Colorectal Cancer Screening Participation: Comparisons With Mammography and Prostate-Specific Antigen Screening

Abstract

Objectives. The relation of personal characteristics, health and lifestyle behaviors, and cancer screening practices to current colorectal cancer (CRC) screening was assessed and compared with those factors' relation to current mammography screening in women and prostate-specific antigen (PSA) screening in men.

Methods. A cross-sectional random-digit-dialed telephone survey of 954 Massachusetts residents aged 50 and older was conducted.

Results. The overall prevalence of current CRC screening was 55.3%. Logistic regression results indicated that family history of CRC (odds ratio [OR] = 1.98; 95% confidence interval [CI] = 1.02, 3.86), receiving a regular medical checkup (OR = 3.07; 95% CI = 2.00, 4.71), current screening by mammography in women and PSA in men (OR = 4.40; 95% CI = 2.94, 6.58), and vitamin supplement use (OR = 1.87; 95% CI = 1.27, 2.77) were significant predictors of CRC screening.

Conclusions. Health and lifestyle behaviors were related to increased current CRC, mammography, and PSA screening. Personal factors independently related to CRC screening were not consistent with those related to mammography and PSA screening. This lack of consistency may reflect different stages of adoption of each type of screening by clinicians and the public.

Despite the availability of several effective screening tests that could substantially reduce colorectal cancer (CRC) morbidity and mortality, participation in screening is low; many Americans are diagnosed with and die from the disease each year.1 Available screening tests include the fecal occult blood test,2^,3 flexible sigmoidoscopy,4^,5 colonoscopy,6 and the double-contrast barium enema.7 The 1997 Behavioral Risk Factor Surveillance System8 indicated that 20% of age-eligible participants reported a fecal occult blood test within the previous year, 30% reported a flexible sigmoidoscopy within the previous 5 years, 41% reported having had one or the other of these tests, and 10% reported having had both. Rates in Massachusetts were slightly higher, with 28% reporting a fecal occult blood test within the previous year and 31% reporting a flexible sigmoidoscopy within the previous 5 years. Prevalence rates for mammography screening are higher than those for CRC screening.9 Among men, prostate-specific antigen (PSA) screening for prostate cancer is widespread,10^,11 despite equivocal recommendations as to the appropriateness of the test.12^,13 Research examining the relationships of common health-related behaviors with other screens, as well as relationships between types of screening, is sparse.

The purpose of this study was to investigate the prevalence of current CRC screening in a population-based sample of Massachusetts adults, and to examine the relationship of demographic variables, family history of CRC, health insurance status, health services use, lifestyle behaviors, and other screening use to current CRC screening. We compare and contrast the prevalence and predictors of CRC screening with the prevalence and predictors of mammography screening in women and of PSA screening in men.

METHODS

Study Design

This cross-sectional study used data from a standardized telephone interview conducted from June to August 1998. We used a random-digit-dialed telephone survey and the sampling method of Kish and Frankel14 to obtain a stratified sample consisting of a basic random sample, a male oversample, and a minority (Black and Hispanic) oversample. Eligible participants were Massachusetts residents 50 years and older who were cognitively able, resided in a home with a working telephone number, and had never been diagnosed with CRC. Men ever diagnosed with prostate cancer and women ever diagnosed with breast cancer were excluded from the sample because they might have been monitored under surveillance rather than screening protocols.

Data Collection and Survey Measures

Interviews were conducted with a computer-assisted telephone system. Interviewers received extensive training in instrument administration as well as technical aspects of the subject matter. As described below, measures were designed or adapted from previous work to capture screening and lifestyle and other personal characteristics. Pilot tests were conducted to refine the instrument.

Cancer Screening Practice Measures

The telephone interview collected information on participants' experiences with the fecal occult blood test kit completed at home and mailed back to the physician's office or laboratory for interpretation, with flexible sigmoidoscopy, with colonoscopy, and with the double-contrast barium enema. For the fecal occult blood test and flexible sigmoidoscopy, participants were asked if they had ever had the test, the length of time since their last test, and whether the test was for screening or diagnostic purposes. Persons who reported never receiving a fecal occult blood test were then asked if they had ever had a stool blood test in the physician's office and the time since their last test. Participants were asked if they had ever received a colonoscopy or double-contrast barium enema for screening or diagnostic purposes.

