Preventive Medicine - Science Direct

Preventive Medicine 32, 118–127 (2001) doi:10.1006/pmed.2000.0798, available online at http://www.idealibrary.com on

Effect of a Community Action Program on Adult Quit Smoking Rates in Rural Australian Towns: The CART Project1 Lynne Hancock, B.Sc.(Hons), Ph.D.,*,†,2 Rob Sanson-Fisher, B.Psych.(Hons), M.Clin.Psych, Ph.D.,† Janice Perkins, B.Psych.(Hons), Ph.D.,*,† Ann McClintock, M.Ed. Admin., B.A., B.Ed., Dip.P.E.Health(FACHPER),‡ Peter Howley, B.Math.(Hons)(Stats),§ and Robert Gibberd, Ph.D.§ *Hunter Centre for Health Advancement, Newcastle, NSW, Australia; †Faculty of Medicine and Health Sciences, University of Newcastle, Australia; ‡NSW Cancer Council, Sydney, NSW, Australia; and §Health Services Research Group, University of Newcastle, Australia Published online January 5, 2001

Background. This article describes one outcome of a randomized controlled trial of community action for cancer prevention. The aims of this article were to (a) explore the effectiveness of a community action program in decreasing community smoking rates in rural Australian towns and (b) describe the relationship between adult smoking quit and uptake rates and demographic variables. Methods. In 1992, 20 towns were selected for randomization. Community action involved formation of community committees and utilization of access point networks to initiate and maintain intervention strategies. At post-test, outcomes were proportion of “quitters” from a cohort of self-described smokers, proportion of “uptakers” from a cohort of selfdescribed nonsmokers, and “net effect." Results. Differences in quit rate, uptake rate, and net effect for intervention compared to control condition favored the intervention in all cases, although mainly nonsignificant. Significantly more male smokers quit in intervention towns than in control towns [7.0% (95% CI: 0.6, 13.5)]. Conclusions. Given that CART utilized and improved upon strategies argued as effective in the literature, the limited success of the project in reducing adult smoking, considered in combination with COMMIT

1 The Cancer Action in Rural Towns (CART) project was a collaborative project jointly funded by the National Health and Medical Research Council (Australia) and the NSW Cancer Council (Australia) (Professor Rob Sanson-Fisher, Principal Investigator). Our sincere appreciation goes to all community members and NSW Cancer Council staff who were involved in the CART project. 2 To whom reprint requests should be addressed at Hunter Centre for Health Advancement, Locked Bag 10, Wallsend 2287 NSW, Australia. Fax: 61-2-49246-209. E-mail: [email protected]. nsw.gov.au.

findings, suggests the need for further innovation in the field. 䉷 2001 American Health Foundation and Academic Press Key Words: smoking; smoking cessation; adult; neoplasm (prevention and control); consumer participation.

INTRODUCTION

After the year 2000, cancer is likely to become the major cause of death in developed countries [1–3]. It is estimated that smoking is responsible for 30% of cancer deaths [1,4] and 90% of cancer of the lung, trachea, and bronchus in Australia [1,4]. Lung cancer has the highest incidence of any cancer for men in Australia, and is the second most common cancer for women [5]. Passive smoking is estimated to contribute to 2.6% of all lung cancer deaths in Australia [5]. Up to 31% of Australians smoke [6]. Men are more likely to report being daily smokers than women, and smoking rates vary according to region [6]. Rural communities in Australia have recently been identified as in particular need of targeted public health programs [7]. Such communities suffer from higher rates of some cancers, and are often not serviced well by existing health care delivery systems and public health agencies. As a consequence, there are lower rates of screening for cancers such as cervical cancer [8] and higher rates of smoking [6]. An effective method of reducing mortality and morbidity from cancer is through primary and secondary prevention. However, despite intensive public education campaigns, community rates of behavior change for cancer prevention remain low [1–3]. One potentially promising strategy to encourage populations to engage in cancer-related preventive behaviors is a community action approach [9,10]. A community action approach

118

0091-7435/01 $35.00 Copyright 䉷 2001 by American Health Foundation and Academic Press All rights of reproduction in any form reserved.

