- Email: [email protected]
Determinants of cholesterol screening and treatment patterns
Determinants of Cholesterol Screening and Treatment Patterns Insights for Decision-Makers Sharon K. Davis, MEd, MPA, PhD, David K. Ahn, PhD, Stephen P. Fortmann, MD, John W. Farquhar, MD Background: Adult cholesterol screening and treatment policies by the National Cholesterol Education Program recommend that physicians screen all adults aged $20. On the other hand, the American College of Physicians recommends that healthy young adult men aged .35 and premenopausal women aged .45 not be screened due to concerns about the cost of and health risks associated with overuse of pharmacologic therapy in lieu of lifestyle modification. Objectives:
The objectives of this study were to determine the type of treatment (lifestyle vs. pharmacologic) that physicians actually prescribe for individuals screened for elevated cholesterol.
Methods:
Self-report data were derived from the 1989 –1990 cross-sectional survey of the Stanford Five-City Project on 1,883 Latino and Anglo men and women aged 20 to 74 years of age. A four-stage sequential design was conducted using multiple stepwise regression analyses with a significance cutpoint of P , .01.
Results:
Young adult men and women were significantly less likely to report ever having been screened (OR 1.02; 95% CI 1.07–1.09). Individuals of low socioeconomic status (SES) were also significantly less likely to report ever being screened (OR, 1.12; CI, 1.08 –1.16), as were Latino men and women, regardless of age (OR 1.57; CI, 1.14 –2.18). There were no significant differences in the pattern of physician care utilization among low SES or Latino individuals during the previous 12-month period. Among those under physician care to lower cholesterol, young adults were more likely to be prescribed lifestyle modification (OR, 0.95; CI, 0.92– 0.98).
Conclusions: Our results suggest that although young adults are less likely to be screened, if screened they are more likely to be prescribed lifestyle modification than pharmacologic treatment for elevated cholesterol. The lower prevalence of screening among low SES and Latino individuals suggests the need for policy discussions to reduce these disparities. Medical Subject Headings (MeSH): public policy, ethnicity, gender, coronary heart disease, primary prevention, cholesterol. (Am J Prev Med 1998;15:178 –186) © 1998 American Journal of Preventive Medicine
Background
A
pproximately 10 years ago public policies allocated significant amounts of resources to the National Heart, Lung, and Blood Institute (NHLBI) to implement the National Cholesterol Education Program (NCEP).1 The overall goal of the program was to increase public knowledge about the From the Harvard School of Public Health, Division of Public Health Practice (Davis); and Stanford University Medical School Center for Research in Disease Prevention (Ahn, Fortmann, Farquhar). Address correspondence to: Sharon K. Davis, Harvard School of Public Health, Division of Public Health Practice, 718 Huntington Ave, Boston, MA 02115.
178
Am J Prev Med 1998;15(3) © 1998 American Journal of Preventive Medicine
adverse effects of high fat diets on cardiovascular health; one specific objective was to increase the prevalence of cholesterol screenings among adults. National education campaigns were disseminated via print and electronic media to heighten public awareness about the medical significance of screening for detection of elevated cholesterol. Practice guidelines were subsequently developed, via the Adult Treatment Panel, to assist physicians in clinical decision-making.2 The Panel recommended that all adults 20 years of age and older be screened, regardless of predisposition for CHD. Treatment recommendations, for individuals with elevated cholesterol, included lifestyle modification (i.e., reduced fat, increased exercise) followed with
0749-3797/98/$19.00 PII S0749-3797(98)00047-6
pharmacologic intervention if altered lifestyle failed to lower cholesterol to a clinically desirable level. Recently revised guidelines included the recommendation that all adults older than 20 years of age know their cholesterol levels.3 Cholesterol detection and treatment guidelines developed under the auspices of the NHLBI have been endorsed or similarly proposed by other medical and public health organizations4,5 and have remained relatively unchallenged until a recent divergence by the American College of Physicians (ACP).6 Unlike NHLBI, the ACP recommends that asymptomatic young adults (specifically, men less than 35 years of age and premenopausal women less than 45 years of age) should not be screened for blood cholesterol because of costs and potential health risks associated with drug treatment if elevated levels are detected.6 –14 Critics argue that the recommendation is based on the presumption that physicians automatically prescribe pharmacologic therapy and not lifestyle modification.8,9 It is also argued that implications of the recommendation may put young adults at long-term risk of developing CHD if high cholesterol is not detected and controlled.8,9 Cholesterol screening has emerged as a controversial health policy issue, with aspects of age and gender at the center of the debate.8 –14 Interestingly, very little is known about the sociodemographic pattern of cholesterol screenings, including the type of treatment prescribed by physicians for patients with elevated cholesterol. The objectives of our study were to elucidate these issues by determining the relative relationship of sociodemographic and health status factors on: (1) the likelihood to have ever been screened for elevated cholesterol; (2) the likelihood of individuals told to have elevated cholesterol to be under physician-care; and (3) the subsequent likelihood of such individuals to be prescribed lifestyle modification or pharmacologic treatment. Population-based data derived from the Stanford Five-City (FCP) provides a unique opportunity to analyze the major parameters of the cholesterol policy debate, age and gender,8 –14 as well as to assess issues associated with socioeconomic status (SES) and ethnicity. Results have significant import for relevant decision-makers in terms of providing empirical information to assist in clarifying ongoing discussions, and policy implications for subsequent discourse.
Methods The Stanford’s FCP was a long-term study (1979 through 1990) that used community-wide programs to reduce cardiovascular disease risk in intervention cities and includes population-based survey data on individuals 12 to 74 years of age. We restricted our study population to individuals aged 20 to 74 in order to gain insight into opposing cholesterol policy recommenda-
tions related to individuals who are 20 years of age and older. Data were derived from the last cross-sectional survey conducted in two treatment and two control cities located in Northern California from April 1989 through May 1990 (the only survey for which cholesterol screening and treatment indices are available). The overall response rate was 63.7%. Nonresponders were slightly younger (aged 35 versus 38), had similar years of education (12 versus 13), and were more likely to smoke (23% versus 18%). Data were pooled because there were no significant interactions between the variables of interest and city condition (treatment versus control). Sociodemographic, health status, and cholesterol-related variables were collected via a selfadministered questionnaire. Cholesterol screening was determined based on yes/no response to the question “Have you ever had your blood cholesterol checked?” Plasma samples were collected by licensed phlebotomists and sent to Stanford for determination of cholesterol level. Overall design and results of the FCP have been previously reported.15,16 Measures that are pertinent to our study are presented in abbreviated form. A complete version of measurement indices, including self-report and physiologic data collection methods, is available from the authors.
Definition and Measurement of Outcome Variables We conducted a four-stage sequential analysis and created the following outcome variables: (1) everscreened for cholesterol measurement; (2) under physician-care to lower cholesterol; (3) type of physicianprescribed treatment (lifestyle modification versus pharmacologic) (Figure 1). “Ever-screened for cholesterol” was evaluated by analyzing survey participants’ yes/no responses to the question “Have you ever had your blood cholesterol checked?” (n 5 1,883). Among respondents that answered “yes” to ever-screened for cholesterol and “yes” to the question “Has a doctor or medical professional ever told you that your cholesterol or blood fat is high?” (n 5 1,001), we then assessed “under physician-care to lower cholesterol” by evaluating yes/no responses to the question “Are you now under the treatment of a doctor to lower your cholesterol or blood fat level?”. We subsequently analyzed “type of physician-prescribed treatment” among those respondents that answered “yes” to being under physician-care (n 5 358) by assessing the subgroup that answered “yes” to the question “Did the doctor advise you to: (1) get more exercise, (2) lose weight, (3) eat less fat, (4) take a prescription medicine?” (n 5 199). Only one answer was possible. Possible responses were dichotomized into lifestyle modification treatment, based on an aggregate measure of responses “1,” “2,” and “3”; pharmacologic treatment was equated to a “4”
Am J Prev Med 1998;15(3)
179
lifestyle modification nor pharmacologic treatment, as well as those who responded “don’t know/unsure.”)
