Background Mass media campaigns and quitlines are both important distinct components of tobacco control programmes around the world. But when used as an integrated package, the effectiveness and cost-effectiveness are not well described. We therefore aimed to estimate the health gain, health equity impacts and cost–utility of the package of a national quitline service and its promotion in the mass media.
Methods We adapted an established Markov and multistate life-table macro-simulation model. The population was all New Zealand adults in 2011. Effect sizes and intervention costs were based on past New Zealand quitline data. Health system costs were from a national data set linking individual health events to costs.
Results The 1-year operation of the existing intervention package of mass media promotion and quitline service was found to be net cost saving to the health sector for all age groups, sexes and ethnic groups (saving $NZ84 million; 95%uncertainty interval 60–115 million in the base-case model). It also produced greater per capita health gains for Māori (indigenous) than non-Māori (2.2 vs 0.73 quality-adjusted life-years (QALYs) per 1000 population, respectively). The net cost saving of the intervention was maintained in all sensitivity and scenario analyses for example at a discount rate of 6% and when the intervention effect size was quartered (given the possibility of residual confounding in our estimates of smoking cessation). Running the intervention for 20 years would generate an estimated 54 000 QALYs and $NZ1.10 billion (US$0.74 billion) in cost savings.
Conclusions The package of a quitline service and its promotion in the mass media appears to be an effective means to generate health gain, address health inequalities and save health system costs. Nevertheless, the role of this intervention needs to be compared with other tobacco control and health sector interventions, some of which may be even more cost saving.
- Social marketing
- Socioeconomic status
Statistics from Altmetric.com
The use of mass media campaigns is an important tobacco control strategy and one which appears to be a good investment from a health economic perspective as per a systematic review of 10 studies.1 A Cochrane Review also reports evidence in favour of tobacco control mass media campaigns, but it found only two studies which indicated that quitline calls increased after the mass media campaigns (see online supplementary material). Quitline support for smoking cessation has also been found to be effective in another Cochrane Review. The review also reported that ‘promotion of quitlines by mass media antismoking campaigns helps to attract callers’ but none of these studies appeared to consider cost-effectiveness. However, other work does suggest that quitlines can be cost-effective interventions in themselves and when combined with nicotine patch dispensing (see online supplementary material).
So what remains relatively unknown is the combined effectiveness and cost-effectiveness of the package of mass media campaigns and a quitline service. New Zealand is a relatively good setting to study this issue as it routinely uses mass media promotion to direct smokers to its national-level quitline service. Furthermore, there is evidence that such media campaigns have successfully engaged with smokers, including for Māori (indigenous) smokers.2–4 The New Zealand quitline service is also notable for its relatively high reach,5 6 including to Māori smokers.5
Given this background, the aim of this study was to identify the health gain, health equity impacts and cost–utility of this particular national quitline service and its affiliated promotion in the mass media compared with the counterfactual of no such programme existing in the studied country (New Zealand).
We adapted an established Markov macro-simulation model using a multistate life-table approach.7–9 This models 16 tobacco-related diseases, using national data by sex, age and ethnicity for the whole New Zealand adult population in 2011 (the base year for all interventions modelled in our epidemiological and health economics modelling work10). We focused on ethnicity as ethnic inequalities are greater than other socioeconomic inequalities in health in this country. The model takes a health system perspective and estimates quality-adjusted life-years (QALYs) gained and net costs over the remainder of the population’s lifetime (with both discounted at 3% per annum).
Effect sizes and intervention costs used in the model were based on past New Zealand quitline data. Health system costs were from a detailed national data set linking individual health events with cost data11 and with recent refinements to our costing approach being detailed in the online supplementary material in Pearson et al.8
More specifically, our modelling involved studying the intervention package of mass media promotion and quitline service, as actually used in New Zealand in 2011. That is, the expenditure of $NZ2.92 million on smoking cessation messages with the quitline number in the mass media (including the campaign management costs), combined with the running of the national quitline costing $6.2 million (see table 1 for details). The comparator was the background quitting behaviour of the New Zealand population with the effect of the quitline-related mass media and quitline service removed.
