Understanding how they really feel: Lesson learned from four approaches to soliciting user preferences for new contraceptive products in development

Perspectives on method characteristics

Description. This first approach gathers information on the characteristics that potential users want to see in new contraceptive products. This information can guide decisions on which attributes and/or levels to target in product development.

In the LAP study, we collected data on desired product characteristics from women through three questions in the PMA2020 survey acceptability module⁶. The questions were framed broadly and not related to any contraceptive product. They included: important characteristics in choosing a method, preferences for method duration, and willingness to use a method causing amenorrhea. Qualitative interviews with women and providers also included a simple ranking exercise⁶. Participants were asked to sort a set of cards with pre-written statements on product characteristics in order of importance to women in choosing a method (Figure 1).

Figure 1. Examples of method characteristic cards that LAP (long-acting products) study participants were asked to rank in order of importance for method decision-making.

Data from the first survey question and the simple ranking exercise provided prioritized lists of important attributes, while survey questions on duration and amenorrhea offered insight into the acceptability of different attribute levels. As an example of the former, effectiveness, duration, and few side effects were the three attributes most commonly cited as important in choosing a method in both Burkina Faso and Uganda. Regarding attribute levels, with half or more women in Burkina Faso and Uganda saying they would like a method that lasts at least one year, the second survey question provides support for long-acting contraceptive products when it comes to duration of protection against pregnancy.

Considerations for design and implementation. The survey and simple ranking provide different types of results and both suffer some limitations. With the survey question on important characteristics, responses are unprompted, and prioritization is based on the distribution of responses across participants. With the simple ranking exercise, participants react to a pre-specified list of items and prioritization is based on analysis of the ordering of the full list across interviews. Since everyone reacts to the same list, the ranking is more systematic, but it is also more limiting. In both cases, the value partly depends on the ability of the researcher to anticipate relevant characteristics, either as pre-coded survey response categories or as statements for the ranking, and on how they are formulated.

There is also a trade-off between the degree of specificity and the practicality of implementation (too many response categories or cards to rank), which both affect data quality. For example, survey results identified “duration” as important, whereas the statement “the method lasts for more than six months” ranked quite low on average in the simple ranking exercise. Yet “duration” is quite broad, and it is only through the second question on preferences for method duration that we acquired more details, with most women favoring methods they would take every year or every few years. These findings suggest that ranking results may have been different if we had used one year as a cut-off instead of six months. Including more detailed response categories or follow-up questions for all characteristics, however, may not always be possible, and some responses may therefore remain open to some degree of interpretation. In this study, this includes “effectiveness”, which may capture a range of responses from “protecting against pregnancy” to more precise specifications in terms of acceptable risk of method failure.

Additionally, while this approach of asking about preferred method characteristics tells us what attributes and/or levels users desire, it does not provide information on how to combine characteristics to achieve optimal demand. For example, although more women mentioned duration as important than few side effects, it does not necessarily mean that women will tolerate any kind of side effects if the method grants protection against pregnancy for a certain duration.

Reactions to products in development

Description. The second approach elicits reactions to specific contraceptive products by providing a description of the product and soliciting the perspectives of participants. It can inform priorities for continued product development by providing insight into which products may be most appealing to potential users, as well as into the desirability of the various characteristics of these products. We used this approach of soliciting reactions to product concepts with a range of methods at different stages of development. Some products like the hormonal IUD already exist, while others like the MAP are still more theoretical. The LAP study allowed some degree of comparison between the six methods, with a view to help prioritize development. The MAP study was focused on a single product, with the goal to inform design decisions.

In the LAP study, we described the six products in terms of their location, duration of effectiveness, potential for early removal or discontinuation, effect on menstruation, and whether they contained hormones⁵. Descriptions included pictures (Figure 2). In the PMA2020 acceptability module, we presented each of the six methods in turn and asked women if they would be interested in using each at any point in the future. We also asked respondents which method they preferred among all the products they said they were interested in, and, if applicable, their current method or a method they had used in the past 12 months. In the qualitative component with all participant groups, interviewers similarly introduced the six methods one-by-one with images and probed to understand perceived advantages and disadvantages.

Figure 2. Examples of images used in the LAP (long-acting products) study to depict new methods in development or to be introduced.

(A) New copper IUD, (B) Hormone-releasing IUD, (C) Five-year single-rod implant, (D) Biodegradable implant, (E) Six-month injectable, (F) Non-surgical permanent contraception.

In the MAP study, qualitative interviews solicited perspectives on advantages and disadvantages of the patch as a platform for contraceptive delivery⁷. In addition, interviews introduced one-by-one a description of seven primary MAP attributes (option of self-application, pain at application, skin reaction, patch location, patch size, wear time, duration of effectiveness), soliciting participants’ perspectives on several possible levels for each characteristic (e.g., three possible patch sizes). We showed participants a close-up image of the potential product with microneedles and mock patches without needles that they could press onto their skin to mimic the experience with the actual product.