Definition and Measurement of Independent Variables
Figure 1. Four-stage sequential research design. Data were derived from the treatment and control cities of the 1989/ 1990 cross-sectional survey of the Stanford Five-City Cardiovascular Disease Risk Reduction Study. Only complete data were used in the models. an 5 1,001 includes respondents that answered “yes” to ever-screened for cholesterol and “yes” to the question “Has a doctor or medical professional ever told you that your cholesterol or blood fat is high?” bSixteen observations were deleted due to missing values.
response and did not include the other possible choices. Thus, lifestyle modification and pharmacologic stratifications were mutually exclusive. (The remaining 159 respondents of the total 358 in Figure 1 that reported to be under physician-care included a combined category of individuals that had neither
180
Socioeconomic status (SES) was determined by level of education, recorded as the highest number of years of school completed and divided into the following categories: ,12 years (less than high school); 12 years (high school graduate); 13–15 (some college or technical school); and $16 years (college graduate or postgraduate training). Age was measured in years and stratified into four age groups: 20 –34, 35– 44, 45– 64, 65–74. For the purposes of our research, we defined individuals aged 20 – 44 as young adults, those aged 45– 64 as middle-aged, and those 65–74 as older adults. Only two ethnic groups were large enough for analysis, Latino, and non-Hispanic white (Anglo). Health status indicators were based on a history of smoking status, hypertension, and diabetes. Participants were classified as current smokers if they reported ever smoking cigarettes on a daily basis and had smoked one or more cigarettes in the last week. Nonsmokers were defined as those who reported never smoking cigarettes on a daily basis. History of hypertension and diabetes was assessed in response to the questions, “Has a doctor or medical professional ever told you that you have high blood pressure?” and “Has a doctor or medical professional ever told you that you have diabetes?” We also assessed presence or absence of reported doctor visits ($1, 0) during the past year as an additional health proxy measure due to the clinical significance of routine physician-care utilization in the detection and prevention of disease. Our primary research objectives were to assess self-reported cholesterol related outcomes and not physiologic effects per se; therefore, actual physiologic measures were not included as predictor variables. We did, however, perform stratified analyses of mean total cholesterol in order to verify pertinent findings.
Analytic Approach Separate stepwise multiple logistic regression models were constructed for each of three dependent variables: ever-screened for cholesterol, under physiciancare to lower cholesterol, and type of physician-prescribed treatment.17 After adjustment for age, all of the independent variables were offered for inclusion in each of the models using a significance level of P # .01. The reported values are from the regression models. All possible first-order interactions with level of education were also added to the models to assess potential SES subgroup effects.
American Journal of Preventive Medicine, Volume 15, Number 3
Table 1. Sample sizes and percent distributions of the study population ever-screened for cholesterol, under physician-care to lower cholesterol, and those prescribed lifestyle modification vs. pharmacologic treatment by sociodemographic and health status indicators, 1989 –1990a
Ever-screened for cholesterolb n
%
N
260 472 574 577
105 253 282 361
40.4 53.6 49.1 62.6
45 108 102 103
20 64 58 57
44.4 59.2 56.8 55.3
20 60 57 54
15 45 49 46
75.0 75.0 85.9 85.1
774 463 441 205
230 252 339 180
29.7 54.4 76.9 87.8
49 64 168 77
20 29 104 46
40.8 45.3 61.9 59.7
18 26 102 45
18 21 85 31
100.0 80.7 83.3 68.8
1,558 325
897 104
57.9 33.9
320 38
181 18
56.5 47.3
174 17
139 16
79.8 94.0
849 1,034
426 575
45.1 54.9
153 205
81 118
52.9 57.5
79 112
65 90
82.2 80.3
406 391 51 1,525
158 296 45 888
38.9 75.7 72.5 58.2
60 146 24 335
31 87 11 195
51.6 59.5 45.8 58.2
30 84 10 187
23 69 7 151
76.6 82.1 70.0 80.7
N Sociodemographicse Education (years) ,12 12 13–15 .16 Age (years) 20–34 35–44 45–64 65–74 Race/ethnicity Anglo Latino Gender Men Women Health status indicators Smoker Hypertensive Diabetic Dr. visit during the past year
Under physician-care to lower cholesterolc n
%
Prescribed lifestyle modification vs. pharmacologic treatmentd N
n
%
a
Data were derived from the treatment and control cities of the 1989 –1990 cross-sectional survey of the Stanford Five City Cardiovascular Disease Risk Reduction Study. b “N” refers to the number of men and women within each subgroup that responded to the cholesterol screening item; “n” refers to the corresponding number that reported “yes” to ever having a cholesterol screening. c “N” refers to the number of men and women within each subgroup that responded “yes” to ever having been screened and “yes” to being told by a doctor or medical professional of having elevated blood cholesterol; “n” refers to the corresponding number that reported “yes” to being under physician-care. d “N” refers to the number of men and women that responded “yes” to being prescribed life-style modification or pharmacologic treatment among the total number of survey participants that reported “yes” to ever having been screened, “yes” to being told by a doctor or medical professional of having elevated blood cholesterol, and “yes” to being under physician-care to lower blood cholesterol; “n” refers to the corresponding number that reported physician-prescribed lifestyle modification. e Mean 6 SD years of education for ever-screened for cholesterol, under physician-care to lower cholesterol, and prescribed lifestyle modification vs. pharmacologic treatment 5 14.2 6 3.4, 14.5 6 3.1, 14.8 6 3.2, respectively; mean 6 SD years of age 5 46.2 6 14.6, 53.6 6 14.0, 55.1 6 13.2, respectively.
Results Our study population included a total of 1,883 men and women. Figure 1 presents the sequential distribution of subgroups according to the study’s outcome measures. Approximately half of the respondents reported being screened for cholesterol. Approximately 36% of those screened and told by a doctor or medical professional to have an elevated cholesterol reported being under physician-care. The mean total cholesterol levels for this sub-group were 228 mg/dL, well above the target level of less than 200. Of those under physician-care, approximately 56% received physician-prescribed treatment. Roughly 22% reportedly were prescribed pharmacologic treatment and about 78% lifestyle modification. Table 1 presents sample sizes and percent distribu-
tions of the outcome measures by sociodemographics and health status indicators. Because very few gender differences were observed, we combined findings related to men and women. The sociodemographic data show that, among the total number of individuals that responded to the cholesterol screening item (i.e., ever screened for cholesterol), those with , 12 years of education reported fewer screenings than those within the higher SES sub-groups. Seventy percent fewer screenings were also reported among individuals in the youngest age-group (aged 20 –34); however, incremental increases were reported among individuals within each of the remaining age-groups. Compared to Anglos, approximately 24% fewer Latino respondents reported ever having been screened. Compared to men, the proportion of women screened was slightly
Am J Prev Med 1998;15(3)
181
Table 2. Odds ratios and 95% confidence intervals for adjusted sociodemographic and health status indicators that significantly predict cholesterol screening, physician-care, and type of physician-prescribed treatment (lifestyle modification versus pharmacologic) from multivariate logistic regression analyses, 1989 –1990a
Sociodemographics Education (years) Age (years) Ethnicity Gender Health status indicators Smoking status Hypertension Diabetes Dr. visit during past year Education 3 Age
Ever-screened for cholesterol
Under physician-care to lower cholesterol
Type of physicianprescribed treatment
OR
95% CI
OR
OR
95% CI
1.12 1.07 1.57 ...