ADULT QUIT SMOKING RATES IN RURAL AUSTRALIAN TOWNS: THE CART PROJECT

to health promotion requires multiple intervention components across multiple access points with community participation integral to program implementation and maintenance. Community action approaches have particular relevance for rural communities, since small integrated communities with strong existing informal networks optimize the effectiveness of community action processes [9]. Moreover, the emphasis on local control is well adapted to the existing ethic of rural communities who frequently prefer to be as independent as possible from distant centralized services, which are often perceived as not being aware of local needs [9,10]. While there is some evidence that community action programs can be effective in decreasing modifiable risk factors and mortality from cardiovascular disease [11,12], there have been very few methodologically adequate studies that have assessed the impact of community action programs on cancer-related health behaviours [13–15]. The COMMIT study, which aimed to evaluate the effect of a range of standardized intervention components on smoking rates in 11 matched pairs of communities throughout the United States and Canada, had only a moderate degree of success in reducing community smoking rates [13,14]. While there were no differences in quit rates for heavy smokers between control and intervention communities, a modest difference in quit rates for light to moderate smokers was achieved [13,14]. This article describes one of the outcomes of a randomized controlled trial of community action for cancer prevention, Cancer Action in Rural Towns (CART). The primary aim of the CART project was to evaluate the effectiveness of a community action program in increasing community rates of preventive and screening behaviors relating to breast, cervical, smoking-related, and skin cancer in rural Australian towns. The aims of this article are to: a. Explore the effectiveness of a community action program in decreasing community adult smoking rates in rural towns of Australia. b. Describe the relationship between adult smoking quit and uptake rates and demographic variables. METHOD

Design Twenty rural Australian towns (with an adult population between 5,001 and 15,000) were selected for inclusion in the CART project. Towns within this population range were chosen for several reasons. First, there needed to be sufficient towns in the population range to allow for appropriate matching of 10 town pairs. There were 43 eligible towns in rural NSW, Australia, in the population range 5,001 to 15,000 adults (that is, not within 50 kilometres of the 3 major urban centres of NSW), allowing a reasonable pool from which to select

119

study towns. Second, Henderson and Thomas have argued that the appropriate catchment area for community programs should be between 6,000 and 20,000 population [16]. Given differences in Australian town sizes compared to those of the United Kingdom; it was decided that 5,001 to 15,000 approximated this population bracket. Third, based on the experience of this research group during a previous community development project, it was felt that this size community would be both large and small enough to allow for effective activation strategies [17]. For the randomized trial, a matched pairs design was used, with towns matched on demographic, infrastructure, and geographical variables. One town from each pair was randomly allocated to either experimental or control condition. The design and matching process are detailed in another publication by the CART group [18]. Baseline data collections were conducted from June 1993 to March 1994. Cohort follow-up occurred from January to March 1997. The CART Intervention The community action intervention involved the formation of community committees and the utilization of access point networks to initiate and maintain intervention strategies within each town. The approach used in the CART intervention was developed with reference to similar projects [11,17,19] and extensive consultation with identified experts in the field, local health promotion workers, and lay people. A main feature of community action is the ability to deliver the same message through many channels, increasing the probability that the program activities will be noticed and acted upon within the community. It was expected that access point networks would provide an ideal vehicle to achieve this multichannel approach [9,10]. Although the project team’s approach to towns was standardized, and extensive materials for activation were provided to encourage similar activities in each town, it was expected from the outset that flexibility would be the key to securing and maintaining community involvement. Consequently, activities and time frames varied considerably among communities. As a starting point, community facilitators (NSW Cancer Council Health Education Officers with experience in community-based cancer reduction strategies) were selected for each intervention town. Facilitators were external mobilization agents whose role was to act as a link between the community and the research team. CART was introduced to each intervention town through an extensively advertised open invitation community meeting, which was chaired by an invited key person from the community (usually the town mayor). During this meeting, members of the community were invited to form a CART Committee. A balance of representatives across key community access points was

120

HANCOCK ET AL.

sought (health care providers, community organizations, media, retailers, schools, and workplaces) [18]. These six key access point networks were the main vehicle of the CART intervention [18]. During the last 6 months of the program period, an attempt was made to accelerate activities by appointing a community liaison officer from within each town, to act as a resource for achieving program targets. Community liaison officers were supervised by facilitators in their day-to-day tasks, and were expected to be much more “hands-on” than facilitators. Standard strategy packages were developed preimplementation, which outlined targets for the program and recommendations for activities in each access point. Resource packages with pamphlets, posters, draft letters, and other resources were provided for each access point, to aid in conducting activities. While these access point packages were progressively refined and abbreviated throughout intervention implementation, in response to community feedback, the standard recommendations for types of strategies considered effective did not change. The introduction of the intervention was staggered across the 10 towns, from June 1993 to March 1994. The active phase of the intervention continued until November 1996. The CART intervention model has been previously described [18]. CART Adult Smoking Activities The target for adult smoking activities was to reduce adult smoking by 5%. The main strategies recommended by the project team for implementation by community committees were to promote smoke free environments in workplaces, restaurants, hotels, motels, sporting clubs, community clubs, and eating places, and through the media; increase the number of workplaces that provided self-help quit smoking materials and information pamphlets and “quit smoking” programs for employees; and increase the number of employers and employees who were aware of the effects of passive smoking on health, and associated legal repercussions. The project team recommended that 50% of the program effort in adult smoking should be directed to workplaces (an identified vulnerable group for smoking) [20– 22], 25% to the media [23], and 25% to community organizations (such as sporting groups). The strategies recommended for implementation in towns were identified through consideration of the relevant literature, past work by this group, and extensive consultation with experts in the field and were seen as having a high probability of achieving success in reducing smoking within a community action program [24]. Figure 1 summarizes the types of adult smoking activities that were conducted in towns. From the tracking of activities conducted, there were no particularly notable variations

