- Email: [email protected]
Exploring the association between perceived risks of smoking and benefits to quitting
Addictive Behaviors 27 (2002) 293 – 307
Exploring the association between perceived risks of smoking and benefits to quitting Who does not see the link? Pauline Lynaa,*, Colleen McBridea,b, Greg Samsab,c, Kathryn I. Pollaka,b a
Duke Comprehensive Cancer Center, Durham, NC, USA b Duke University Medical Center, Durham, NC, USA c Durham Veteran’s Administration Medical Center, Durham, NC, USA
Abstract This report explored associations between different measures of smokers’ perceived risks of smoking and benefits to quitting and the extent to which these associations varied by demographic and other characteristics for 144 smokers. We hypothesized greater perceived risk of smoking would be associated with greater perceived benefits to quitting and would be strongest among smokers who were concerned about health effects of smoking and motivated to quit. Results indicated smokers’ perceived themselves at risk for lung cancer regardless if they continued or quit smoking and was strongest for smokers who were older and minimized the importance of reducing lung cancer risk. There was a weak correlation between perceived risk for lung cancer when compared to nonsmokers and perception that quitting smoking would reduce lung cancer risk and was weakest for African Americans, lighters smokers, and smokers with higher intrinsic relative to extrinsic motivation for cessation. In conclusion, these subgroup differences in the relationship between perceptions of risks and benefits could be important to consider to increase the relevance and motivational potency of smoking cessation interventions. D 2001 Elsevier Science Ltd. All rights reserved. Keywords: Smoking cessation; Perceived risks; Health behavior; Motivation
* Corresponding author. Duke University Medical Center, DUMC, Hanes House, Trent Drive, Box 2949, Durham, NC 27710-2949, USA. Tel.: +1-919-681-4559; fax: +1-919-681-4785. E-mail address: [email protected] (P. Lyna). 0306-4603/02/$ – see front matter D 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 3 0 6 - 4 6 0 3 ( 0 1 ) 0 0 1 7 5 - 7
294
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
1. Introduction Cigarette smoking is the leading preventable cause of death in the United States and is forecasted to be an ongoing public health problem. Following two decades of decline, rates of smoking have remained relatively stable, hovering around 25% for the past 10 years. National surveys indicate that 70% or more of smokers report that smoking is hazardous to health and some reports suggest that smokers even overestimate the health risks of smoking (Viscusi, 1990). Smokers cite concerns about the harms of smoking as primary motivators for them to consider cessation (Duncan, Cummings, Hudes, Zahnd, & Coates, 1992; Halpern & Warner, 1993). Yet, many continue to smoke despite these concerns. How is it that smokers can be aware of the substantial harms to smoking and yet continue to smoke? This apparent discordance in awareness and behavior is of particular concern because the harms of smoking and benefits of quitting are a major motivational focus of most smoking cessation programs. Thus, a better understanding of this discordance could be informative for improving cessation interventions. Popular conceptual models (Strecher & Rosenstock, 1997) suggest a cascade of factors that influence smoking cessation. Generally, these models posit that for smoking cessation to occur, an individual must perceive personal vulnerability to related negative outcomes (Bandura, 1977; Janz & Becker, 1984; Rosenstock, 1974), must see the outcomes as severe, and believe that quitting smoking will reduce the likelihood of their occurrence. Further, it has been argued that concepts of perceived vulnerability to health effects, and severity of risk are relatively complex domains (Weinstein, 1998). For example, perceptions of risk can vary based on whether one is comparing oneself to a specific reference group (e.g., similar in age, race, or smoking status), the time frame that is being considered (lifetime, 10 years, etc.), and conditional on whether one quits or continues to smoke (Klein & Weinstein, 1997; Ronis, 1992; Weinstein, 1980; Weinstein & Klein, 1995). Thus, it has been argued that smokers may acknowledge risks of smoking to generalized others, yet at the same time not fully acknowledge their own personal vulnerability to these risks (Rothman, Klein, & Weinstein, 1996; Strecher, Kreuter, & Kobrin, 1995; Weinstein, 1988). The majority of these studies have focused on the accuracy of smokers’ risk perceptions rather than the extent to which these perceptions vary with demographic and other smoking-related characteristics. Consideration of the characteristics of smokers who do and do not see the link between their smoking and potential health harms could be informative in guiding the development of effective smoking cessation programs. For example, if a light smoker perceives less harm to smoking and in turn, fewer benefits to quitting, regardless of the reference group than a heavier smoker, then health messages for this population may be best focused on countering these misperceptions. In turn, heavier smokers who may see great harm to smoking but few benefits to quitting may benefit from more optimistic representations of the benefits of smoking cessation. Thus, understanding the multidimensionality of the link between perceived risks and benefits and how these vary among subgroups of smokers could influence how risk messages are communicated to smokers. Smokers’ perceived benefits of smoking cessation also may be influenced by other factors such as their level of concern or worry about smoking-related health outcomes and their confidence that they could quit if they tried. For example, the Health Belief and other models
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
295
(Bandura, 1977; Strecher & Rosenstock, 1997) would suggest that perceived risks and benefits might be most strongly associated among those who are especially concerned about suffering the health effects of smoking than among those who are relatively unconcerned. Exploring conceptually meaningful associations among subgroups of smokers could give greater insight to the development of motivational interventions for smokers. To date, few studies have included the broad array of measures of perceived risks, benefits, and related concerns that are needed to better understand possible interrelationships. With the above in mind, this report asked two questions. First, is there an association between comparative and conditional risk for lung cancer and perceived benefits of quitting smoking for reducing these risks? We hypothesized that higher perceived risk regardless of the conditional comparison would be associated with greater perceived benefits to quitting. The second question asked, whether associations among risk perceptions and perceived benefits of quitting were influenced by demographics, smoking patterns, motivation, and self-efficacy for cessation or concerns related to the disease outcome? We hypothesized that associations among perceived risks and benefits would be greatest among smokers who were most concerned about health effects, and those who were most motivated and most confident they could quit.
2. Methods 2.1. Study recruitment As part of a pilot study conducted in preparation for a randomized intervention trial, smokers were recruited via newspaper and public service radio advertisements and recruitment tables located at various community and medical center locations. A detailed description of the pilot study is provided elsewhere (McBride et al., 2000). Study advertisements stated that smokers were being recruited to evaluate a ‘‘new test for smokers’’ and stated specifically that participants would not be asked to quit smoking. Interested participants called the Duke Risk Communication Laboratory (RCL) and were screened for eligibility. Participants were eligible if they had smoked at least 100 cigarettes in their lifetime and five or more cigarettes in the prior 7 days, were age 18 or older, were not receiving care at the site of the planned intervention trial, spoke and read English, and were willing to provide breath and blood samples. African American smokers were oversampled because they were the target group for the planned intervention trial. Eligible participants completed a battery of self-administered questionnaires during their first visit to the RCL. Eligible participants were paid US$65 for their participation in this pilot project. 2.2. Measures 2.2.1. Socio-demographics Age, gender, race, education, marital status, actual income, perceived adequacy of income, and health insurance status were assessed.