Owing to multiple screening options, it is difficult to determine CRC screening status. Persons were considered currently screened on the basis of recent guidelines15 if they reported having had 1 or more of the following: (1) a fecal occult blood test within the previous year; (2) flexible sigmoidoscopy within the previous 5 years; (3) colonoscopy within the previous 10 years; (4) a double-contrast barium enema within the previous 10 years. This definition included those who received a CRC test for screening as well as for diagnostic reasons. Those who reported being screened but beyond the time periods recommended by guidelines, and those who reported never being screened, were considered not current. Those who reported only receiving a fecal occult blood test performed by a physician during an office visit were considered not current, because this is not an acceptable screen according to guidelines.

For mammography and PSA screening, women and men were first asked if they had ever received the test appropriate for their sex. Those who answered yes reported the length of time since their most recent test. On the basis of accepted guidelines, for both procedures, being current was defined as having been screened within the previous year.16^,17

Personal Measures

Demographic information collected included sex, education, race/ethnicity, income, age, and marital status. Family histories of CRC, and of breast cancer in women and prostate cancer in men, were each classified as yes or no. Insurance and entitlement coverage was hypothesized to be a potential predictor of CRC screening, as it is for mammography.18^,19 Insurance status was classified as private, non–health maintenance organization (HMO) insurance with or without supplemental insurance; private, HMO insurance with or without supplemental insurance; Medicare, non-HMO; Medicare, HMO; Medicaid or other insurance; and uninsured.20

Health and Lifestyle Behavior Measures

Health service use and lifestyle behaviors were hypothesized to be associated with CRC screening. Access to a regular source of medical care enables other preventive services.21^,22 Participants were asked if they received a regular medical checkup, and if so, how often. A regular checkup was defined as a checkup every year or more often. Smokers are less likely to use many preventive services.23^–25 Participants were asked if they had ever smoked 100 cigarettes, and if so, if they were current or former smokers. Smoking status was classified as never, former, and current. Certain dietary supplements may have chemoprevention potential for some cancers.26 Although evidence is equivocal, the public has adopted considerable use of supplements,27 and participants were asked if they currently used any type of vitamin supplement (yes or no).

Statistical Analysis

Univariate analyses were used to describe the study sample and screening prevalence. Both in the total sample and in the analyses stratified by sex, bivariate χ² statistics and odds ratios, with 95% confidence intervals, were used to document the crude relationship between each independent variable and CRC screening status. Analyses were repeated with mammography and PSA as the outcome variables.

Logistic regression was used to model the association between each outcome measure and health and lifestyle behavior. Personal characteristics were considered potential confounders if they were marginally associated with the outcome or the exposure variables in bivariate analyses or were strongly associated in prior literature. Variables demonstrating sex differences in relation to CRC screening were tested for statistical interaction. Covariates were retained in the model if they were independent predictors of the outcome variable or if their inclusion in the model changed the odds ratio associated with an exposure variable by more than 15%. Data manipulation and frequency estimates were conducted with SAS Version 6.12.28 All other analyses were conducted with SUDAAN Version 7.52.29 To account for the stratified sampling scheme, we weighted analyses to represent the sex, race/ethnicity, and education distributions of Massachusetts.

RESULTS

Among 1747 eligible persons contacted, 1119 (64%) completed the survey, a response rate consistent with comparable population-based surveys.27^,30 We excluded from this sample 36 women with a personal history of breast cancer, 18 men with a personal history of prostate cancer, and 111 respondents who reported screening by either colonoscopy or double-contrast barium enema but did not specify time since screening, resulting in a final sample of 954. The sample is described in Table 1 ▶.

