Requestsforreprintsshould besent to Dr. Richard B. Warnecke, Survey Research Lab- oratory, Universityof Illinois at Chicago, P.O.. Box 6905, M/C 336, Chicago, ...
The Second Chicago Televised Smoking Cessation Program: A 24-Month Follow-up
Richard B. Wamecke, PhD, Patricia Langenberg PhD, Siu Chi Wong, MS, Bnan R Flay, DPhil and Thomas D. Cook, PhD
Introduction Data obtained in the National Health Interview Survey through 1985 show that about 30% of the adult US population over the age of 20 years smokes.' The goal of the US Public Health Service and the National Cancer Institute is to reduce smoking among that population to 20% by the year 2000.2-3 Other data compiled from the survey indicate that the smoking population of the future is likely to comprise hard-core smokers who either are unmotivated to quit or have failed to succeed at past efforts.4-6 These remaining smokers have not responded to traditional smoking interventions that have been successfully offered in the past to the general population.7-9. Our ongoing research and that ofothers indicate that these continuing smokers are likely to have needs that limit their capacity to respond to such interventions. Thus, new interventions will probably have to incorporate different strategies to reach these smokers. For example, those who continue to smoke may be most likely to respond to interventions that do not call for attendance at face-to-face meetings and that require minimal reading or provide reading materials that are written at a very low literacy level.4,10 Televised smoking cessation interventions that incorporate models to whom these smokers can relate are often used because they reach many smokers. Although some community-based programs have not found televised interventions to be effective without direct, face-to-face support from community health workers,"1'2 others have found otherwise, especially when such interventions are combined with other self-help materials.9,13 Elsewhere'0 we reported that those at risk
to continue smoking were likely to view credible televised programs, such as those offered as part of a news broadcast, and could recall 6 or more of the 20 segments that were broadcast. However, analysis of participation in the intervention also indicated that those who recalled segments of the broadcasts were not the same as those who referred to the accompanying manual. Thus, if the effectiveness of intervention is contingent both on viewing and recalling televised segments and on referring to a manual, the intervention may not fully reach those likely to be the smokers in the future. As we show here, however, the combination of referring to a manual and viewing the programs daily predicts the greatest success with the program. The principal purpose of this paper is to report long-term (24-month) results from the minimal self-help smoking cessation intervention that combined the American Lung Association manual, Freedom from Smoking in 20 Days, with a daily series of 20 televised news segments designed to supplement the daily steps described by the manual. These results compare the 24-month multiple point
Richard B. Warnecke is with the Survey Research Laboratory, Brian R. Flay is with the Prevention Research Center, and Siu Chi Wong is with the Epidemiology-Biostatistics Program at the University of Illinois at Chicago. Patricia Langenberg is with the Department of Epidemiology and Preventive Medicine at the University of Maryland at Baltimore, and Thomas D. Cook is with the Sociology Department at Northwestern University in Evanston, IL. Requests for reprints should be sent to Dr. Richard B. Warnecke, Survey Research Laboratory, University of Illinois at Chicago, P.O. Box 6905, M/C 336, Chicago, IL 60680. This paper was submitted to the Journal December 26, 1990, and accepted with revisions November 27, 1991.
American Journal of Public Health 835
Wanedke et aL
ing the intervention, adjusting for differin demographics and baseline
ences
Population Interview Wave
Preintervention Post+ 50 days intervention --_--__. .X
6 mos.
(18 mos.)*
smoking, and to model smoking cessation patterns absent any intervention, we used the weighted, combined population data,
which is described Post-
Participant Interview Wave
intervention -_- - -
50 ~ays/
6 mos.
12 mos.
24 mos.
6 mros.
121MS
24 MOS.
IPostintervention "Zero"
Time Line for "Match"
*Status determined from 24-month interview.