296
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
2.2.2. Smoking-related characteristics Participants were asked the number of cigarettes smoked per day, the minutes to their first cigarette of the day, level of desire to quit smoking (10-point scale, 1 — not at all strong to 10 — extremely strong), confidence in ability to quit smoking (5-point Likert scale, 1 — not at all to 5 — extremely), and perceived addiction to cigarettes (10-point scale, 1 — not at all to 10 — extremely. 2.2.3. Perceived risk Perceived risk for lung cancer was assessed with four questions: (1) ‘‘What do you think is your chance of getting lung cancer in your lifetime if you were to continue to smoke?’’ (2) ‘‘What do you think is your chance of getting lung cancer in your lifetime if you were to quit smoking?’’ (3) ‘‘Compared to [others/African Americans] your age and sex who smoke as much as you do, what do you think is your chance of getting lung cancer in your lifetime?’’ (4) ‘‘Compared to [others/African Americans] your age and sex who do NOT smoke, what do you think is your chance of getting lung cancer in your lifetime?’’ Each question was asked on a five-point Likert scale (very unlikely, somewhat unlikely, 50–50 or same chance as others, somewhat likely and very likely). 2.2.4. Perceived benefits of quitting Two questions assessed the perceived benefits of quitting: (1) ‘‘In your opinion, how much would quitting smoking reduce your chances of getting lung cancer?’’ (2) ‘‘How much would quitting smoking reduce your chances of getting other smoking-related diseases such as emphysema, stroke, and heart disease?’’ Both questions were asked on a four-point Likert scale (not at all to very much). 2.2.5. Concerns about lung cancer risk Two questions assessed concerns: (1) ‘‘How concerned are you about getting lung cancer in your lifetime (four-point Likert scale, 1 — not at all to 4 — very much)?’’ (2) ‘‘How important is it to reduce your chances for getting lung cancer in your lifetime (five-point Likert scale, 1 — not at all important to 5 — extremely important)?’’ 2.2.6. Motivation to quit Smokers’ levels of motivation to quit were assessed with the ‘‘reasons for quitting scale’’ (RFQ) (Curry, Wagner, & Grothaus, 1990). The 12-item RFQ includes six items that assess the intrinsic domain of motivation and six items assess the extrinsic domain (Curry, Grothaus, & McBride, 1997). Each item is phrased as a statement regarding a possible reason for considering smoking cessation. Respondents were instructed to rate ‘‘how true each statement is for you right now’’ on a scale from not at all true (0) to extremely true (4). Examples of intrinsic motivation were ‘‘to prove to others you can quit’’ and for extrinsic ‘‘people close to you will be mad at you if you don’t quit.’’
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
297
2.3. Statistical analysis Spearman correlations tested the associations among the four risk perception measures and the two health-related benefits to quitting smoking. Differences in correlations among perceived risk of smoking and benefits to quitting were tested by demographic and smoking-related variables. All socio-demographic variables were dichotomized prior to conducting subgroup comparisons of correlations of risks and benefits. The F test statistic was used to evaluate the significance of the differences in correlations among subgroups of smokers (see Appendix A for detailed description of method used).
3. Results 3.1. Sample description Three hundred and nine smokers called the RCL and completed the telephone screener. Forty-eight were ineligible, nine did not smoke enough, six had participated in a previous study, one was pregnant, and 32 were receiving care at the planned trial site. Of the 261 eligible participants, 17 declined to participate at screening, and 100 eligible smokers who agreed to participate did not keep their appointment at the RCL; 144 smokers participated in the study. Overall, the participants were relatively young with a mean age of 39 years (S.D. = 10.4). Approximately half of the participants were African American, male, and reported having an annual income of less than US$30,000. The majority had at least a high school education and reported having health insurance (Table 1). Table 1 Demographic and smoking characteristics of the sample Characteristics
Percentage (n = 144)
Age, mean (S.D.) % Male % African American % More than HS % Married % Income US$30,000 + % Extra money % Has insurance Cigarettes smoked per day, mean (S.D.) % First cigarette within 30 min of waking Desire to quit, mean (S.D.) Self-efficacy, mean (S.D.) Addiction, mean (S.D.) Intrinsic motivation to quit, mean (S.D.) Extrinsic motivation to quit, mean (S.D.) Intrinsic – extrinsic motivation to quit, mean (S.D.)
39.1 46 55 67 34 46 42 84 19.0 65 5.9 3.0 7.9 2.6 1.4 1.2
(10.4)
(12.2) (2.6) (1.1) (2.0) (0.9) (0.8) (0.9)
298
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
3.2. Smoking characteristics Participants reported smoking an average of 19 cigarettes/day and 65% reported smoking their first cigarette within 30 min of waking up. Participants reported a high level of addiction to cigarettes, expressed a modest desire to quit smoking, and were only moderately confident that they could quit if they tried (Table 1). Overall, level of motivation both intrinsic and extrinsic was relatively low. 3.3. Risk perceptions, concerns about health effects, and benefits to quitting Seventy-six percent of subjects estimated that they were somewhat (42%) to very likely (34%) to get lung cancer in their lifetime if they continued to smoke. Accordingly, the majority of smokers perceived themselves to be somewhat (35%) to very likely (40%) to get lung cancer compared to nonsmokers. When asked to compare their risk for lung cancer to peers who smoked as much as they did, 53% of the participants reported that they were at the same risk; 34% felt they were at greater risk and 13% felt that they were at less risk of getting lung cancer. The majority of smokers (83%) reported that quitting smoking would reduce (i.e., somewhat to very much) their risk for lung cancer, and their risk of other conditions such as emphysema, stroke, and heart disease (91%). However, almost half of the participants (47%) reported they were somewhat to very likely to get lung cancer in their lifetime even after quitting. A majority of the smokers (97%) reported that it was important to reduce their chances of getting lung cancer. In addition, 87% were somewhat to very concerned about getting lung cancer in their lifetime. 3.4. Correlations among risks and benefits Initial analyses indicated that smokers who perceived that quitting smoking would reduce their risk for lung cancer also perceived that quitting would reduce their risk for other smoking-related conditions (r =.51, P =.0001). Thus, the correlation analyses focus on risk and benefits related to reducing lung cancer risk only. Six correlations were significant (Table 2). Three of the significant correlations were consistent with the initial general hypothesis that greater perceived risk of smoking would be Table 2 Correlation* for measures of risk perceptions and benefits to quitting Measure
Mean
S.D.
1
2
3
4
1. 2. 3. 4. 5.
3.9 3.0 3.3 4.0 3.3
1.1 1.3 0.9 1.1 0.8
–
0.48**
0.31** 0.41**
0.37** 0.15 0.06
If smoke If quit Compared to smokers Compared to nonsmokers Quitting reduces risk * Spearman correlation. ** P values .05.
5 0.10 0.35** 0.08 0.22** –
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
299
Table 3 Comparison of Spearman coefficients for lung cancer risk perceptions by smoker characteristics
Lung cancer risk ‘‘if smoke’’ with ‘‘compared to smokers’’
Lung cancer risk ‘‘compared to nonsmokers’’ with ‘‘quitting reduces risk’’
r =.41
r =.22
.31 .49
.18 .27
.48 .33
.12 .35
.42 .36
.47* .06*
.25 .47
.23 .25
.42 .41
.03 .43
.39 .39
.30 .14
.29 .53
.25 .11
.27 .52
.26 .16
.58 .40
.33 .50
.21 .22
.52 .41
.48 .24
.23 .23
.53 .42
.42 .40
.46* .03*
.48 .47
.30 .51
.38* .04*
.52 .42
.36 .50
.42* .01*
Lung cancer risk ‘‘if smoke’’ with ‘‘if quit’’ Socio-demographics r =.48 Age 39 .33* > 39 .62* Gender Female .46 Male .52 Race White/other .42 African American .59 Education High school or less .38 More than high school .52 Number of cigarettes per day 15 .52 > 15 .46 Self-reported addiction (1 – 10) 8 .51 >8 .38 Concerns about lung cancer Important to reduce risk of lung cancer (1 – 5) Not at all to very important ( 4) .62* Extremely important (5) .31* Concern about getting lung cancer (1 – 4) Not at all to somewhat ( 3) .58 Very (4) .33 Desire to quit Desire to quit (1 – 10) 6 >6 Self-efficacy (1 – 5) Not at all to somewhat (1 – 3) Very to extremely (4 – 5) Motivation to quit Intrinsic motivation (six items) Below the mean ( < 2.6) At or above the mean ( 2.6) Extrinsic motivation (six items) Below the mean ( < 1.4) At or above the mean ( 1.4) Intrinsic – extrinsic Below the mean ( < 1.2) At or above the mean ( 1.2)
* Correlation coefficients were significantly different at P .05 using the F test statistic.