TABLE 1—

Characteristics of Participants (n = 954) in Study of Cancer Screening Practices

	n (%)
Personal Characteristics
Age, y
50–64	497 (52.6)
65–74	262 (27.8)
75–84	185 (19.6)
Race/ethnicity
White	863 (91.0)
Other	85 (9.0)
Educational level
< High school	312 (32.7)
High school graduate, 1–3 years	481 (50.5)
of college/trade school
College graduate or higher	160 (16.8)
Family history of colorectal cancer
Yes	103 (10.8)
No	851 (87.2)
Family history of breast cancer (women)
Yes	87 (16.0)
No	453 (84.0)
Family history of prostate cancer (men)
Yes	35 (8.3)
No	379 (91.7)
Sex
Male	414 (43.4)
Female	540 (56.6)
Marital status
Married/cohabitating	546 (57.4)
Divorced/separated	140 (14.7)
Widowed	219 (23.0)
Never married	47 (4.9)
Annual household income, $
<25 000	357 (37.4)
25 000–49 999	207 (21.7)
≥ 50 000	253 (26.6)
Missing data	136 (14.3)
Health insurance coverage
Private, non-HMO	117 (12.3)
+/– supplemental insurance
Private, HMO	283 (29.7)
+/– supplemental insurance
Medicare, non-HMO	240 (25.2)
Medicare, HMO	183 (19.2)
Medicaid/other	90 (9.4)
Uninsured	40 (4.2)
Health and Lifestyle Behaviors
Checkup at least every year
Yes	700 (73.7)
No	249 (26.3)
Vitamin supplement use
Yes	549 (57.6)
No	405 (42.4)
Smoking status
Current	157 (16.5)
Former	437 (45.9)
Never	359 (37.7)
Cancer Screening Practices
Colorectal cancer screening (total sample)
Current	527 (55.3)
Not current	427 (44.7)
Colorectal cancer screening (women)
Current	307 (56.9)
Not current	233 (43.1)
Colorectal cancer screening (men)
Current	220 (53.2)
Not current	194 (46.8)
Mammography screening (women)
Current	398 (73.9)
Not current	141 (26.1)
PSA screening (men)
Current	217 (52.4)
Not current	197 (47.6)

Note. +/– = with or without; PSA = prostate-specific antigen. Numbers may not add to 954 due to missing data.

Screening Prevalence and Relationship of Screens

Overall, 55.3% of participants were currently screened for CRC (i.e., current for 1 or more tests). Details of the screening experience of all respondents are reported elsewhere.31 Briefly, 36.2% of the sample reported current fecal occult blood test screening and 27.5% reported current flexible sigmoidoscopy screening, with 11.9% reporting both. Participation in colonoscopy and double-contrast barium enema was lower, with 8.5% and 6.3% reporting use in the past 10 years, respectively. Included in the notcurrently-screened group are the 9.7% of the sample that reported only receiving a stool blood test in their physician's office. Another 13.3% of respondents reported being screened in the past but not within the recommended interval, and 21.7% reported never having been screened for CRC. The majority of women (73.9%) and men (52.4%) reported mammography and PSA screening, respectively, within the recommended intervals.

Women who had had a mammogram in the past year were 4 times as likely (odds ratio [OR] = 4.01; 95% confidence interval [CI] = 2.26, 7.12) to be current for CRC screening as women who had not, while men who were current for PSA screening were almost 7 times as likely (OR = 6.73; 95% CI = 3.91, 11.57) to be current for CRC screening as men who were not (Table 2 ▶).

TABLE 2—

Relationship of Other Screening With Current Colorectal Cancer (CRC) Screening

	CRC Screening, n (%)	Odds Ratio (95% CI)	P
Mammography screening (women)
Current	261 (65.7)	4.01 (2.26, 7.12)	< .001
Not current	46 (32.3)	1.0
PSA screening (men)
Current	161 (74.2)	6.73 (3.91, 11.57)	< .001
Not current	59 (29.2)	1.0
Other screening (total sample)
Current	422 (68.7)	5.33 (3.7, 766)	< .001
Not current	104 (30.9)	1.0

Note. CI = confidence interval; PSA = prostate-specific antigen.

Factors Associated With Screening in Women

The bivariate associations of women's personal characteristics and health and lifestyle behaviors with current CRC and mammography screening are presented in Table 3 ▶. Although no personal characteristic achieved statistical significance, family history of the disease, lower education, and being a Medicare HMO member were associated with higher CRC screening rates. Receiving a regular checkup (OR = 4.22; 95% CI = 2.41, 7.41) and using vitamin supplements (OR = 2.12; 95% CI = 1.31, 3.41) were significantly associated with increased CRC screening. While smoking status did not reach statistical significance, current smokers had the lowest screening rate.