FIGURE -Lagged analysis plan
prevalence of cessation of a cohort of proparticipants with that of a sample drawn from the smoking population targeted by the intervention. A second purpose is to descnibe how level of exposure to the intervention affected rate of sustained cessation and timing of the cessation event. gram
Meho&ds Design of the Intervention Details of the intervention and characteristics of the participants are described in detail elsewhere.10 Briefly, the televised portion of the intervention was broadcast over 20 days as part of the local evening news at 4 PM and 10 PM on WLSTV, the local ABC-owned station in Chicago. During the 3-week registration period, 10- and 30-second spots, featuring a mime, "kicking the habit," and donated by WLS-TV, were broadcast telling smokers how to register. Through collaboration with the Chicago Lung Association, True Value Hardware stores, and PruCare Health Maintenance Organization, the American Lung Association manual, Freedom from Smoking in 20 Days, was distributed at no cost when smokers registered for the program or requested the manual. In all, 9182 smokers registered for the program, and an estimated 75 000 additional manuals were distnbuted through the stores to those smokers who requested them.
least 3 days a week and were not participants. Participants were registrants who, at the first postintervention interview, reported referring to the manual at least once and/or viewing at least one segment of the program. In October 1986 (6 months before the intervention was broadcast), two population panels were selected through telephone screening from the population of smokers in the Chicago metropolitan statistical area. Their purpose was to provide baseline estimates of population smoking behavior prior to the intervention and to model changes in smoking behavior in that population absent any intervention. Because the two panels were selected to be equal in number, a disproportionate, stratified design was usedwith directory-based random-digit dialing. We screened 35 541 telephone numbers to obtain 2398 com-
pleted interviews. The population panels were selected from all smokers 18 years of age and older who resided in the six-county area comprising the Chicago metropolitan statistical area, smoked at least seven cigarettes a week, and viewed the late afternoon or evening news at least three times per week (including weekends). The two panels were distinguished by their channel of preference for viewing the evening news as determined by their responses to the initial interview and to a second preintervention interview 3 months later. Smokers who did not watch the evening news were dropped from the panels.
Study Samples The results reported here are from a sample of 2089 participants and a sample of 2398 smokers in the general population who reported viewing the evening news at
836 Amencan Journal of Public Health
Statistical Methods To compare the smoking cessation patterns among the participant and population samples over the 24 months follow-
panel.
as
the population
"lagged" analysis compared changes in smoking behavior among subjects in the participant sample and population panel at 50 days and at 6, 12, and 24 months past the "zero" point, defined as the time offirst inteview for the population panel and the date ofthe start of the iWervention for the participants. The matching point was arbitrarily selected at 50 days A
because the median time interval between the intervention zero point and the first postintervention interview was 50 days (see Figure 1). At zero point, all subjects were smoking. At each interview, the exact (or as nearly exact as possible) date of "quit" was ascertained. Because the final interviews were conducted after a 12month interval, there was no interview at 18 months for the participants or at 24 months for the population. (The population interviews began 6 months before the intervention.) The 24- and 30-month interviews
included carefully designed probes
for smoking status at 18 months for participants and at 24 months for the population. These data were used to estimate with reasonable confidence the quitting status at the relevant points of comparison for the 6-month point for both samples. The analysis focused on examining and comparing cessation patterns in the participant sample and the weighted, combined population panel, overall and within intervention exposure groups. Survival
analysis (event history) methodology was
used to model and plot time to cessation from the zero point. Because respondents were interviewed at different times, they were either quit at the time of interview or censored (final outcome not determined) at the date of the interview. Analyses were made at each interview wave so that each analysis answers the question: "For a respondent who was not smoking at a given interview point, when did the last quit without subsequent relapse occur?" The respondent had to have been quit for 3 or more days at the first postintervention interview and for 7 or more days at subsequent postintervention interviews to be considered a nonsmoker. Modeling was performed using the Cox proportional hazards method to adjust for demographic and smoking history variables on which the samples differed in preliminary crosstabulations. Because the proportional June 1992, Vol. 82, No. 6
24-Month Follow-up of a Teei
Smoki CessationmProgm
hazards assumptions were not met for exposure subgroups, this variable was entered as a stratfying variable. Cumulative cessation rate curves (complement of survival curves) were plotted for exposure groups.
