Patient Smoking Cessation Advice by Health Care Providers: The Role of Ethnicity, Socioeconomic Status, and Health

Abstract

Objectives. We assessed differences by ethnicity in ever receiving advice from providers to quit smoking. We evaluated whether socioeconomic status and health status were moderators of the association.

Methods. We used 2000 Behavioral Risk Factor Surveillance Survey data, a population-based cross-sectional survey.

Results. After adjusting for complex survey design, 69% of the 14089 current smokers reported ever being advised to quit by a provider. Hispanics (50%) and African Americans (61%) reported receiving smoking counseling less frequently compared with Whites (72%, P<.01 for each). Ethnic minority status, lower education, and poorer health status remained significantly associated with lower rates of advice to quit after adjustment for number of cigarettes, time from last provider visit, income, comorbidities, health insurance, gender, and age. Smoking counseling differences between African Americans and Whites were greater among those with lower income and those without health insurance. Compared with Whites, differences for both Hispanics and African Americans were also greater among those with lower education.

Conclusion. We found lower rates of smoking cessation advice among ethnic minorities. However, we also found complex interactions of ethnicity with socioeconomic factors.

Smoking is the most preventable cause of death in developed countries and has been described as the leading behavioral health problem of our time, with a projected 8.4 million total smoking-related deaths by the year 2020.1 On the basis of recent data from the Behavioral Risk Factor Surveillance Survey (BRFSS), the overall smoking rate has not declined in the 1990s and has actually increased in some Southern states.2

The US Department of Health and Human Services has published clear guidelines on screening and counseling smoking cessation in clinical practice based on extensive review of the literature.3 The implementation of these guidelines has been shown to be cost-effective.4 Despite these guidelines, the frequency of smoking cessation interventions delivered by physicians has not changed substantially in the 1990s. A large-scale survey of physicians showed an increase in advice to quit between 1991 (20%) and 1993 (38%) and then a decrease in 1995 (29%).5

Smokers who are female, older, smoke more cigarettes per day, use the health system more frequently, and have a history of relevant medical problems are more likely to receive advice to quit smoking.5^–8 However, previous clinic-based studies are inconsistent with regard to income and other socioeconomic characteristics.7^–13 A recent study using data from the 1997 BRFSS found that, among 3818 smokers who had seen a provider within the past year, Hispanics reported tobacco counseling less frequently (63.4%) than non-Hispanic Whites (70.2%).14 This difference was not statistically significant after adjustment for age, gender, and educational status.14

The independent associations of patient factors with tobacco cessation counseling need to be explored in large, population-based samples of adult smokers. We used data from all smokers in the 2000 BRFSS to further the previous research. Our primary purpose was to assess differences in smoking cessation counseling rates by ethnicity when further accounting for health service use factors, including health insurance and time of most recent provider visit. Models of health care access, such as Andersen’s behavioral model of health care use, hypothesize that predisposing factors (e.g., age, race), enabling factors (e.g., health insurance, income), and health factors have complex interactions.15^–17 In previous health services research, differences by ethnicity were sometimes moderated by variables such as socioeconomic status or illness severity.18^–21 Thus, we also wanted to assess the moderating effect of socioeconomic and health status on differences in counseling by ethnicity.

METHODS

Study Design

The Centers for Disease Control and Prevention BRFSS is an ongoing, cross-sectional, telephone survey designed to collect uniform, state-based data on preventive health practices and risk behaviors linked to chronic diseases and injuries in the US population. This project represents a secondary analysis using the 2000 BRFSS dataset.22

Sample and Survey Content

For the BRFSS, states use telephone-based surveys to collect data from a representative sample of households with telephones. Surveys are administered in Spanish for Spanish-speaking participants. Most states use either a disproportionate stratified sample or a Mitofsky-Waksberg–type sample design to draw a random sample from the set of all possible telephone numbers based on area codes and prefixes.23 The 2000 BRFSS included a set of core questions asked of all adult (older than 18 years) respondents (N = 183 341). Individuals were identified as current smokers by the core question “Do you now smoke cigarettes every day, some days, or not at all?” Overall, 22.5% (N = 41 259) of adults reported current smoking.