1.08, 1.16 1.07, 1.09 1.14, 2.18 ...
... 1.02 ... ...
... 1.0, 1.03 ... ...
... 0.95 ... ...
... 0.92, 0.98 ... ...
2.07 1.77 ... 2.82 ...
1.58, 2.70 1.29, 2.40 ... 1.29, 3.74 ...
... ... ... 6.17 ...
... ... ... 2.04, 18.64 ...
... ... ... ... ...
... ... ... ... ...
95% CI
a
Data were derived from treatment and control cities of the 1989 –1990 cross-sectional survey of the Stanford Five City Cardiovascular Disease Risk Reduction Study. OR 5 odds ratio, CI 5 confidence intervals. The following coding scheme was used as predictor variables in the analyses: education (in years), age (in years), ethnicity (1 5 Anglo, 0 5 Latino), gender (1 5 male, 0 5 female), smoking status (1 5 current smoker, 0 5 nonsmoker), history of hypertension (1 5 yes, 0 5 no), history of diabetes (1 5 yes, 0 5 no), Dr. visit during past year (1 5 $ 1, 0 5 0). All sociodemographic and health status variables were offered in the models. Also included were all first-order interactions with education as a measure of socioeconomic status (including education 3 age) at P # .01. Ellipses indicate that the variable was not selected because P . .0100.
higher. Health status indicators show that hypertensives, diabetics, and respondents who visited a doctor during the past year were more likely to report ever having a cholesterol screening compared to smokers. The proportion of the study population reportedly under physician-care to lower cholesterol among the total number that responded to having a screening and were told by a physician or medical professional of an elevated cholesterol was fairly evenly distributed across SES. There were, however, age-group differentials with higher proportions among middle-aged (45– 64) and older (65–74) respondents. The distribution of Anglo and Latino respondents receiving care was fairly similar, as was that of men and women. Respondents reporting the various health status indicators were also equally likely to report being under physician-care to lower cholesterol. The data from the subsequent and final subgroup of respondents that reportedly were prescribed lifestyle modification or pharmacologic treatment show that the majority were prescribed lifestyle modification. For lower and higher SES individuals the proportion distribution was between 75% and 85%. However, consistent with the previous outcome measures, there were observable age-group gradient differentials. All respondents in the 20 –34 age group (n 5 18) reported being prescribed lifestyle modification compared to decreasing proportions among each of the remaining older age groups. The proportion of Anglo and Latino respondents reporting physician-prescribed lifestyle modification was fairly similar, as was that of men and women and among those according to the study’s health status indicators.
182
Outcome Findings Table 2 presents the results of the stepwise multiple logistic regression models. The data show that SES (i.e., level of education in years), age, and ethnicity significantly influenced the likelihood of a reported cholesterol screening; smoking status, history of hypertension, and physician-care utilization in the past year were significant health status predictors (all P # .01). Specifically, higher SES respondents were more likely to report ever being screened, as were older individuals, and Anglo respondents. Those that smoked and those that were hypertensive were almost twice as likely to report being screened compared to nonsmoking and normtensive respondents. Respondents that saw a physician within the year were almost three times more likely to report ever being screened for cholesterol compared to those that did not visit a physician during the past 12 months. There were no interaction effects with educational level, thereby suggesting the absence of an SES sub-group relationship. Age (odds ratio of 1.02) and doctor visit during the past year (odds ratio of 6.17) were the only significant factors influencing the likelihood of respondents with elevated cholesterol to report being under physiciancare to lower cholesterol (P # .01, respectively). Age (odds ratio of 0.95) was also the only significant predictor affecting type of physician-prescribed treatment among respondents under physician-care (P # .01). As Figure 2 illustrates all respondents aged 20 –34 reported being prescribed lifestyle modification instead of pharmacologic intervention; this proportion decreases with increasing age, but even the oldest age
American Journal of Preventive Medicine, Volume 15, Number 3
Figure 2. Percent distribution of age in years as a sociodemographic factor that significantly influenced the likelihood of having physician-prescribed lifestyle modification versus pharmacologic treatment, derived from multiple logistic regression analysis (P # .01, Tables 1 and 2).
group (65–74) was most likely to be prescribed lifestyle change.
Discussion The threefold goal of our study was to assess the relative impact of sociodemographic and health status factors on the likelihood of having a screening to detect evidence of elevated cholesterol, the likelihood among individuals with elevated levels to be under physiciancare for treatment, and the subsequent likelihood to be prescribed lifestyle modification or pharmacologic intervention. After controlling for gender and age, our findings revealed no significant differences in cholesterol screenings between men and women. There were, however, widening age-group differentials between younger and older individuals, thereby suggesting that younger men and women are equally less likely to be screened for elevated cholesterol. Our results also showed that age was a significant predicator of the likelihood to be under physician-care to lower cholesterol. Among those told to have elevated cholesterol by a medical professional, only 24% of young adults ages 20 – 44 reported to be under care for treatment, compared to 75% among middle-aged and older individual ages 45–74 (Table 1). This observation may be explained in part by the relationship between advancing age and subsequent plasma cholesterol level. Logistic regression analyses revealed middle-aged and older individuals were less likely to have cholesterol
levels lower than ,200 mg/dL, but rather more likely to have levels between 200 –239 mg/dL or .240 mg/dL greater (P , .01, P , .01, respectively). Age combined with differences in the distribution of borderline (i.e., 200 –239 mg/dL) and elevated (.240 mg/dL) cholesterol levels may have also influenced type of physician treatment within our study population. Approximately 38% of younger individuals had borderline levels among those reportedly under physician-care and prescribed treatment; 10% had levels .240 mg/dL. Fifty percent of older individuals, on the other hand, had borderline levels, and 32% had levels .240 mg/dL. Physicians may, therefore, be more inclined to prescribe pharmacologic therapy to older individuals because of an increased risk of coronary disease associated with age and higher prevalence of elevated cholesterol (Figure 2). Our analyses revealed a significant disparity in cholesterol screenings among respondents according to level of SES. Sixty percent of respondents with less than 12 years of education reported never having a cholesterol screening compared to 63% among those with $16 years of education (Figure 3). The lack of an SES interaction effect with physician visit during the past year suggest that upper, middle, and lower SES individuals were equally exposed to the possibility of a screening and thereby precludes lack of access as an explanatory factor for the difference. Other factors may include health insurance reimbursement issues associated with preventive health care or general lack of
Am J Prev Med 1998;15(3)
183
Figure 3. Percent distribution of socioeconomic status via level of education in years as a sociodemographic factor that significantly influenced the likelihood of having a cholesterol screening, derived from multiple logistic regression analysis (P # .01, Tables 1 and 2).