in overall level of activity between towns for this target behavior. Evaluation Sample and Procedure The main post-test outcome measures were proportion of “quitters” from a cohort of self-described smokers and proportion of “uptakers” from a cohort of self-described nonsmokers (including ex-smokers). Permission for conduct of both baseline and follow-up surveys was obtained through the University of Newcastle Human Research Ethics Committee. A simple random sampling approach was used for the baseline survey. One thousand households from each rural town were selected for phone contact using random telephone number lists. The procedure for baseline data collection is fully detailed in a previous publication [25]. Only adults of ages between 18 and 70 years were included in the smoking survey sample, as other outcomes of the CART study were inappropriate outside this age limit (for example, cervical and breast cancer), and samples were standardized across the different study outcomes. Only people consenting at baseline for recontact were followed-up. All consenting smokers identified from baseline, and a random sample of 300 nonsmokers per town were mailed information letters and contacted for resurvey, 3 years later. At least 1 week following the information letter mail-out, participants were phoned and a computer-assisted telephone interview was conducted with consenting people. An extensive participant tracking protocol was utilized, given that followup was 3 years after baseline. If participants were no longer resident at the original contact phone number, and no follow-up address or phone number could be given, then a relative whose name and phone number was collected at baseline was contacted. If participants could not be contacted through this relative, then Internet Australian Telephone White Pages were searched by name for the town, for NSW, and then for Australia. All matching names were contacted. Some people were still not contactable despite this tracking protocol (see Table 1). Measures The questionnaire schedule included demographic and smoking status questions. Given that adult selfreport of smoking is known to be relatively reliable, and the deception rate could be expected to be similar for baseline and follow-up [26,27], external validation of self-report was not conducted. Items used were similar to those used in past published studies of this type, and underwent an iterative critical review process with clinicians and identified experts in this field [13,26,27]. The questions of interest to this article that were asked at baseline and follow-up were:


121

FIG. 1. Summary of antismoking activities conducted in CART towns.

• demographics (standard questions on age, gender, country of birth, language spoken at home and aboriginality) • self-description of smoking status (regular smoker, occasional smoker, ex-smoker, never smoked) Statistical Analyses Smokers were defined as all self-described smokers (regular or occasional). The proportion of people quitting and taking up smoking from pre- to post-test and the net effect of these two behaviors at post-test were examined across treatments (intervention, control), by town pair (1 to 10), gender, and age (18–29 years, 30–44 years, 45–70 years). The overall proportion of persons quitting or taking up smoking in each treatment group was estimated. An unweighted proportion was used since the size of the population in towns was similar. To estimate the

standard errors of these proportions, the two-stage cluster sampling formula was used [28], where the random selection of the town was the first stage, and the selection of adults, the second stage. The differences in the proportions of people quitting and taking up smoking between paired towns (intervention/control) were obtained by subtraction. Calculation of the standard errors of the difference between the paired towns was based on that for differences in proportions. A similar approach was used to obtain the “net effect” proportion. The smokers cohort only was used to calculate quit rates. Quit rate was defined as number of smokers at Time 1 who had become ex-smokers by Time 2, over the number of smokers contacted at Time 1. The nonsmokers cohort only was used to calculate uptake rates. Uptake rate was defined as number of nonsmokers at Time 1 who had become smokers by Time 2, over the number of nonsmokers contacted at Time 1. Quit and

122

HANCOCK ET AL.

