Discrete choice experiment for dyadic data collection: eliciting preferences of couple-based smoking cessation interventions

Article information

Womens Health Nurs. 2024;30(1):9-17

Publication date (electronic) : 2024 March 29

doi : https://doi.org/10.4069/whn.2024.03.08.1

Seung Hee Choi, Thomas Templin

College of Nursing, Wayne State University, USA

Corresponding author: Seung Hee Choi College of Nursing, Wayne State University, 5557 Cass Ave #350, Detroit, MI 48202, USA Tel: +1-313-577-8945 E-mail: seunghee.choi@wayne.edu

Received 2024 January 15; Revised 2024 March 5; Accepted 2024 March 8.

Introduction

A discrete choice experiment (DCE) is a validated method of eliciting preferences regarding healthcare services and health behavior interventions, including smoking cessation programs. This approach has been widely used within the field of health economics [1,2]. Grounded in economic theory, DCEs posit that any product or service can be defined by its attributes along with the levels, or subcategories, of those attributes. Additionally, these experiments assume that individuals’ choices regarding a service are influenced by the degree to which they value these attributes and levels [3]. Compared to traditional ratings or ranking scales, DCEs have been shown to more accurately reflect decision-making processes and better clarify potential trade-offs [1].

A DCE survey consists of several choice sets, each containing two or more profiles and a potential opt-out option (Figure 1). Each profile includes a specific combination of attribute levels representing a hypothetical intervention. Participants evaluate each choice set and select their preferred option. These answers provide insights into the respondent’s perceptions of the benefits and trade-offs among attributes [3]. For instance, if there is a choice between smoking cessation interventions with two attributes (delivery method and intervention dose), the attribute levels for the former might include in-person counseling and mobile/web applications, while the levels for the latter may be daily sessions over 6 weeks and monthly sessions over 6 months. Each choice set then presents a combination of these attribute levels; for example, Option A might offer in-person counseling in the form of monthly sessions for 6 months, whereas Option B could provide a web-based program with daily sessions for 6 weeks. Ultimately, individuals make an informed decision between Option A or B, based on their perceived value of each (Figure 1).

Figure 1.

Choice set, attributes, attribute levels, and profiles.

The use of DCEs is gaining popularity in the literature on smoking cessation, particularly in designing interventions tailored to specific populations. For instance, Katz et al. [4] employed a DCE to determine veterans’ preferences regarding smoking cessation interventions and nicotine replacement therapies (NRTs). Similarly, Salloum et al. [5] conducted a DCE among 191 university students in Lebanon to evaluate their willingness to make trade-offs concerning health risks, distance to treatment facilities, treatment costs, and types of treatment when choosing NRTs. Additionally, DCEs have helped shape tobacco control policies. Pesko and colleagues utilized a DCE of 1,200 adult smokers to assess the potential impact of regulations on electronic nicotine delivery systems (ENDS). Their findings suggest that imposing higher taxes and implementing severe warning labels on ENDS may discourage their use [6].

Despite the widespread utilization of DCEs in the tobacco control literature, no prior study has employed this methodology to explore the preferences of smoker couples for couple-based smoking cessation interventions, nor to understand their desires for partner support (framed as the smoker’s support “demand” and the partner’s “supply”) through the collection of dyadic data. Here, “smoker couples” are married or cohabiting pairs in which at least one partner smokes. For these couples, positive, emotional, and instrumental support from the partner, along with collaborative messages, have been associated with successful cessation [7,8]. Couple-based interventions that include partners in the process have been shown to enhance social support and encourage behavioral changes, such as reducing problem drinking [9] and increasing physical activity [10]. Partner support can act as a buffer against stress, boost motivation and social pressure to change behavior, and foster effective coping strategies [11]. Thus, couple-based smoking cessation interventions are highly promising. Despite the demonstrated success of couple-based interventions for other health behaviors [9,10], a recent review of 11 smoking cessation studies found limited evidence supporting the effectiveness of these interventions. In most of the included studies, long-term quit rates were low, and the interventions did not lead to increased partner support [12]. If interventions overly rely on changes in partner support to satisfy the smoker’s “demand” without considering the partner’s understanding of and willingness to adopt the support role (“supply”), then partner support may become more of a hindrance than a help [13,14]. Gaining deeper in­sight into participants’ preferences for couple-based interven­tions is essential for enhancing the effectiveness of these pro­grams and fostering effective partner support.

To understand the preferences of smoker couples regarding the format and content of couple-based smoking cessation interventions, we conducted an online DCE survey study. We also evaluated their preferences for partner support in smoking cessation, examining both the perspectives of the smoker’s “demand“ and the partner’s “supply“. In this article, we detail our procedures, including (1) the development of the survey and (2) the collection of dyadic data during individual online meetings. During these sessions, couples engaged in a think-aloud method, verbalizing their thoughts step by step as they completed the survey. This approach has been found effective in exploring the cognitive processes behind decision-making and improving the validity of DCE data [15]. Here, we present preliminary findings on the preferences for the format, content, and partner support of the described interventions, with a full report available elsewhere (unpublished data).