knowledge among lower SES individuals about the relationship of elevated cholesterol to cardiovascular disease.18 We also detected a significant gap in cholesterol screenings between Latino and Anglo respondents; 40% and 58%, respectively (Figure 4). These differences remained after controlling for SES. We also assessed physician utilization during the past year according to ethnicity and found no significant differ-
ences. This variable is a proxy measure for physician care, which may or may not reflect primary care visits for preventive care. Nonetheless, the lack of variation in patterns of utilization makes it less likely that disparities were due to a general lack of health care access. This suggests that Latino individuals, regardless of SES, and health care access, were less likely to be screened for detection of elevated cholesterol than Anglo individuals. Moreover, the findings imply the need for further
Figure 4. Percent distribution of ethnicity (Anglo/Latino) as a sociodemographic factor that significantly influenced the likelihood of having a cholesterol screening, derived from multiple logistic regression analysis (P # .01, Tables 1
184
American Journal of Preventive Medicine, Volume 15, Number 3
research to investigate other issues associated with disparities in health care utilization among Latinos. Of clinical importance was the observation that individuals who were smokers, unlike those with a history of hypertension, were also less likely to have been screened. Such a finding may reflect the general risktaking behavior of smokers and the need for specific intervention efforts to ameliorate the prevalence of multiple risk factors.
Caveats There are a number of caveats that require discussion regarding our study results. Because of the ethnic makeup and geographic location of the study population, our findings for instance may be tempered because they are based primarily on Anglo and Latino adults residing in four Northern California cities. However, there is little reason to believe that exposure in our study population was significantly different from other ethnically similar communities in the United States since most cholesterol education information is received through television, pamphlets, magazines, and other nationally distributed mass media.1 Approval of Lovastatin, the first HMG CO-A reductase inhibitor, occurred approximately 2–3 years after our 1989/1990 survey was conducted. Older cholesterol lowering medications were more difficult to prescribe and less tolerated by patients. However, advances in pharmacological technology over the past 5 years have introduced a variety of new drugs with less adverse side-effects. Therefore, our findings based on 1989 – 1990 information may not accurately reflect treatment intervention choices currently available to physicians. In addition, recent 1992 survey data reflects overall increases in the prevalence of cholesterol screening since implementation of the NCEP in 1985 and since our 1989 –1990 survey.19 However, unlike our results, information does not report on SES or race/ethnicity differentials. Further research is needed to substantiate and corroborate our findings. Another potential limitation is that the study variables were mostly self-reported questionnaire items. A major strength of the FCP database is the opportunity it offers to conduct secondary analyses on large numbers of population-based men and women for whom sociodemographic and physiologic measurements are available. Thus, the ability to verify self-reported data with physiologic measures mitigates potential reporting bias and enhances the generalizability of our findings.
Conclusions The National Cholesterol Education Program and the American Heart Association recommend that all adults aged 20 and older be screened for cholesterol.2–5
Guidelines recently developed by the American College of Physicians recommend screening only high-risk men aged 20 –34 and women aged 20 – 44 due to concern with costs, pharmacologic treatment, and potential health risks.6,7 The objective of our research was to provide insight into this ongoing cholesterol policy debate8 –11 by analyzing sociodemographic patterns of cholesterol screenings and the type of treatment prescribed by physicians. Our findings suggest that in 1990, approximately 5 years after implementation of the NCEP and 6 years before the recent conflicting opinions appeared, young adult men and women aged 20 – 44 were significantly less likely to have been screened for elevated cholesterol than middle-aged and older individuals. And among those under treatment, the majority, regardless of age, were prescribed lifestyle modification. In summary, our analysis of population-based data suggest that the ACP concerns may be unwarranted, or at least amenable to physician education efforts. They also suggest that NCEP’s policy objectives have been less effective among young adults. The lower prevalence of cholesterol screening among low SES and Latino respondents, on the other hand, may suggest the need for subsequent policy discussions regarding physician-level intervention efforts in order to ameliorate the potential continuation of health care utilization disparities. This work was supported by Public Health Service grant R01-HL-21906 from the National Heart, Lung, and Blood Institute to Stephen P. Fortmann, MD, and a Minority Scientist Career Development Award from the American Heart Association to Sharon K. Davis, PhD. We thank Helena Kramer, PhD, for statistical assistance and Leslie Power for preparation of tables and graphs.
References 1. Lefant C. A new challenge for America: The National Cholesterol Education Program. Circulation 1986;73: 855– 6. 2. National Cholesterol Education Program. Report of the Expert Panel on detection, evaluation, and treatment of high cholesterol in adults. Arch Intern Med 1988;148: 36 –9. 3. Summary of the Second Report of the National Cholesterol Education Program (NCEP): Expert Panel on detection, evaluation, and treatment of high blood cholesterol in adults (Adult Treatment Panel II). JAMA 1993; 269:3015–23. 4. National Research Council. Diet and health: implications for reducing chronic disease risk. Washington, DC: National Academy Press; 1989. 5. American Heart Association Task Force on Cholesterol Issues. The cholesterol facts: a summary of the evidence relating to dietary fats, serum cholesterol, and coronary
Am J Prev Med 1998;15(3)
185
6.
7.
8.
9. 10. 11.
12.
13.
186
heart disease: a joint statement by the American Heart Association and the National Heart, Lung, and Blood Institute. Circulation 1990;81:5. American College of Physicians. 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; 124:515–17. Garber AM, Brown WS, Hully SB. Cholesterol screening in asymptomatic adults, revisited. Ann Intern Med 1996; 124:518 –31. Task Force on Risk Reduction, American Heart Association: Cholesterol screening in asymptomatic adults: no cause to change. Circulation 1996;93:1067– 8. LaRosa JC. Cholesterol agnostics. Ann Intern Med 1996; 124:505– 8. Davidoff F. Evangelists and snail redux: the case of cholesterol screening. Ann Intern Med 1996;24:513–14. Toronto Working Group on Cholesterol Policy. Asymptomatic hypercholesteroemia: a clinical policy review. J Clin Epidemiol 1990;43:1028 –1121. Hulley SB, Newman TB, Grady D, Garber AM, Baron RB, et al. Should we be measuring blood cholesterol levels in younger adults? JAMA 1993;269:1416 –19. Nayor CD. Dyslipidaemias and the primary prevention of
14. 15.
16.
17. 18.
19.
coronary heart disease: reflections on some unresolved policy issues. In: Gold P, Grover S, Roncari DAK, eds. Cholesterol and coronary heart disease: the great debate. Park Ridge, New Jersey: Parthenon; 1992:401–25. LaRosa JC. Cholesterol and public policy. Atherosclerosis 1994;108:s137–s141. Farquhar JW, Fortmann SP, Maccoby N, et al. The Stanford Five-City Project: design and methods. Am J Epidemiol 1985;122:323–34. Farquhar JW, Fortmann SP, Flora JA, et al. Effects of communitywide education on cardiovascular disease risk factors. The Stanford Five-City Project. JAMA 1990;264: 359 – 65. Neter J, Wasserman W, Kutner MH. Applied linear statistical models. Boston: Richard D. Irwin, Inc.; 1990. Davis SK, Winkleby MA, Farquhar JW. Increasing disparity in knowledge of cardiovascular disease risk factors and risk-reduction strategies by socioeconomic status: implications for policymakers. Am J Prev Med 1995;11:318 –23. Pieper RM, Arnett DK, McGovern PG, Shahar E, Blackburn H, Luepker RV. Trends in cholesterol knowledge and screening and hyperchoesterolemia awareness and treatment, 1980 –1992. The Minnesota Heart Survey. Arch Intern Med 1997;157:2326 –32.
