DISCUSSION
This study provides some weak evidence for the utility of automated interventions as a population-based cessation strategy when considered in the context of other studies. We failed to find clear significant effects between the intervention conditions and the control, due at least in part to the low rates of intervention uptake and the contaminating effects of use of alternative interventions by controls. As we fell marginally short of our planned sample size, inadequate power could also account for this. However, the magnitude of the effects we found for the QuitCoach were consistent with those reported in recent meta-analyses of internet-based cessation programs [3–5] and the short-term effects for onQ (1.5–2.0,depending on the basis for the estimate) were comparable to those reported in the few studies of text-messaging support [9]. It is therefore reasonable to conclude that these interventions are of similar efficacy to those evaluated in other studies. Use of either the onQ textmessaging program or the QuitCoach provided a similar effect, but we found no evidence that combining the two improved outcomes (however, nor did it interfere). Offering a choice of interventions increased the likelihood that one would be used, but did not improve outcomes. The findings are not consistent with the evidence of an additive effect of text-messaging and interactive webbased help [11], which we note is based on indirect comparisons. This may be because there is no benefit, or that there was simply no benefit in this case, perhaps because the extent of integration of our two programs was not sufficient and thus they did not complement each other as we thought they would. More detailed exploration of how multi-mode interventions are used is required. There was some evidence that QuitCoach was more effective (relative to onQ) among information-seekers,whereas onQ was of greater benefit among the coldcontacted subsample. By contrast, recruitment modality (telephone or web) was related strongly to intervention uptake in the opposite direction, with those recruited by web (predominantly the cold-contacted sample) more likely to use a web-based intervention (QuitCoach), whereas those recruited by telephone were more likely to accept onQ (see Table 6). The interaction might be due to a combination of less motivated participants using the modality consistent form of help (by accepting it passively when not really interested) and more motivated ones using the form they had to make additional effort to receive. Alternatively, it could be that structure (i.e. onQ) is more useful for those with low initial motivation who are prepared to give it a try, while more detailed advice might be more useful to those with a higher initial commitment and thus less need for constant reminders. Compared to the close to 100% difference in the use of help between the intervention and control groups normally engineered in a study, there was only an approximate 40% difference in the use of cognitive–behavioural cessation assistance, including the interventions trialled here. We did not try to control for use of help in the control condition in order not to create expectancy biases favouring the intervention groups (or more probably disadvantaging the controls), but as a consequence the main analysis was under powered to find effects of the magnitude we observed. None the less, when we attempted to control for actual use by estimating effect size in the control group for those who might have used,we found a significant intervention effect. Moreover, if those in the control group who actually used a behavioural intervention were more motivated (as is assumed in the need for a RCT design), then the fact that these were numerically less successful than the greater number who used an intervention in the intervention groups is evidence that our strategy has not over-controlled for potential bias. The overall increase in success among those offered help was approximately 3%; however, the estimated effect among those taking up one of the interventions was approximately 6%. This means that approximately 33 people need to be offered such an intervention to gain one additional quitter, while it takes only approximately 16 to actually use an intervention to achieve the same goal. Once computer-based, automated interventions have been developed they cost very little to maintain, and can
be provided at an ever-decreasing net cost per user. Ongoing delivery costs are negligible in the case of the QuitCoach (but base costs are higher, as it is more complex and thus more difficult to maintain), and are calculated for onQ to be less than A$20 per user. As textmessaging costs decline, and/or they become effectively free (i.e. within paid-for data download limits), the cost-effectiveness of this service will become even more attractive, particularly if it is used by large numbers. For inexpensive (per smoker reached) interventions of this kind, any benefit is likely to be of considerable public