Response Rates In all, 74% of the participants were retained through 24 months, and 51% of the population panel members were retained through 30 months. Rather extensive attrition occurredwithin the study period, particularly in the population panel. Because the sample was selected using random-digit dialing, in which telephone numbers are selected randomly and include unlisted and directory-assisted numbers, respondents were often lost to follow-up when they changed residence. Although we collected information about a person who would know how to reach the respondent, often this secondary source could not be located or refused to provide the information to locate the respondent. As shown in Table 1, males, Blacks, low-income, and less-educated respondents were overrepresented among respondents lost to follow-up compared with those who remained over the entire period. Although, as expected in large samples, many of the differences in the table are statistically significant, most of the actual differences are relatively small. Of more importance, the baseline smoking rates did not differ in those followed up compared with the rates in those lost to follow-up. The distribution of smoking behavior at the time of loss to follow-up did not change among those who remained and those who were lost (not shown in the table). Respondents who were not recontacted in a given wave but were recontacted at a later wave were questioned about their smoking status for the missing wave and were retained in the continuous cohort when their smoking behavior for the missing wave could be ascertained.
Results First, the intervention attracted a disproportionate number of heavy smokers (30 or more cigarettes per day) relative to the population (see Table 1). Female and Black smokers were also overrepresented among participants compared with the population.10 We examined point prevalence of cessation following the intervention and multiple point prevalence of cessation at subsequent observation points. Multiple June 1992, Vol. 82, No. 6
point prevalence is defined here as the cessation rate at any subsequent wave of data collection among those participants and members of the population panel who reported being abstinent for at least 3 days at 50 days post zero point and who reported being abstinent for at least 7 days at each intermediate contact. It is our estimate of continuous cessation; however, as in most studies where it is used, multiple point prevalence is biased because it includes as abstinent any persons who may have slipped between measurement points. The overall point prevalence of cessation among participants at 50 days post zero, adjusted for baseline smoking level, was 16%; the comparable point prevalence rate in the population panel was 4%. At 6 months, the adjusted multiple point prevalence of cessation was 9% for the participants and 3% for the population; at 12 months, the rates were 6% and 3%, respectively; and at 24 months, they were 6% and 2%, respectively. Relapse during the 6 months following the intervention was greater for the participants than for the population; nevertheless, the 24month multiple point prevalence rate
among the participants was three times that observed in the population. Table 2 presents these data broken down by average number of cigarettes smoked daily at baseline. Several things are worth noting in the table. First, the point prevalence of cessation at post intervention for each level of baseline smoking is higher among participants than among the population panel, consistent with an intervention effect. Second, light smokers at baseline were more likely than heavy smokers to be abstinent at 50 days post zero. However, although the heaviest smokers among the participants quit at two thirds the rate ofthe light smokers, the rate of cessation among the heaviest smokers in the population was only 25% ofthe rate among light smokers. Third, the rate of relapse at 24 months postintervention among light smokers was 70%o among the participants and 62% among the population; among the heaviest smokers, it was 60% among the participants and zero among the population. These comparisons suggest that the heaviest smokers in both groups were somewhat better able to sustain their decision to quit than the light Amencan Joumal of Pubhc Health 837
Warnedke et al.