A number of modular questions were asked of a subset of adult smokers. To assess smoking cessation advice by a health care provider, 14 623 adult current smokers were asked: “Has a doctor or other health professional ever advised you to quit smoking?” This question was asked in the states of Alaska, Colorado, Delaware, Indiana, Louisiana, Mississippi, Missouri, Montana, Nebraska, New Jersey, North Carolina, Ohio, Oklahoma, South Carolina, Texas, Virginia, West Virginia, Wisconsin, and Wyoming, and in the District of Columbia. Overall, 14 376 (98%) responded to the question. We excluded individuals of ethnic minorities other than Hispanic, Native American, or African American because of their small numbers (Asian = 91, Other = 126, missing race = 70). Thus, for our analysis, 14 089 current smokers were included.

The dataset also included a number of variables that may influence advice to quit smoking: comorbidities (e.g., asthma, diabetes mellitus) and health insurance, time from last health care provider visit, and self-reported health status.

Analysis

All means and proportions used for analyses were weighted to adjust for the complex survey design using the Stata statistical package (Stata Corp, College Station, Tex). First, we used Wald χ² tests for trend adjusted for complex sample design to assess overall prevalence of advice to quit smoking. We also used these tests to examine univariate relationships between advice to quit smoking and each variable of interest.

We then developed multivariable models to assess differences in the prevalence of smoking cessation advice by ethnicity, socioeconomic status, and health status by sequentially introducing covariates individually into logistic regression models. We based the order of variable entry on the basis of principles of hierarchical modeling described in Cohen and Cohen.24 After the variance associated with each variable was assessed, we assembled the variables into 3 incremental models. In hierarchical models, variables occurring earlier in time are introduced first, which follows a theoretical flow of causality.24 To adjust for confounding by recall bias, all models were adjusted for number of cigarettes smoked per day and time since last provider visit. Model 1 included these variables plus gender, age, and education. Model 2 included variables in Model 1 plus health (e.g., general health status and comorbid conditions, including asthma and diabetes) and socioeconomic factors (e.g., income, health insurance). Finally, Model 3 further included our a priori independent variable: ethnicity. Differences in explained variance between models were assessed using the c statistic (area under the receiver operating characteristic curve) and the −2 log likelihood statistic.25^,26

We then introduced interaction terms to assess for interaction between each covariate and the main independent variable (ethnicity) in Model 3. Final models were weighted for the complex survey design used in the BRFSS.

RESULTS

Overall, after adjusting for the complex survey design, among these 14089 current smokers, 69% reported ever being advised to quit smoking by a health care professional. Hispanics (50%) and African Americans (61%) reported ever being advised to quit less frequently than Whites (72%, P < .01 for each) (Table 1 ▶). Women and older patients were also more likely to receive counseling. Differences in health status, health services use, amount of smoking, and health insurance are summarized in Table 2 ▶. Hispanics and African Americans had poorer general health status and lower rates of health insurance coverage compared with Whites. A higher proportion of African Americans and Native Americans had diabetes. A higher proportion of Whites visited a health care provider within the past year, and Whites also smoked more cigarettes per day, compared with the ethnic minorities.

TABLE 1—

Smoking Cessation Advice, by Demographic Characteristics, Among US Smokers (n = 14 089): 2000 Behavioral Risk Factor Surveillance Survey

	Total No.	Received Advice No. (Weighted Percentage, %)a
Ethnicity*
Non-Hispanic White	11 360	8 310 (72)
African American	1 450	934 (61)
Hispanic	794	452 (50)
Native American	485	319 (72)
Age**
< 30	3 345	2 110 (59)
30–45	5 306	3 808 (70)
46–65	4 382	3 324 (75)
> 65	1 056	773 (76)
Gender**
Male	6 043	3 947 (63)
Female	8 046	6 068 (75)
Education**
Less than high school	2 256	11 544 (65)
High school	9 572	6 868 (69)
College	2 261	1 603 (71)
Incomeb
< $20 000	3 398	2 432 (68)
$20 000–$34 999	4 077	2 903 (67)
$35 000–$50 000	2 280	1 647 (71)
> $50 000	2 890	2 104 (71)

^a Weighted percentages adjusted for complex survey design.