health importance. This study highlights some of the challenges of studying the effects of cognitive–behavioural interventions. The types of behavioural interventions used here can work [3,8], but their effectiveness is, to a large extent, dependent upon how potential users engage with them. The challenges are even greater, as in this case,where the interventions are delivered without face-to-face contact and in a context where alternative equally or more effective interventions are available. Because double-blinded trials are not possible with interventions with a psychological component, we believe there is a need for a mix of studies such as this one, where no expectancies are set up of between-group differences in what is received and studies that make participation conditional on taking up
the offer provided, where between-group expectancies cannot be avoided. However, if this is to be supported by the scientific community, methods are needed for comparing the results equitably. In this paper we suggest one way of adjusting the effect estimates for ‘offer’ type studies. Not unexpectedly, higher success rates were found in the more motivated information-seeker sample in all groups, including the control, but the magnitude of the effect was surprising. This study highlights the importance of considering the nature of the population from which samples are drawn in comparing studies, as the overall success rate can vary markedly as a function of where and how the sample is recruited. We found that there was little difference by sample in the marginal success rate, but the ORs for success were markedly different, as there was a more than 10-fold difference in the base quit rate. We think it is important to estimate the marginal improvement in success rates, for which number-needed-to-treat is a useful measure. We conclude that smokers interested in quitting who were assigned randomly to an offer of either the Quit-Coach internet-based support program and/or the onQ text-messaging program had non-significantly greater odds of quitting for at least 6 months than those randomized to an offer of a simple information website. Taken in conjunction with other research, this study provides modest additional evidence for the effectiveness of both forms of intervention, but not for additive effects. Among smokers prepared to use them, the effect sizes are likely to be greater. Offering such programs widely has the potential to have a modest population-level impact on smoking cessation, especially if more smokers can be encouraged to use them.
สนทนาThis study provides some weak evidence for the utility of automated interventions as a population-based cessation strategy when considered in the context of other studies. We failed to find clear significant effects between the intervention conditions and the control, due at least in part to the low rates of intervention uptake and the contaminating effects of use of alternative interventions by controls. As we fell marginally short of our planned sample size, inadequate power could also account for this. However, the magnitude of the effects we found for the QuitCoach were consistent with those reported in recent meta-analyses of internet-based cessation programs [3–5] and the short-term effects for onQ (1.5–2.0,depending on the basis for the estimate) were comparable to those reported in the few studies of text-messaging support [9]. It is therefore reasonable to conclude that these interventions are of similar efficacy to those evaluated in other studies. Use of either the onQ textmessaging program or the QuitCoach provided a similar effect, but we found no evidence that combining the two improved outcomes (however, nor did it interfere). Offering a choice of interventions increased the likelihood that one would be used, but did not improve outcomes. The findings are not consistent with the evidence of an additive effect of text-messaging and interactive webbased help [11], which we note is based on indirect comparisons. This may be because there is no benefit, or that there was simply no benefit in this case, perhaps because the extent of integration of our two programs was not sufficient and thus they did not complement each other as we thought they would. More detailed exploration of how multi-mode interventions are used is required. There was some evidence that QuitCoach was more effective (relative to onQ) among information-seekers,whereas onQ was of greater benefit among the coldcontacted subsample. By contrast, recruitment modality (telephone or web) was related strongly to intervention uptake in the opposite direction, with those recruited by web (predominantly the cold-contacted sample) more likely to use a web-based intervention (QuitCoach), whereas those recruited by telephone were more likely to accept onQ (see Table 6). The interaction might be due to a