TABLE 3—

Bivariate Relationship of Women's Personal Characteristics and Health and Lifestyle Behaviors With Current Colorectal Cancer (CRC) and Mammography Screening

	CRC Screening, Women (n = 540)			Mammography Screening (n = 540)
	Screened, n (%)	OR (95% CI)	P	Screened, n (%)	OR (95% CI)	P
Age, y
50–64	139 (56.3)	1.0	.96	189 (76.9)	1.0	.04
65–74	93 (58.2)	1.08 (0.62, 1.87)		130 (80.8)	1.26 (0.64, 2.50)
≥ 75	72 (57.1)	1.04 (0.56, 1.92)		74 (59.6)	0.44 (0.23, 0.86)
Race/ethnicity
White	272 (56.1)	1.0	.46	351 (72.4)	1.0	.03
Other	34 (63.2)	1.34 (0.62, 2.92)		46 (87.0)	2.55 (1.10, 5.90)
Marital status
Married/cohabitating	144 (57.7)	1.0	.65	195 (77.9)	1.0	.35
Divorced/separated	51 (63.0)	1.25 (0.65, 2.39)		61 (76.4)	0.92 (0.43, 1.95)
Widowed	97 (54.2)	0.87 (0.50, 1.50)		118 (66.4)	0.56 (0.30, 1.05)
Never married	14 (49.5)	0.72 (0.32, 1.62)		23 (77.4)	0.97 (0.37, 2.53)
Annual household income, $
<25 000	130 (56.1)	1.0	.57	167 (72.2)	1.0	.02
25 000–49 999	63 (60.4)	1.20 (0.63, 2.27)		77 (74.2)	1.11 (0.55, 2.22)
≥50 000	54 (51.0)	0.82 (0.47, 1.43)		90 (85.7)	2.32 (1.16, 4.61)
Missing data	60 (61.2)	1.24 (0.62, 2.48)		64 (64.8)	0.71 (0.33, 1.52)
Educational level
< High school	112 (62.8)	1.80 (0.92, 3.53)	.20	122 (68.6)	0.67 (0.32, 1.41)	.53
High school graduate, 1–3 years of college/trade school	160 (55.2)	1.31 (0.84, 2.05)		231 (76.4)	1.00 (0.59, 1.69)
College graduate or higher	34 (48.4)	1.0		53 (76.4)	1.0
Health insurance coverage
Private, non-HMO +/– supplemental insurance	31 (49.7)	1.0	.68	52 (82.7)	1.0	.38
Private, HMO +/– supplemental insurance	71 (54.4)	1.20 (0.55, 2.68)		100 (76.8)	0.69 (0.25, 1.93)
Medicare, non-HMO	84 (55.2)	1.25 (0.55, 2.83)		97 (64.1)	0.37 (0.13, 1.05)
Medicare, HMO	77 (66.3)	1.99 (0.82, 4.83)		91 (78.5)	0.77 (0.25, 2.36)
Medicaid/other	28 (46.2)	0.98 (0.36, 2.66)		43 (75.0)	0.63 (0.17, 2.25)
Uninsured	12 (56.8)	1.33 (0.26, 6.91)		15 (70.7)	0.50 (0.10, 2.60)
Family history of screen-specific cancer
Yes	37 (65.7)	1.51 (0.73, 3.15)	.25	63 (73.0)	0.95 (0.46, 1.99)	.88
No	234 (55.8)	1.0		335 (74.0)	1.0
Checkup at least every year
Yes	235 (65.6)	4.22 (2.41, 7.41)	< .001	333 (82.4)	4.95 (2.73, 8.97)	< .001
No	35 (31.1)	1.0		66 (48.6)	1.0
Smoking status
Current	42 (48.3)	0.78 (0.38, 1.58)	.14	57 (65.0)	0.68 (0.32, 1.41)	.21
Former	122 (64.1)	1.48 (0.89, 2.46)		150 (78.9)	1.36 (0.73, 2.50)
Never	142 (54.6)	1.0		191 (73.4)	1.0
Vitamin supplement use
Yes	218 (63.5)	2.12 (1.31, 3.41)	.003	268 (77.8)	1.74 (1.01, 3.02)	.05
No	85 (40.2)	1.0		130 (66.8)	1.0

Note. OR = odds ratio; CI = confidence interval; +/– = with or without.