^b Income available only for 12 653 (90%) of 14 089 subjects.

*P < .001 for each ethnic minority compared with Whites (Wald test adjusted for complex survey design).

**P < .01 (Wald test for trend, adjusted for complex survey design).

TABLE 2—

Health Status, Comorbidities, Number of Physician Visits, and Amount of Smoking by Ethnicity Among US Smokers (n = 14 089): 2000 Behavioral Risk Factor Surveillance Survey

	No. (Weighted Percent, %)a
	Non-Hispanic Whites	African Americans	Hispanics	Native Americans
Total no.b	11 360	1450	794	485
General health statusc,d
Poor	620 (5)	93 (7)	49 (6)	37 (5)
Fair	1 381 (11)	243 (17)	145 (22)	80 (13)
Good	3 820 (35)	500 (32)	290 (37)	181 (34)
Very good/excellent	5 528 (49)	613 (44)	308 (35)	186 (47)
Comorbid conditions
No asthma	10 031 (88)	1259 (86)	713 (91)	426 (81)
Asthma	1 318 (12)	191 (14)	80 (9)	58 (19)
No diabetes	10 885 (95)	1318 (90)	764 (97)	432 (88)
Diabetesc,e	467 (5)	130 (10)	29 (3)	50 (12)
Visited health care providerc
Within past year	7 129 (62)	1138 (77)	482 (57)	335 (67)
Within 1–2 y	1 451 (14)	140 (10)	119 (18)	66 (10)
Within 3–5 y	1 032 (9)	70 (5)	82 (9)	27 (9)
Within > 5 y	1 604 (14)	89 (7)	102 (16)	49 (13)
Amount of smokingc
≥ ½ pack per day	4 206 (36)	904 (65)	490 (69)	263 (47)
½–1 pack per day	4 988 (46)	416 (29)	222 (23)	155 (37)
> 1 pack per day	1 971 (18)	75 (5)	67 (8)	53 (16)
Health insurancec,d
No	2 328 (19)	375 (26)	277 (42)	135 (27)
Yes	9 005 (81)	1072 (74)	514 (58)	346 (73)

^a Weighted (column) percentages adjusted for complex survey design.

^b Total n varies slightly due to small number of missing data (less than 3%).

^cP < .01 (African American vs White) (Wald test, adjusted for survey design).

^dP < .01 (Hispanic vs White) (Wald test, adjusted for survey design).

^eP < .01 (Native American vs White) (Wald test, adjusted for survey design).

The multivariable models identified several independent effects associated with receipt of smoking cessation counseling (Table 3 ▶). Compared with Model 1, the introduction of health and socioeconomic variables in Model 2 increased the explained variance of smoking cessation counseling, as evidenced by the −2 log likelihood statistic and the c statistic (area under the receiver operating characteristic curve). In Model 2, those with the poorest health status were more likely to report ever receiving smoking cessation (odds ratio [OR] = 0.51; 95% confidence interval [CI] = 0.36, 0.72). Higher education was associated with reported smoking cessation advice.

TABLE 3—

Association of Smoking Advice Counseling, Ethnicity, Socioeconomic Factors, and Health Adjusting for Multiple Covariates in Incremental Logistic Regression Models