TABLE 1 CART Baseline Samples and Eligible, Contactable Follow-up Samples Follow-up (eligible and contactablea)

Baseline Town

Smokers (n)

Nonsmokers (n)

1 3 5 7 9 11 13 15 17 19

138 166 159 120 136 114 109 102 87 149

418 479 417 434 398 403 361 344 387 343

2 4 6 8 10 12 14 16 18 20 Total

121 145 112 120 122 101 72 117 112 81 2383

333 435 324 382 369 368 293 341 336 347 7512

a b

Smokers (n)

Contact rate (%)

Nonsmokersa (n)

Contact rateb (%)

62 69 69 73 68 74 66 71 72 72

185 167 199 173 194 161 173 149 157 193

44 35 48 40 49 40 48 43 42 56

62 68 71 66 68 80 68 69 69 63 69

184 202 193 170 180 155 153 167 152 150 3457

55 46 60 45 49 42 52 49 45 43 46

Intervention 86 115 109 88 93 84 72 72 63 94 Control 75 98 80 79 83 81 49 81 77 57 1636

While all smokers were followed-up, only a sample of nonsmokers was selected for follow-up. From whole Time 1 sample.

uptake rates were calculated by treatment (intervention/control), town pair, gender, and age. The difference between rates was calculated by treatment. Confidence intervals and standard errors were used to examine whether significant differences occurred between treatments [28]. To ensure that the potential confounding variables of age, sex, and town pair did not influence the estimate of the intervention effect, logistic regression was used to determine the statistical significance of the treatment, while including the confounders [28]. A single statistic was used to estimate the net impact of smoking status changes pre- to post-test. Given that at Time 1 about 25% of people were smokers and 75% were non-smokers, the statistic to consider the net effect was calculated as: 0.25 ∗ propnquit ⫺ 0.75 ∗ propnuptake.

Again, confidence intervals using standard errors based on the two stage sampling formula were used to examine whether significant differences occurred between treatments [28].

RESULTS

Sample At baseline, 20,350 households were selected to participate in the survey, of which 4,835 (24%) were noncontactable ("return to sender” mail, vacant houses, changed or disconnected telephones, no answer to telephone during the survey period), and 2,825 (14%) were ineligible (infirm, older than 70 years). In the contacted eligible households, 10,256 (81%) randomly selected participants (the “next birthday” person in each household) consented to take part. A contact rate of 50% was achieved. The baseline sample had a higher proportion of women, a lower proportion of the younger age groups, and a higher proportion of the oldest age group when compared to the 1991 Census population, as reported in detail in a previous publication [25]. At follow-up, 2,383 smokers and 5,996 nonsmokers were sent information letters: 191 (8%) smokers and 294 (5%) of the contacted nonsmokers sample either refused to participate or did not complete the interview; 158 (6%) smokers and 175 (3%) of the nonsmokers sample were ineligible (died, infirm, moved out of area 2 or more years ago); 398 (17%) smokers and 2,070 (34%) nonsmokers were noncontactable ("return to sender”


mail, disconnected telephone numbers, no known address, not available during survey period). Of the contacted eligible participants, 1,636 smokers (90%) and 3,457 nonsmokers (92%) completed the interview. A contact rate of 69% for smokers and 58% for nonsmokers (46% if the original sample size, rather than only those who were followed up are considered) was achieved. Table 1 details sample sizes across towns. There were some demographic differences between the Time 1 and Time 2 smoker samples. While about 70% of most smoker age groups were recontacted, only 48% of the youngest age group (18 to 24 years) were contacted at follow-up. More women smokers (71%) were contacted at Time 2 than were male smokers (66%), although this difference was not large. Baseline Smoking Rates Table 2 provides the proportion of smokers at baseline, according to treatment, overall, and by town pair, gender, and age group. Overall Differences Quit rates Overall, 20.4% of smokers in intervention towns had quit at Time 2, and 16.9% of smokers in control towns. The difference in the proportion quitting (intervention ⫺ control) was 3.5%, which although in the right direction, was not statistically significant (95% CI: ⫺0.3, ⫹7.2). Uptake rates. Overall, 3.8% of nonsmokers in intervention towns became smokers at Time 2, and 4.2% in

control towns. The difference in the proportion taking up smoking was ⫺0.3%, which, although again in the right direction, was not statistically significant (95% CI: ⫺2.2, ⫹1.6). Net effect. The net effect of quitting less uptake was 2.2% for intervention towns and 1.1% for control towns. The difference in net effect between treatments was 1.1% (95% CI: ⫺0.6, ⫹2.8). Age Differences Quit rates. Table 3 details quit rates, uptake rates and net effect by age group and gender, for treatment. The rate of quitting appeared to be affected by age, with the oldest age group experiencing the greatest proportion of quitters. This occurred in both the control and the intervention towns, as shown in Fig. 2. Although the intervention towns, as a group, had higher proportions of quitters, no age group showed a significant difference due to the intervention. Uptake rates. The rate of smoking uptake also depended upon age, with the youngest age group experiencing the greatest uptake rate and the rate decreasing with increasing age. This trend is evident in both the control and the intervention towns (Fig. 2). Although the intervention towns, as a group, had lower uptake rates, no age group showed a significant difference due to the intervention. Net effect. For all age groups, the net effect of quitting less uptake was in the direction of an intervention effect, but differences were not statistically significant. Gender Differences