For mammography, higher income (OR = 2.32; 95% CI = 1.16, 4.61), being 65 to 74 years of age (OR = 0.44; 95% CI = 0.23, 0.86), and non-White race (OR = 2.55; 95% CI = 1.10, 5.90) were associated with higher rates of screening. Persons who received regular checkups (OR = 4.95; 95% CI = 2.73, 8.97) and vitamin supplement users (OR = 1.74; 95% CI = 1.01, 3.02) were more likely to be currently screened.

Factors Associated With Screening in Men

Men aged 65 to 74 were more likely to be currently screened for CRC than men aged 50 to 65 (OR = 2.57; 95% CI = 1.39, 4.74) (Table 4 ▶). Lower educational level was associated with decreased likelihood of current CRC screening (OR = 0.40; 95% CI = 0.20, 0.81). Members of Medicare HMO plans had nonsignificantly higher CRC screening rates than men with private insurance. Men with a family history of CRC (OR = 2.78; 95% CI = 1.34, 5.77) and those who received a regular checkup (OR = 5.47; 95% CI = 3.06, 9.77) had higher screening rates. Vitamin supplement users had a nonsignificantly higher rate of screening. CRC screening was less likely among current smokers than among never smokers (OR = 0.36; 95% CI = 0.16, 0.82).

TABLE 4—

Bivariate Relationship of Men's Personal Characteristics and Health and Lifestyle Behaviors With Current Colorectal Cancer (CRC) and Prostate-Specific Antigen (PSA) Screening

	CRC Screening, Men (n = 414)			PSA Screening (n = 414)
	Screened, n (%)	OR (95% CI)	P	Screened, n (%)	OR (95% CI)	P
Age, y
50–64	117 (46.5)	1.0	.009	117 (45.9)	1.0	.008
65–74	70 (69.1)	2.57 (1.39, 4.74)		70 (69.5)	2.61 (1.42, 4.82)
≥75	33 (55.5)	1.44 (0.63, 3.26)		49 (49.4)	1.12 (0.50, 2.50)
Race/ethnicity
White	202 (53.3)	1.0	.89	201 (53.1)	1.0	.88
Other	17 (54.5)	1.05 (0.53, 2.09)		16 (51.8)	0.95 (0.48, 1.89)
Marital status
Married/cohabitating	159 (54.0)	1.0	.90	161 (54.7)	1.0	.74
Divorced/separated	28 (47.4)	0.77 (0.36, 1.66)		26 (43.8)	0.65 (0.30, 1.39)
Widowed	23 (56.7)	1.12 (0.43, 2.89)		20 (50.0)	0.83 (0.34, 2.05)
Never married	10 (55.4)	1.06 (0.44, 2.57)		10 (53.5)	0.95 (0.40, 2.30)
Annual household income, $
<25 000	62 (49.6)	1.0	.20	64 (51.7)	1.0	.77
25 000–49 999	49 (47.2)	0.91 (0.44, 1.87)		59 (57.5)	1.27 (0.61, 2.61)
≥50 000	87 (55.9)	1.62 (0.86, 3.05)		79 (50.8)	0.98 (0.52, 1.85)
Missing data	18 (47.1)	0.98 (0.37, 2.58)		17 (45.9)	0.79 (0.30, 2.10)
Educational level
< High school	57 (42.7)	0.40 (0.20, 0.81)	.02	63 (47.3)	0.87 (0.44, 1.72)	.38
High school graduate, 1–3 years of college/trade school	105 (55.1)	0.67 (0.44, 1.01)		108 (56.8)	1.27 (0.85, 1.91)
College graduate or higher	58 (64.8)	1.0		45 (50.8)	1.0
Health insurance coverage
Private, non-HMO +/– supplemental insurance	29 (53.0)	1.0	.03	30 (54.3)	1.0	.03
Private, HMO +/– supplemental insurance	77 (50.2)	0.89 (0.42, 1.92)		72 (46.8)	0.74 (0.34, 1.59)
Medicare, non-HMO	53 (59.7)	1.31 (0.54, 3.17)		54 (61.8)	1.36 (0.56, 3.30)
Medicare, HMO	47 (70.0)	2.07 (0.76, 5.67)		43 (64.7)	1.54 (0.57, 4.17)
Medicaid/other	14 (42.7)	0.66 (0.35, 1.88)		17 (53.7)	0.98 (0.34, 2.83)
Uninsured	1 (5.9)	0.06 (0.01, 0.26)		1 (4.0)	0.04 (0.01, 0.20)
Family history of screen-specific cancer
Yes	34 (74.0)	2.78 (1.34, 5.77)	.008	20 (55.6)	1.15 (0.52, 2.56)	.73
No	186 (50.6)	1.0		198 (52.1)	1.0
Checkup at least every year
Yes	188 (63.8)	5.47 (3.06, 9.77)	< .001	192 (65.2)	6.64 (3.61, 12.20)	< .001
No	28 (24.4)	1.0		25 (22.0)	1.0
Smoking status
Current	26 (37.1)	0.36 (0.16, 0.82)	.07	31 (44.5)	0.74 (0.32, 1.69)	.56
Former	133 (54.2)	0.73 (0.42, 1.27)		135 (54.9)	1.12 (0.64, 1.97)
Never	61 (61.9)	1.0		51 (52.0)	1.0
Vitamin supplement use
Yes	121 (59.1)	1.60 (0.97, 2.65)	.07	118 (57.9)	1.54 (0.93, 2.54)	.09
No	99 (47.4)	1.0		99 (47.1)	1.0