American Journal of Preventive Medicine, Volume 15, Number 3
The objectives of this study were to determine the type of treatment (lifestyle vs. pharmacologic) that physicians actually prescribe for individuals screened for elevated cholesterol.
Methods:
Self-report data were derived from the 1989 –1990 cross-sectional survey of the Stanford Five-City Project on 1,883 Latino and Anglo men and women aged 20 to 74 years of age. A four-stage sequential design was conducted using multiple stepwise regression analyses with a significance cutpoint of P , .01.
Results:
Young adult men and women were significantly less likely to report ever having been screened (OR 1.02; 95% CI 1.07–1.09). Individuals of low socioeconomic status (SES) were also significantly less likely to report ever being screened (OR, 1.12; CI, 1.08 –1.16), as were Latino men and women, regardless of age (OR 1.57; CI, 1.14 –2.18). There were no significant differences in the pattern of physician care utilization among low SES or Latino individuals during the previous 12-month period. Among those under physician care to lower cholesterol, young adults were more likely to be prescribed lifestyle modification (OR, 0.95; CI, 0.92– 0.98).
Conclusions: Our results suggest that although young adults are less likely to be screened, if screened they are more likely to be prescribed lifestyle modification than pharmacologic treatment for elevated cholesterol. The lower prevalence of screening among low SES and Latino individuals suggests the need for policy discussions to reduce these disparities. Medical Subject Headings (MeSH): public policy, ethnicity, gender, coronary heart disease, primary prevention, cholesterol. (Am J Prev Med 1998;15:178 –186) © 1998 American Journal of Preventive Medicine
Background
A
pproximately 10 years ago public policies allocated significant amounts of resources to the National Heart, Lung, and Blood Institute (NHLBI) to implement the National Cholesterol Education Program (NCEP).1 The overall goal of the program was to increase public knowledge about the From the Harvard School of Public Health, Division of Public Health Practice (Davis); and Stanford University Medical School Center for Research in Disease Prevention (Ahn, Fortmann, Farquhar). Address correspondence to: Sharon K. Davis, Harvard School of Public Health, Division of Public Health Practice, 718 Huntington Ave, Boston, MA 02115.
178
Am J Prev Med 1998;15(3) © 1998 American Journal of Preventive Medicine
adverse effects of high fat diets on cardiovascular health; one specific objective was to increase the prevalence of cholesterol screenings among adults. National education campaigns were disseminated via print and electronic media to heighten public awareness about the medical significance of screening for detection of elevated cholesterol. Practice guidelines were subsequently developed, via the Adult Treatment Panel, to assist physicians in clinical decision-making.2 The Panel recommended that all adults 20 years of age and older be screened, regardless of predisposition for CHD. Treatment recommendations, for individuals with elevated cholesterol, included lifestyle modification (i.e., reduced fat, increased exercise) followed with
0749-3797/98/$19.00 PII S0749-3797(98)00047-6
pharmacologic intervention if altered lifestyle failed to lower cholesterol to a clinically desirable level. Recently revised guidelines included the recommendation that all adults older than 20 years of age know their cholesterol levels.3 Cholesterol detection and treatment guidelines developed under the auspices of the NHLBI have been endorsed or similarly proposed by other medical and public health organizations4,5 and have remained relatively unchallenged until a recent divergence by the American College of Physicians (ACP).6 Unlike NHLBI, the ACP recommends that asymptomatic young adults (specifically, men less than 35 years of age and premenopausal women less than 45 years of age) should not be screened for blood cholesterol because of costs and potential health risks associated with drug treatment if elevated levels are detected.6 –14 Critics argue that the recommendation is based on the presumption that physicians automatically prescribe pharmacologic therapy and not lifestyle modification.8,9 It is also argued that implications of the recommendation may put young adults at long-term risk of developing CHD if high cholesterol is not detected and controlled.8,9 Cholesterol screening has emerged as a controversial health policy issue, with aspects of age and gender at the center of the debate.8 –14 Interestingly, very little is known about the sociodemographic pattern of cholesterol screenings, including the type of treatment prescribed by physicians for patients with elevated cholesterol. The objectives of our study were to elucidate these issues by determining the relative relationship of sociodemographic and health status factors on: (1) the likelihood to have ever been screened for elevated cholesterol; (2) the likelihood of individuals told to have elevated cholesterol to be under physician-care; and (3) the subsequent likelihood of such individuals to be prescribed lifestyle modification or pharmacologic treatment. Population-based data derived from the Stanford Five-City (FCP) provides a unique opportunity to analyze the major parameters of the cholesterol policy debate, age and gender,8 –14 as well as to assess issues associated with socioeconomic status (SES) and ethnicity. Results have significant import for relevant decision-makers in terms of providing empirical information to assist in clarifying ongoing discussions, and policy implications for subsequent discourse.
Methods The Stanford’s FCP was a long-term study (1979 through 1990) that used community-wide programs to reduce cardiovascular disease risk in intervention cities and includes population-based survey data on individuals 12 to 74 years of age. We restricted our study population to individuals aged 20 to 74 in order to gain insight into opposing cholesterol policy recommenda-
tions related to individuals who are 20 years of age and older. Data were derived from the last cross-sectional survey conducted in two treatment and two control cities located in Northern California from April 1989 through May 1990 (the only survey for which cholesterol screening and treatment indices are available). The overall response rate was 63.7%. Nonresponders were slightly younger (aged 35 versus 38), had similar years of education (12 versus 13), and were more likely to smoke (23% versus 18%). Data were pooled because there were no significant interactions between the variables of interest and city condition (treatment versus control). Sociodemographic, health status, and cholesterol-related variables were collected via a selfadministered questionnaire. Cholesterol screening was determined based on yes/no response to the question “Have you ever had your blood cholesterol checked?” Plasma samples were collected by licensed phlebotomists and sent to Stanford for determination of cholesterol level. Overall design and results of the FCP have been previously reported.15,16 Measures that are pertinent to our study are presented in abbreviated form. A complete version of measurement indices, including self-report and physiologic data collection methods, is available from the authors.
Definition and Measurement of Outcome Variables We conducted a four-stage sequential analysis and created the following outcome variables: (1) everscreened for cholesterol measurement; (2) under physician-care to lower cholesterol; (3) type of physicianprescribed treatment (lifestyle modification versus pharmacologic) (Figure 1). “Ever-screened for cholesterol” was evaluated by analyzing survey participants’ yes/no responses to the question “Have you ever had your blood cholesterol checked?” (n 5 1,883). Among respondents that answered “yes” to ever-screened for cholesterol and “yes” to the question “Has a doctor or medical professional ever told you that your cholesterol or blood fat is high?” (n 5 1,001), we then assessed “under physician-care to lower cholesterol” by evaluating yes/no responses to the question “Are you now under the treatment of a doctor to lower your cholesterol or blood fat level?”. We subsequently analyzed “type of physician-prescribed treatment” among those respondents that answered “yes” to being under physician-care (n 5 358) by assessing the subgroup that answered “yes” to the question “Did the doctor advise you to: (1) get more exercise, (2) lose weight, (3) eat less fat, (4) take a prescription medicine?” (n 5 199). Only one answer was possible. Possible responses were dichotomized into lifestyle modification treatment, based on an aggregate measure of responses “1,” “2,” and “3”; pharmacologic treatment was equated to a “4”
Am J Prev Med 1998;15(3)
179
lifestyle modification nor pharmacologic treatment, as well as those who responded “don’t know/unsure.”)