	Model 1	Model 2	Model 3 (Full Model)
	Adjusted OR (95% CI)	Adjusted OR (95% CI)	Standardized Coefficient	Adjusted ORa (95% CI)	t (P)
Gender
Men	Reference	Reference	Reference	Reference
Women	1.70 (1.51, 1.92)	1.75 (1.54, 2.00)	0.142	1.67 (1.47, 1.91)	7.7 (0.001)
Age (Per Year)	1.02 (1.01, 1.02)	1.01 (1.00, 1.02)	0.094	1.011 (1.01, 1.02)	4.6 (0.0001)
Education
< High school	Reference	Reference	Reference	Reference	2.8 (0.005)
High school	1.21 (1.02, 1.43)	1.26 (1.04, 1.52)	0.044	(0.97, 1.43)
College	1.47 (1.19, 1.80)	1.52 (1.20, 1.93)	0.07	(1.10, 1.79)
Last physician check-up
Within the past year	Reference	Reference	Reference	Reference	−8.0 (0.001)
Within 1–2 y	0.74 (0.62, 0.89)	0.73 (0.61, 0.88)	−0.047	0.75 (0.51, 1.11)
Within 3–5 y	0.71 (0.58, 0.87)	0.76 (0.61, 0.94)	−0.06	0.54 (0.38, 0.76)
Within > 5 y	0.45 (0.38, 0.54)	0.47 (0.39, 0.57)	−0.145	0.5 (0.35, 0.72)
Cigarettes per day
≤ ½ pack	Reference	Reference	Reference	Reference	11 (0.001)
½–1 pack	2.1 (1.8, 2.4)	2.2 (1.9, 2.5)	0.189	2.0 (1.74, 2.13)
>1 pack	2.9 (2.4, 3.5)	3.0 (2.5, 3.7)	0.201	2.73 (2.21, 3.37)
Income (per $10 000)		1.04 (1.00, 1.07)	0.027	1.02 (0.98, 1.06)	1.26 (0.21)
Health insurance
No health insurance		Reference	Reference	Reference	1.92 (0.055)
Health insurance		1.22 (1.03, 1.45)	0.039	1.19 (1.00, 1.43)
General health status
Poor		Reference	Reference	Reference	−4.2 (0.001)
Fair		0.73 (0.50, 1.07)	−0.052	0.76 (0.51, 1.10)
Good		0.54 (0.38, 0.77)	−0.161	0.54 (0.38, 0.76)
Very good or excellent		0.51 (0.36, 0.72)	−0.188	0.5 (0.36, 0.72)
Diabetes
No		Reference	Reference	Reference	2.4 (0.018)
Yes		1.58 (1.09, 2.29)	0.053	1.56 (1.08,2.26)
Asthma
No		Reference		Reference	4.7 (0.001)
Yes		1.74 (1.40, 2.16)	0.093	1.69 (1.36, 2.10)
Ethnicity
White			Reference	Reference
African American			− 0.056	0.72 (0.58, 0.89)	−3.21 (0.001)
Hispanic			−0.062	0.62 (0.47, 0.79)	−3.81 (0.001)
Native American			0.013	1.14 (0.68, 1.93)	0.5 (0.61)
c statistic	0.654	0.675		0.680
−2 log likelihood statistic	15498	13554		13507

Note. OR = odds ratio; CI = confidence interval.

^a Adjusted odds of receiving smoking cessation advice (adjusted ORs from incremented logistic regression weighted for complex survey design).

After adjustment for age, gender, education, income, health insurance, time of most recent physician visit, number of cigarettes per day, general health status, and comorbidities, African Americans (OR = 0.72; 95% CI = 0.58, 0.89) and Hispanics (OR = 0.61; 95% CI = 0.47, 0.79) were less likely to report ever having received smoking cessation advice compared with Whites (Model 3, Table 3 ▶). However, the absolute values of the standardized coefficients for ethnicity were less than those for health status, suggesting that ethnicity does not contribute as strongly to differences as health status. Also, note that the −2 log likelihood statistic and c statistic were essentially unchanged after introduction of ethnicity to the model.

We conducted tests of interaction between ethnicity and other socioeconomic and health factors. To further explore potential mechanisms for the differences in reporting receiving smoking cessation counseling, we conducted a series of analyses stratified by variables noted to have significant interactions with ethnicity (e.g., income, education, and insurance) (Table 4 ▶). On the basis of the adjusted odds ratios from these analyses, differences in report of smoking cessation counseling between African Americans and Whites were greater among those who did not have health insurance (OR = 0.45; 95% CI = 0.31, 0.67) and were no longer significant among those who did have health insurance (OR = 0.81; 95% CI = 0.63, 1.04). In addition, a trend toward wider differences was seen for African Americans compared with Whites as income decreased. In contrast, wider differences were seen at higher incomes for Hispanics. As education strata decreased from college to less than high school, a trend toward wider disparities was seen for both African Americans and Hispanics compared with Whites.