Modelling and key parameters
Parameters around the intervention effect size and the comparator
The parameters were informed by a literature review (see online supplementary information). Of particular note, however, is that New Zealand quitline callers appear to be fairly similar to those smokers who do not use the quitline. One study found no statistically significant differences by age group, sex, two different measures of deprivation, ‘strength of intention to quit’ or for quitting self-efficacy (in a multivariate analysis5). Nevertheless, Māori smokers were significantly more likely to call the quitline as were those under financial stress and those with a past diagnosis of a mental health disorder. This comparability—or exchangeability in epidemiological terms—of smokers using and not using the quitline is a critical assumption to our estimation below of the intervention effect size of the New Zealand package of quitline service and its mass media promotion.
We obtained net annual smoking cessation rate data for different population groups from a study that examined smoking trends from census data (which collects smoking data), as per online supplementary table A2.12 These ‘net’ rates reflect long-term cessation outcomes and so incorporate the fact that many smokers will quit and relapse multiple times before finally succeeding.
The values in table 2 (and online supplementary table A2) reflect our calculations of the impact of the intervention package (mass media promotion and service), but also non-quitline-related quitting. To estimate the latter (which was our comparator with the intervention package effect stripped out), we considered the following equation where:
NCRsp is the actual observed net annual cessation rate in the specific population groups (as per online supplementary table A2). That is, the number of complete long-term quitters in a year, over the prevalent population of smokers (in that age/sex/ethnic group).
NCRp is the net annual cessation rate for those participants who respond to the mass media promotion and call the quitline (in each age/sex/ethnic group). That is the number of complete long-term quitters in a year in this group, over the population of smokers who respond to the mass media promotion and call the quitline (ie, as a result of the intervention package and for no other reason).
NCRnp is the net annual cessation rate for those non-participants who don’t respond to the mass media promotion and don’t call the quitline in response to it (in each age/sex/ethnic group). That is, the number of complete long-term quitters in a year in this group, over the total population of smokers who don’t respond to the intervention package. (These people quit as a result of other reasons or separate tobacco control interventions and represent our intended ‘comparator’ population, namely a ‘fictional New Zealand’ where there was no quitline promotion in the mass media and no quitline service).
P is the proportion of the smoker population who respond to the mass media promotion and call the quitline (ie, as a result of the intervention package).
The net annual cessation rate in the specific population groups is defined in eq 1 below:
Solving for :
Note that was used as the ‘reset’ comparator population quit rate in this evaluation, and the (actual) as the intervention arm quit rate.
We also calculate the relative risk (RR):
The RR is the estimated increased quit rate among people actually using the New Zealand quitline with affiliated with mass media compared with what (counterfactually) their quit rate would have been had no such programme existed in this country (with the values shown in online supplementary table A2).
The model has been subject to a model validation exercise by comparing it with a separately designed model built in different software (TreeAge) and using slightly different base-case disease incidence rates and base-case case-fatality rates. In this model comparison work, there was an acceptable 6% difference in QALYs gained for the same cardiovascular disease intervention: 103 000 versus 110 000 QALYs, as detailed elsewhere.13
Sensitivity and scenario analyses
Equity analysis scenario
We performed an ‘equity analysis’ scenario so as to ‘value’ potential health gains from preventing tobacco-related diseases similarly between Māori and non-Māori. It does this by assuming that Māori in this analysis had the background morbidity (prevalent years lived with disability) and mortality rates of non-Māori.14 This type of analysis effectively equalises the envelope of potential health gain between the two ethnic groups (for the same starting ages) and therefore does not penalise Māori in the analysis for having a greater burden of comorbid conditions and lower life expectancy.
Scenario A (50% reduced effectiveness)
The central estimates of RRs in table 2 range from 3.38 to 4.66, implying that the New Zealand quitline service and its promotion in the mass media increase quit rates among those people contacting the quitline by around fourfold compared with a counterfactual world where the intervention package never existed. This is considerably higher than the Cochrane Review estimate of 1.37 (online supplementary table A2); however, this Cochrane Review included studies with diverse levels of media promotion of the quitline. Also, the New Zealand quitline programme is also more popular with smokers than in other countries studied,6 provides multiple call-backs and offers heavily subsidised nicotine replacement therapy. Also, the evidence we have suggests that the quitline and non-quitline users are comparable (or exchangeable; see above). However, our RRs are deduced observational associations only. Therefore, we halved the effect size (RRs ≈ 2.5; or effect size ≈ 0.5% increase in smoking population average annual net cessation rates; see table 2 for specific values).