TABLE 2 Baseline Percentages of Smokers for Control and Intervention Treatments Overall, by Town Pair, by Gender, and by Age Proportion smoking

Overall Town pair 1 2 3 4 5 6 7 8 9 10 Gender Male Female Age group 18–29 30–44 45–70

123

Control

Intervention

23.8

24.3

26.7 25.0 25.7 23.9 24.8 21.5 19.7 25.5 25.0 18.9

24.8 25.7 27.6 21.7 25.5 22.1 23.2 22.9 18.4 30.3

26.9 22.1

26.2 23.1

34.3 28.7 18.2

30.2 28.6 19.9

Quit rates. Table 3 details quit rates, uptake rates, and net effect by gender and treatment. For males, there was a significant difference in the proportion of smokers quitting at Time 2 (intervention ⫺ control) of 7.0% (95% CI: 0.6–13.5). For females, while the difference in the proportion quitting smoking was in the right direction, it was not statistically significant. Uptake rates. For males and females, while the difference in the proportion taking up smoking was always in the right direction, there were no statistically significant differences. Net effect. For males and females, the net effect of quitting less uptake was in the direction of an intervention effect, but differences were not statistically significant. Predictors of Quitting and Uptake The logistic regression model exploring the predictors of quitting failed to indicate the treatment group as a significant predictor (P ⫽ 0.09) once age was included

124

HANCOCK ET AL.

TABLE 3 Quit Rates, Uptake Rates, and Net Effecta by Age and Gender, by Treatment Characteristic

Control

Intervention

Difference

Standard error

Lower 95% confidence limit

Upper 95% confidence limit

Percentage Age 18–29 years Quit rate Uptake rate Net effecta 30–44 years Quit rate Uptake rate Net effecta 45–70 years Quit rate Uptake rate Net effecta Gender Male Quit rate Uptake rate Net effecta Female Quit rate Uptake rate Net effecta a b

18.3 12.7 ⫺5.0

19.6 7.7 ⫺0.9

1.3 ⫺5.0 4.1

6.3 ⫺5.0 3.9

⫺11.1 ⫺14.2 ⫺3.5

⫹13.7 ⫹4.2 ⫹11.7

12.8 5.6 ⫺1.1

17.1 5.9 ⫺0.1

4.3 0.3 0.9

2.7 1.6 1.4

⫺0.9 ⫺2.9 ⫺1.8

⫹9.5 ⫹3.4 ⫹3.6

21.1 2.2 3.6

23.9 2.0 4.5

2.8 ⫺0.2 0.9

3.1 0.8 1.0

⫺3.4 ⫺1.8 ⫺1.0

⫹8.9 ⫹1.3 ⫹2.8

14.0 3.4 0.9

21.0 3.1 2.8

7.1 ⫺0.3 2.0

3.3 1.1 1.2

⫹0.6 ⫺2.4 ⫺0.3

⫹13.5b ⫹1.9 ⫹4.2

18.7 4.5 1.3

20.0 4.3 1.8

1.3 ⫺0.3 0.5

2.4 1.1 1.0

⫺3.3 ⫺2.4 ⫺1.4

⫹5.9 ⫹1.8 ⫹2.5

0.25 ∗ propnquit ⫺ 0.75 ∗ propnuptake. Significant difference.

in the model. Age group was the only statistically significant variable (P ⫽ 0.001), with the oldest age group (45–70 years) experiencing the greatest proportion of quitters and the youngest age group (18–29 years) showing the next greatest proportion of quitters (see Fig. 2). The odds in favor of a 30- to 44-year-old quitting were 0.76 (95% CI: 0.50, 1.14) times the odds of an 18to 29-year-old quitting, while the odds of a 45- to 70year-old quitting were 1.25 (95% CI: 0.84, 1.86) times the odds of an 18- to 29-year-old quitting. This shows the major difference in proportions quitting occurs between the 30- to 44-year-old and the 45- to 70-year-old age groups. For uptake of smoking, the statistically significant variables were age group (P ⬍ 0.001) and town pair (P ⫽ 0.009). Again, treatment group was not shown to have a significant effect (P ⬎ 0.5). The youngest age group (18–29 years) was more likely to take up smoking than the other age groups (see Fig. 2). The odds of a 30- to 44-year-old taking up smoking were 0.50 (95% CI: 0.31, 0.81) times the odds of an 18- to 29-year-old commencing smoking, while the odds of a 45- to 70year-old commencing smoking were 0.17 (95% CI: 0.10, 0.29) times the odds of an 18- to 29-year-old beginning. The main effect of town pair was attributable to pair 7 (towns 13 and 14), as town 13 (an intervention town) had the highest uptake of all towns. Exclusion of pair 7 yielded town pair an insignificant predictor (P ⫽ 0.2).