TABLE 4—

Association of Ethnicity With Receiving Smoking Cessation Advice: Stratified Analyses^a

		Adjusted Odds of Receiving Smoking Cessation Advice (Odds Ratio [95% Confidence Interval])
	No.b	African American	Hispanic	Native American	White
Health insurance (Medicare or private insurance)
Has insurance	10 937	0.81 (0.63, 1.04)	0.61 (0.45, 0.82)	1.30 (0.67, 2.45)	Reference
No insurance	3 114	0.45 (0.31, 0.67)	0.60 (0.37, 0.93)	0.95 (0.38, 2.38)	Reference
Income
> $50 000	4 334	0.74 (0.47, 1.14)	0.56 (0.35, 0.87)	1.11 (0.44, 2.82)	Reference
$20 000 to $50 000	6 357	0.64 (0.48, 0.84)	0.49 (0.35, 0.68)	0.90 (0.49, 1.63)	Reference
< $20 000	3 398	0.52 (0.37, 0.72)	0.74 (0.48, 1.13)	0.74 (0.35, 1.76)	Reference
Education
< High school	2 256	0.61 (0.38, 0.98)	0.44 (0.26, 0.72)	2.53 (0.91, 7.06)	Reference
High school	9 572	0.65 (0.50, 0.83)	0.61 (0.45, 0.83)	0.93 (0.50, 1.73)	Reference
College	2 261	1.04 (0.59, 1.83)	1.04 (0.42, 2.53)	1.29 (0.59, 2.81)	Reference

^a Each row contains adjusted odds ratios from a separate stratified multivariable logistic regression analysis adjusted for complex survey design. Covariates for all equations included age, gender, education, number of cigarettes per day, and comorbidities. Income, general health status, health insurance, and time of most recent physician visit were included when they were not the variable of stratification.

^b Total no. varies because of small amount of missing data.

DISCUSSION

Data from the BRFSS previously have been used to track trends in state and national rates of smoking.2 We used the BRFSS to conduct a multivariable analysis to further assess the factors independently associated with receipt of advice to quit smoking. African Americans and Hispanics were significantly less likely to receive smoking cessation advice from their health care providers. Self-reported poor health status and comorbid conditions were independently associated with increasing prevalence of tobacco cessation advice. These differences were robust to adjustment for multiple potential confounders, including time since last physician visit, health insurance coverage (or lack of it), age, and other sociodemographic variables.

Thus, our results provide additional evidence that differences exist in patient-reported receipt of advice to quit smoking. Our stratified analyses suggest that lack of health insurance is a strong moderator of this difference by ethnicity for African Americans but not for Hispanics. Educational attainment strongly affected the relationship between ethnicity and tobacco cessation counseling; the widest differences were seen for Hispanics with less than a high school education compared with Whites.

Significant medical conditions have been related to prevalence of tobacco advice.5^,7 Our study provides additional evidence that worse self-reported general health status is also independently associated with higher rates of tobacco cessation advice, even after adjustment for related factors, including frequency of health care provider visits and health insurance coverage. Calculated standardized coefficients showed that the impact of health status was greater than that of ethnicity. One perspective on our findings is that providers are better at counseling at higher rates in the secondary prevention arena and thus in those patients who need it most. Certainly in patients with known cardiovascular disease, for example, the effect of smoking cessation on subsequent mortality is high.27 However, smoking advice is believed to be cost-effective in the general clinical population.3 Unfortunately, tobacco cessation advice in healthier patients could be improved.

The lowest prevalence of advice for any demographic subgroup was seen among Hispanics. An absolute difference of 16% was seen between non-Hispanic Whites and Hispanics. This is arguably a clinically significant difference. The overall prevalence found in our study is consistent with the data reported by Denny et al. from 1997 BRFSS data.14 They also suggested a trend, although non-significant, toward lower rates of tobacco counseling among 199 Hispanic smokers who had seen a provider within the past year.14 Because our outcome was “ever advised to quit,” our sample of 14089 smokers (more than 700 Hispanics) included those who had seen a provider more than 1 year ago. We then adjusted for time since last provider visit. Thus, we have additional power to detect differences in prevalence.