300
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
Fig. 1. Perceived risk of lung cancer if continue to smoke or quit by age.
associated with greater perceived benefit of quitting. Accordingly, smokers who perceived greater lung cancer risk even if they quit were less likely to endorse the benefits of risk reduction with quitting (r = .35, P =.0001). Smokers who perceived themselves to be at greater risk for lung cancer if they continued to smoke also perceived themselves to have a greater likelihood of lung cancer compared to other smokers (r =.31, P =.0001) and to other nonsmokers (r =.37, P =.0001). Three correlations were counter to the initial hypothesis. Smokers who perceived a greater likelihood of getting lung cancer if they continued to smoke also perceived a greater likelihood of getting lung cancer if they quit smoking (r =.48, P =.0001). Smokers who thought they were at greater risk for lung cancer than other smokers also perceived their risk to be high if they quit smoking (r =.41, P =.0001). Lastly, smokers’ perceptions of greater lung cancer risk compared to nonsmoking peers was only weakly associated with the perception that quitting smoking would reduce their risk for lung cancer (r =.22, P =.007). 3.5. Subgroup comparisons of correlations of risks and benefits Differences in correlations by demographic, smoking, and attitudinal variables were tested to explore subgroup differences in the associations between risk and benefits. These
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
301
Fig. 2. Perceived risk of lung cancer if continue to smoke or quit by importance of reducing risk.
analyses focused on the three significant correlations that were counter to our hypotheses; the two strong and unexpected positive correlations between smokers’ perceptions of being at greater risk regardless of whether they continued or quit smoking and the relatively weak correlation between perceived risk for lung cancer as compared to nonsmokers and perceptions that quitting smoking would reduce risk for lung cancer (Table 3). Smokers who perceived lung cancer risk to be likely regardless of whether they continued or quit smoking were older (.62 vs. .33, P < .05) and less likely to consider it important to reduce risk for lung cancer (.62 vs. .31, P < .05). Histograms of these associations for age and perceived importance of reducing risk illustrate these differences. Significantly more smokers who were older perceived themselves to be at high risk for lung cancer whether they continued to smoke and if they quit compared to younger smokers (53% vs. 37%). By contrast, younger smokers were more likely to perceive high risk if they continued to smoke but low risk if they quit (40% vs. 21%, see Fig. 1). Smokers who thought it was very important to reduce their risk for lung cancer were more likely to perceive themselves at high risk whether they continued to smoke or quit compared to those who did not think it was very important to reduce risk (53% vs. 37%, see Fig. 2). By contrast, those who thought it was not very important to reduce risk were much more likely to perceive themselves to be at low risk for lung cancer whether they
302
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
Fig. 3. Perceived risk of lung cancer compared to nonsmokers and that risk would be reduced if quit by race.
smoked or quit than those who thought it was very important to reduce risk (37% vs. 10%). There were no factors that significantly influenced the correlation of smokers’ perceptions of risk for lung cancer if they quit smoking with their perceived risk compared to other smokers. The association between perceived risk of getting lung cancer compared to nonsmoking peers and perceptions that quitting would reduce lung cancer risk was weakest among African Americans ( .06 vs. .47, P < .05), lighter smokers (.03 vs. .43, P < .05), those who reported high levels of intrinsic (.03 vs. .46) and extrinsic (.04 vs. .39) motivation to quit smoking and higher intrinsic relative to extrinsic motivation (.01 vs. .41). Again, histograms illustrate the differences by race, amount smoked, and intrinsic relative to extrinsic motivation. African American smokers were more likely to perceive themselves to be at low risk by comparison to their peers who did not smoke and to perceive that quitting would reduce risk than White smokers (18% vs. 3%, see Fig. 3). By contrast, White smokers were more likely to see themselves at high risk compared to nonsmoking peers and to perceive greater benefit from quitting (51% vs. 23%). Heavier smokers were less likely to see themselves at low risk and high benefit if they quit than lighter smokers (6% vs. 13%, see Fig. 4). Instead, heavier smokers were more likely to perceive themselves at low risk and low benefit than lighter smokers (21% vs. 12%). Lastly, with respect to motivation, those who had lower intrinsic relative to extrinsic motivation were more likely to see themselves at low risk and low benefit than those who had greater intrinsic motivation (22% vs. 9%, see Fig. 5). By contrast, those
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
303
Fig. 4. Perceived risk of lung cancer compared to nonsmokers and that risk would be reduced if quit by amount smoked.
with greater intrinsic relative to extrinsic motivation perceived lower risk but higher benefit than those with lower intrinsic motivation (14% vs. 6%).
4. Discussion Consistent with a large literature, smokers in this study were aware of health risks and for the most part rated their risks to be higher than nonsmokers. In addition, a majority of smokers acknowledged that quitting would reduce their chances of getting lung cancer. However, the picture becomes far more complicated when associations between these perceptions are explored further. Smokers who perceive negative health effects of smoking may not necessarily see related benefits to quitting and for some groups of smokers, in particular lighter smokers, and those who are older or African American, these perceptions may not be related at all. As others have shown (Orleans, Jepson, Resch, & Rimer, 1994), older smokers may feel pessimistic about the benefits of quitting smoking when they have
304
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
Fig. 5. Perceived risk of lung cancer compared to nonsmokers and that risk would be reduced if quit by level of intrinsic to extrinsic motivation.
smoked for many years. However, the mean age for our sample was relatively young (age 39), suggesting that smokers’ pessimism about the benefits of quitting smoking may start relatively early in the smoking career. Similarly, while one might posit that lighter smokers might see fewer harms to smoking relative to nonsmokers and, in turn, see fewer benefits to quitting, this association was not supported either. In fact, lighter smokers while optimistically biased about their own risk as compared to other smokers still believed that smoking cessation would reduce their risk for lung cancer. This was also more true for African American than White smokers and for those who were more intrinsically than extrinsically motivated to quit smoking. Thus, while smokers may hold stereotypes about the type of smoker or smoking patterns that yield the greatest harms and perceive their own smoking to be less harmful, for some smokers, this belief system may be empowering in that it is associated with greater intrinsic relative to extrinsic motivation that has been associated with successful cessation (Curry, McBride, Grothaus, Louie, & Wagner, 1995). Unfortunately, this group was in the relative minority of smokers (approximately 14%). Considering differences among subgroups of smokers in the association between their perceptions of risks and benefits could be important in increasing the relevance and motivational potency of smoking cessation interventions. The advent of computer tailoring enables consideration of smoker’s characteristics and beliefs when delivering printed health education materials (Kreuter, Strecher, & Glassman, 1999; Rimer & Glassman, 1998). These customized cessation messages could take into account relatively complex and often contradictory associations between smokers’ perceptions of risks and benefits. For example, messages could be individually customized to address concerns about the harms of smoking and pessimistic
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
305
beliefs related to the benefits of quitting, while simultaneously considering the smoker’s objective risk profile, level, and type of motivation to quit, race, and age and as such increase the salience and potential impact of smoking cessation materials. Moreover, customization could be used to increase smokers’ awareness of the social comparisons they are making and how these comparisons might undermine or bolster motivation for smoking cessation. The results reported here should be considered carefully. This was a small sample of smokers who were paid to participate in this study. However, these smokers were not participating in a smoking cessation intervention trial that should have eliminated potential response bias relating to risks and benefits of quitting. In conclusion, health communication strategies that consider the complexities of smokers perceptions regarding the health harms of smoking and the advantages of quitting are needed to make smoking cessation messages salient to broader and heterogeneous populations of smokers.
