Investigators ran a web-based conjoint survey between August 14 and November 22, 2024, inviting clinicians, laboratory professionals, researchers, and implementation scientists to evaluate hypothetical point-of-care (POC) HIV NAAT profiles. Recruitment used professional networks and listservs—including Clinton Health Access Initiative, FIND, UNAIDS—and permitted snowball distribution; a probability sample was not possible because no sampling frame exists for the target population. Two hundred seventy-eight individuals accessed the questionnaire and 103 completed it, yielding an approximate 37% completion rate and an average completion time of 11 minutes. Respondents represented 33 countries, with 74.8% based in Africa. The survey collected respondent demographics, current access to NAAT platforms, prioritized use case (initial diagnosis or initiating/monitoring), and preferred health-system level for deployment (community/home, health center, district hospital or higher).
The conjoint task presented 12 balanced test profiles derived from 144 possible combinations of six attributes: limit of detection (LOD), sample-collection method, turnaround time (TAT), result format, price, and power source. Using linear regression on the ratings, investigators estimated attribute importance weights. Across the entire sample, LOD accounted for 36.5% of decision influence, TAT for 28.7%, and sample-collection method for 17.3%. Price, result format, and power source registered much smaller importance weights and were not statistically significant drivers of adoption in these models. The study team observed that stakeholders were willing to trade off among attributes rather than prioritizing every desirable feature simultaneously.
The survey team stratified analyses by role, geography, preferred use case, and health-system level to detect variation. Laboratory respondents placed primary emphasis on LOD and TAT. Non-laboratory respondents also valued sample collection alongside LOD and TAT. Geographic splits showed African respondents assigned greater relative importance to TAT than non-African respondents. When respondents prioritized initial diagnosis as the use case, TAT and LOD dominated; when initiating or monitoring ART was prioritized, LOD and sample-collection method rose in importance. Setting mattered: in community and home contexts, respondents favored fingerstick sample collection and faster TAT even at the expense of sensitivity, whereas at health centers and hospitals high sensitivity regained top priority alongside rapid TAT.
Part-worth utility coefficients quantified preferences within each attribute. The reference levels used in the regression (and thus the zero utility baseline) were 200 copies/mL for LOD, fingerstick for sampling, mains power with battery backup, $8 for price, quantitative result format, and a 15-minute TAT. Across most groups the highest-ranked combination—theoretical ‘ideal’—included an LOD of 200 copies/mL, fingerstick sampling, a 15-minute TAT, mains power with battery backup, a quantitative readout, and a lower test price. The authors cautioned that this ideal mix may be technically infeasible; for example, achieving a 200 copies/mL detection threshold using small capillary blood volumes poses practical challenges. Part-worths also identified acceptable alternative configurations that stakeholders might prefer if some attributes cannot be met simultaneously.
The study team framed the results around necessary trade-offs for test developers: no single device can maximize sensitivity, speed, minimal sample volume, and low cost without engineering and cost constraints. Investigators acknowledged limitations in the survey design and sampling: recruitment via professional networks and listservs, the inability to draw a probability sample, and a modest sample size of 103 respondents may limit generalizability. The authors provided the full survey instrument and supplemental data in the publication. They also noted that while attribute importance weights and part-worth utilities indicate stakeholder priorities, the article did not report specific technical pathways or product-level feasibility assessments to reconcile the most preferred attribute combinations with current assay technology. Remaining unanswered questions documented by the authors include how manufacturers will balance these trade-offs in real-world product development and which specific performance thresholds are achievable at scale given sample-volume and detection-limit constraints.