smokers but that the benefits of the program were most obvious in the high initial quit rates among heavy smokers who participated. That is, among light smokers the ratio of initial point prevalence of cessation of participants to that of the population was about 3 to 1 in favor of participants; the comparable ratio among heavy smokers was 7.5 to 1, also in favor of participants. Figure 2 presents the cumulative cessation rate curves at 50 days post zero and at 6, 12, and 24 months following the intervention, adjusted for gender, race, education, smoking rates at baseline, and self-reported health status. Separate curves were plotted for various levels of participation in the intervention. Participation (n = 1550)was defined at three levels for the plots presented here: high participation (10%) accounts for those who watched the televised segments daily and referred to the manual daily; moderate partic4pation (37%) accounts for those who watched the televised segments at least once a week and referred to the manual at least two to six times a week; and lowparicpation (53%) accounts for those who participated to any lesser extent. In Figure 2, at each interview wave, even at 24 months, the cumulative proportion quit among high participants was considerably greater than that among moderate participants, low participants, and the population panel. Consistent with the data in Table 2, the greatest contrast is between high participants and the population panel at every point. Also consistent with Table 2, the initial increases in slope for the high and moderate participants indicate that, among those who were abstinent at 24 months, most of the cessation occurred as part of the intervention and clustered around day 18, the quit 838 American Journal of Public Health
day specified by the program. The gradual increase in the slope of the low participants is similar both to that of the population and to what would be expected with no intervention. In Table 3, we present results of the Cox proportional hazards method, predicting cessation across waves of intervention as a function of demographic characteristics, baseline smoking behavior, and overall health status. The variables above the dotted line, which are the variables associated with cessation on which the participants and the population panel differed, are the covariates that were controlled in the Cox proportional hazards analysis. Race and education are sufficiently correlated so that neither appears to have an independent, significant association with quitting. However, having a college education is significantly and positively associated with cessation at postintervention compared with having less than a high school education. The association appears to reverse at 12 months. Race is significant only at 24 months, although non-Blacks were more likely than Blacks to be consistently abstinent across all waves. Males were more prone than females to have quit at all points, although the relationship is not statistically significant at 24 months. Health is also an important factor. Those in good health were more likely to be abstinent at each observation; however, the relationship is not significant at the 12month observation. Only baseline smoking level is consistently associated with cessation through 24 months: lighter smokers were more likely to quit than heavier smokers. Below the dotted line in Table 3 are the risk ratios associated with participation level. Because the proportional haz-
ards assumptions were not met for the participationvariable, the confidence interval
estimates should be viewed with caution as they tend to average across time. The inferences support the findings from the cumulative cessation plots that those in the high and moderate participation groups were more likely to quit and remain quit than were those in the population panel, although the effects are not always significant.
Dicussion Participants who followed the program with the greatest regularity were more likely to quit smoking and to sustain their abstinence over 24 months than were those who participated less intensively or individuals in the population who quit without the inteivention. Moreover, as seen from the curves in Figure 2, there was a dose-response relationship between level of participation and maintenance of abstinence over 24 months, with larger risk ratios for higher participation categories. Furthermore, the cumulative cessation rate curves show clearly that most of those who were abstinent at 24 months quit initially during the intervention period. Finally, the longterm effects of participation in the intervention were especially evident for heavy smokers, given that those who participated were more likely to quit than were those in the population panel, who did not have access to the intervention (see Table 2). Moreover, among the participants, although light smokers were more likely
to
quit postintervention, heavy
smokers who quit were somewhat more likely to be abstinent at 24 months. Thus, June 1992, Vol. 82, No. 6
24-Month Folbw-up of a Teised
,CPgm
q
6 Months
Postintervention 045
.Hub _
,
_
0.40
0.45
I 035
0.40 0.35
0.30
0.30
0.25
0.25
0
Q20
&20
0
015
015
-
0.10
.20g a .9
0
C.)