Acknowledgments This work was supported by a grant from the National Cancer Institute (CA72099 and CA76945), Biomarker and Partner. In addition, the authors would like to acknowledge the contributions of Dr. Bercedis Peterson for help with statistical analyses.
Appendix A. Many elementary statistical texts give formulas that can be used to test whether the correlation of two variables is different between two or more independent subpopulations (e.g., age, race, etc.) and while controlling for covariates. (A) The code below shows how SAS can be used to test whether the Pearson correlation of two variables ( Y1 and Y2) is different between two independent subpopulations (SUBPOP). The test of Y2*SUBPOP is the test of interest. proc standard mean = 0 std = 1; by SUBPOP; var Y1 Y2; proc glm; class SUBPOP; model Y1 = Y2 Y2*SUBPOP/noint. (B) This code can easily be applied to tests of Spearman correlations by first using the RANK procedure to rank order Y1 and Y2 within levels defined by SUBPOP: proc ranks; rank Y1 Y2; by SUBPOP; proc standard mean = 0 std = 1; by SUBPOP; var Y1 Y2; proc glm; class SUBPOP; model Y1 = Y2 Y2*SUBPOP/noint. (C) The SAS code can easily be extended to test whether the partial (i.e., covariateadjusted) Pearson correlation of two variables ( Y1 and Y2) is different between two
306
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
independent subpopulations (SUBPOP). Let X indicate a list of covariates to be partialled out of the correlation between Y1 and Y2. The SAS code below takes advantage of the fact that the partial correlation between Y1 and Y2 is identical to the correlation between (1) the residuals from the regression model in which Y1 is regressed on X and (2) the residuals from the regression model in which Y2 is regressed on X. The test of resid2*SUBPOP in the SAS code below is the test of interest: data mine; set mine; if Y1=. or Y2=. or X=. then delete; proc reg; model Y1 = X; output out = temp1 residual = resid1; by SUBPOP; proc reg; model Y2 = X; output out = temp2 residual = resid2; by SUBPOP; data temp; merge temp1 temp2; proc standard data = temp out = temp mean = 0 std = 1; by SUBPOP; var resid1 resid2; proc glm data = temp; class SUBPOP; model resid1 = resid2 resid2*SUBPOP/noint. (D) If SUBGRP has k levels, the code is identical to that given above, but the test statistic of interest will have k 1 degrees of freedom. The ESTIMATE statement can then be used to test whether any two correlations are different from one another. (E) To check the accuracy of one’s SAS code, the first-time user of this SAS code might want to print out the solution to the regression equation in (A) above (just include the ‘‘s’’ option on the model statement). If the user’s code is correct, the regression coefficient on Y2 will be equal to the correlation within one of the SUBPOP groups, while the regression coefficient on Y2*SUBPOP will be the difference in the correlation coefficients between the two subpopulations. (F) A test whether the correlation of two variables is significantly different between two dependent subpopulations (e.g., the correlation of two variables taken at two different time points) can also be accomplished with SAS. The MIXED procedure with the REPEATED statement will be needed, and this is left as an exercise for the reader.
References Bandura, A. (1977). Social learning theory. Englewood Cliffs: Prentice-Hall. Curry, S. J., Grothaus, L. C., & McBride, C. M. (1997). Reasons for quitting: intrinsic and extrinsic motivation for smoking cessation in a population-based sample of smokers. Addictive Behaviors, 22, 727 – 739. Curry, S. J., McBride, C., Grothaus, L. C., Louie, D., & Wagner, E. H. (1995). A randomized trial of self-help materials, personalized feedback, and telephone counseling with nonvolunteer smokers. Journal of Consulting and Clinical Psychology, 63, 1005 – 1014. Curry, S. J., Wagner, E. H., & Grothaus, L. C. (1990). Intrinsic and extrinsic motivation for smoking cessation. Journal for Consulting and Clinical Psychology, 58, 310 – 316. Duncan, C. L., Cummings, S. R., Hudes, E. S., Zahnd, E., & Coates, T. J. (1992). Quitting smoking: reasons for quitting and predictors of cessation among medical patients. Journal of General Internal Medicine, 7, 398 – 404.
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
307
Halpern, M. T., & Warner, K. E. (1993). Motivations for smoking cessation: a comparison of successful quitters and failures. Journal of Substance Abuse, 5, 247 – 256. Janz, N. K., & Becker, M. H. (1984). The health belief model a decade later. Health Education Quarterly, 11, 1 – 47. Klein, W. M., & Weinstein, N. D. (1997). Social comparison and unrealistic optimism about personal risk. In: B. P. Gibbons, & F. X. Gibbons (Eds.), Health, coping, and well-being: perspectives from social comparison theory ( pp. 25 – 62). Mahwah, NJ: Lawrence Erlbaum. Kreuter, M. W., Stretcher, V. J., & Glassman, B. (1999). One size does not fit all: the case for tailoring print materials. Annals of Behavioral Medicine, 21, 276 – 283. McBride, C. M., Halabi, S., Bepler, G., Lyna, P., McIntyre, L., Lipkus, I., & Albright, J. (2000). Maximizing the motivational impact of feedback of lung cancer susceptibility on smokers desire to quit. Journal of Health Communications, 5, 229 – 241. Orleans, C. T., Jepson, C., Resch, N., & Rimer, B. K. (1994). Quitting motives and barriers among older smokers. The 1986 adult use of tobacco survey revisited. Cancer, 74, 2055 – 2061. Rimer, B. K., & Glassman, B. (1998). Tailoring communication for primary care setting. Methods of Information in Medicine, 37, 171 – 177. Ronis, D. L. (1992). Conditional health threats: health beliefs, decisions and behaviors among adults. Health Psychology, 11, 127 – 134. Rosenstock, I. M. (1974). The health belief model: origins and correlates. Health Education Monograph, 2, 336 – 353. Rothman, A. J., Klein, W. M., & Weinstein, N. D. (1996). Absolute and relative biases in estimation of personal risk. Journal of Applied Social Psychology, 26, 1213 – 1236. Strecher, V. J., Kreuter, M. W., & Kobrin, S. C. (1995). Do cigarette smokers have an unrealistic perceptions of their heart attack, cancer and stroke risks? Journal of Behavior Medicine, 18, 45 – 54. Strecher, V. J., Rosenstock I. M. (Eds.) (1997). The health belief model (2nd ed., pp. 41 – 59). San Francisco: Jossey-Bass. Viscusi, W. K. (1990). Do smokers underestimate risk? Journal of Political Economy, 98, 1253 – 1269. Weinstein, N. D. (1988). The precaution adoption process. Health, 7, 355 – 386. Weinstein, N. D. (1980). Unrealistic optimism about life events. Journal of Personal Social Psychology, 39, 806 – 820. Weinstein, N. D. (1998). Accuracy of smokers’ risk perceptions. Journal of Behavior Medicine, 20, 135 – 140. Weinstein, N. D., & Klein, W. M. (1995). Resistance of personal risk perceptions to debasing interventions. Health Psychology, 14, 132 – 140.
