Preferences in Clinical Care of Individuals With Differences of Sex Development

This study uses an online best–worst scaling survey to measure relative preferences for process and outcome attributes associated with care decisions for individuals with DSD. Preferences, a concept that comes from economic theory, can be measured based on the choices and tradeoffs people make given a set of alternatives. Preferences can be observed (revealed preferences), as in the case of consumer behavior, or assessed under hypothetical conditions in surveys (stated preference).²⁷  Best–worst scaling is a type of stated preference method that has been used in many areas, including health care decision-making,²⁸  to measure preferences and tradeoffs based on the different qualities or attributes of goods and services.²⁹  In best–worst scaling, respondents are asked to make tradeoffs by selecting the best and worst attributes in a set. By providing 2 choices, best–worst scaling elicits more information about a respondent’s preferences than ranking and rating exercises or other conjoint analyses, allowing for the calculation of a quantitative relative value weighting across attributes.

The survey was developed following best practices for stated preference methods.³⁰  First, in-depth qualitative interviews with DSD stakeholders (n = 110) were conducted to understand stakeholder perspectives regarding optimal care delivery and outcomes for individuals with DSD. Information from these interviews, which have been described previously,²⁴  were used to identify and define 14 attributes related to the process of achieving successful DSD care (process attributes) and 16 attributes related to outcomes associated with successful DSD care (outcome attributes; Table 1 and 2). Additional qualitative interviews were conducted to refine, confirm, and prioritize attributes before being included in the best-worst scaling survey. Second, a paper-based survey draft was iteratively pretested among DSD stakeholders (n = 17). Refinements were made to the survey based on the feedback collected from stakeholders through cognitive interviews. In the final stage, an online version of the best–worst scaling survey was coded in Qualtrics (Qualtrics, Provo, UT) and pilot-tested among additional DSD stakeholders (n = 44) to ensure functionality.

The best–worst scaling survey instrument included 4 sections: introduction, participation consent, 8 best–worst scaling tasks, and a final section with questions about respondents’ demographics, beliefs, and attitudes. The introduction included descriptions of the 14 process and 16 outcome attributes as well as a sample best–worst scaling task (Fig 1). In answering each task, participants were instructed to imagine the process of obtaining treatment of a child with DSD. (In the final survey and in recruitment materials, we used the term “Conditions Affecting Reproductive Development” instead of DSD because our qualitative research revealed that some stakeholders preferred this term.) Participants were asked to select the “most important” and “least important” attributes for 8 best–worst scaling tasks: 4 for process attributes and 4 for outcome attributes. Each task included a subset of 6 process or outcome attributes for the participant to consider. The subset of attributes included in each question set were determined based on a balanced incomplete block design methodology,²⁹  generated using Stata (StataCorp College Station, TX). In the final section, participants were presented with questions on sociodemographic characteristics, attitudes and beliefs on DSD care (which were pretested and pilot-tested), and experiences completing the survey. Respondents were asked about their racial and ethnic identity to understand the diversity of our sample.

Participants, recruited between October 2020 and May 2022, were a sample of 520 DSD stakeholders, including adolescents and adults with DSD, parents of a child with DSD, multidisciplinary health care providers, and allied professionals. Adolescents and adults with DSD and parents with a DSD-affected child were recruited through 3 US pediatric medical centers (n = 133) or through patient support and advocacy organizations (n = 41). Professionals, including health care providers and other stakeholders, were recruited based on evidence of their connection to DSD, similar to a process followed in a previous study (n = 323 from medical centers and n = 23 from advocacy organizations).²⁴  The institutional review board of the University of Michigan Medical School approved the research.

Relative importance scores were calculated for each of the 14 process and 16 outcome attributes. The number of times each attribute was chosen as least important was subtracted from the number of times each attribute was chosen as most important, and the difference was divided by the number of times each attribute was available to be chosen by the participants.²⁹  Relative importance scores range from –1 to 1, and their magnitude and direction represent the relative value associated with the attribute. Higher positive scores denote attributes that are selected more often as most important by the participants, whereas lower negative scores denote attributes that are selected more often as least important. Scores close to 0 denote relative indifference to the attribute. We also stratified relative importance scores by selected participant characteristics to explore potential heterogeneity across subgroups in which we had sufficient representation. Primary analysis included all observations, and secondary analyses excluded (1) respondents who did not select both best and worst options or (2) respondents who indicated difficulty answering the questions.

To gain further insight into the relative significance of the attributes, we employed conditional logit regression analysis with effects coding.²⁹  The regression model assumes a sequential best–worst assumption, where respondents are assumed to select the most important attribute before selecting the least important attribute in a list. Coefficients from the regression analyses, which can be negative or positive, were transformed to ratio-scaled preference weights ranging from 0 to 100 using a procedure described elsewhere (Supplemental Information).³¹ 

We also conducted latent class analysis (LCA) to explore heterogeneity in preferences. LCA is a probabilistic modeling approach that identifies unobserved or “hidden” group membership based on patterns in preferences.^32 ,33  We used conditional logit models to conduct the LCA, and model fit was assessed by minimizing the Bayesian Information Criterion and consistent Akaike Information Criterion.³⁴  All analyses were conducted in Stata.