combination of less motivated participants using the modality consistent form of help (by accepting it passively when not really interested) and more motivated ones using the form they had to make additional effort to receive. Alternatively, it could be that structure (i.e. onQ) is more useful for those with low initial motivation who are prepared to give it a try, while more detailed advice might be more useful to those with a higher initial commitment and thus less need for constant reminders. Compared to the close to 100% difference in the use of help between the intervention and control groups normally engineered in a study, there was only an approximate 40% difference in the use of cognitive–behavioural cessation assistance, including the interventions trialled here. We did not try to control for use of help in the control condition in order not to create expectancy biases favouring the intervention groups (or more probably disadvantaging the controls), but as a consequence the main analysis was under powered to find effects of the magnitude we observed. None the less, when we attempted to control for actual use by estimating effect size in the control group for those who might have used,we found a significant intervention effect. Moreover, if those in the control group who actually used a behavioural intervention were more motivated (as is assumed in the need for a RCT design), then the fact that these were numerically less successful than the greater number who used an intervention in the intervention groups is evidence that our strategy has not over-controlled for potential bias. The overall increase in success among those offered help was approximately 3%; however, the estimated effect among those taking up one of the interventions was approximately 6%. This means that approximately 33 people need to be offered such an intervention to gain one additional quitter, while it takes only approximately 16 to actually use an intervention to achieve the same goal. Once computer-based, automated interventions have been developed they cost very little to maintain, and canได้เคยลดลงสุทธิต้นทุนต่อผู้ใช้ ต้นทุนการจัดส่งอย่างต่อเนื่องเป็นระยะในกรณี QuitCoach (แต่ต้นทุนพื้นฐาน ใจ สูง เป็นมันซับซ้อนจึงยากต่อการรักษา), และมีคำนวณสำหรับ onQ จะ น้อยกว่าแบบ $20 ต่อผู้ใช้ เป็น textmessaging ต้นทุนลดลง หรือพวกเขาเป็นอิสระได้อย่างมีประสิทธิภาพ (เช่นภายในขีดจำกัดชำระสำหรับข้อมูลดาวน์โหลด), ประหยัดค่าใช้จ่ายของบริการนี้จะยิ่งน่าสนใจ โดยเฉพาะอย่างยิ่งถ้าจะใช้ตัวเลขขนาดใหญ่ ในราคาไม่แพง (ต่อการสูบบุหรี่ถึง) การรักษาชนิดนี้ ผลประโยชน์ใดมีแนวโน้มของประชาชนจำนวนมากความสำคัญของสุขภาพ การศึกษานี้เน้นบางประการของการศึกษาผลกระทบของการรับรู้ – พฤติกรรมการรักษา ชนิดของการแทรกแซงพฤติกรรมใช้ที่นี่ได้ [3,8], แต่ประสิทธิภาพ จะ ขนาดใหญ่ ขึ้นเมื่อผู้ใช้ที่มีศักยภาพการมีส่วนร่วมกับพวกเขา ความท้าทายยิ่ง ในกรณีนี้ ซึ่งงานจะจัดส่งแบบพบปะติดต่อ และ ในบริบทอื่นอย่างเท่าเทียมกัน หรือการแทรกแซงที่มีประสิทธิภาพมีการ เนื่องจากมองไม่เห็นห้องทดลองไม่ได้กับการแทรกแซงด้วยส่วนประกอบทางจิตวิทยา เราเชื่อว่า มีความจำเป็นสำหรับการซื้อคละกันของการศึกษาเช่นนี้ ที่ expectancies ไม่มีตั้งค่าของความแตกต่างระหว่างกลุ่มที่ได้รับการศึกษาที่ทำให้มีส่วนร่วมแบบมีเงื่อนไขในการขึ้นthe offer provided, where between-group expectancies cannot be avoided. However, if this is to be supported by the scientific community, methods are needed for comparing the results equitably. In this paper we suggest one way of adjusting the effect estimates for ‘offer’ type studies. Not unexpectedly, higher success rates were found in the more motivated information-seeker sample in all groups, including the control, but the magnitude of the effect was surprising. This study highlights the importance of considering the nature of the population from which samples are drawn in comparing studies, as the overall success rate can vary markedly as a function of where and how the sample is recruited. We found that there was little difference by sample in the marginal success rate, but the ORs for success were markedly different, as there was a more than 10-fold difference in the base quit rate. We think it is important to estimate the marginal improvement in success rates, for which number-needed-to-treat is a useful measure. We conclude that smokers interested in quitting who were assigned randomly to an offer of either the Quit-Coach internet-based support program and/or the onQ text-messaging program had non-significantly greater odds of quitting for at least 6 months than those randomized to an offer of a simple information website. Taken in conjunction with other research, this study provides modest additional evidence for the effectiveness of both forms of intervention, but not for additive effects. Among smokers prepared to use them, the effect sizes are likely to be greater. Offering such programs widely has the potential to have a modest population-level impact on smoking cessation, especially if more smokers can be encouraged to use them.
การแปล กรุณารอสักครู่..