Exploring the association between perceived risks of smoking and benefits to quitting Who does not see the link? Pauline Lynaa,*, Colleen McBridea,b, Greg Samsab,c, Kathryn I. Pollaka,b a
Duke Comprehensive Cancer Center, Durham, NC, USA b Duke University Medical Center, Durham, NC, USA c Durham Veteran’s Administration Medical Center, Durham, NC, USA
Abstract This report explored associations between different measures of smokers’ perceived risks of smoking and benefits to quitting and the extent to which these associations varied by demographic and other characteristics for 144 smokers. We hypothesized greater perceived risk of smoking would be associated with greater perceived benefits to quitting and would be strongest among smokers who were concerned about health effects of smoking and motivated to quit. Results indicated smokers’ perceived themselves at risk for lung cancer regardless if they continued or quit smoking and was strongest for smokers who were older and minimized the importance of reducing lung cancer risk. There was a weak correlation between perceived risk for lung cancer when compared to nonsmokers and perception that quitting smoking would reduce lung cancer risk and was weakest for African Americans, lighters smokers, and smokers with higher intrinsic relative to extrinsic motivation for cessation. In conclusion, these subgroup differences in the relationship between perceptions of risks and benefits could be important to consider to increase the relevance and motivational potency of smoking cessation interventions. D 2001 Elsevier Science Ltd. All rights reserved. Keywords: Smoking cessation; Perceived risks; Health behavior; Motivation
* Corresponding author. Duke University Medical Center, DUMC, Hanes House, Trent Drive, Box 2949, Durham, NC 27710-2949, USA. Tel.: +1-919-681-4559; fax: +1-919-681-4785. E-mail address: [email protected] (P. Lyna). 0306-4603/02/$ – see front matter D 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 3 0 6 - 4 6 0 3 ( 0 1 ) 0 0 1 7 5 - 7
294
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
1. Introduction Cigarette smoking is the leading preventable cause of death in the United States and is forecasted to be an ongoing public health problem. Following two decades of decline, rates of smoking have remained relatively stable, hovering around 25% for the past 10 years. National surveys indicate that 70% or more of smokers report that smoking is hazardous to health and some reports suggest that smokers even overestimate the health risks of smoking (Viscusi, 1990). Smokers cite concerns about the harms of smoking as primary motivators for them to consider cessation (Duncan, Cummings, Hudes, Zahnd, & Coates, 1992; Halpern & Warner, 1993). Yet, many continue to smoke despite these concerns. How is it that smokers can be aware of the substantial harms to smoking and yet continue to smoke? This apparent discordance in awareness and behavior is of particular concern because the harms of smoking and benefits of quitting are a major motivational focus of most smoking cessation programs. Thus, a better understanding of this discordance could be informative for improving cessation interventions. Popular conceptual models (Strecher & Rosenstock, 1997) suggest a cascade of factors that influence smoking cessation. Generally, these models posit that for smoking cessation to occur, an individual must perceive personal vulnerability to related negative outcomes (Bandura, 1977; Janz & Becker, 1984; Rosenstock, 1974), must see the outcomes as severe, and believe that quitting smoking will reduce the likelihood of their occurrence. Further, it has been argued that concepts of perceived vulnerability to health effects, and severity of risk are relatively complex domains (Weinstein, 1998). For example, perceptions of risk can vary based on whether one is comparing oneself to a specific reference group (e.g., similar in age, race, or smoking status), the time frame that is being considered (lifetime, 10 years, etc.), and conditional on whether one quits or continues to smoke (Klein & Weinstein, 1997; Ronis, 1992; Weinstein, 1980; Weinstein & Klein, 1995). Thus, it has been argued that smokers may acknowledge risks of smoking to generalized others, yet at the same time not fully acknowledge their own personal vulnerability to these risks (Rothman, Klein, & Weinstein, 1996; Strecher, Kreuter, & Kobrin, 1995; Weinstein, 1988). The majority of these studies have focused on the accuracy of smokers’ risk perceptions rather than the extent to which these perceptions vary with demographic and other smoking-related characteristics. Consideration of the characteristics of smokers who do and do not see the link between their smoking and potential health harms could be informative in guiding the development of effective smoking cessation programs. For example, if a light smoker perceives less harm to smoking and in turn, fewer benefits to quitting, regardless of the reference group than a heavier smoker, then health messages for this population may be best focused on countering these misperceptions. In turn, heavier smokers who may see great harm to smoking but few benefits to quitting may benefit from more optimistic representations of the benefits of smoking cessation. Thus, understanding the multidimensionality of the link between perceived risks and benefits and how these vary among subgroups of smokers could influence how risk messages are communicated to smokers. Smokers’ perceived benefits of smoking cessation also may be influenced by other factors such as their level of concern or worry about smoking-related health outcomes and their confidence that they could quit if they tried. For example, the Health Belief and other models
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
295
(Bandura, 1977; Strecher & Rosenstock, 1997) would suggest that perceived risks and benefits might be most strongly associated among those who are especially concerned about suffering the health effects of smoking than among those who are relatively unconcerned. Exploring conceptually meaningful associations among subgroups of smokers could give greater insight to the development of motivational interventions for smokers. To date, few studies have included the broad array of measures of perceived risks, benefits, and related concerns that are needed to better understand possible interrelationships. With the above in mind, this report asked two questions. First, is there an association between comparative and conditional risk for lung cancer and perceived benefits of quitting smoking for reducing these risks? We hypothesized that higher perceived risk regardless of the conditional comparison would be associated with greater perceived benefits to quitting. The second question asked, whether associations among risk perceptions and perceived benefits of quitting were influenced by demographics, smoking patterns, motivation, and self-efficacy for cessation or concerns related to the disease outcome? We hypothesized that associations among perceived risks and benefits would be greatest among smokers who were most concerned about health effects, and those who were most motivated and most confident they could quit.
2. Methods 2.1. Study recruitment As part of a pilot study conducted in preparation for a randomized intervention trial, smokers were recruited via newspaper and public service radio advertisements and recruitment tables located at various community and medical center locations. A detailed description of the pilot study is provided elsewhere (McBride et al., 2000). Study advertisements stated that smokers were being recruited to evaluate a ‘‘new test for smokers’’ and stated specifically that participants would not be asked to quit smoking. Interested participants called the Duke Risk Communication Laboratory (RCL) and were screened for eligibility. Participants were eligible if they had smoked at least 100 cigarettes in their lifetime and five or more cigarettes in the prior 7 days, were age 18 or older, were not receiving care at the site of the planned intervention trial, spoke and read English, and were willing to provide breath and blood samples. African American smokers were oversampled because they were the target group for the planned intervention trial. Eligible participants completed a battery of self-administered questionnaires during their first visit to the RCL. Eligible participants were paid US$65 for their participation in this pilot project. 2.2. Measures 2.2.1. Socio-demographics Age, gender, race, education, marital status, actual income, perceived adequacy of income, and health insurance status were assessed.