We used descriptive statistics to summarize respondents’ sociodemographic characteristics and their attitudes and beliefs on DSD care.

The characteristics of the 520 survey participants are shown in Table 3 and Supplemental Table 4. Most respondents (n = 272 or 52.3%) were health care providers with expertise in, among others, pediatric urology or surgery (n = 54 or 10.4%), pediatric endocrinology (n = 54 or 10.4%), and psychology or psychiatry (n = 48 or 9.2%). Individuals with DSD (n = 87) or family members of individuals with DSD (n = 89) comprised 33.9% (n = 176) of respondents, followed by other professionals (n = 50 or 9.6%) and advocates (n = 28 or 5.4%). The sample was predominantly White (n = 415 or 79.8%), non-Hispanic (n = 440 or 84.6%), cisgender women (n = 360 or 69.2%), straight/heterosexual (n = 385 or 74.0%), and individuals with postgraduate education (n = 347 or 66.7%).

Among the 14 process attributes (Fig 2A), good communication between care team and patient/family, care team educates the patient/family about the condition, and care team incorporates the values of patient/family received the highest relative importance scores. The lowest-scoring process attribute was providers make final treatment decisions, followed by care team provides the patient/family with skills in sharing with others outside of the family, care delivered by a team of specialists in the same location, and parents make the final decision, which had similar scores.

The highest-scored outcome attributes (Fig 2B) were patient satisfaction, patient mental health, and treatment maintains physical health. The lowest-scoring outcome attributes were patient is raised by the family either as a boy or as a girl, typical appearance of body, patient is given specific cause of their condition, and typical appearance of genitals. Results were similar even when respondents with incomplete responses or who reported difficulty completing the best–worst scaling tasks were excluded from the analysis.

Between individuals with DSD and parents of affected individuals, we found that both groups had general agreement about the most and least important process-related attributes but showed divergent preferences for some outcome-related attributes (Supplemental Fig. 4). Relative importance scores by age, gender, religiosity, and patient or family member status (versus provider or other professional status) are found in Supplemental Figs. 5–8. Subgroups generally had similar attributes with the highest and lowest relative importance scores, and most variation was observed among the attributes ranked in the middle (scores between –0.1 and 0.1).

Results from the conditional logit regression analysis using data from all respondents identified similar, albeit slightly different rankings of process and outcomes attributes (Supplemental Tables 5 and 6) compared with the relative importance scores. The ratio scale preference weights calculated using the regression coefficients suggest that the most important process and outcomes attributes are at least 20 times as influential as the least important attributes (Supplemental Fig. 9).

Most respondents (53.8%) agreed with the definition of DSD as a medical condition or disorder of development, whereas just more than one third of respondents said that DSD is a variation of development that should not be medicalized (Supplemental Table 7).

When asked about surgery involving gonads or atypical genitals, 61.3% of respondents indicated that these decisions should be driven by patients themselves when they are old enough to weigh the pros and cons (Supplemental Table 7). A similar proportion of respondents (61.2%) indicated that atypical genitals should be addressed both surgically and psychologically, as opposed to only psychologically (18.8%), only surgically (6.0%), or not addressed (3.8%).

When respondents were asked to quantify relative importance of 5 attributes in a standard rating task, patient mental health received the highest allocation and typical appearance of genitalia received the lowest allocation (Supplemental Table 8).

Respondents who selected “Typical appearance of genitalia” as the least important attribute in any of the best–worst scaling tasks (n = 222) were asked additional questions. When asked whether this attribute was still the least important attribute among all the attributes shown to them, more than half (58.1% or 129) answered “yes” (Supplemental Table 8). Among respondents who answered “no” or “not sure” (n = 74), most respondents (71.6% or 53) said that the other attributes that were presented to them had equal importance (Supplemental Table 8).

In 2 additional questions administered to all participants, using a scale of 1 (no effects) to 10 (significant negative effects), the median rating given by respondents regarding the effects of atypical genitals on a person’s romantic/sexual intimacy and overall quality of life was relatively negative at 7 in both cases (Supplemental Table 8).

The LCA revealed latent groups of respondents for both sets of attributes (Fig 3). (The sociodemographic characteristics of respondents that fall into each latent group are found in Supplemental Tables 9 and 10). For the process attributes, 3 respondent groups were identified. Most respondents (46%) fell into the “Patient autonomy and support” group, which showed preferences for attributes such as delaying nonurgent and irreversible treatments, care team educating and communicating with the patient/family about the condition, and patients making the final decision (Fig 3A). The second (“Education and care transitions”; 18%) and third (“Shared decision-making”; 36%) respondent groups also valued education and communication from the care team but placed lower emphasis on delaying nonurgent treatments compared with the “Patient autonomy and support” group. Respondents from the “Education and care transitions” group showed stronger preferences for transitions to adult care and receiving skills from the care team on sharing information about the child’s medical condition with others outside the family. The “Shared decision-making” group showed stronger preferences for care delivery in the same location and equal weighting of parents or patients making the final decision.