Considerations for design and implementation. One advantage of the series of PMA2020 questions is that it can inform modeling of potential demand. Because there are other data on participant characteristics in the larger PMA2020 female questionnaire, it is also possible to cross-reference responses with participant characteristics. As an example, colleagues used data from the PMA2020 acceptability module to formulate assumptions on key inputs related to switching rates from other methods and uptake among women with unmet need for a demand forecast model for a six-month injectable (not publicly released). Qualitative interviews in both studies offer more details on what potential users like and do not like about products, which can provide guidance for subsequent development efforts or messaging.

One key challenge is the description of the methods. Some words, like “hormones,” do not easily translate into local languages, and other wordings such as “causing changes in the body” or “interfering with body processes” may not elicit the same reaction. A related challenge is the amount and format of the information being provided. In the LAP study, we used five characteristics to describe each method (location, duration, possibility of early removal, effect on period, and whether the method contained hormones). Particularly in the context of the PMA2020 acceptability module, it can be a lot of information for respondents to absorb, especially when switching rapidly from one method to the next. It is also critical to ensure that the delivery by data collectors is standardized so that all respondents react to the same information. Our strategy to ease the cognitive burden for the respondent while also streamlining interviewing was to structure method descriptions according to a small number of characteristics presented in a systematic way. Yet this approach risks leaving out other attributes that may be important to potential users, while conversely potentially focusing attention on attributes that respondents may otherwise not have considered. For example, in describing the LAI, we explained that side effects could not be reversed until the end of the six months as a counterpoint to saying that implants and IUDs could be removed at any time. In qualitative interviews, many participants emphasized lingering side effects as a disadvantage of the LAI; however, when they were discussing early removal of other methods, the focus tended to be on pregnancy intentions rather than side effects. The way methods are described and possible ensuing biases must be kept in mind when interpreting results.

Other limitations must be acknowledged that affect predictive validity. We cannot rule out some level of social desirability bias. Moreover, our estimates may not reflect demand for several reasons. First, we asked survey participants if they may be interested in using each product at any point in the future. In qualitative interviews, some participants who said they would consider using a product clearly qualified their response (e.g., when I no longer want children). The potentially time-dependent nature of demand may warrant additional attention for better projections. Second, demand is likely to be affected by other factors related to product introduction in real settings, such as provider capacity or pricing, that were not considered in the LAP study. Third, our goal in the LAP study was to help prioritize development efforts. Consequently, when we asked women about their preferred method in the survey, the choices were any of the six methods the respondent had expressed interest in and her current/recent method, as applicable. In real-world settings, demand for each method would be affected by the method mix, which would likely not include all six methods but would include other currently available products.

Trade-offs between various method characteristics

Description. The third approach of gauging user preferences generates information on the relative importance of product characteristics. It can inform potential adjustments to product design by guiding trade-offs between method characteristics.

The qualitative component of the MAP study included a proportional piling exercise. We presented participants with a set of cards with pre-written statements on MAP characteristics, gave them a set number of beads and asked to distribute them across cards to reflect the relative importance of each statement to them in deciding whether to use the MAP. In the second phase of the study, we conducted a discrete choice experiment (DCE) to elicit potential users’ preferences for the MAP design⁸. A DCE is a quantitative technique that involves asking respondents to state their preferred choice over alternative hypothetical scenarios. Here we presented women with 10 choice pairs of hypothetical MAPs described as bundles of six attributes (pain, skin reaction, location, size, duration, and effect on period) that vary in their levels (e.g., foot, kneecap, or wrist for location). Analysis of response data provides quantitative information on the strength of preference for each attribute level, as well as trade-offs. Prior to implementing the DCE, we conducted 10 cognitive interviews in India and eight in Nigeria to ensure proper implementation and test correct understanding of the exercise and method options by respondents. The survey was implemented on tablets, using both text and pictures (Figure 3).

Figure 3. Example of a MAP (microarray patch platform) choice pair shown to respondents as part of the MAP DCE survey.

Considerations for design and implementation. The proportional piling exercise is more sensitive than simple ranking because the bead count provides a measure of difference between items. However, it is a more limited expression of the trade-off approach than the DCE because only a few statements can be included, with each statement representing a specific value of one attribute level. In contrast, a DCE allows for permutation of attributes and levels to create many possible method descriptions. For example, our DCE had six attributes of two to three levels each, resulting in a total number of 486 possible combinations (method descriptions) and 235,710 possible pairs that can be shown to participants. The ability to elicit preferences for products that do not exist or for a much wider range of products than could actually be developed and made available is a well-recognized advantage of DCEs¹⁰. The differentiating strength of this technique is that it allows the determination of how important each attribute is relative to others (relative importance by attribute), as well as understanding what specific levels within an attribute drive preferences (level part-worths). While these outputs may not be easily interpreted by all stakeholders, it is possible to run market simulations between two or more products to estimate the percentage of respondents who would prefer each and help prioritize specific combinations of attributes and levels for continued development.