296
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
2.2.2. Smoking-related characteristics Participants were asked the number of cigarettes smoked per day, the minutes to their first cigarette of the day, level of desire to quit smoking (10-point scale, 1 — not at all strong to 10 — extremely strong), confidence in ability to quit smoking (5-point Likert scale, 1 — not at all to 5 — extremely), and perceived addiction to cigarettes (10-point scale, 1 — not at all to 10 — extremely. 2.2.3. Perceived risk Perceived risk for lung cancer was assessed with four questions: (1) ‘‘What do you think is your chance of getting lung cancer in your lifetime if you were to continue to smoke?’’ (2) ‘‘What do you think is your chance of getting lung cancer in your lifetime if you were to quit smoking?’’ (3) ‘‘Compared to [others/African Americans] your age and sex who smoke as much as you do, what do you think is your chance of getting lung cancer in your lifetime?’’ (4) ‘‘Compared to [others/African Americans] your age and sex who do NOT smoke, what do you think is your chance of getting lung cancer in your lifetime?’’ Each question was asked on a five-point Likert scale (very unlikely, somewhat unlikely, 50–50 or same chance as others, somewhat likely and very likely). 2.2.4. Perceived benefits of quitting Two questions assessed the perceived benefits of quitting: (1) ‘‘In your opinion, how much would quitting smoking reduce your chances of getting lung cancer?’’ (2) ‘‘How much would quitting smoking reduce your chances of getting other smoking-related diseases such as emphysema, stroke, and heart disease?’’ Both questions were asked on a four-point Likert scale (not at all to very much). 2.2.5. Concerns about lung cancer risk Two questions assessed concerns: (1) ‘‘How concerned are you about getting lung cancer in your lifetime (four-point Likert scale, 1 — not at all to 4 — very much)?’’ (2) ‘‘How important is it to reduce your chances for getting lung cancer in your lifetime (five-point Likert scale, 1 — not at all important to 5 — extremely important)?’’ 2.2.6. Motivation to quit Smokers’ levels of motivation to quit were assessed with the ‘‘reasons for quitting scale’’ (RFQ) (Curry, Wagner, & Grothaus, 1990). The 12-item RFQ includes six items that assess the intrinsic domain of motivation and six items assess the extrinsic domain (Curry, Grothaus, & McBride, 1997). Each item is phrased as a statement regarding a possible reason for considering smoking cessation. Respondents were instructed to rate ‘‘how true each statement is for you right now’’ on a scale from not at all true (0) to extremely true (4). Examples of intrinsic motivation were ‘‘to prove to others you can quit’’ and for extrinsic ‘‘people close to you will be mad at you if you don’t quit.’’
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
297
2.3. Statistical analysis Spearman correlations tested the associations among the four risk perception measures and the two health-related benefits to quitting smoking. Differences in correlations among perceived risk of smoking and benefits to quitting were tested by demographic and smoking-related variables. All socio-demographic variables were dichotomized prior to conducting subgroup comparisons of correlations of risks and benefits. The F test statistic was used to evaluate the significance of the differences in correlations among subgroups of smokers (see Appendix A for detailed description of method used).
3. Results 3.1. Sample description Three hundred and nine smokers called the RCL and completed the telephone screener. Forty-eight were ineligible, nine did not smoke enough, six had participated in a previous study, one was pregnant, and 32 were receiving care at the planned trial site. Of the 261 eligible participants, 17 declined to participate at screening, and 100 eligible smokers who agreed to participate did not keep their appointment at the RCL; 144 smokers participated in the study. Overall, the participants were relatively young with a mean age of 39 years (S.D. = 10.4). Approximately half of the participants were African American, male, and reported having an annual income of less than US$30,000. The majority had at least a high school education and reported having health insurance (Table 1). Table 1 Demographic and smoking characteristics of the sample Characteristics
Percentage (n = 144)
Age, mean (S.D.) % Male % African American % More than HS % Married % Income US$30,000 + % Extra money % Has insurance Cigarettes smoked per day, mean (S.D.) % First cigarette within 30 min of waking Desire to quit, mean (S.D.) Self-efficacy, mean (S.D.) Addiction, mean (S.D.) Intrinsic motivation to quit, mean (S.D.) Extrinsic motivation to quit, mean (S.D.) Intrinsic – extrinsic motivation to quit, mean (S.D.)
39.1 46 55 67 34 46 42 84 19.0 65 5.9 3.0 7.9 2.6 1.4 1.2
(10.4)
(12.2) (2.6) (1.1) (2.0) (0.9) (0.8) (0.9)
298
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
3.2. Smoking characteristics Participants reported smoking an average of 19 cigarettes/day and 65% reported smoking their first cigarette within 30 min of waking up. Participants reported a high level of addiction to cigarettes, expressed a modest desire to quit smoking, and were only moderately confident that they could quit if they tried (Table 1). Overall, level of motivation both intrinsic and extrinsic was relatively low. 3.3. Risk perceptions, concerns about health effects, and benefits to quitting Seventy-six percent of subjects estimated that they were somewhat (42%) to very likely (34%) to get lung cancer in their lifetime if they continued to smoke. Accordingly, the majority of smokers perceived themselves to be somewhat (35%) to very likely (40%) to get lung cancer compared to nonsmokers. When asked to compare their risk for lung cancer to peers who smoked as much as they did, 53% of the participants reported that they were at the same risk; 34% felt they were at greater risk and 13% felt that they were at less risk of getting lung cancer. The majority of smokers (83%) reported that quitting smoking would reduce (i.e., somewhat to very much) their risk for lung cancer, and their risk of other conditions such as emphysema, stroke, and heart disease (91%). However, almost half of the participants (47%) reported they were somewhat to very likely to get lung cancer in their lifetime even after quitting. A majority of the smokers (97%) reported that it was important to reduce their chances of getting lung cancer. In addition, 87% were somewhat to very concerned about getting lung cancer in their lifetime. 3.4. Correlations among risks and benefits Initial analyses indicated that smokers who perceived that quitting smoking would reduce their risk for lung cancer also perceived that quitting would reduce their risk for other smoking-related conditions (r =.51, P =.0001). Thus, the correlation analyses focus on risk and benefits related to reducing lung cancer risk only. Six correlations were significant (Table 2). Three of the significant correlations were consistent with the initial general hypothesis that greater perceived risk of smoking would be Table 2 Correlation* for measures of risk perceptions and benefits to quitting Measure
Mean
S.D.
1
2
3
4
1. 2. 3. 4. 5.
3.9 3.0 3.3 4.0 3.3
1.1 1.3 0.9 1.1 0.8
–
0.48**
0.31** 0.41**
0.37** 0.15 0.06
If smoke If quit Compared to smokers Compared to nonsmokers Quitting reduces risk * Spearman correlation. ** P values .05.
5 0.10 0.35** 0.08 0.22** –
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
299
Table 3 Comparison of Spearman coefficients for lung cancer risk perceptions by smoker characteristics
Lung cancer risk ‘‘if smoke’’ with ‘‘compared to smokers’’
Lung cancer risk ‘‘compared to nonsmokers’’ with ‘‘quitting reduces risk’’
r =.41
r =.22
.31 .49
.18 .27
.48 .33
.12 .35
.42 .36
.47* .06*
.25 .47
.23 .25
.42 .41
.03 .43
.39 .39
.30 .14
.29 .53
.25 .11
.27 .52
.26 .16
.58 .40
.33 .50
.21 .22
.52 .41
.48 .24
.23 .23
.53 .42
.42 .40
.46* .03*
.48 .47
.30 .51
.38* .04*
.52 .42
.36 .50
.42* .01*
Lung cancer risk ‘‘if smoke’’ with ‘‘if quit’’ Socio-demographics r =.48 Age 39 .33* > 39 .62* Gender Female .46 Male .52 Race White/other .42 African American .59 Education High school or less .38 More than high school .52 Number of cigarettes per day 15 .52 > 15 .46 Self-reported addiction (1 – 10) 8 .51 >8 .38 Concerns about lung cancer Important to reduce risk of lung cancer (1 – 5) Not at all to very important ( 4) .62* Extremely important (5) .31* Concern about getting lung cancer (1 – 4) Not at all to somewhat ( 3) .58 Very (4) .33 Desire to quit Desire to quit (1 – 10) 6 >6 Self-efficacy (1 – 5) Not at all to somewhat (1 – 3) Very to extremely (4 – 5) Motivation to quit Intrinsic motivation (six items) Below the mean ( < 2.6) At or above the mean ( 2.6) Extrinsic motivation (six items) Below the mean ( < 1.4) At or above the mean ( 1.4) Intrinsic – extrinsic Below the mean ( < 1.2) At or above the mean ( 1.2)
* Correlation coefficients were significantly different at P .05 using the F test statistic.