W
Low P
MO
Lowb pUh.k"
LAW ~Modr_ _II -11 on E :O_dm
-
L,
_uIh1pn 0.1P
0.05
0.05
0.00
9.09 20
40
so
60
0
150
Days Elapsed
12 Months
24 Months
0.45
0.45
Q 0.40
0.40
a 0 ., 0.30
0.35 0.30
&
0.25
125
2
0.20
0.20
a
120
40
Days Elapsed
0.35
Hb -
r_-
*.15 a
O
0.10
{0.I
0A05
0.s
0.00
0.00
P0
IS0
360
270
0
90
0SO
Days Elapsed High Participation
=
Watc
270
30
450
540
40
720
Days Elapsed
tesed segments
diWly and referred to manual daily
Moderate Participation = Watched televised segments at least once a week and referred to manual at least two to Sx times a week Low Participatm = Parficipated to any lesser extent
FIGURE 2-Cu
na
cmson rates
L. a
all the data are consistent with an intervention effect and suggest particular benefit for heavy smokers. The target population in this study was defined as those smokers who watched the evening news on the intervention television channel. The demographic characteristics of the viewing audience indicate that the news audience for this channel contained a disproportionate number of those who are likely to be the smokers in the year 2000. These include heavy smokers, low-socioeconomic females, Blacks, and blue-collarworkers.4"6 Therefore, the fact that the participants drawn from this audience did so much better than the quitters from this audience who attempted to quit without the intervention is also quite encouraging, as are the positive effects of the program on heavy smokers who participated. This latter finding is particularly noteworthy,
June 1992, Vol. 82, No. 6
given that heavy smokers were represented disproportionately relative to their distribution in the population. Based on the risk ratios in Table 3, however, light smokers at each wave of observation were most likely to be abstinent. Although the relationship is not consistently significant, males appear to have had more
with stopping than fedid smokers in good health. Non-Blacks were more likely than Blacks to quit, and more education favored cessation during the immediate postintervention period. Data from this study reported elsewhere indicate that the heaviest users of the manual were highly educated females.10 These data also indicate that, although less-educated, Black females were consistent viewers of the televised intervention, they did not read the manual. Males were unlikely to have participated in
males,
as
success
any aspect of the intervention, but those who did were most likely to use the manual. Thus, although these data are encouraging about the potential of such interventions, they need to be read with some caution. High and even moderate participation were correlated with higher rates of cessation than was low participation; in fact, those who registered and participated only slightly did worse at some waves than did the population panel, as can be seen from the bottom line in Table 3. On the other hand, this intervention may be reaching the same population that has responded to other interventions. This would imply that more attention must be given to developing materials that support the televised component and will make the intervention attractive to and used by those segments of the smoking population who are not being reached by current inAmerican Joumal of Public Health 839
Warnecke et aL
References 1. Thornberry OT, Wilson RE, Golden PM. Health promotion data for the 1990 objectives. Advance data from J4tal and Health Statistics, No. 126. Hyattsville, Md: Nahb* 145*121*1~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . .11..25* tional Center for Health Statistics; 1986. 0Imc~~~~~~~~~~~~~~Ic 0790)7 Q.U 0.79'~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~........... 2. US Departnent of Health and Human Services, Public Health Service. Promoting 1iV~~~~~~~~~~~~~~~~~~~........ ......0)4 Health/Peventig Disease: Objectives for 0)4 the Nation Washhigton, DC: DHHS; 1980. .......5 ......7.......... 3. Greenwald P, Sondik EJ. Cancer Control Objectives for the Nation: 1985-2000. Bethesda, Md: National Cancer Institute; 1986. NCI Monographs, National Insti...) tutes of Health publication 86-2880. 2C1-2va 0-19 0)2' 0.78' 0.75* 0.71~. ......... 4. Pierce JP, Fiore MJ, Novotny TE, Hatzian....4...............i,0 H10~~~~~~~~ 7A5 Z00' 127~~.................. dreu EJ, Davis RM. Trends in cigarette .......................W...........1...(..... .2...... smoking in the United States: projec..........1.... tions to the year 2000. JAMA 1989;261:61(115~~~~~~~~~~~~~~~~~.2...................0~5~ ~3 65. 4p