DISCUSSION

Community-wide interventions present a particular challenge to evaluaters. Ensuring rigorous scientific evaluation involves consideration of several domains, including study design, sampling and control procedures, and analysis issues [29,30]. This article describes one outcome of a randomized controlled trial of community action for cancer prevention, CART, which aimed to evaluate the effectiveness of a community action program in decreasing community smoking rates in rural Australian towns, using a standardized approach and scientific evaluation procedures. Several methodological issues should be considered before discussing the results of this evaluation. One of the major issues of evaluation of community-wide interventions is to ensure that an adequate number of towns are involved, and that control procedures are adequate. Randomization was an important element of the CART design, which involved randomization to intervention or control. The study design was ambitious, and the cut-point of 10 towns per condition was mediated by cost as well as advice from biostatisticians that 10 towns per condition should be considered the lower limit [30]. However, the results of this study demonstrate that to show a significant difference of 3.5% in quit rates over a 4-year period, 40 towns overall would be required for the study. Besides doubling the


125

FIG. 2. Quit and uptake rates by age group for control and intervention treatments.

cost of the study, the public health benefit of detecting this small difference is arguable. Obviously, studies of this type will always be constrained by the costs of evaluation, where the evaluation team is located at distance from the multiple study communities (for CART, up to 600 km distant in some cases). Costs of the CART intervention were defrayed by the support of the NSW Cancer Council in providing free resources (such as posters, pamphlets), funding for the development of other resource materials (such as training manuals and CART-specific resources), staff time (town facilitators), and expert advice on community outreach. CART attempted to maximize the chances of program success by using small towns (population 5,001– 15,000), which have been suggested to be more amenable to community action interventions [16]. However, this decision had implications for the sample size at second stage. The initial contact sample size for baseline data collection was 1,000 households per town. Given the size of the towns, this would have represented a significant proportion of the population for some towns (up to 40%, for a town of 5,000, assuming two members per household). Thus, the initial contact could not have been much more extensive, particularly in control towns, where saturation sampling could lead to concern over a potential Hawthorne effect [31]. At follow-up, the sample size for the smokers cohort dictated how many nonsmokers were recontacted. With loss to follow-up (at 21% for smokers and 42% for nonsmokers), the final sample sizes were in some cases less than optimal, with consequent reduced statistical power. While overall results were not compromised by the sample size, more detailed analyses such as by sex and age, and by individual smoking status (occasional or regular), in some cases involved quite small samples within individual towns (see Table 1). The recontact rate of only 48% for the younger age group was of particular concern, given that this demographic was more likely to take up smoking and less likely to quit than the older age groups. It is a plausible hypothesis that

the younger nonrespondents would be more likely to be continuing smokers and taking up smoking, which could have meant a more negative result for this age group, had they responded to the survey [32]. It must be acknowledged that the study did not conduct external validation of smoking self-report. There were several reasons for this decision to rely on selfreport alone. First, there is good evidence that adult self-report of smoking is relatively reliable [27], and that external validation is not needed. Second, past studies by this group have shown that the deception rate for adult smoking is small [33,34]. Third, it was assumed that the deception rate could be expected to be similar for baseline and follow-up [26,27]. Last, the considerable cost of external validation in the field across 20 communities was not justifiable in view of the evidence for validity of adult smoking self-report. Another issue for this study was the length of the intervention, which was between 3 and 3.5 years, as length did vary slightly between towns due to the staggered implementation of the intervention. While an intervention of this length may be reasonable for many types of programs, it must be acknowledged that community activation can have a very long lead time. Additionally, lead time can vary between communities, depending upon the available resources and community motivation. For CART, the project team did in fact feel that more time may have benefited some towns, but overall, most towns expressed that they were ready to move on from the project by the agreed program end. It is doubtful that continuing the intervention in this community climate would have produced an increased treatment effect. One recommendation from our experience with lead times would be to encourage other practitioners to ensure that programs of this type target already motivated communities, rather than attempt to motivate less interested communities. CART was of course constrained by randomization from choosing motivated towns.

126

HANCOCK ET AL.