Scenario B (75% reduced effectiveness)
As per the above (scenario A), but taking an even more conservative approach with a quartering of the effect size (table 2).
Scenario C (20-year intervention)
This scenario involved a sustained (20-year) intervention of the base-case, although using the same intervention effect and costs as per the single-year intervention. We did not have data on cross-year spillover effects of the intervention package so we assumed there was none. Yet it is actually plausible that smokers may become more motivated to quit over time as smoking becomes more denormalised (and indeed there is no evidence for the ‘hardening hypothesis’ in this country15). Also using the quitline may provide skills that will help them with future quit attempts. In contrast, however, other smokers might be discouraged from using the quitline again if they have not been successful in quitting when using it. Overall, it appears that the New Zealand quitline has continued to sustain a relatively high level of engagement with smokers throughout its >25-year history with a mix of different mass media strategies and new services.16
Scenario D (considering impact on smoking uptake)
It is likely that tobacco control mass media campaigns have some spillover effects on reducing smoking uptake by youth. An Australian study found a reduction in adolescent smoking prevalence with ‘predominantly adult-oriented campaigns’.17 Also, adult-focused advertisements are noticed by adolescents and are perceived by them to be effective in changing their behaviour based on Australian18 and US studies.19 20 Given lack of details on the size of this impact, we assumed a 5% reduction in youth uptake in the intervention year with a 95% uncertainty interval (UI) consistent with a 2%–20% range, beta distribution (mean=5%, SD=18%). We also adjusted the comparator for this scenario accordingly by removing the assumed effect on smoking uptake (so as to more accurately model the impact of the full intervention impact).
Scenarios E and F (varying the time horizon)
We varied the time horizon for assessing health gain and health costs to just a 10-year period and a 20-year period.
The health gain (in QALYs) and health system costs saved for the intervention package compared with the comparator New Zealand population with no intervention for the base-case (single-year intervention at 3% discounting), by sex, age and ethnicity are shown in table 3. QALY gains occurred particularly for Māori at 35% of the total (1470/4200), despite Māori being only 15% of the population in 2011. That is, there would be four times as much health gain per capita for Māori compared with non-Māori (2.19/0.729 per 1000 population, a ratio of 3.6 with age-standardisation). Health gains were largest for those aged <65 years in 2011. Health system savings were $NZ84 million (95% UI $NZ60–115 million) over the remaining life of the 2011 cohort.
Table 3 also shows a scenario of an ‘equity analysis’ for Māori in which non-Māori background mortality and morbidity were used (so as to ‘value’ potential health gains from preventing tobacco-related diseases similarly between Māori and non-Māori14). This resulted in a 42% increase in the health gains for Māori (1470 to 2100 QALYs).
Other sensitivity and scenario analyses are reported in table 4 and with more detailed results in the online supplementary material tables A3–A7). All of these still produced net health gain and cost savings for the health system, including when the intervention effect size was quartered (to account for possible residual confounding) and when the discount rate was raised to 6%. As expected, the smallest gains were seen with a 10-year time horizon and the largest when the intervention was extended from 1 to 20 years, resulting in 13 times greater health gains (discounted QALYs: 54 100) and $NZ1.07 billion (discounted) in health system savings (equivalent to US$0.72 billion).
The details of how the input parameters determined uncertainty in the results are shown in tornado plots in the supplementary figures A1–A4). Uncertainty around QALYs gained was mainly driven by the level of uncertainty in the net cessation rate of the participants (which largely reflects the intervention effect size). This parameter was also the main driver for uncertainty in costs.
The results from this study are compared in table 5 with various other tobacco control interventions using the same multistate life-table model. These indicate that the health gain for this intervention package for 1 year was 7% of that for a modelled multiyear tobacco tax intervention and 16% for a modelled tobacco retail outlet reduction intervention. But if this package was implemented for 20 years (as per scenario C), the health gain was closer to that of the tobacco tax intervention (94% of it), but still far less than two modelled endgame interventions (table 5). The relative health gain for Māori versus non-Māori from the intervention package was similarly favourable as per these other interventions (table 5).