Note. OR = odds ratio; CI = confidence interval; +/– = with or without.

With respect to PSA screening, men aged 65 to 74 years were more likely to be currently screened (OR = 2.61; 95% CI = 1.42, 4.82). Receiving a regular checkup (OR = 6.64; 95% CI = 3.61, 12.20) was positively associated with screening, and there was a nonsignificant increase in screening among vitamin supplement users.

Multivariate Analyses

CRC screening.

The final multivariate model is presented in Table 5 ▶. While neither sex nor education had an independent effect, an interaction between education and sex emerged. Compared with men with a college degree or higher, those with less than a high school diploma (OR = 0.11; 95% CI = 0.04, 0.32) and men with high school diplomas or some college or trade school (OR = 0.31; 95% CI = 0.15, 0.64) were less likely to be currently screened. It should be noted that odds ratios for the main effects of sex and education are not interpretable independently of the interaction odds ratio.32

TABLE 5—

Multiple Logistic Regression Model Predicting Current Colorectal Cancer (CRC) Screening

Overall Model	OR	95% CI	P
Sexa
Male	4.12	2.28, 7.43	< .001
Female	1.0
Educationa
<High school	2.77	1.25, 6.13	.01
High school graduate, 1–3 years of college/trade school	1.59	0.93, 2.73	.09
College graduate or higher	1.0
Education-by-sex
Male × <high school	0.11	0.04, 0.32	.001
Male × high school graduate, 1–3 years of college/trade school	0.31	0.15, 0.64	.002
Male × college graduate or higher	1.0
Health insurance coverage
Private, non-HMO +/– supplemental insurance	1.0
Private, HMO +/– supplemental insurance	1.13	0.63, 2.02	.69
Medicare, non-HMO	1.38	0.73, 2.61	.33
Medicare, HMO	2.25	1.13, 4.46	.02
Medicaid/other	0.80	0.35, 1.84	.59
Uninsured	0.63	0.21, 1.94	.42
Family history of CRC
Yes	1.98	1.02, 3.86	.04
No	1.0
Current screening for other cancer
Yes	4.40	2.94, 6.58	< .001
No	1.0
Checkup at least every year
Yes	3.07	2.00, 4.71	< .001
No	1.0
Vitamin supplement use
Yes	1.87	1.27, 2.77	.02
No	1.0

Note. OR = odds ratio; CI = confidence interval; +/– = with or without.