The main post-test outcome measures were proportion of “quitters” from a cohort of self-described smokers, proportion of “uptakers” from a cohort of self-described nonsmokers (including ex-smokers), and “net effect” of quitting and uptake. These measures were explored overall, and by town pair, age group, and gender. When differences in quit rate, uptake rate, and net effect were compared for intervention versus control condition, in all cases the difference was in the right direction (favoring the intervention). However, the effect was not large enough to achieve statistical significance except for male quitters. As can be seen from Fig. 2, there were age effects for quit and uptake of smoking. Not surprisingly, smoking and ex-smoking appears to range along a dynamic continuum. It is well documented within the literature that smokers may make several attempts before successfully quitting [35]. The results of the CART study show that quitting seems to follow a U-shaped curve according to age. The younger and older age groups quit at higher rates than did the middle age groups. The younger age groups also had greater uptake rates than the other age groups, again demonstrating the mobility of smoking status in this age group. This finding makes intrinsic sense. It is known from past research that younger smokers are more likely to be “social” smokers, so their smoking behavior is likely to be more labile, with both high quitting and high uptake rates. For the middle years, smokers are most likely to be addicted smokers whose status remains more stable. In the later years (from 45 years), smokers are potentially suffering the beginnings of serious health effects from smoking [2]. At this stage, a good proportion of smokers would begin to quit. This postulated scenario is further supported by the uptake rates data, where uptake was high for the young age group, but reduced as age increased. Similar to the findings of the COMMIT study, there was an overall trend in favor of an intervention effect, which failed to reach significance [13,14]. In contrast to COMMIT, which exhibited more effect for female smokers, male quitters were the only group where the CART intervention effect was statistically significant. However, this result is easily explained for CART, where the access points for intervention concentration were workplaces, community organizations, clubs, and hotels, which are traditionally more male oriented, particularly in rural Australia. This focus was deliberate, given that adult male smoking rates were consistently higher than for females at baseline (see Table 2). COMMIT did not consider uptake rates or net effect of quit and uptake. As seen from the CART results, however, uptake rates should be a consideration for evaluations of this type. As discussed above, smoking and ex-smoking ranges along a dynamic continuum and consideration of quitting only in this type of evaluation would indeed only tell half the story.

The results of this evaluation are somewhat disappointing for a project that consumed so much community and project team energy and time. While a trend in the intervention direction is clear and consistent across analyses, this effect was too small to reach significance in most cases, both from a statistical and a public health viewpoint. While there are a number of potential methodological weaknesses with this study, such as limited number of communities randomized and reliance on self-report, these results should regardless be cause for reflection, when considered in combination with the recent COMMIT trial findings. In considering this limited effect, it should be noted that best-practice was used: the intervention involved the utilization of access points that have been argued to reach the target groups [20,21], and utilized, and in some ways improved upon, strategies argued as effective in the literature and seen as effective by experts in the field [24]. The limited success of the project suggests the need for further innovation in the field. REFERENCES 1. Australian Cancer Society. National Cancer Prevention Policy 1993. Sydney: The Australian Cancer Society, 1993. 2. U.S. Preventive Services Task Force. Guide to clinical preventive services: an assessment of the effectiveness of 169 interventions. Baltimore: Williams & Wilkins, 1989. 3. Canadian Taskforce on the Periodic Health Examination. The Canadian Guide to Clinical Preventive Health Care. Ottawa: Canada Communication Group-Publishing, 1994. 4. Coates M, Smith D, Taylor R, McCredie M. Cancer incidence and mortality by local government area and health region New South Wales 1985–1989. Sydney, NSW, 1993: Cancer Epidemiology Research Unit, NSW Cancer Registry, NSW Cancer Council. 5. Department of Community Services and Health. The quantification of drug caused morbidity and mortality in Australia 1988. Canberra: Australian Government Publishing Service, 1990. 6. NSW Department of Health. NSW health survey 1997—current smoking. [http://www.health.nsw.gov.au/public-health/hs97/ smk-sm1.htm]. 7. Reid M, Solomon S. Improving Australia’s rural health and aged care services. National Health Strategy Background Paper No. 11. Canberra: Commonwealth Department of Health, Housing and Community Services, 1992. 8. Mills AE, Simpson JM, Shelley JM, Turnbull DA. Evaluation of the New South Wales Cancer Council Pap test registry service. Aust J Pub Health 1994;18(2):170–5. 9. Green LW, Kreuter MW. Health promotion planning: an educational and environmental approach. Mountain View: Mayfield, 1991. 10. Kinne S, Thompson B, Chrisman NJ, Hanley JR. Community organization to enhance the delivery of preventive health services. Am J Prev Med 1989;5(4):225–9. 11. Puska P. Community based prevention of cardiovascular disease: The North Karelia project In: Matarazzo JD, editor.A handbook of health enhancement and disease prevention. New York: Wiley, 1984. 12. Puska P, Nissinen A, Tuomilehto J, Salonen JT, Koskela K, McAlister A, et al. The community-based strategy to prevent coronary heart disease: conclusions from ten years of the North Karelia Project. Annu Rev Public Health 1985;6:147–93.