300
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
Fig. 1. Perceived risk of lung cancer if continue to smoke or quit by age.
associated with greater perceived benefit of quitting. Accordingly, smokers who perceived greater lung cancer risk even if they quit were less likely to endorse the benefits of risk reduction with quitting (r = .35, P =.0001). Smokers who perceived themselves to be at greater risk for lung cancer if they continued to smoke also perceived themselves to have a greater likelihood of lung cancer compared to other smokers (r =.31, P =.0001) and to other nonsmokers (r =.37, P =.0001). Three correlations were counter to the initial hypothesis. Smokers who perceived a greater likelihood of getting lung cancer if they continued to smoke also perceived a greater likelihood of getting lung cancer if they quit smoking (r =.48, P =.0001). Smokers who thought they were at greater risk for lung cancer than other smokers also perceived their risk to be high if they quit smoking (r =.41, P =.0001). Lastly, smokers’ perceptions of greater lung cancer risk compared to nonsmoking peers was only weakly associated with the perception that quitting smoking would reduce their risk for lung cancer (r =.22, P =.007). 3.5. Subgroup comparisons of correlations of risks and benefits Differences in correlations by demographic, smoking, and attitudinal variables were tested to explore subgroup differences in the associations between risk and benefits. These
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
301
Fig. 2. Perceived risk of lung cancer if continue to smoke or quit by importance of reducing risk.
analyses focused on the three significant correlations that were counter to our hypotheses; the two strong and unexpected positive correlations between smokers’ perceptions of being at greater risk regardless of whether they continued or quit smoking and the relatively weak correlation between perceived risk for lung cancer as compared to nonsmokers and perceptions that quitting smoking would reduce risk for lung cancer (Table 3). Smokers who perceived lung cancer risk to be likely regardless of whether they continued or quit smoking were older (.62 vs. .33, P < .05) and less likely to consider it important to reduce risk for lung cancer (.62 vs. .31, P < .05). Histograms of these associations for age and perceived importance of reducing risk illustrate these differences. Significantly more smokers who were older perceived themselves to be at high risk for lung cancer whether they continued to smoke and if they quit compared to younger smokers (53% vs. 37%). By contrast, younger smokers were more likely to perceive high risk if they continued to smoke but low risk if they quit (40% vs. 21%, see Fig. 1). Smokers who thought it was very important to reduce their risk for lung cancer were more likely to perceive themselves at high risk whether they continued to smoke or quit compared to those who did not think it was very important to reduce risk (53% vs. 37%, see Fig. 2). By contrast, those who thought it was not very important to reduce risk were much more likely to perceive themselves to be at low risk for lung cancer whether they
302
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
Fig. 3. Perceived risk of lung cancer compared to nonsmokers and that risk would be reduced if quit by race.
smoked or quit than those who thought it was very important to reduce risk (37% vs. 10%). There were no factors that significantly influenced the correlation of smokers’ perceptions of risk for lung cancer if they quit smoking with their perceived risk compared to other smokers. The association between perceived risk of getting lung cancer compared to nonsmoking peers and perceptions that quitting would reduce lung cancer risk was weakest among African Americans ( .06 vs. .47, P < .05), lighter smokers (.03 vs. .43, P < .05), those who reported high levels of intrinsic (.03 vs. .46) and extrinsic (.04 vs. .39) motivation to quit smoking and higher intrinsic relative to extrinsic motivation (.01 vs. .41). Again, histograms illustrate the differences by race, amount smoked, and intrinsic relative to extrinsic motivation. African American smokers were more likely to perceive themselves to be at low risk by comparison to their peers who did not smoke and to perceive that quitting would reduce risk than White smokers (18% vs. 3%, see Fig. 3). By contrast, White smokers were more likely to see themselves at high risk compared to nonsmoking peers and to perceive greater benefit from quitting (51% vs. 23%). Heavier smokers were less likely to see themselves at low risk and high benefit if they quit than lighter smokers (6% vs. 13%, see Fig. 4). Instead, heavier smokers were more likely to perceive themselves at low risk and low benefit than lighter smokers (21% vs. 12%). Lastly, with respect to motivation, those who had lower intrinsic relative to extrinsic motivation were more likely to see themselves at low risk and low benefit than those who had greater intrinsic motivation (22% vs. 9%, see Fig. 5). By contrast, those
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
303
Fig. 4. Perceived risk of lung cancer compared to nonsmokers and that risk would be reduced if quit by amount smoked.
with greater intrinsic relative to extrinsic motivation perceived lower risk but higher benefit than those with lower intrinsic motivation (14% vs. 6%).
4. Discussion Consistent with a large literature, smokers in this study were aware of health risks and for the most part rated their risks to be higher than nonsmokers. In addition, a majority of smokers acknowledged that quitting would reduce their chances of getting lung cancer. However, the picture becomes far more complicated when associations between these perceptions are explored further. Smokers who perceive negative health effects of smoking may not necessarily see related benefits to quitting and for some groups of smokers, in particular lighter smokers, and those who are older or African American, these perceptions may not be related at all. As others have shown (Orleans, Jepson, Resch, & Rimer, 1994), older smokers may feel pessimistic about the benefits of quitting smoking when they have
304
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
Fig. 5. Perceived risk of lung cancer compared to nonsmokers and that risk would be reduced if quit by level of intrinsic to extrinsic motivation.
smoked for many years. However, the mean age for our sample was relatively young (age 39), suggesting that smokers’ pessimism about the benefits of quitting smoking may start relatively early in the smoking career. Similarly, while one might posit that lighter smokers might see fewer harms to smoking relative to nonsmokers and, in turn, see fewer benefits to quitting, this association was not supported either. In fact, lighter smokers while optimistically biased about their own risk as compared to other smokers still believed that smoking cessation would reduce their risk for lung cancer. This was also more true for African American than White smokers and for those who were more intrinsically than extrinsically motivated to quit smoking. Thus, while smokers may hold stereotypes about the type of smoker or smoking patterns that yield the greatest harms and perceive their own smoking to be less harmful, for some smokers, this belief system may be empowering in that it is associated with greater intrinsic relative to extrinsic motivation that has been associated with successful cessation (Curry, McBride, Grothaus, Louie, & Wagner, 1995). Unfortunately, this group was in the relative minority of smokers (approximately 14%). Considering differences among subgroups of smokers in the association between their perceptions of risks and benefits could be important in increasing the relevance and motivational potency of smoking cessation interventions. The advent of computer tailoring enables consideration of smoker’s characteristics and beliefs when delivering printed health education materials (Kreuter, Strecher, & Glassman, 1999; Rimer & Glassman, 1998). These customized cessation messages could take into account relatively complex and often contradictory associations between smokers’ perceptions of risks and benefits. For example, messages could be individually customized to address concerns about the harms of smoking and pessimistic
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
305
beliefs related to the benefits of quitting, while simultaneously considering the smoker’s objective risk profile, level, and type of motivation to quit, race, and age and as such increase the salience and potential impact of smoking cessation materials. Moreover, customization could be used to increase smokers’ awareness of the social comparisons they are making and how these comparisons might undermine or bolster motivation for smoking cessation. The results reported here should be considered carefully. This was a small sample of smokers who were paid to participate in this study. However, these smokers were not participating in a smoking cessation intervention trial that should have eliminated potential response bias relating to risks and benefits of quitting. In conclusion, health communication strategies that consider the complexities of smokers perceptions regarding the health harms of smoking and the advantages of quitting are needed to make smoking cessation messages salient to broader and heterogeneous populations of smokers.