Definition and Measurement of Independent Variables
Figure 1. Four-stage sequential research design. Data were derived from the treatment and control cities of the 1989/ 1990 cross-sectional survey of the Stanford Five-City Cardiovascular Disease Risk Reduction Study. Only complete data were used in the models. an 5 1,001 includes respondents that answered “yes” to ever-screened for cholesterol and “yes” to the question “Has a doctor or medical professional ever told you that your cholesterol or blood fat is high?” bSixteen observations were deleted due to missing values.
response and did not include the other possible choices. Thus, lifestyle modification and pharmacologic stratifications were mutually exclusive. (The remaining 159 respondents of the total 358 in Figure 1 that reported to be under physician-care included a combined category of individuals that had neither
180
Socioeconomic status (SES) was determined by level of education, recorded as the highest number of years of school completed and divided into the following categories: ,12 years (less than high school); 12 years (high school graduate); 13–15 (some college or technical school); and $16 years (college graduate or postgraduate training). Age was measured in years and stratified into four age groups: 20 –34, 35– 44, 45– 64, 65–74. For the purposes of our research, we defined individuals aged 20 – 44 as young adults, those aged 45– 64 as middle-aged, and those 65–74 as older adults. Only two ethnic groups were large enough for analysis, Latino, and non-Hispanic white (Anglo). Health status indicators were based on a history of smoking status, hypertension, and diabetes. Participants were classified as current smokers if they reported ever smoking cigarettes on a daily basis and had smoked one or more cigarettes in the last week. Nonsmokers were defined as those who reported never smoking cigarettes on a daily basis. History of hypertension and diabetes was assessed in response to the questions, “Has a doctor or medical professional ever told you that you have high blood pressure?” and “Has a doctor or medical professional ever told you that you have diabetes?” We also assessed presence or absence of reported doctor visits ($1, 0) during the past year as an additional health proxy measure due to the clinical significance of routine physician-care utilization in the detection and prevention of disease. Our primary research objectives were to assess self-reported cholesterol related outcomes and not physiologic effects per se; therefore, actual physiologic measures were not included as predictor variables. We did, however, perform stratified analyses of mean total cholesterol in order to verify pertinent findings.
Analytic Approach Separate stepwise multiple logistic regression models were constructed for each of three dependent variables: ever-screened for cholesterol, under physiciancare to lower cholesterol, and type of physician-prescribed treatment.17 After adjustment for age, all of the independent variables were offered for inclusion in each of the models using a significance level of P # .01. The reported values are from the regression models. All possible first-order interactions with level of education were also added to the models to assess potential SES subgroup effects.
American Journal of Preventive Medicine, Volume 15, Number 3
Table 1. Sample sizes and percent distributions of the study population ever-screened for cholesterol, under physician-care to lower cholesterol, and those prescribed lifestyle modification vs. pharmacologic treatment by sociodemographic and health status indicators, 1989 –1990a
Ever-screened for cholesterolb n
%
N
260 472 574 577
105 253 282 361
40.4 53.6 49.1 62.6
45 108 102 103
20 64 58 57
44.4 59.2 56.8 55.3
20 60 57 54
15 45 49 46
75.0 75.0 85.9 85.1
774 463 441 205
230 252 339 180
29.7 54.4 76.9 87.8
49 64 168 77
20 29 104 46
40.8 45.3 61.9 59.7
18 26 102 45
18 21 85 31
100.0 80.7 83.3 68.8
1,558 325
897 104
57.9 33.9
320 38
181 18
56.5 47.3
174 17
139 16
79.8 94.0
849 1,034
426 575
45.1 54.9
153 205
81 118
52.9 57.5
79 112
65 90
82.2 80.3
406 391 51 1,525
158 296 45 888
38.9 75.7 72.5 58.2
60 146 24 335
31 87 11 195
51.6 59.5 45.8 58.2
30 84 10 187
23 69 7 151
76.6 82.1 70.0 80.7
N Sociodemographicse Education (years) ,12 12 13–15 .16 Age (years) 20–34 35–44 45–64 65–74 Race/ethnicity Anglo Latino Gender Men Women Health status indicators Smoker Hypertensive Diabetic Dr. visit during the past year
Under physician-care to lower cholesterolc n
%
Prescribed lifestyle modification vs. pharmacologic treatmentd N
n
%
a
Data were derived from the treatment and control cities of the 1989 –1990 cross-sectional survey of the Stanford Five City Cardiovascular Disease Risk Reduction Study. b “N” refers to the number of men and women within each subgroup that responded to the cholesterol screening item; “n” refers to the corresponding number that reported “yes” to ever having a cholesterol screening. c “N” refers to the number of men and women within each subgroup that responded “yes” to ever having been screened and “yes” to being told by a doctor or medical professional of having elevated blood cholesterol; “n” refers to the corresponding number that reported “yes” to being under physician-care. d “N” refers to the number of men and women that responded “yes” to being prescribed life-style modification or pharmacologic treatment among the total number of survey participants that reported “yes” to ever having been screened, “yes” to being told by a doctor or medical professional of having elevated blood cholesterol, and “yes” to being under physician-care to lower blood cholesterol; “n” refers to the corresponding number that reported physician-prescribed lifestyle modification. e Mean 6 SD years of education for ever-screened for cholesterol, under physician-care to lower cholesterol, and prescribed lifestyle modification vs. pharmacologic treatment 5 14.2 6 3.4, 14.5 6 3.1, 14.8 6 3.2, respectively; mean 6 SD years of age 5 46.2 6 14.6, 53.6 6 14.0, 55.1 6 13.2, respectively.
Results Our study population included a total of 1,883 men and women. Figure 1 presents the sequential distribution of subgroups according to the study’s outcome measures. Approximately half of the respondents reported being screened for cholesterol. Approximately 36% of those screened and told by a doctor or medical professional to have an elevated cholesterol reported being under physician-care. The mean total cholesterol levels for this sub-group were 228 mg/dL, well above the target level of less than 200. Of those under physician-care, approximately 56% received physician-prescribed treatment. Roughly 22% reportedly were prescribed pharmacologic treatment and about 78% lifestyle modification. Table 1 presents sample sizes and percent distribu-
tions of the outcome measures by sociodemographics and health status indicators. Because very few gender differences were observed, we combined findings related to men and women. The sociodemographic data show that, among the total number of individuals that responded to the cholesterol screening item (i.e., ever screened for cholesterol), those with , 12 years of education reported fewer screenings than those within the higher SES sub-groups. Seventy percent fewer screenings were also reported among individuals in the youngest age-group (aged 20 –34); however, incremental increases were reported among individuals within each of the remaining age-groups. Compared to Anglos, approximately 24% fewer Latino respondents reported ever having been screened. Compared to men, the proportion of women screened was slightly
Am J Prev Med 1998;15(3)
181
Table 2. Odds ratios and 95% confidence intervals for adjusted sociodemographic and health status indicators that significantly predict cholesterol screening, physician-care, and type of physician-prescribed treatment (lifestyle modification versus pharmacologic) from multivariate logistic regression analyses, 1989 –1990a
Sociodemographics Education (years) Age (years) Ethnicity Gender Health status indicators Smoking status Hypertension Diabetes Dr. visit during past year Education 3 Age
Ever-screened for cholesterol
Under physician-care to lower cholesterol
Type of physicianprescribed treatment
OR
95% CI
OR
OR
95% CI
1.12 1.07 1.57 ...
