ESTUDIO DE MERCADO
3.2 INVESTIGACIÓN DE MERCADOS
3.3.1 Administration of Discrete Choice Experiment
A Qualtrics research panel was used to gather a sample of American adults (≥ 18 years) for the online administration of the discrete choice experiment. Online
administration of DCEs is increasingly common in health economics,213 and patient preferences did not significantly differ between those responding to the health state valuation instrument face-to-face and online.239 In order to proceed to the survey,
participants must have answered “yes” to the question “Have you filled a prescription at a pharmacy, other than a mail-order pharmacy, within the last 12 months.” This screening criteria was instituted to exclude those who had not recently filled a prescription at a brick-and-mortar pharmacy. A gender quota was implemented to ensure that the proportions of males and females did not exceed a 60/40 split for either gender. While the Qualtrics panel is opt-in such that participants are not randomly selected, the panel has been shown to have an acceptable level of national representativeness.240 Furthermore, compared with the demographic composition of other opt-in panels, Qualtrics has more representative proportions of older adults, racial minorities, those with low levels of educational attainment, and those living in urban and rural areas.240 This study was approved by the Virginia Commonwealth University Institutional Review Board.
Prior to the discrete choice experiment, participants were provided information about the forthcoming choice tasks and the pharmacy quality measures. The specific wording and information presented was selected based on expert opinion, ISPOR
guidelines for discrete choice experiments, existing literature, and feedback from the pilot
whether a higher or lower number is better for a specific quality indicator,53 participants were supplied with specific information that the ratings would appear on a scale of one to five stars, where more stars were better. A scale of the star levels, along with their
evaluative word labels (i.e. “much below average” to “much above average”) was also presented prior to the start of the experiment. The addition of word labels to the star ratings reflects the real-world presentation format of quality metrics on Medicare’s Hospital Compare and Nursing Home Compare websites. The addition of world labels is also associated with improved understanding of quality stars.65,69 Based on pilot testing feedback, a statement was added to the survey instructions noting, “An overall rating is computed based on a number of scores on specific aspects of pharmacy practice.
Accordingly, an overall quality rating may differ from any single, specific rating.” While past studies have reported that the presentation of an overall performance measure in addition to specific quality metrics aids in the identification of high-quality providers,14,53,71 pilot test participants commonly felt confused when the two scores differed.
The instructions prior to each choice task presented the hypothetical scenario under which participants would be making their choice (i.e. having moved to a new location) and the key assumptions of this DCE, namely that price and location were held constant. Finally, an example choice task was provided. The three primary elements included in the pre-survey introduction – the context of the study scenario, a description of the quality attributes and levels, and the example choice task – are consistent with ISPOR recommendations.166
3.3.2 Sample Size
There is no gold standard for determining the sample size for a DCE.166 Sample size recommendations vary considerably and include both set sample sizes (e.g. 100
respondents) and sample sizes calculated based on parametric approaches.166,241 Formulas for sample size calculation may require input of population proportions, variance, expected parameter values, statistical power and/or the number of parameters estimated, choice sets, and alternatives.241 Considerable variability exists in the sample sizes used in
published healthcare-related DCEs, and very few (6%) report use of parametric approaches for sample size estimation.241 Given the lack of consensus on optimal approaches for
sample size calculations, the need for parameter value estimates for sample size
calculations, and the limited use of parametric approaches in health economics thus far, a calculation-based approach to sample size calculation was not used in this study.180,241
This study targeted a sample size of 500. The target sample size was decided upon after considering expert recommendations, common practices for healthcare-related DCEs, and the a priori plan to conduct a latent class analysis. Two reports from the ISPOR Good Research Practices for Conjoint Analysis Task Force cite recommendations that conjoint analyses include at least 300 study participants.166,180 A review article of discrete choice experiments in healthcare stated that the mean sample size of conjoint analysis studies in health care was 259 respondents.166 Furthermore, Monte-Carlo simulations suggest that sample sizes over 100 are acceptable for latent class analysis, with sample sizes of 500 resulting in precise parameter estimates even under conditions of low data quality.242 Accordingly, recent latent class analyses in healthcare report sample sizes of
approximately 200-500.209,210,243
Additional consideration was given to the a priori analysis plan to evaluate the community pharmacy preferences of the subgroup of patients taking chronic medications.
An estimated 50% of all Americans live with a chronic condition. 244–246 With the use of chronic medications come concerns about high levels of nonadherence247 and an increased risk for drug-drug interactions between two prescribed medications248 or a prescribed and an over-the-counter (OTC) medication.249 A sample size of at least 250 patients taking chronic medications was targeted in order to have a sufficient number of patients for a valid and meaningful subgroup analysis. All study participants, per the inclusion criteria, must have reported filling a prescription at a non-mail order pharmacy in the last 12 months. In an average month, the vast majority (>80%) of prescriptions filled for adult patients are chronic medications.250 Additionally, the prescribing rate for antibiotics, the most commonly prescribed acute medications, among non-elderly adults is approximately 350 prescriptions per 1,000 patients.251 The assumption was thus made that at least half of the 500 study participants would be taking at least one chronic medication, yielding a predicted sample size of 250 for the subgroup analysis.