Main findings and interpretation
This study estimated that the intervention package of a quitline service and its mass media promotion generated substantial health gain and was also cost saving for the health system. This same pattern was also seen when modelling other tobacco control interventions using the same multistate life-table model in the same setting (table 5). Indeed, if this package was implemented for 20 years (as per scenario C), the health gain was fairly similar to that of the tobacco tax intervention (table 5). However, it was still far less than two modelled endgame interventions, especially the sinking lid on tobacco sales.
The relative health gain for Māori versus non-Māori from the intervention package was also similarly favourable as per other types of tobacco control interventions (table 5). The greater health gains generated for Māori from the intervention package in part reflect higher smoking prevalence and the fact that in every age group the proportion of Māori smokers using the quitline is higher than for non-Māori, for example, 19% versus 13% among 20–34-year age groups. Indeed, the mass media used by the quitline have prioritised using media channels with a relatively high Māori audience (eg, Māori TV) and the quitline service has worked on ensuring culturally appropriate service delivery.
Our results for cost-effectiveness are compatible with the international literature around the cost-effectiveness of tobacco control mass media campaign interventions and quitlines when considered separately.1 21 However, this appears to be the first study to find the mass media promotion/quitline service intervention package was cost saving. This is possibly because relatively few studies include the costs of tobacco-related diseases in the analyses (which become (large) cost savings when tobacco consumption is reduced), but it could also relate to specific aspects of the New Zealand situation (eg, the high population reach of the quitline5 6). Tobacco control models can also differ in design, for example, where costs of residential care for older people costs are included, then this can reduce the apparent cost-effectiveness of interventions (eg, as per some modelling work from the Netherlands which studied obesity and smoking-related lifetime costs22).
It is reassuring that the cost-saving aspects of the intervention package were maintained for all sensitivity and scenario analyses, for example, at a discount rate of 6% and when the intervention effect size was quartered (given the possibility of residual confounding in our estimates of smoking cessation). Nevertheless, for policymakers focusing on the short term it is notable that the bulk of the health gains are quite far into the future (shown graphically in related modelling work on tobacco tax7), and when considering that the 10-year time horizon scenario resulted in the smallest health gain.
Study strengths and limitations
This study used very detailed epidemiological and cost data, which were the key model inputs, for a whole country. The study also benefited from detailed data on the cost of running the existing quitline service. However, this model has residual structural limitations since it did not include some tobacco-related health impacts (eg, adverse second-hand smoke impacts, and some tobacco-related diseases such as asthma and the cardiovascular diseases that were not coronary heart disease or stroke, eg, peripheral vascular disease). As a result, the health gain from this intervention may be somewhat underestimated.
This analysis also does not capture some additional likely impacts around the intervention effect size. That is, for smokers who are stimulated to quit by the mass media campaign promoting the quitline, but who do not attribute their behaviour to exposure to such promotion. Similarly, some smokers who do use the quitline service do not use the telephone counselling component but rather the website and the text messaging services that it also provides. As we had no detailed information on the size of such potential impacts, it does suggest that this analysis might be underestimating the benefit of the intervention package.
On the other hand, a small proportion of quitline promotion is likely to be from the telephone number in the health warning on all tobacco packaging sold in New Zealand,23 24 and from some signage paid for by District Health Boards.25 Also not costed was the ancillary pharmacotherapy which is used by some quitline users. It is estimated that pharmacotherapy is typically used in 25.1% of quit attempts by New Zealand smokers and most of this (61.7%) is partially subsidised by the health system (New Zealand Health Survey 2012/2013 data26). Although such pharmacotherapy is partly paid for by smokers themselves and partly by the state agency PHARMAC, both quitcard provision and telephone-delivered guidance on its use by quitline workers were still captured in quitline service costs included in this model.
New Zealand data and our model allows a comprehensive assessment of heterogeneity of impact by ethnicity. However, we had to assume that the RRs for current versus never-smokers for each of the tobacco-related diseases was the same for Māori and non-Māori. Work we have just completed finds that the RRs for cardiovascular disease, lung cancer and all-cause mortality are considerably less among Māori in the 2011–2016 period. This would mean we have somewhat (maybe 10%–20%) overestimated the gains for Māori. Further work is warranted on this; however, we are very confident that most tobacco control interventions will still have a (much) greater impact for Māori compared with non-Māori as a result of the much higher prevalence of smoking and higher overall rates of tobacco-related diseases.
