Digital healthcare is often presented as a contest between traditional medical consultations and increasingly sophisticated artificial intelligence. Yet a choice experiment involving 322 people with chronic kidney disease in Sichuan, China, suggests that patients may place much greater weight on the practical cost of ongoing care. In the researchers’ statistical model, monthly out-of-pocket cost accounted for 50.79% of the relative importance assigned to the service features tested, while an AI conversation service accounted for 9.64%.
The peer-reviewed study, published on 9 October 2026 in Frontiers in Public Health, examined how patients trade off different features of digital kidney care. Its findings matter for healthcare providers developing remote follow-up, virtual consultations and AI-assisted services. They also caution against treating the availability of an AI feature as proof that patients will find a digital programme affordable or attractive.
Why chronic kidney care is different from a one-off appointment
Chronic kidney disease usually requires repeated monitoring and management over time. Depending on the individual, care can include blood pressure monitoring, laboratory tests, medication reviews, dietary advice and specialist assessment. Patients may need to coordinate multiple appointments while balancing travel, work and household costs. Digital services can make some forms of follow-up easier, but they can also introduce subscription fees, connectivity requirements and uncertainty about who will respond to questions.
These considerations make the design of digital kidney services especially important. A technically impressive service is not necessarily the service a patient prefers. The relevant question is how patients value cost, clinician involvement, consultation arrangements and AI assistance when those attributes are presented together.
How the researchers measured patient preferences
Lifang Wei, Yueyang Huang, Difei Duan and colleagues conducted a discrete choice experiment among 322 patients with chronic kidney disease in Sichuan Province. Rather than simply asking respondents whether they liked AI or telemedicine, the researchers presented hypothetical digital care options with different combinations of features and asked participants to choose between them.
The study assessed six attributes of digital care services. The analysis included monthly out-of-pocket cost, provider type, consultation frequency and whether an AI conversation service was included. The remaining features were also incorporated into the experimental design, allowing the researchers to estimate how strongly each attribute influenced the pattern of choices within the tested alternatives.
Discrete choice experiments are useful because patients often value several characteristics simultaneously. A respondent may prefer frequent consultations but not at any price. Someone else may accept less frequent contact if the service is delivered by a trusted specialist team. Statistical models estimate these trade-offs from repeated hypothetical decisions; they do not record real purchases or prove how someone would behave when faced with an actual medical bill.
Cost dominated the modelled choices
The clearest result was the importance of price. Monthly out-of-pocket cost accounted for 50.79% of the modelled relative importance across the six attributes. Provider type ranked second at 17.21%, followed by consultation frequency at 9.95% and the AI conversation service at 9.64%.
These numbers require careful interpretation. They are not percentages of patients who selected a particular feature. Nor do they mean that 50.79% of participants were unable to afford care or that 9.64% wanted AI. They express the relative contribution of attributes to choices within the specific levels and ranges included in the experiment. If the researchers had tested a different price range or a different set of service options, the relative-importance figures could change.
Nevertheless, the difference between the tested attributes is substantial. Cost’s 50.79% relative importance was almost three times provider type’s 17.21% and more than five times the 9.64% assigned to an AI conversation service. In this sample and experimental design, affordability was the most influential feature of the digital care package.
Patients valued a specialist care team
Provider type was the second-most-important tested attribute. The research reported particularly favourable preferences for a combined nephrologist and specialist nurse team. That result fits the practical demands of chronic disease management: patients may need both medical expertise and accessible ongoing support.
A nephrologist can evaluate kidney function, treatment changes and complications. Specialist nurses can support education, monitoring and continuity between appointments. The study does not demonstrate that a combined team improves clinical outcomes compared with every alternative model. It does suggest that the identity and composition of the care team matter to patients when choosing among hypothetical digital services.
This distinction is especially relevant for organisations that advertise an AI chatbot as the centrepiece of a new service. Patients may see value in automated answers, but the presence of a recognisable clinical team can be a more influential part of the overall offering.
Where AI fits into the preference picture
The AI conversation service contributed 9.64% of modelled relative importance. This shows that AI was part of the decision structure, but it did not outweigh cost, provider type or consultation frequency. The study should not be summarised as patients rejecting AI. Its finding is more precise: AI was not the leading attribute when respondents compared the complete service packages tested.
AI-enabled communication may still be useful for appointment reminders, basic education, symptom reporting or navigation, subject to clinical oversight and appropriate privacy protections. However, an AI feature does not substitute for reliable medical review, affordable access or a clear route to a qualified professional when a patient’s condition changes.
For health technology developers, the lesson is to evaluate AI as one component of a care pathway. An algorithm may work accurately in a demonstration yet add little practical value if patients cannot afford the programme or cannot reach a clinician when they need one.
Implications for service design
The findings point to three practical design questions. First, what will patients pay each month, and is that amount predictable? Second, who will provide the care, and how will specialist nurses and physicians coordinate their roles? Third, how often can patients receive meaningful consultations or follow-up?
These questions should be answered before treating AI functionality as the main selling point. Providers could test several service packages with patients, including lower-cost options, different consultation schedules and clearly defined access to a specialist team. A transparent explanation of which tasks are handled by software and which require a clinician may also help patients make informed choices.
The study does not calculate the cost of delivering these models or show which would be financially sustainable for hospitals. Nor does it establish that lowering prices by a particular amount would produce a predictable increase in actual enrolment. Those questions require implementation studies and real-world uptake data.
What the evidence cannot tell us
The research involved 322 participants in one Chinese province. Healthcare financing, digital literacy, specialist availability and attitudes towards remote care differ across regions and countries. Preferences may also vary with disease severity, age, previous experience of telehealth and the extent to which patients already have a trusted clinical team.
Because participants chose between hypothetical packages, the findings measure stated preferences rather than observed service subscriptions or medical outcomes. Relative-importance percentages depend on the attributes and levels researchers selected. The results do not establish whether an AI tool is clinically safe, improves kidney function or reduces hospital admissions. Those outcomes require separate evaluations.
It is also important not to interpret preference as an endorsement of replacing in-person care. Some patients may benefit from digital follow-up while still needing physical examinations, laboratory testing or urgent assessment. A useful digital service should support those pathways, not obscure them.
The broader lesson
For patients living with chronic illness, innovation is valuable when it solves a real problem. In this study, the most influential problem in the choice model was cost, followed by who delivered care. AI assistance was relevant but comparatively less important. Digital kidney programmes may therefore gain more from making care affordable, trustworthy and continuous than from adding automated conversation features without addressing the fundamentals.
Source Information
Original study: Peer-reviewed discrete choice experiment examining patient preferences for digital chronic kidney disease management services.
Authors: Lifang Wei, Yueyang Huang, Difei Duan and colleagues.
Journal: Frontiers in Public Health.
Publication date: 9 October 2026.
Study design: Discrete choice experiment involving 322 patients with chronic kidney disease in Sichuan, China.
Key reported findings: Relative importance of monthly out-of-pocket cost 50.79%, provider type 17.21%, consultation frequency 9.95% and AI conversation service 9.64%.
DOI: 10.3389/fpubh.2026.1905738.
