Developing Attributes and Levels for a Discrete Choice Experiment on Internet Hospital Service Preferences Among Chinese Patients with Chronic Diseases: A Mixed-Methods Study

Abstract

Background

Internet hospitals are increasingly used to support chronic disease care in China. Although previous research has examined online healthcare preferences, evidence on the systematic development of context-specific attributes and levels for Chinese patients with chronic diseases remains limited.

Objective

This study aimed to develop candidate attributes and levels for a future discrete choice experiment (DCE) on Internet hospital preferences among Chinese patients with chronic diseases.

Methods

A mixed-methods approach was applied to systematically develop the relevant attributes and levels. Potential attributes were first identified through a systematic literature review. Next, qualitative interviews with 24 patients with chronic diseases were conducted to elicit contextual attributes. These attributes were then evaluated by 12 experts, and the expert ranking method was applied to prioritize the most relevant ones. Finally, the attributes and levels were determined in accordance with the established principles for developing DCE attributes and levels.

Results

The literature review included 11 studies and generated 34 potential attributes. Patient interviews identified 29 candidate attributes across four dimensions: platform and institutional characteristics, physician service quality, specific services, and drug delivery and related costs. After expert consultation and ranking, seven candidate attributes with 18 levels were retained for a future DCE: out-of-pocket cost, ease of use, physician’s professional title, mode of consultation, upload of self-monitoring data and automatic alerts for abnormal readings, access to electronic medical records and test results, and patient ratings.

Conclusion

This study developed a preliminary attribute-level framework for a future DCE. The framework provides a structured basis for pilot testing and subsequent preference elicitation, and may inform patient-centered Internet hospital service design after further validation.

Plain Language Summary

People living with long-term conditions often need regular follow-up, repeat prescriptions, and ongoing support. Internet hospitals are digital care platforms that can provide remote consultations, online prescribing, and follow-up care, which may save time and make care easier to access. However, we still know little about which parts of these services matter most to people with chronic diseases.

To prepare for a future survey that measures these preferences, we first built a clear list of service features to test. We reviewed published studies and collected 34 possible features. We then interviewed 24 patients with chronic diseases to learn, in their own words, what they consider when choosing an Internet hospital, which added further service features. Next, 12 experts reviewed the list and ranked the most important items. After combining similar ideas and prioritizing what was most relevant, we identified seven key features that are likely to shape choices: out-of-pocket cost, ease of use, physician’s professional title, mode of consultation, upload of self-monitoring data and automatic alerts for abnormal readings, access to electronic medical records and test results, and patient ratings. We also defined 18 practical levels for these features.

The next step is to test whether patients understand these features and levels and then use them in a full discrete choice experiment. That future study will estimate how patients compare different service features when choosing Internet hospitals.

Introduction

Chronic diseases are a major global public health challenge. Noncommunicable diseases, including cardiovascular diseases, cancer, diabetes, and chronic respiratory diseases, cause approximately 41 million deaths each year and account for nearly 74% of global deaths.¹ In China, chronic diseases account for 88.5% of all deaths.² Their long-term and recurrent nature creates a continuing need for standardized treatment, regular monitoring, follow-up care, and medication management.

Internet hospitals offer a potential way to support these needs. In China, they are regulated digital care platforms established by or affiliated with licensed medical institutions, rather than stand-alone teleconsultation websites. They are subject to requirements for institutional access, physician qualifications, online prescribing, information security, and service supervision.^3,4 Services may include online follow-up, prescription renewal, access to test results, payment, reimbursement, and drug delivery. Compared with digital care models in some other countries, Chinese Internet hospitals are more closely connected with physical hospitals, public medical insurance, and regional regulatory systems.

National policies have supported the development of Internet hospitals and digital chronic disease management. The 2018 “Internet Plus Healthcare” policy promoted online medical services, prescription circulation, insurance supervision, and digital identity systems.⁵ More recently, the Blue Book on Comprehensive Management of Chronic Diseases under the Healthy China Strategy (2025) called for a more refined chronic disease management system supported by digital health.⁶ The scale of Internet hospital development has also expanded. By February 2024, 30 provinces had established provincial-level Internet medical regulatory platforms, and more than 2,700 Internet hospitals had been approved nationwide.⁷ In 2025, Internet medical platforms served approximately 5,000 medical institutions and delivered more than 55 million service episodes annually.⁸

Discrete choice experiments (DCEs) provide a useful approach for examining healthcare service preferences. Based on random utility theory, DCEs ask respondents to choose between hypothetical service options described by different attributes and levels, thereby allowing researchers to estimate how people value service characteristics and make trade-offs.^9,10 It is also consistent with patient-centered care and technology acceptance perspectives, which emphasize patients’ needs, ease of use, trust, and perceived usefulness.^11,12 In the context of Internet hospitals, these perspectives suggest that patients’ choices may be shaped by both healthcare service characteristics and digital platform features.