1.08, 1.16 1.07, 1.09 1.14, 2.18 ...
... 1.02 ... ...
... 1.0, 1.03 ... ...
... 0.95 ... ...
... 0.92, 0.98 ... ...
2.07 1.77 ... 2.82 ...
1.58, 2.70 1.29, 2.40 ... 1.29, 3.74 ...
... ... ... 6.17 ...
... ... ... 2.04, 18.64 ...
... ... ... ... ...
... ... ... ... ...
95% CI
a
Data were derived from treatment and control cities of the 1989 –1990 cross-sectional survey of the Stanford Five City Cardiovascular Disease Risk Reduction Study. OR 5 odds ratio, CI 5 confidence intervals. The following coding scheme was used as predictor variables in the analyses: education (in years), age (in years), ethnicity (1 5 Anglo, 0 5 Latino), gender (1 5 male, 0 5 female), smoking status (1 5 current smoker, 0 5 nonsmoker), history of hypertension (1 5 yes, 0 5 no), history of diabetes (1 5 yes, 0 5 no), Dr. visit during past year (1 5 $ 1, 0 5 0). All sociodemographic and health status variables were offered in the models. Also included were all first-order interactions with education as a measure of socioeconomic status (including education 3 age) at P # .01. Ellipses indicate that the variable was not selected because P . .0100.
higher. Health status indicators show that hypertensives, diabetics, and respondents who visited a doctor during the past year were more likely to report ever having a cholesterol screening compared to smokers. The proportion of the study population reportedly under physician-care to lower cholesterol among the total number that responded to having a screening and were told by a physician or medical professional of an elevated cholesterol was fairly evenly distributed across SES. There were, however, age-group differentials with higher proportions among middle-aged (45– 64) and older (65–74) respondents. The distribution of Anglo and Latino respondents receiving care was fairly similar, as was that of men and women. Respondents reporting the various health status indicators were also equally likely to report being under physician-care to lower cholesterol. The data from the subsequent and final subgroup of respondents that reportedly were prescribed lifestyle modification or pharmacologic treatment show that the majority were prescribed lifestyle modification. For lower and higher SES individuals the proportion distribution was between 75% and 85%. However, consistent with the previous outcome measures, there were observable age-group gradient differentials. All respondents in the 20 –34 age group (n 5 18) reported being prescribed lifestyle modification compared to decreasing proportions among each of the remaining older age groups. The proportion of Anglo and Latino respondents reporting physician-prescribed lifestyle modification was fairly similar, as was that of men and women and among those according to the study’s health status indicators.
182
Outcome Findings Table 2 presents the results of the stepwise multiple logistic regression models. The data show that SES (i.e., level of education in years), age, and ethnicity significantly influenced the likelihood of a reported cholesterol screening; smoking status, history of hypertension, and physician-care utilization in the past year were significant health status predictors (all P # .01). Specifically, higher SES respondents were more likely to report ever being screened, as were older individuals, and Anglo respondents. Those that smoked and those that were hypertensive were almost twice as likely to report being screened compared to nonsmoking and normtensive respondents. Respondents that saw a physician within the year were almost three times more likely to report ever being screened for cholesterol compared to those that did not visit a physician during the past 12 months. There were no interaction effects with educational level, thereby suggesting the absence of an SES sub-group relationship. Age (odds ratio of 1.02) and doctor visit during the past year (odds ratio of 6.17) were the only significant factors influencing the likelihood of respondents with elevated cholesterol to report being under physiciancare to lower cholesterol (P # .01, respectively). Age (odds ratio of 0.95) was also the only significant predictor affecting type of physician-prescribed treatment among respondents under physician-care (P # .01). As Figure 2 illustrates all respondents aged 20 –34 reported being prescribed lifestyle modification instead of pharmacologic intervention; this proportion decreases with increasing age, but even the oldest age
American Journal of Preventive Medicine, Volume 15, Number 3
Figure 2. Percent distribution of age in years as a sociodemographic factor that significantly influenced the likelihood of having physician-prescribed lifestyle modification versus pharmacologic treatment, derived from multiple logistic regression analysis (P # .01, Tables 1 and 2).
group (65–74) was most likely to be prescribed lifestyle change.
Discussion The threefold goal of our study was to assess the relative impact of sociodemographic and health status factors on the likelihood of having a screening to detect evidence of elevated cholesterol, the likelihood among individuals with elevated levels to be under physiciancare for treatment, and the subsequent likelihood to be prescribed lifestyle modification or pharmacologic intervention. After controlling for gender and age, our findings revealed no significant differences in cholesterol screenings between men and women. There were, however, widening age-group differentials between younger and older individuals, thereby suggesting that younger men and women are equally less likely to be screened for elevated cholesterol. Our results also showed that age was a significant predicator of the likelihood to be under physician-care to lower cholesterol. Among those told to have elevated cholesterol by a medical professional, only 24% of young adults ages 20 – 44 reported to be under care for treatment, compared to 75% among middle-aged and older individual ages 45–74 (Table 1). This observation may be explained in part by the relationship between advancing age and subsequent plasma cholesterol level. Logistic regression analyses revealed middle-aged and older individuals were less likely to have cholesterol
levels lower than ,200 mg/dL, but rather more likely to have levels between 200 –239 mg/dL or .240 mg/dL greater (P , .01, P , .01, respectively). Age combined with differences in the distribution of borderline (i.e., 200 –239 mg/dL) and elevated (.240 mg/dL) cholesterol levels may have also influenced type of physician treatment within our study population. Approximately 38% of younger individuals had borderline levels among those reportedly under physician-care and prescribed treatment; 10% had levels .240 mg/dL. Fifty percent of older individuals, on the other hand, had borderline levels, and 32% had levels .240 mg/dL. Physicians may, therefore, be more inclined to prescribe pharmacologic therapy to older individuals because of an increased risk of coronary disease associated with age and higher prevalence of elevated cholesterol (Figure 2). Our analyses revealed a significant disparity in cholesterol screenings among respondents according to level of SES. Sixty percent of respondents with less than 12 years of education reported never having a cholesterol screening compared to 63% among those with $16 years of education (Figure 3). The lack of an SES interaction effect with physician visit during the past year suggest that upper, middle, and lower SES individuals were equally exposed to the possibility of a screening and thereby precludes lack of access as an explanatory factor for the difference. Other factors may include health insurance reimbursement issues associated with preventive health care or general lack of
Am J Prev Med 1998;15(3)
183
Figure 3. Percent distribution of socioeconomic status via level of education in years as a sociodemographic factor that significantly influenced the likelihood of having a cholesterol screening, derived from multiple logistic regression analysis (P # .01, Tables 1 and 2).
knowledge among lower SES individuals about the relationship of elevated cholesterol to cardiovascular disease.18 We also detected a significant gap in cholesterol screenings between Latino and Anglo respondents; 40% and 58%, respectively (Figure 4). These differences remained after controlling for SES. We also assessed physician utilization during the past year according to ethnicity and found no significant differ-
ences. This variable is a proxy measure for physician care, which may or may not reflect primary care visits for preventive care. Nonetheless, the lack of variation in patterns of utilization makes it less likely that disparities were due to a general lack of health care access. This suggests that Latino individuals, regardless of SES, and health care access, were less likely to be screened for detection of elevated cholesterol than Anglo individuals. Moreover, the findings imply the need for further
Figure 4. Percent distribution of ethnicity (Anglo/Latino) as a sociodemographic factor that significantly influenced the likelihood of having a cholesterol screening, derived from multiple logistic regression analysis (P # .01, Tables 1
184
American Journal of Preventive Medicine, Volume 15, Number 3
research to investigate other issues associated with disparities in health care utilization among Latinos. Of clinical importance was the observation that individuals who were smokers, unlike those with a history of hypertension, were also less likely to have been screened. Such a finding may reflect the general risktaking behavior of smokers and the need for specific intervention efforts to ameliorate the prevalence of multiple risk factors.