^aOdds ratios for these variables must be modified to account for the interaction term present in the model and should not be considered independently.

Family history of CRC was an independent predictor of screening (OR = 1.98; 95% CI = 1.02, 3.86). Although crude estimates of the effect of family history were stronger in men, there was not a statistically significant sex interaction. Compared with those with private, non-HMO insurance, members of Medicare HMO plans were significantly more likely to be currently screened (OR = 2.25; 95% CI = 1.13, 4.46).

Men and women who were currently screened for PSA and mammography, respectively, were more likely to be screened for CRC than those who were not (OR = 4.40; 95% CI = 2.94, 6.58). Those who had a regular checkup (OR = 3.07; 95% CI = 2.00, 4.71) and vitamin supplement users (OR = 1.87; 95% CI = 1.27, 2.77) had higher screening rates. Sex-stratified models predicting current CRC screening were also fit; because no further differences were observed, these models are not presented.

Mammography and PSA screening.

For mammography, women 75 years or older had a lower likelihood of current screening (OR = 0.45; 95% CI = 0.19, 1.02). Having a regular checkup was also associated with increased screening (OR = 5.67; 95% CI = 3.03, 10.61). There was a suggestion of increased likelihood of screening among vitamin supplement users and those having an annual household income of greater than $50 000.

For PSA, men aged 64 to 75 years were more likely to be screened than younger men (OR = 2.61; 95% CI = 1.45, 4.70). Receiving a regular checkup was significantly associated with increased screening (OR = 6.35; 95% CI = 3.30, 12.20). Although not statistically significant, vitamin supplement use was also associated with current PSA screening.

DISCUSSION

The diffusion of CRC screening technologies represents an interesting contrast to that of mammography and PSA screening. Many factors have contributed to wide diffusion, adoption, and institutionalization of mammography screening by health plans, clinicians, and the public over the last decade. The designation of breast cancer as a national public health priority,33 a major cancer control research initiative,34 promotion by public advocacy groups, studies confirming the effectiveness of screening,35 development of supportive state policies and legislation,36 wide promulgation of practice guidelines,37 and widely promoted health plan and physician performance standards have all played important roles.38

In contrast, the issue of PSA screening for prostate cancer has been more controversial and driven largely by patient demand39^,40 amid concern about prostate cancer as a major health issue, particularly for African Americans.41 While the effectiveness of PSA screening in improving prostate cancer survival continues to be debated, low cost, ease of performance, and fear of cancer have contributed to the popularity of the test.

Credible guidelines have only recently been developed and promoted for CRC screening, and the recommendations are complex, with numerous testing options. Health plan and provider performance is more difficult to measure, given the number of screening options and the fact that a person may be “currently screened” on the basis of a test conducted several years earlier for diagnostic purposes. While health plans aggressively educate members about mammography and have developed sophisticated in-reach (e.g., tracking and physician cuing) and outreach programs, similar efforts are not yet applied to CRC and prostate cancer detection in most plans. Patient perceptions of CRC screening are frequently negative, particularly in contrast to the ease of acceptance of the PSA blood test. Findings from this study must be viewed in the context of the distinct histories and characteristics of the 3 cancer screening efforts we investigated.

This study is among the first few to investigate the prevalence of CRC screening according to recent guidelines.15 Overall, the prevalence of current CRC screening was modest (55.3%). The prevalence of PSA screening among men in the sample was similar to the rate of CRC screening (52.4%), while the prevalence of mammography screening was considerably higher (73.9%). Estimates of rates of fecal occult blood testing, flexible sigmoidoscopy, and mammography in this study were similar to those observed in the Massachusetts Behavioral Risk Factor Surveillance System.8^,9 No prevalence estimate is currently available for PSA screening in Massachusetts; however, other population-based studies10^,11^,42 have reported PSA screening rates of 25% to 31%. Given the absence of a consensus on PSA screening in the medical community,12^,13 our findings suggest a strong consumer demand for the test.

We observed interesting similarities and differences among screens. As previously noted, a strong association was found between each cancer screen and receiving a regular checkup.10^,43 Two explanatory factors are probably operating: similar attitudes and predisposing factors motivate people to get checkups and screening, and checkups provide clinicians with the opportunity to promote screening.