While the rigor of DCEs is appealing, this approach is complex to design and implement. Preliminary qualitative work as was done in the MAP study is increasingly recommended in the DCE literature to inform attribute and level selection^11–14. DCE design is guided by important decisions balancing statistical implications with the cognitive burden on the respondent¹⁵. While DCE results allow for determination of combinations of attributes and levels for optimal acceptability, findings also need to be reconciled with what is technically feasible. Feasibility should already be considered when deriving the possible levels of each attribute; however, not all combinations of possible levels may be realistic. It is also important to note that findings on attribute importance and level values depend on the range of attribute and levels included in the DCE. For example, our DCE included three levels for the attribute pain at application: none, feeling like a light pinprick when the needles go in, and feeling like a hard/deep pinprick when the needles go in. Pain ranked fourth in order of importance in the analysis of results from India, with an importance value of 9%. With a wider range of levels including more acute pain, pain may have been more important. Similarly, an attribute importance is relative to the other attributes being included. In Nigeria, we implemented two versions of the DCE: one with all six attributes and one without effect on period but with the remaining five attributes. In the six-attribute version, the importance of duration relative to other attributes (including effect on period) was 25%; in the five-attribute version, it was 44%.

Finally, the predictive validity of DCEs is not ascertained¹⁶. Due to their hypothetical nature, DCEs are not constrained by what is actually currently available, contrary to other approaches such as randomized controlled experiments. The flip side of this is that DCEs reveal stated preferences among hypothetical choices, and these may not be the same as revealed preferences that may be observed if respondents were asked to make a real choice.

Blue-sky ideas on novel contraceptive technologies

Description. The fourth approach captures users’ ideas about novel contraceptive products. It can help spur innovation in ways consistent with users’ preferences. In FGDs with women and with men in the LAP study, we asked participants to pair and brainstorm their ideal contraceptive method^17,18. We instructed them to consider its delivery system, its duration of effectiveness, whether it would be used by women or men, and any other characteristics they thought were important. Each pair then reported on their method to the larger group, and the group had to agree on one method, which could be a method proposed by one of the pairs or a new idea spurred by the discussion. This exercise was also conducted individually with providers.

This approach to collecting user input follows a user-centered design process. It starts with potential users and engages them in a creative process to end up with contraceptive methods that are tailored to their needs. Because it places participants at the start of the creative process – rather than seek their reaction to methods already designed by scientists – it provides an opportunity to innovate in new directions. With this orientation, the blue-sky exercise we implemented borrows elements from human-centered design (HCD), a product development process long-used in the private sector that has become increasingly popular in global health in recent years. The full HCD process involves learning from potential users to understand their needs and circumstances, then ideating and conducting prototyping of possible solutions. Although earlier sections of the interviews helped with learning, the blue-sky exercise focused on creating ideas for new solutions in the context of a one-time interaction without any subsequent prototyping.

Considerations for design and implementation. Although our intent was to identify potential novel designs, we found that many methods proposed by participants were based on common existing delivery mechanisms, namely injections, implants, and pills. Only 13% of proposed methods in Burkina Faso and 27% in Uganda were novel designs, such as creams, consumables, or wearables. In addition, over half of proposed methods were female-centered, and where methods were for use by men and women, there were several instances where it was clear from interview transcripts that participants added men as a potential target group after being queried by the moderator. One possible explanation for the high degree of similarity to existing methods may be that participants had difficulty abstracting from their actual experiences, particularly in the context of a short exercise within a larger interview that had already covered perspectives on and experiences with long-acting methods. One example supporting this explanation is that in Burkina Faso, a few participants justified proposing a particular delivery mechanism because they found the side effect profile of the corresponding current method tolerable. In addition, although they were instructed not to, moderators gave examples of existing delivery mechanisms to clarify instructions in a few FGDs. These limitations apply to both countries; however, it is worth noting that there was more innovation in Uganda than in Burkina Faso overall, possibly suggesting there may be cultural differences in terms of openness to thinking outside of the box. For example, although he eventually obliged and suggested his ideal method, a Burkinabe provider initially said that proposing an entirely new method was a job for scientists and “too much for him.”

The blue-sky exercise is structured to rapidly collect a range of user-generated ideas. When using this approach, other considerations must be addressed to sort and prioritize concepts for actual product development. User-generated ideas will need to be evaluated by scientists for the feasibility of their individual or combined attributes. Another inherent challenge is to build confidence that any idea may hold appeal beyond the participants who suggested it, including other participants who may have proposed other methods or those not directly involved in the research. After an initial assessment of feasibility, additional research may be needed to gather user input to justify continued development of the most promising feasible ideas. While the scope of the blue-sky exercise was constrained by the cross-sectional design and the other approaches being simultaneously implemented, others have recently implemented a more in-depth process in Kenya and Nigeria to ideate new contraceptive methods. This process included in-country market assessments engaging women, male partners, healthcare providers, and other stakeholders, followed by ideation events with a multidisciplinary group of local and global experts spanning product developers, tech start-ups, program managers, providers, and international procurement agencies¹⁹.