ADULT QUIT SMOKING RATES IN RURAL AUSTRALIAN TOWNS: THE CART PROJECT 13. COMMIT Research Group. Community intervention trial for smoking cessation (COMMIT): I. Cohort results from a four-year community intervention. Am J Public Health 1995;85:183–92. 14. COMMIT Research Group. Community intervention trial for smoking cessation (COMMIT): II. Changes in adult cigarette smoking prevalence. Am J Public Health 1995;85(2):193–200. 15. Hancock L, Sanson-Fisher RW, Redman S, Burton R, Burton L, Butler J, et al. Community action for health promotion: A review of methods and outcomes 1990–1995. Am J Prev Med 1997; 13(4):229–39. 16. Henderson P, Thomas DN. Skills in neighbourhood work. Allen & Unwin: London, 1980. 17. Knight J, Bowman J, Considine R, Sanson-Fisher R, Redman S. The adolescent project: reducing adolescent smoking and unsafe drinking in rural towns in the Hunter area of NSW. Health Prom J 1994;4(1):54–6. 18. Hancock L, Sanson-Fisher R, Redman S, Burton R, Burton L, Butler J, et al. Community action for cancer prevention: Overview of the Cancer Action in Rural Towns (CART) project, Australia. Health Prom Int 1996;11(4):277–90. 19. COMMIT Research Group. Community intervention trial for smoking cessation (COMMIT): summary of design and intervention. J Natl Cancer Inst 1991;83:1620–8. 20. Dawley LT, Dawley HH, Glasgow RE, Rice J, Correa P. Worksite smoking control, discouragement, and cessation. Int J Addictions 1993;28(8):719–33. 21. Hocking B, Borland R, Owen N, Kemp G. A total ban on workplace smoking is acceptable and effective. J Occup Med 1991;33(2): 163–7. 22. Wakefield MA, Wilson D, Owen N, Roberts L. Workplace smoking restrictions, occupational status, and reduced cigarette consumption. J Occup Med 1992;34(7):693–7. 23. Popham WJ, Potter LD, Bal DG, Johnson MD, Duerr JM, Quinn V. Do anti-smoking media campaigns help smokers quit? Public Health Rep 1993;108(4):510–3. 24. Paul CL, Sanson-Fisher RW. Experts’ agreement on the relative

25.

26.

27.

28. 29.

30.

31. 32.

33. 34.

35.

127

effectiveness of 29 smoking reduction strategies. Prev Med 1996;25:517–26. Hancock L, Sanson-Fisher R, Redman S, Reid A, Tripodi T. Knowledge of cancer risk reduction practices in rural towns of New South Wales. Aust NZ J Public Health 1996;20(5):529–37. Hill D, Gray N. Australian patterns of tobacco smoking and related health beliefs in 1983. Community Health Stud 1984; 8(3):307–15. Patrick DL, Cheadle A, Thompson DC, Diehr P, Koepsell T, Kinne S. The validity of self-reported smoking: a review and metaanalysis. Am J Public Health 1994;84(7):1086–93. Cochran, W.G. Sampling techniques. 3rd edition. New York: Wiley, 1977. Koepsell TD, Martin DC, Diehr PH, Psaty BM, Wagner EW. Data analysis and sample size issues in evaluation of communitybased health promotion and disease prevention programs: a mixed-model analysis of variance approach. J Clin Epidemiol 1991;44(7):701–13. Sanson-Fisher R, Redman S, Hancock L, Halpin S, Clarke P, Schofield M, et al. Developing methodologies for evaluating community-wide health promotion. Health Prom Int 1996;11(3): 227–36. Adair J. The Hawthorne effect: a reconsideration of the methodological artifact. J Appl Psychol 1984;69:334–45. Hancock L, Sanson-Fisher R, Perkins J, McClintock A, Howley P, Gibberd R. The effect of community action program on adolescent smoking rates in rural Australian towns: The CART project (in preparation). Slama K, Redman S, and Sanson-Fisher R. Examining smokers’ beliefs about smoking and health (in preparation). Nagle, A.L. Impact of hospitalisation on patient smoking: current practice and potential for nurse-provided smoking cessation care 1996. Thesis, Univ of Newcastle. Ritvo PG, Irvine MJ, Lindsay EA, Kraetschmer N, Blair N, Shnek ZM. A critical review of research related to family physicianassisted smoking cessation interventions. Cancer Prev Control 1997;1(4):289–303.