Methods

Survey development

A two-part DCE survey was developed specifically for this study. Part 1 addressed the format and content of couple-based smoking cessation interventions. Part 2 assessed the type of partner support desired by the smokers (Part 2A) and the partners’ willingness and comprehension regarding providing support (Part 2B). For Parts 1 and 2, 6 two-level attributes were identified each, drawing from the existing literature on smoking cessation interventions and partner support in this context. The attributes in Part 1 included convenience, delivery method, partner involvement, topics, duration and frequency of treatment, and interaction style (Table 1). The attributes of Part 2 encompassed the categories of partner support, positive versus negative support, decision-making, presence of partner involvement, strategies versus habits, and frequency of partner support (Table 2).

Table 1.

Six attributes and levels pertaining to the format and content of a couple-based smoking cessation intervention (Part 1)

Table 2.

Six attributes and levels pertaining to partner support (Part 2)

Both partners selected between two hypothetical scenarios concerning the format and content of a couple-based smoking cessation intervention. They jointly completed Part 1, and then individually filled out Part 2. In Part 2A, primary smokers were presented with scenarios that varied in desired partner support. In Part 2B, partners were given scenarios that differed in the ways they could support the primary smokers. Part 2A and Part 2B featured the same attributes and levels but were tailored to the perspective of each partner. For example, regarding positive versus negative partner support, primary smokers completing Part 2A were shown the statement “Negative support from my partner helps me more to quit smoking,” while partners working through Part 2B saw “Negative support helps my partner more to quit smoking.” In couples with only one partner who smoked, that individual was considered the primary smoker and completed both Part 1 and Part 2A as such, while the nonsmoking partner completed Part 1 and Part 2B. For couples in which both partners smoked, the individual who filled out the screening survey and contacted the research team was deemed the primary smoker. This person answered questions regarding the smoker’s perspective, or demand (Part 1 and Part 2A), while their smoking partner responded to questions from the partner’s viewpoint, or supply (Part 1 and Part 2B).

These differing choice sets and profiles represented a distinction from traditional surveys, in which all participants typically answer the same set of questions. In the present research, each survey attribute can be compared to a factor in a factorial analysis of variance design. A survey with 6 dichotomous attributes represents a 2⁶ factorial design, amounting to 64 cells. However, it would be impractical to ask participants to evaluate 64 combinations. Therefore, a DCE employs a modified fractional factorial design to reduce this number. For this study, the Choice Designs platform within JMP Pro 14 (SAS Institute, Cary, NC, USA) [18], a statistical software package, was used to generate the combinations of attributes and their levels, which were then used to construct the profiles and choice sets. Each choice set consisted of a pair of profiles, with 8 choice sets each in Part 1 and Part 2. To better reflect real-world decision-making regarding interventions, an opt-out option was included. This allowed participants to select “Option A,” “Option B,” or “Neither Option A nor Option B.” The choice sets were designed to be comprehensible at a ninth-grade reading level. Each survey was individually uploaded to Qualtrics, an online survey platform (https://www.qualtrics.com). Figure 2 presents an example of a choice set.

Figure 2.

Sample choice set from the discrete choice experiment regarding the format and content of a couple-based smoking cessation intervention.

Collection of dyadic data during individual online meetings

Participants and procedures

Smoker couples were recruited through various channels, including community organizations and Facebook ads. To be eligible for inclusion, individuals had to (a) be at least 18 years old, (b) have been married to or living as married with an partner for a minimum of 6 months, and (c) have access to a computer or tablet with an internet connection. Additionally, at least one partner had to (d) be a current smoker and (e) respond positively to the question, “Are you currently attempting to quit or considering quitting smoking within the next 6 months?” The exclusion criteria encompassed individuals who (a) could not understand spoken or written English or (b) were significantly cognitively impaired.

Among the 186 responses received from the screening survey, 73.7% of eligible respondents (n=137) were invited to participate in the study. Of these 137 invited couples, 81 did not respond to the invitation, failed to attend, or chose not to participate, resulting in a final sample of 56 smoker couples (Figure 3). Detailed participant information is presented elsewhere (unpublished data). In brief, most primary smokers were male (n=44, 78.6%) and non-Hispanic Black or African American (n=41, 73.2%), with their partners predominantly female (n=44, 78.6%) and similarly non-Hispanic Black or African American (n=41, 73.2%). The mean age of the primary smokers was 32.1 (±8.2) years, while for their partners, it was 29.6 (±7.9) years. Nearly half of the couples (n=24, 42.8%) had an annual income of US $59,999 or less, and the average duration of marriage or cohabitation was 6.3 (±5.5) years. Over 50% of the couples contained only one smoking partner (n=32 couples, 57.1%), while the remainder were dual-smoker couples (n=24, 42.9%). Nearly half of the primary smokers began smoking before the age of 17, and the average duration of smoking was 8.05 (±12.7) years. A total of 23 primary smokers (41.1%) exhibited high nicotine dependence (Fagerstrom Test for Nicotine Dependence ≥6), with a mean score of 4.8 (±2.0). After completing the survey, participants received a US $40 Amazon gift card as a token of appreciation for their participation.