This study also has a likely limitation around the choice of the comparator (ie, background quitting minus the effect of the intervention package studied). That is, it is possible that if the quitline did not exist then the private market for quitting activity would expand. This could include the expansion of pharmaceutical industry marketing of smoking cessation products and health professionals marketing quitting support counselling and access to prescribed pharmaceuticals. Nevertheless, we note that the level of uncertainty around the NCRnp variable in the tornado plots is relatively small and so it may not influence the results substantially if this was changed with a large private market in quitting activity.
We ran our analysis from a health perspective, which did not include benefits such as financial savings for smokers who quit (from reduced expenditure on tobacco), and also costs associated with unemployment as a result of tobacco-related disease (eg, as calculated for stroke in New Zealand27). If these costs were considered in a societal analysis, then the intervention would probably be even more beneficial to society as a whole.
Possible further research
This intervention may not be a high-priority area for further research in New Zealand given that it seems to be such a cost-saving intervention for all the sensitivity and scenario analyses and when considering the uncertainty intervals for the base-case intervention. But given the diversity of the media environments and design of quitlines in different settings, other countries should consider their own analyses. In particular, greater use of online advertising for quitline services offers scope for enhanced efficiencies, as found in one study.28 Head-to-head comparisons (for effectiveness and cost-effectiveness) could also be made between quitline services and with mobile phone-based smoking cessation interventions as per a recent Cochrane Review.29 Similarly with smoking cessation apps.
Possible policy implications
This study, along with the international literature referred to in the section ‘Introduction’, should provide some reassurance to policymakers in high-income countries that a combined mass media promotion and quitline service package is a relatively good health sector investment for a government to make. But if a government wished to make rapid progress towards a smokefree nation goal (or at least a very low smoking prevalence target), then it might still prioritise other interventions detailed in table 5 that have potentially larger impacts. These include regular tobacco tax increases7 or more novel endgame interventions such as phasing down tobacco retail outlets8 or a sinking lid on tobacco supply.9 30
The package of mass media promotion and quitline service as currently used in this high-income country appears to be an effective means to generate health gain, address health inequalities and save health system costs. Nevertheless, the role of this intervention needs to be compared with other tobacco control and health sector interventions, some of which appear to be even more cost saving.
What this paper adds
Mass media campaigns and quitlines are both important distinct components of tobacco control programmes around the world. But when used as a package (quitline service and its promotion via mass media), the effectiveness, cost-effectiveness and health equity impacts are not well described.
This study found that this intervention package appears to be an effective means to generate health gain (measured in quality-adjusted life-years), compared with the counterfactual of no such programme in the studied country (New Zealand).
The modelled intervention also generated four times the per capita health gains for the indigenous population (comparing Māori vs non-Māori populations).
The intervention produced net savings in health system costs and all sensitivity and scenario analyses indicated the intervention remained cost saving (even when the effect size was quartered, given the possibility of residual confounding).
This intervention package appears to be a good investment for governments to make, but ideally its cost–utility should be compared with other tobacco control interventions (such as increased tobacco taxes) that can be even more effective.
The authors acknowledge colleagues who have helped with other aspects of building the tobacco control model used in this study: June Atkinson, Linda Cobiac, Giorgi Kvizhinadze, Melissa McLeod and Amber Pearson.
Contributors NW and TB conceived and designed the study. NW, NNg, WL and FSvdD contributed to data collection and model parameterisation. NNg built model updates. CC contributed to model validity testing. NNg generated all the model results. NW led the writing of the manuscript and all authors contributed to manuscript drafts and approved the final version.
Funding This work was supported by funding from the Ministry of Business, Innovation and Employment (MBIE), grant number: UOOX1406. Work on the original model was supported by a grant from the Health Research Council of New Zealand (grant 10/248).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The authors can be contacted for additional data and this will be provided pending agreement from the agency providing access to epidemiological and costing data (the New Zealand Ministry of Health).
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.