Two respondent groups were identified for the outcome attributes (Fig 3B): “Preserving function and appearance” (59%) and “Patient health and satisfaction” (41%). Both respondent groups valued overall patient satisfaction, satisfaction with gender, and mental and physical health. However, respondents from the “Preserving function and appearance” group showed stronger preferences for preservation of reproductive and sexual function, as well as successful treatment with psychological intervention, medicine, or surgery. By contrast, the “Patient health and satisfaction” group valued patient knowledge of the condition and integration into the society. More affected individuals and family members (41.6% vs 15.7%, P < .001) were likely to be in the “Preserving function and appearance” group, whereas more health care providers were in the “Patient health and satisfaction” group (73.1% vs 38.4%, P < .001; Supplemental Table 10).

This survey study explored the preferences of different stakeholders for attributes related to DSD care using best–worst scaling, a type of stated preference method. The top 3 most important process attributes of 14 related to good communication and education from the care team and the inclusion of patient and family values. Among the 16 outcome attributes, respondents identified patient-focused attributes such as patient satisfaction and mental and physical health as the top 3 most important. The small discrepancy between relative importance scores and conditional logit coefficients implies that DSD stakeholders’ preferences are partially dependent on the set of options they must choose from and the tradeoffs they have to make.

LCA identified latent groups for both process and outcome attributes. Although respondent groups shared several similar preferences, especially for the highest valued attributes, key differences between these groups were also observed. For example, although respondents in the “Patient autonomy and support” group ranked delaying nonurgent and irreversible treatments as the most important process attribute, the 2 remaining groups—“Education and care transitions” and “Shared decision-making”—ranked this attribute as the 12th and fourth most important, respectively. The possible presence of latent groups of respondents with diverse characteristics and who differentially value specific attributes of DSD care emphasizes the importance of incorporating an individual’s and family’s preferences in the decision-making process. Providers should also be aware of their preferences; the LCA found that providers are likely to differ with affected individuals and families in their preferences for outcome-related attributes. Thus, even if clinical recommendations in DSD were based on stronger evidence of long-term outcomes, disparities would persist because of the gap between “best care” in the aggregate versus values-informed decision making. This gap can be addressed through shared decision making, which necessitates a 2-way discussion exploring what ultimately matters most to the patient and their family.³⁵ 

Our survey revealed that “typical appearance of genitals” ranked 13th of 16 outcome attributes of successful DSD care in the main analysis. Similarly, the 2 outcome-related latent groups ranked this attribute 14th and 11th most important, and both individuals with DSD and parents ranked this among the least important attributes. Instead, overall patient satisfaction, mental and physical health, and satisfaction with gender were rated as the most important outcome attributes of successful DSD care. These observations were unexpected given the amount of attention and controversy surrounding the topic of genital surgery and its effect on another attribute, satisfaction with gender.³⁶  It is possible that when respondents are forced to prioritize important outcomes of care, as in best–worst scaling, they are able to imagine patient satisfaction and well-being in the presence of atypical genital appearance. Notwithstanding these observations, respondents’ median rating of the effect of atypical genitals on a person’s overall quality of life and romantic life/sexual intimacy was 7 of 10 (where 10 equals “significant negative effects”; Supplemental Table 8). This apparent difference in the weighting of attributes according to item format underscores the influence of context when individuals consider the importance of attributes of care. For example, the “focusing illusion” is a cognitive bias that occurs when individuals overestimate the importance of 1 specific factor and believe it has a greater impact on their well-being or happiness than it actually does.^37 ,38  The illusion arises from a tendency to focus on a particular aspect of or event in our lives, which results in excessive weight in judgments and a distorted perception of reality. In the clinical setting, when caregivers are first learning about their child’s DSD, they may easily focus on “fixing” atypical genital appearance, fearing that this characteristic will weigh heavily on the child’s future adaptation and well-being. Under these circumstances, consideration of other important factors may not receive adequate or any consideration. To guard against this bias, clinicians can assist decision-makers in considering both harms and benefits associated with every reasonable option available, such as through the use of decision aids.³⁹ 

This study has several strengths and limitations. We included a limited set of attributes of DSD care in this study. Although these attributes were identified through a multistep qualitative assessment,²⁴  it is possible that other factors important to DSD care may have been omitted. Participants were predominantly non-Hispanic (85%) and White (80%) and may not reflect the full racial and ethnic diversity of DSD stakeholders. However, the investigators, who were from 3 academic medical centers with established interdisciplinary DSD programs and who serve different demographic populations, were able to recruit a large sample of participants representing different DSD stakeholder groups from across the United States.

The study identified the relative value of attributes associated with the process and outcomes of DSD care from a large sample of DSD stakeholders. Although respondents showed general agreement about the most important attributes of care such as patient satisfaction and health and good communication and education from the care team, LCA showed preference heterogeneity for specific attributes. Findings from this study emphasize the importance of incorporating an individual’s and family’s preferences in shared decision-making and increasing awareness about providers’ own preferences that might impact satisfaction with and quality of DSD care.

The authors thank Norah Crossnohere and Abram Wagner for their input on the analysis plan.

DSD

differences of sex development

LCA

latent class analysis