Figure 3.

Flowchart of participant recruitment.

Individual online meetings

Eligible and consenting participants were invited to individual Zoom meetings with research staff. Here, participants engaged with a practice choice set (related to house-buying preferences) before proceeding to the study choice sets, and they received explanations about the attributes and their respective levels prior to making their selections. In Part 1, participants were encouraged to engage in discussion and make decisions collaboratively. Additionally, they were asked to share their screens throughout the session to allow research staff to promote the think-aloud method. If participants neglected to verbalize their thoughts using this method, they were gently reminded by the research staff after completing two or three choice sets. Staff also monitored participants for any indications of cognitive fatigue, such as frustration, irritation, or boredom.

All participants successfully completed the DCE survey under this protocol. The average duration of the interviews was 59.55 (±19.29) minutes, with a range of 20.17 to 124.09 minutes. The two most common reasons for interviews exceeding 60 minutes were technical difficulties, including internet connectivity problems and device malfunctions. No respondents exhibited the abovementioned signs of cognitive fatigue during the decision-making process.

Think-aloud method

All participants employed the think-aloud method, which facilitated an in-depth understanding of their preferences regarding smoking cessation interventions and partner support. This method also heightened participants’ awareness of their choices, thereby supporting the validity of the DCE data [15]. Furthermore, the think-aloud dialogue provided a platform for smoker couples to engage in discussions about smoking cessation interventions with their partners and to share their preferences. For instance, one participant said, “I don’t think that [the duration/frequency of treatment] is super important to me, what about you?” (Couple ID 6). Another remarked, “I like the nicotine replacement therapy, what do you think?” (Couple ID 8).

The following two examples illustrate constructive dialogues between partners during decision-making.

Partner A: “I don’t think [partner involvement] is a high priority. Do you want to be involved?”

Partner B: “Yeah, I don’t think that’s necessarily important.”

Partner A: “Um, what about willingness to support?”

Partner B: “I mean, either way, the partner is going to support.”

Partner A: “Have you made up your mind? Well, I think Option A is more important to me. What about you? I’m asking for your opinion.” (Couple ID 6)

Partner A: “Two-way interaction where you can have interactive conversations with the intervention, ask questions, get feedback...I like that better. [Don’t] you think?”

Partner B: “…you are the one who is still smoking, so whatever benefits you more…so, we [choose] B for this one.” (Couple ID 8)

After completing several choice sets, couples became familiar with each other’s preferences and were able to make decisions more swiftly. For instance, one partner remarked, “Again, no travel. We don’t even have to [discuss] that. I already know what he’s going to say.” (Couple ID 10)

Preliminary discrete choice experiment results

Preliminary DCE data revealed six attributes favored by smoker couples in the context of this study. These couples preferred interventions that engaged both partners, were interactive and personalized, encouraged positive support and open discussions between partners, and involved the smoker’s partner in the cessation process. Additionally, smokers expressed a preference for receiving empathy, love, and trust from their partners (Table 3).

Table 3.

Preliminary discrete choice experiment findings regarding preferences for couple-based interventions and partner support

Conclusion

Our study is the first to explore preferences for couple-based smoking cessation interventions and partner support from the perspectives of both partners, utilizing an online DCE survey administered during individual online meetings. The paper provides a detailed account of the development of the DCE survey and the dyadic data collection process employed during these meetings. We found the survey to be a valid and feasible method for gathering dyadic data on the partners’ preferences. Furthermore, the think-aloud technique offered not only a deeper understanding of the participants’ choices but also a valuable opportunity for the partners to reciprocally express their preferences and expectations of partner support. This discussion and collaboration enhanced the validity of the DCE data. However, using the think-aloud technique may also increase the cognitive burden on participants [20]. Therefore, researchers should weigh the advantages and disadvantages of this method when considering its incorporation into their studies.

Despite the increasing popularity and benefits of DCEs, researchers must also be mindful of their limitations and challenges, which include complex survey design, intricate statistical analyses, and the potential for cognitive fatigue among participants. To navigate the survey design and statistical intricacies of DCEs, researchers should seek guidance from experts in the field and consider collaboration. To lessen the cognitive burden on participants, it is essential to employ straightforward language, offer adequate explanations of tasks, and restrict the number of attributes and choice sets. In our study, we introduced a sample choice set (related to purchasing a house) before the study sets and provided clear definitions for each attribute. Consequently, all participants were able to complete the DCE survey without any reports of cognitive fatigue.

This study represents a unique contribution to nursing research through the use of DCE methodology in the design of complex behavioral interventions. Conducted by nurse scientists, this approach enhanced our understanding of couples’ preferences regarding smoking cessation interventions and partner support. Moreover, our findings provide guidance for the development of future couple-based smoking cessation programs. These programs can be customized to meet the specific needs of couples in terms of intervention format and content, as well as the nature of support given and received. Such tailoring is anticipated to promote effective partner support and increase the likelihood of successful smoking cessation among couples with at least one smoker.

Article information Continued