Caveats There are a number of caveats that require discussion regarding our study results. Because of the ethnic makeup and geographic location of the study population, our findings for instance may be tempered because they are based primarily on Anglo and Latino adults residing in four Northern California cities. However, there is little reason to believe that exposure in our study population was significantly different from other ethnically similar communities in the United States since most cholesterol education information is received through television, pamphlets, magazines, and other nationally distributed mass media.1 Approval of Lovastatin, the first HMG CO-A reductase inhibitor, occurred approximately 2–3 years after our 1989/1990 survey was conducted. Older cholesterol lowering medications were more difficult to prescribe and less tolerated by patients. However, advances in pharmacological technology over the past 5 years have introduced a variety of new drugs with less adverse side-effects. Therefore, our findings based on 1989 – 1990 information may not accurately reflect treatment intervention choices currently available to physicians. In addition, recent 1992 survey data reflects overall increases in the prevalence of cholesterol screening since implementation of the NCEP in 1985 and since our 1989 –1990 survey.19 However, unlike our results, information does not report on SES or race/ethnicity differentials. Further research is needed to substantiate and corroborate our findings. Another potential limitation is that the study variables were mostly self-reported questionnaire items. A major strength of the FCP database is the opportunity it offers to conduct secondary analyses on large numbers of population-based men and women for whom sociodemographic and physiologic measurements are available. Thus, the ability to verify self-reported data with physiologic measures mitigates potential reporting bias and enhances the generalizability of our findings.
Conclusions The National Cholesterol Education Program and the American Heart Association recommend that all adults aged 20 and older be screened for cholesterol.2–5
Guidelines recently developed by the American College of Physicians recommend screening only high-risk men aged 20 –34 and women aged 20 – 44 due to concern with costs, pharmacologic treatment, and potential health risks.6,7 The objective of our research was to provide insight into this ongoing cholesterol policy debate8 –11 by analyzing sociodemographic patterns of cholesterol screenings and the type of treatment prescribed by physicians. Our findings suggest that in 1990, approximately 5 years after implementation of the NCEP and 6 years before the recent conflicting opinions appeared, young adult men and women aged 20 – 44 were significantly less likely to have been screened for elevated cholesterol than middle-aged and older individuals. And among those under treatment, the majority, regardless of age, were prescribed lifestyle modification. In summary, our analysis of population-based data suggest that the ACP concerns may be unwarranted, or at least amenable to physician education efforts. They also suggest that NCEP’s policy objectives have been less effective among young adults. The lower prevalence of cholesterol screening among low SES and Latino respondents, on the other hand, may suggest the need for subsequent policy discussions regarding physician-level intervention efforts in order to ameliorate the potential continuation of health care utilization disparities. This work was supported by Public Health Service grant R01-HL-21906 from the National Heart, Lung, and Blood Institute to Stephen P. Fortmann, MD, and a Minority Scientist Career Development Award from the American Heart Association to Sharon K. Davis, PhD. We thank Helena Kramer, PhD, for statistical assistance and Leslie Power for preparation of tables and graphs.
References 1. Lefant C. A new challenge for America: The National Cholesterol Education Program. Circulation 1986;73: 855– 6. 2. National Cholesterol Education Program. Report of the Expert Panel on detection, evaluation, and treatment of high cholesterol in adults. Arch Intern Med 1988;148: 36 –9. 3. Summary of the Second Report of the National Cholesterol Education Program (NCEP): Expert Panel on detection, evaluation, and treatment of high blood cholesterol in adults (Adult Treatment Panel II). JAMA 1993; 269:3015–23. 4. National Research Council. Diet and health: implications for reducing chronic disease risk. Washington, DC: National Academy Press; 1989. 5. American Heart Association Task Force on Cholesterol Issues. The cholesterol facts: a summary of the evidence relating to dietary fats, serum cholesterol, and coronary
Am J Prev Med 1998;15(3)
185
6.
7.
8.
9. 10. 11.
12.
13.
186
heart disease: a joint statement by the American Heart Association and the National Heart, Lung, and Blood Institute. Circulation 1990;81:5. American College of Physicians. 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; 124:515–17. Garber AM, Brown WS, Hully SB. Cholesterol screening in asymptomatic adults, revisited. Ann Intern Med 1996; 124:518 –31. Task Force on Risk Reduction, American Heart Association: Cholesterol screening in asymptomatic adults: no cause to change. Circulation 1996;93:1067– 8. LaRosa JC. Cholesterol agnostics. Ann Intern Med 1996; 124:505– 8. Davidoff F. Evangelists and snail redux: the case of cholesterol screening. Ann Intern Med 1996;24:513–14. Toronto Working Group on Cholesterol Policy. Asymptomatic hypercholesteroemia: a clinical policy review. J Clin Epidemiol 1990;43:1028 –1121. Hulley SB, Newman TB, Grady D, Garber AM, Baron RB, et al. Should we be measuring blood cholesterol levels in younger adults? JAMA 1993;269:1416 –19. Nayor CD. Dyslipidaemias and the primary prevention of
14. 15.
16.
17. 18.
19.
coronary heart disease: reflections on some unresolved policy issues. In: Gold P, Grover S, Roncari DAK, eds. Cholesterol and coronary heart disease: the great debate. Park Ridge, New Jersey: Parthenon; 1992:401–25. LaRosa JC. Cholesterol and public policy. Atherosclerosis 1994;108:s137–s141. Farquhar JW, Fortmann SP, Maccoby N, et al. The Stanford Five-City Project: design and methods. Am J Epidemiol 1985;122:323–34. Farquhar JW, Fortmann SP, Flora JA, et al. Effects of communitywide education on cardiovascular disease risk factors. The Stanford Five-City Project. JAMA 1990;264: 359 – 65. Neter J, Wasserman W, Kutner MH. Applied linear statistical models. Boston: Richard D. Irwin, Inc.; 1990. Davis SK, Winkleby MA, Farquhar JW. Increasing disparity in knowledge of cardiovascular disease risk factors and risk-reduction strategies by socioeconomic status: implications for policymakers. Am J Prev Med 1995;11:318 –23. Pieper RM, Arnett DK, McGovern PG, Shahar E, Blackburn H, Luepker RV. Trends in cholesterol knowledge and screening and hyperchoesterolemia awareness and treatment, 1980 –1992. The Minnesota Heart Survey. Arch Intern Med 1997;157:2326 –32.
American Journal of Preventive Medicine, Volume 15, Number 3