We found that family history of CRC is a powerful stimulus to screening for both men and women.44 Although receiving mammography screening was associated with family history of breast cancer in earlier studies,45 we did not find a significant association. This suggests a ceiling on the effect of family history when screening rates approach 70% among women without a family history. The absence of a substantial effect of a family history of prostate cancer on PSA screening may be due to the fact that family history has not been widely promoted as a risk factor.

In the bivariate analysis, higher income level was associated with mammography and PSA use, but not with CRC screening. The lack of significance of income in the multivariate models is probably due to the complex relationships among income, education, employment, and insurance coverage, which cannot be modeled precisely without a very large sample. The sample size was also inadequate to detect differences between racial/ethnic subgroups. Future study of these population subgroups is needed to assess possible differences.

We found that those 65 years or older in Medicare HMOs were most likely to be screened for CRC. Bivariate analyses demonstrated similarly high rates of screening for PSA for the Medicare HMO groups. This may reflect differences in patient mix among systems of care,46 organized administrative systems that enable screening,47 or benefits and copayment policies.48

Among men, age differences in CRC screening were attributable to high enrollment among men aged 65 to 74 years in Medicare HMO plans. For PSA, however, men aged 65 to 74 years had increased screening, possibly highlighting consumer-driven interest in this disease among this age group. The relatively low rate of PSA screening in the oldest group may reflect the general consensus among proponents of PSA that screening men older than 75 is not likely to improve survival.

Although there were no age differences among women with respect to CRC screening, the oldest women were less likely to be screened by mammography, a finding demonstrated previously.49 Among the oldest women in this study, there was a particularly strong association between participation in mammography and CRC screening. This, along with the similarity of screening prevalence for mammography and CRC among the oldest women, suggests one group of older women who adopt new screening technologies quickly and another group of women unlikely to accept screening even after years of exposure to promotional messages. If the diffusion of CRC screening follows the pattern of mammography, we might expect the rate among younger women to rise considerably and the rate in the oldest group to remain stable or increase more slowly.

In the overall sample, men and women had similar rates of CRC screening and there was no significant association between CRC screening and educational level. However, among men, there were differences in screening rates according to educational level, with decreasing educational level associated with less screening. No such association was observed among women. This issue is worthy of more investigation, as it may reflect patterns of care-seeking behavior or patient–clinician interactions that vary by combinations of sex and educational level.50

The findings related to health and lifestyle behavior suggest an overall constellation of health-promoting behavior, including supplement use and care seeking. We also found, as did Patterson et al.,27 that supplement use was an independent predictor of mammography and of PSA and CRC screening. Nonsmoking has been associated with mammography screening24^,51; we identified modest but statistically insignificant higher rates of screening among nonsmokers for all 3 screening tests. Smoking behavior has been associated with less use of medical services.25^,52 We found that whereas 76% of nonsmokers received regular checkups, only 64% of smokers did. However, future study that more precisely measures smoking intensity is needed to sufficiently address the question of how smoking status influences CRC screening behavior.

As with any random-digit-dialed survey, there are acknowledged limitations to this study. Relying on self-reported data yields possible bias due to underestimating the length of time since screening examinations and checkups,53^–55 as well as inaccuracies in reports of type of insurance (e.g., HMO vs non-HMO).56 Social desirability could also result in bias in reporting lifestyle characteristics. As with other studies, we found it difficult to achieve a high response rate in this era of intrusive telephone marketing, answering machines, and public distrust.30

Overall, within sex groups, health and lifestyle behaviors affecting screening prevalence were similar for CRC and mammography and for CRC and PSA screening. Interest in prevention as measured by checkups, nonsmoking, and use of supplements defines a group of consistently health-conscious individuals very likely to adhere to screening. Differences based on personal characteristics may be attributable to differences based on the history of diffusion of the test or on peculiarities of the promotion of the test. Implications of our findings for public health and clinical practice are that screening may “beget” screening and that efforts to improve access to preventive care in general should be promoted. Working with individuals to accept one screening test may lead to a shift in attitudes, making them open to other tests.