Data obtained through the National Health Interview Survey in 2000 showed that Hispanic adults have a slightly lower prevalence of smoking (18.6%) compared with Whites (24.1%), African Americans (23.2%), and Native Americans (36%).28 Also, only 61.9% of Hispanic smokers reported wanting to quit compared with 71.1% of Whites, 68.4% of African Americans, and 69.8% of Native Americans.28 Perhaps the lower frequency of smoking cessation advice among Hispanics reflects differences in health care providers’ experiences or perceptions of the effectiveness or need for cessation advice in this ethnic subgroup. Further research is needed to address health care providers’ practices and the variables that determine their decision to counsel some smokers and not others.

With regard to income, we did not find any direct association with provider smoking counseling, although it did affect the association of ethnicity and smoking counseling. Previous results have been mixed. Taira et al. found that physicians tend to discuss diet and exercise with high-income patients and smoking with low-income patients.9 Frank et al. found a significantly positive trend in which higher income patients were more likely to receive advice to quit.7 However, Hymowitz et al. failed to detect income level differences with regard to health care providers’ delivery of smoking cessation intervention.8 These varying results suggest that socioeconomic status may have a complex association with other covariates and will need further study to confirm that income does not influence providers’ advice to quit smoking.

The most important limitation of this study is that the observational, cross-sectional nature of the data does not allow for demonstration of causality. Further, our main outcome, receipt of smoking cessation advice, was self-reported. This study focused on participants’ report only. As evidenced in the literature, there has been a great discrepancy between patients’ report of receiving advice to quit smoking from their providers and the providers’ report in providing such advice.6^,7^,29^–37 Specifically, gender (women are more accurate), smoking status (smokers are more accurate), and interval since doctor visit (more recent, more accurate) affect the accuracy of the patient recall.37 To reduce the influence of recall bias, we adjusted for length of time since last clinic visit in our multivariate analysis. However, it is still possible that differences in recall, not actual counseling, may account for our results. Although the BRFSS was administered in Spanish and has been documented to be reliable in Spanish-speaking samples,38 accurate translation does not necessarily create a culturally appropriate, and thus comparable, response.39 Thus, it is possible that a different interpretation of the questions among Hispanics resulted in the reported differences in counseling between Hispanics and non-Hispanics. Additional methods such as chart review or audiotapes should be used to further validate these differences in self-reported counseling. Unfortunately, our power to detect differences in counseling between Native Americans and Whites was limited by the small number of Native Americans in our sample. Additional studies should be targeted to further explore the experiences of this group.

The BRFSS assessed whether participants have received smoking cessation advice by a health care provider, making it impossible to distinguish between physicians, nurses, dentists, and the like. Future studies should address this issue, stratified by types of providers, so that tailored training through professional associations can be developed to address this gap. In addition, this study assessed whether participants received smoking cessation advice—not the amount, quality, or effectiveness of such advice. Nevertheless, we believe that this study has several strengths. First, the population-based nature of the BRFSS and the large number of current smokers available, which allowed for considerable adjustment for important covariates, are the major strengths of this study. Second, the study addresses a very important topic in the efforts to decrease health disparities in the United States. That is, this study identified an important segment of smokers who may not be getting smoking cessation advice from health care providers (e.g., healthier, less-educated, ethnic minority patients).

In conclusion, our results confirm previous studies that suggest that even after development and dissemination of smoking cessation guidelines in clinical practice, more than one quarter of adult smokers reported never having received smoking cessation advice from their health care providers. Lower health status and higher education, but not income, were associated with increasing prevalence of reports of receiving smoking cessation advice. The reasons for the differences in smoking cessation advice by ethnicity are unclear. Future studies should further examine language barriers and evaluate the interactions between income, education, health insurance, and ethnicity to determine whether the relationship between these variables and smoking cessation advice is similar across ethnicity. Other factors influencing multicultural differences, such as patient attitudes, discrimination, or stereotyping, also should be assessed. If clinicians, researchers, and policymakers are to make an impact on reducing health disparities, our results suggest that health care providers must specifically target smoking cessation efforts among younger healthy smokers, patients with low literacy, and ethnic minorities. Future interventions designed to improve health care providers’ practice patterns also should emphasize that these groups may be less likely to have received previous cessation advice by a health care provider and thus should be considered at-risk groups.