Acknowledgments This work was supported by a grant from the National Cancer Institute (CA72099 and CA76945), Biomarker and Partner. In addition, the authors would like to acknowledge the contributions of Dr. Bercedis Peterson for help with statistical analyses.
Appendix A. Many elementary statistical texts give formulas that can be used to test whether the correlation of two variables is different between two or more independent subpopulations (e.g., age, race, etc.) and while controlling for covariates. (A) The code below shows how SAS can be used to test whether the Pearson correlation of two variables ( Y1 and Y2) is different between two independent subpopulations (SUBPOP). The test of Y2*SUBPOP is the test of interest. proc standard mean = 0 std = 1; by SUBPOP; var Y1 Y2; proc glm; class SUBPOP; model Y1 = Y2 Y2*SUBPOP/noint. (B) This code can easily be applied to tests of Spearman correlations by first using the RANK procedure to rank order Y1 and Y2 within levels defined by SUBPOP: proc ranks; rank Y1 Y2; by SUBPOP; proc standard mean = 0 std = 1; by SUBPOP; var Y1 Y2; proc glm; class SUBPOP; model Y1 = Y2 Y2*SUBPOP/noint. (C) The SAS code can easily be extended to test whether the partial (i.e., covariateadjusted) Pearson correlation of two variables ( Y1 and Y2) is different between two
306
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
independent subpopulations (SUBPOP). Let X indicate a list of covariates to be partialled out of the correlation between Y1 and Y2. The SAS code below takes advantage of the fact that the partial correlation between Y1 and Y2 is identical to the correlation between (1) the residuals from the regression model in which Y1 is regressed on X and (2) the residuals from the regression model in which Y2 is regressed on X. The test of resid2*SUBPOP in the SAS code below is the test of interest: data mine; set mine; if Y1=. or Y2=. or X=. then delete; proc reg; model Y1 = X; output out = temp1 residual = resid1; by SUBPOP; proc reg; model Y2 = X; output out = temp2 residual = resid2; by SUBPOP; data temp; merge temp1 temp2; proc standard data = temp out = temp mean = 0 std = 1; by SUBPOP; var resid1 resid2; proc glm data = temp; class SUBPOP; model resid1 = resid2 resid2*SUBPOP/noint. (D) If SUBGRP has k levels, the code is identical to that given above, but the test statistic of interest will have k 1 degrees of freedom. The ESTIMATE statement can then be used to test whether any two correlations are different from one another. (E) To check the accuracy of one’s SAS code, the first-time user of this SAS code might want to print out the solution to the regression equation in (A) above (just include the ‘‘s’’ option on the model statement). If the user’s code is correct, the regression coefficient on Y2 will be equal to the correlation within one of the SUBPOP groups, while the regression coefficient on Y2*SUBPOP will be the difference in the correlation coefficients between the two subpopulations. (F) A test whether the correlation of two variables is significantly different between two dependent subpopulations (e.g., the correlation of two variables taken at two different time points) can also be accomplished with SAS. The MIXED procedure with the REPEATED statement will be needed, and this is left as an exercise for the reader.
References Bandura, A. (1977). Social learning theory. Englewood Cliffs: Prentice-Hall. Curry, S. J., Grothaus, L. C., & McBride, C. M. (1997). Reasons for quitting: intrinsic and extrinsic motivation for smoking cessation in a population-based sample of smokers. Addictive Behaviors, 22, 727 – 739. Curry, S. J., McBride, C., Grothaus, L. C., Louie, D., & Wagner, E. H. (1995). A randomized trial of self-help materials, personalized feedback, and telephone counseling with nonvolunteer smokers. Journal of Consulting and Clinical Psychology, 63, 1005 – 1014. Curry, S. J., Wagner, E. H., & Grothaus, L. C. (1990). Intrinsic and extrinsic motivation for smoking cessation. Journal for Consulting and Clinical Psychology, 58, 310 – 316. Duncan, C. L., Cummings, S. R., Hudes, E. S., Zahnd, E., & Coates, T. J. (1992). Quitting smoking: reasons for quitting and predictors of cessation among medical patients. Journal of General Internal Medicine, 7, 398 – 404.
P. Lyna et al. / Addictive Behaviors 27 (2002) 293–307
307
Halpern, M. T., & Warner, K. E. (1993). Motivations for smoking cessation: a comparison of successful quitters and failures. Journal of Substance Abuse, 5, 247 – 256. Janz, N. K., & Becker, M. H. (1984). The health belief model a decade later. Health Education Quarterly, 11, 1 – 47. Klein, W. M., & Weinstein, N. D. (1997). Social comparison and unrealistic optimism about personal risk. In: B. P. Gibbons, & F. X. Gibbons (Eds.), Health, coping, and well-being: perspectives from social comparison theory ( pp. 25 – 62). Mahwah, NJ: Lawrence Erlbaum. Kreuter, M. W., Stretcher, V. J., & Glassman, B. (1999). One size does not fit all: the case for tailoring print materials. Annals of Behavioral Medicine, 21, 276 – 283. McBride, C. M., Halabi, S., Bepler, G., Lyna, P., McIntyre, L., Lipkus, I., & Albright, J. (2000). Maximizing the motivational impact of feedback of lung cancer susceptibility on smokers desire to quit. Journal of Health Communications, 5, 229 – 241. Orleans, C. T., Jepson, C., Resch, N., & Rimer, B. K. (1994). Quitting motives and barriers among older smokers. The 1986 adult use of tobacco survey revisited. Cancer, 74, 2055 – 2061. Rimer, B. K., & Glassman, B. (1998). Tailoring communication for primary care setting. Methods of Information in Medicine, 37, 171 – 177. Ronis, D. L. (1992). Conditional health threats: health beliefs, decisions and behaviors among adults. Health Psychology, 11, 127 – 134. Rosenstock, I. M. (1974). The health belief model: origins and correlates. Health Education Monograph, 2, 336 – 353. Rothman, A. J., Klein, W. M., & Weinstein, N. D. (1996). Absolute and relative biases in estimation of personal risk. Journal of Applied Social Psychology, 26, 1213 – 1236. Strecher, V. J., Kreuter, M. W., & Kobrin, S. C. (1995). Do cigarette smokers have an unrealistic perceptions of their heart attack, cancer and stroke risks? Journal of Behavior Medicine, 18, 45 – 54. Strecher, V. J., Rosenstock I. M. (Eds.) (1997). The health belief model (2nd ed., pp. 41 – 59). San Francisco: Jossey-Bass. Viscusi, W. K. (1990). Do smokers underestimate risk? Journal of Political Economy, 98, 1253 – 1269. Weinstein, N. D. (1988). The precaution adoption process. Health, 7, 355 – 386. Weinstein, N. D. (1980). Unrealistic optimism about life events. Journal of Personal Social Psychology, 39, 806 – 820. Weinstein, N. D. (1998). Accuracy of smokers’ risk perceptions. Journal of Behavior Medicine, 20, 135 – 140. Weinstein, N. D., & Klein, W. M. (1995). Resistance of personal risk perceptions to debasing interventions. Health Psychology, 14, 132 – 140.