Research articleWeb-Based Smoking-Cessation Programs: Results of a Randomized Trial
Introduction
More than a decade of research evaluating the effectiveness of computer-tailored smoking-cessation materials has led to a new class of intervention tools that are well into the early phases of large-scale dissemination. Meta-analyses and reviews of print-based, tailored smoking-cessation interventions have demonstrated a generally positive set of findings, both in significance over comparison conditions and in magnitude of change.1, 2 Concepts utilized in these interventions are now migrating to the World Wide Web (“web”), a medium capable of reaching very large numbers of smokers, at any time, for a fraction of the cost of tailored print-based materials. Recent evaluations of web-based tailored smoking-cessation programs have demonstrated outcomes similar to those found in tests of tailored print materials,3, 4, 5 although variation in efficacy among web-based interventions has also been found.6
This first generation of research examining the effectiveness of computer-tailored smoking-cessation materials focused on whether “black-box” interventions—what the investigators considered to be an optimal algorithm of various psychosocial and communications components—had a greater impact on cessation than either untailored smoking-cessation materials or the absence of smoking-cessation materials. However, the generally positive but variable results from these interventions suggested the need for a second generation of research that identified the active components of tailored interventions.
In an effort to systematically identify active components of a web-based smoking-cessation intervention, this study used an approach adapted from a framework that has been used successfully in engineering for many years.7 This process, termed Multiphase Optimization Strategy (MOST) by Collins and colleagues,8, 9 involves (1) screening potentially active components of a black-box intervention, (2) refining knowledge of the effects of the most relevant components identified in the screening stage, and (3) confirming the optimized set of the components through a trial of the resulting intervention.
In the screening phase, a relatively large number of potentially important components (their selection guided by theory and existing empirical research) are usually examined. These components are typically evaluated through the use of either a factorial or a fractional factorial design. A fractional factorial design employs a systematic approach to reduce the number of study arms to allow a more manageable study, at the cost of allowing only main effects and a pre-specified set of interactions to be tested. In the screening phase, this trade-off is usually acceptable.
This article presents the screening phase of the study, examining five intervention components (factors) using a fractional factorial design. Two of the factors selected for this randomized trial, outcome and efficacy expectations (i.e., self-efficacy), are derived from social cognitive theory10 and are central to most major theories of health-related behavior. For example, within the health belief model,11 perceived-threat and perceived-benefits constructs are associated with outcome expectations, while the perceived-barriers construct is associated with efficacy expectations. Within the transtheoretical model,12 the pros-and-cons constructs are associated with outcome expectations, while the temptations construct is associated with efficacy expectations.
The other three factors—success stories, message source, and message exposure—may be viewed as methods of conveying outcome and efficacy expectation content and are classic constructs examined in the study of persuasive communications.13, 14 Success stories communicate a variety of messages: In this study, they communicated both outcome and efficacy expectation messages in a narrative format as opposed to an advice-driven format. Message source focused on the degree of personalization given to the source, which was the HMO. Message exposure focused on the temporal distribution of the web-based sections of the program: with the sections either grouped into a large single entity or distributed over time, as in a correspondence course.
A second issue relevant to computer-tailored smoking-cessation materials is the degree of tailoring depth required to produce a significant outcome. Tailoring depth refers to the degree to which assessment data and connections among data have been utilized to produce the message. In other words, has sufficient feedback been produced from the assessment and utilized in the tailored message? The impact of tailoring depth, or “granularity,” has been discussed frequently in the literature and at scientific conferences,15, 16 but to date has not been tested. The questions becomes: How many versions of self-help materials are needed? Adding depth to computer-tailored messages requires more effort to develop concept and message and greater technologic capabilities (e.g., tailoring software), so an understanding of the point at which these efforts can be relaxed would be important to both researchers and developers.
The two aims of this research included (1) identifying active psychosocial and communication components of smoking-cessation interventions and (2) examining the impact of increasing the tailoring depth in the web-based intervention. To accomplish these aims, a fractional factorial design was used to screen and identify promising intervention components from a set of potentially active web-based smoking-cessation components. This design allowed the examination of both main effects of the intervention components as well as pre-specified interactions among intervention components and participant characteristics.
