selective screening seems in order. In a second component of this study, active selective screening using a self-administered

CHD risk appraisal questionnaire sent to all practice members age 20 to 69 with no documented history of cholesterol screening was evaluated as a method of improving the allocation of screening resources [20]. Our population-based approach to performance appraisal of selective primary care screening is applicable to other targeted screening activities such as Pap smears or to targeted preventive interventions such as smoking cessation counselling. This approach permits an assessment of the extent to which the use of preventive care resources is aligned with population needs. 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. This project was supported by a research grant from the Ontario Ministry of Health. Drs. Hutchison and Birch are supported as National Health Research Scholars by Health Canada

References 1. Reader BA, Dagenais GR, Johansen H, Lauzon R, Mao Y, Nair C, et al. Cardiovascular Disease in Canada 1993. Ottawa: Heart and Stroke Foundation of Canada; 1993: 3. 2. OECD. Program OECD Health Data (Version 1.01, August 1991). Paris: OECD; 1991. 3. Lipid Research Clinics Program. The Lipid Research Clinics Coronary Primary Prevention Trial results. 1. Reduction in incidence of coronary heart disease. JAMA 1984; 251: 351– 364. 4. Frick MH, Elo O, Haapa K, Heinonen OP, Heinsalmi P, Helo P, et al. Helsinki Heart Study: Primary prevention trial with genfibrozil in middle-aged men with dyslipidemia. N Engl J Med 1987; 317: 1237–1245. 5. Shepherd J, Cobbe SM, Ford I, Isles CG, Lorimer AR, MacFarlane PW, et al. Prevention of coronary heart disease with pravastatin in men with hypercholesterolemia. N Engl J Med 1995; 333: 1301–1307. 6. Canadian Task Force on the Periodic Health Examination. Periodic health examination, 1993 update: 2. Lowering the blood total cholesterol level to prevent coronary heart disease. Can Med Assoc J 1993; 148: 521–538. 7. Kannel WB, Castelli WP, Gordon T. Cholesterol in the prediction of atherosclerotic disease. New perspectives based on the Framingham Study. Ann Intern Med 1979; 90: 1985– 1991. 8. Multiple Risk Factor Intervention Trial Research Group. Multiple risk factor intervention trial. Risk factor changes and mortality results. JAMA 1982; 248: 1465–1467. 9. Coronary Prevention Group. Risk assessment in the prevention of coronary heart disease: A policy statement. Br J Gen Pract 1990; 40: 467–469. 10. European Atherosclerosis Society Study Group. Strategies for the prevention of coronary heart disease. Eur Heart J 1987; 8: 77–88. 11. International Task Force for the Prevention of Coronary Health Disease. Prevention of coronary heart disease: Scientific background and new clinical guidelines. Recommendations of the European Atherosclerosis Society. Nutr Metab Cardiovasc Dis 1992; 2: 113–156. 12. National Cholesterol Education Program Expert Panel. Report of the National Cholesterol Education Program Ex-

Screening for Hypercholesterolemia

13.

14. 15.

16. 17.

18.

19.

20.

21.

pert Panel on detection, evaluation, and treatment of high blood cholesterol in adults. Arch Intern Med 1988; 148: 36–69. National Cholesterol Education Program Expert Panel. Second report of the National Cholesterol Education Program Expert Panel on detection, evaluation, and treatment of high blood cholesterol in adults (adult treatment panel II). JAMA 1993; 269: 3015–3023. Canadian Consensus Conference on Cholesterol. Final report. Can Med Assoc J 1988; 139 (Suppl): 1–8. Toronto Working Group on Cholesterol Policy. Detection and Management of Asymptomatic Hypercholesterolemia: A Policy Document. Toronto: Ontario Ministry of Health; 1989. Toronto Working Group on Cholesterol Policy. Asymptomatic hypercholesterolemia: A clinical policy review. J Clin Epidemiol 1990; 43: 1029–1121. American College of Physicians. Clinical guidelines, part 1. Guidelines for using serum cholesterol, high-density lipoprotein cholesterol, and triglyceride levels as screening tests for preventing coronary heart disease in adults. Ann Intern Med 1996; 124: 515–517. Pyo¨rala¨ K, DeBacker G, Graham I, Poole-Wilson P, Wood D. Prevention of coronary heart disease in clinical practice: Recommendations of the Task Force of the European Society of Cardiology, European Atherosclerosis Society and European Society of Hypertension. Eur Heart J 1994; 15: 1300– 1331. Royal College of General Practitioners. Guidelines for the Management of Hyperlipidaemia in General Practice: Towards the Primary Prevention of Coronary Heart Disease. Occasional paper 55. Exeter: The Royal College of General Practitioners; 1992. Hutchison B, Birch S, Evans CE, Goldsmith LJ, Markham BA, Frank J, Paterson M. Screening for hypercholesterolemia in primary care: Randomized controlled trial of postal questionnaire appraising risk of coronary heart disease. Br Med J 1998; 316: 1208–1213. Gardner SB, Winter PD, Gardner M. Confidence Interval

825

22. 23. 24. 25.

26.

27. 28. 29. 30.

31. 32.

Analysis (CIA) Version I.O. London: British Medical Journal; 1989. Smith GD, Song F, Sheldon TA. Cholesterol lowering and mortality: The importance of initial level of risk. Br Med J 1993; 306: 1367–1373. Law MR, Thompson SC, Wald NJ. Assessing possible hazards of reducing serum cholesterol. Br Med J 1994; 308: 373–379. Gould AL, Rossouw JE, Santanello NC, Heyse JF, Furberg DC. Cholesterol reduction yields clinical benefit. A new look at old data. Circulation 1995; 91: 2274–2282. van der Weijden T, Dansen A, Schouten BJ, Knottnerus JA, Grol RPTM. Comparison of appropriateness of cholesterol testing in general practice with the recommendations of national guidelines: An audit of patient records in 20 practices. Quality Health Care 1996; 5: 218–222. Ovhed I, Odeberg H, Ra¨stam L. Opportunistic screening for hypercholesterolemia with participants selected by the general practitioner. Inclusion and drop-out rate. Family Practice 1991; 8: 360–366. Hahn DL, Berger MG. Implementation of a systematic health maintenance protocol in a private practice. J Family Practice 1990; 31: 492–504. McMenamin JP. Screening for coronary heart disease risk among men in a general practice. N Z Med J 1995; 108: 167– 168. OXCHECK Study Group. Prevalence of risk factors for heart disease in OXCHECK trial: Implications for screening in primary care. Br Med J 1991; 302: 1057–1060. Mann JI, Lewis B, Shepherd J, Winder AF, Fenster S, Rose L, Morgan B. Blood lipid concentrations and other cardiovascular risk factors: Distribution, prevalence, and detection in Britain. Br Med J 1988; 296: 1702–1706. Sacks G, Marsden R. Evaluation of a practice-based programme of health checks: Financial cost and success at risk detection. J R C Gen Practit 1989; 39: 369–372. Jones J, Davies DH, Dove JR, Collinson MA, Brown PMR. Identification and treatment of risk factors for coronary heart disease in general practice: A possible screening model. Br Med J 1988; 296: 1711–1714.