Intervention components developed with high-depth tailoring (e.g., high-depth efficacy expectations, outcome expectations, and success stories) were hypothesized to produce higher subsequent rates of smoking cessation than low-depth tailored versions. Greater personalization of the message source was hypothesized to produce higher cessation rates than a less-personalized source, and distributing the web-based program sections over five weekly installments was hypothesized to result in higher cessation rates than combining the sections into one large installment.
A number of interactions among intervention components, and among intervention components and participants' characteristics, were also hypothesized. Since self-efficacy is a relatively consistent predictor of subsequent health-related behavior change,10 the efficacy expectation intervention component was of particular interest; the fractional factorial design was constructed to maximize the ability to explore 2-way interactions between a high- versus low-depth tailored efficacy expectation component and other intervention components. It was hypothesized that high-depth efficacy expectation messages—when paired with either a more personalized message source, with high-depth outcome expectation messages, or with high-depth success stories—would produce particularly high rates of cessation. Also examined were potential interactions between the efficacy expectation component and participants' baseline self-efficacy, between the outcome expectation component and participants' baseline motivation, and with the success story component according to participants' baseline level of education; it was hypothesized that more deeply-tailored messages would have particularly strong effects among participants with lower baseline levels of self-efficacy, motivation, and education, respectively.
Section snippets
Participants
Participants were recruited from the memberships of two HMOs participating in the National Cancer Institute's Cancer Research Network: Group Health of Seattle WA and the Henry Ford Health System (HFHS) of Detroit MI. Both Group Health and HFHS are not-for-profit healthcare delivery systems. An individual was eligible to participate if he or she (1) had smoked at least 100 cigarettes in his or her lifetime, currently smoked at least 10 cigarettes per day, and had smoked in the past 7 days; (2)
Project Quit Recruitment and Follow-Up Response
During an 11-month recruitment period, 3256 people from both HMOs visited the website; 2651 (81% of website visitors) were screened for eligibility; 2011 (62% of website visitors) were eligible; and 1866 enrolled and were randomized to one of the 16 study arms (57% of website visitors). The primary reasons for ineligibility were: did not smoke enough (26%); medical contraindications for NRT (23%); already enrolled in another smoking-cessation program (16%); lack of adequate Internet/e-mail
Discussion
This study used a randomized experimental design to address two aims: (1) identifying active psychosocial and communications elements of a web-based smoking-cessation intervention and (2) testing the impact of tailoring depth on smoking cessation. The study tested these aims in two large, generalizable health delivery systems using recruitment strategies similar to those used in the health promotion programming of these systems.
Two intervention components demonstrated particular promise for
References (24)
Computer-tailored smoking cessation materials: a review and discussion
Patient Educ Couns
(1999)- et al.
The multiphase optimization strategy (MOST) and the sequential multiple assignment randomized trial (SMART): new methods for more potent eHealth interventions
Am J Prev Med
(2007) - et al.
Managed care members talk about trust
Soc Sci Med
(2002) - et al.
Self-help interventions for smoking cessationCochrane Database of Systematic Reviews 1998(4)
- et al.
A randomised control study of a fully automated Internet based smoking cessation programme
Tob Control
(2006) - et al.
Randomized controlled trial of a web-based computer-tailored smoking cessation program as a supplement to nicotine patch therapy
Addiction
(2005) Comparing the efficacy of two Internet-based, computer-tailored smoking cessation programs: a randomized trial
J Med Internet Res
(2005)- et al.
American Cancer Society's QuitLink: randomized trial of Internet assistance
Nicotine Tob Res
(2007) - et al.
Statistics for experimenters: an introduction to design, data analysis, and model building
(1978) - et al.
A strategy for optimizing and evaluating behavioral interventions
Ann Behav Med
(2005)
Health promotion by social cognitive means
Health Educ Behav
Social learning theory and the health belief model
Health Educ Q
Cited by (205)
Distributive randomization: a pragmatic fractional factorial design to screen or evaluate multiple simultaneous interventions in a clinical trial
2024, BMC Medical Research MethodologyTechnology-based interventions for tobacco smoking prevention and treatment: a 20-year bibliometric analysis (2003–2022)
2024, Substance Abuse: Treatment, Prevention, and Policy
The full text of this article is available via AJPM Online at www.ajpm.online.net; 1 unit of Category-1 CME credit is also available, with details on the website.