Previous DCEs have examined preferences for telemedicine, online consultations, Internet hospitals, and disease-specific digital care.^13–18 Common attributes include cost, waiting or response time, physician qualifications, consultation mode, hospital level, insurance reimbursement, and patient ratings. Some studies have described attribute development through literature reviews, qualitative interviews, focus groups, expert consultation, or pilot testing, but often provide limited detail on how attributes were generated, merged, excluded, and translated into levels.^19–21 Broader qualitative and digital health studies have also identified trust, privacy, usability, communication quality, service boundaries, and continuity of care as relevant factors,^22–24 although these findings are not always converted into attributes and levels suitable for a DCE.

Therefore, further work is needed to develop a transparent and feasible attribute-level framework for patients with chronic diseases in the Chinese Internet hospital context. This is particularly relevant because chronic disease management often involves repeated service use, long-term medication, regular monitoring, and follow-up care. This study used a mixed-methods design combining literature review, qualitative interviews with patients, and expert consultation to develop candidate attributes and levels for a future DCE on Internet hospital service preferences among Chinese patients with chronic diseases.

Materials and Methods

A mixed-methods approach was adopted (Figure 1). First, a systematic literature review was conducted to preliminarily identify conceptual attributes related to the preferences of patients with chronic diseases in their choice of Internet hospitals. Second, these conceptual attributes were used as the basis for developing an initial interview framework, through which qualitative interviews and inductive analysis were conducted to identify contextual attributes valued by patients. Third, expert consultation and importance ranking were conducted to refine and limit the number of attributes to a manageable range, ensuring that the final attributes and levels adhered to the established principles for developing DCE attributes and levels. The qualitative component was reported according to the COREQ checklist²⁵ (Supplementary File 1).

Figure 1.

Process and methods for developing candidate attributes and levels for a future DCE on Internet hospital preferences among patients with chronic diseases.

Literature Review

A systematic literature review was conducted in PubMed, Web of Science, Embase, CNKI, Wanfang, and VIP databases, covering all studies published from the inception of each database to December 16, 2025. In addition to database searching, we identified further records through reference list screening, grey literature searches, and manual searching. Both subject terms and free-text terms were used in the search strategy, including “chronic disease”, “patient”, “Internet hospital”, “discrete choice experiment”, and “DCE”. Detailed search strategies for each database are provided in Tables S1–S6 (Supplementary File 2).

Eligibility criteria were established as follows. Studies were required to meet at least one of the following population criteria and the methodological criterion simultaneously.

Population Criteria

Primarily included studies eliciting preferences for Internet hospital services among patients with chronic diseases; when chronic disease–specific evidence was limited, we additionally considered studies conducted in general populations (eg, patients with other conditions or the general public).

Methodological Criterion

The study employed a DCE as the preference-elicitation method.

Language Restriction

Only studies published in English or Chinese were included.

Exclusion Criteria

(i) studies outside the above research scope, such as those examining general patients’ preferences for conventional (non-Internet) healthcare services or patients with chronic diseases’ preferences for specific treatment regimens; (ii) records without full-text availability, including conference abstracts, posters, and letters to the editor.

The screening process was conducted independently by two researchers (DQ and XY) in two stages—title and abstract screening, followed by full-text review. Any discrepancies were resolved through discussion with a third researcher (YG). Finally, data from the included studies were systematically extracted and summarized to identify potential attributes and levels relevant to patients’ choices of Internet hospitals. Extracted information included the first author, country, study population, included attributes, and the methods used for attribute and level development. A DCE attribute-level database was then established to inform subsequent qualitative interviews and expert consultation.

Qualitative Interviews

Based on the literature review and DCE attribute development principles, we developed an initial semi-structured interview guide. The guide was refined through two to three rounds of internal discussion among the research team, with repeated revisions to improve clarity, remove leading wording, and ensure consistency with the study objectives (Supplementary File 3).

Participants were recruited between March and May 2026 from chronic disease outpatient clinics and inpatient wards of a tertiary general hospital in Shanghai that receives referrals from across China, and from a community health service center where patients were predominantly local residents. Participants were purposively sampled from patients with chronic diseases, including both those with and those without previous experience using Internet hospital services. Eligible patients were identified through clinician referral and on-site screening. All interviewers (DQ, WD, SD, and RL) had backgrounds in health policy or health services research and were trained in qualitative interviewing. The researchers explained the study purpose, provided an information sheet, and obtained written informed consent.

Interviews were conducted in a private or quiet area of the participating institutions. Only the participant and interviewer were present. Interviews included basic sociodemographic and clinical information, followed by open-ended questions on how participants evaluated Internet hospital services and which features they considered important when making choices. Neutral prompts were used when needed. Interviews were audio-recorded with permission, and field notes were written during or immediately after each interview. No repeat interviews were conducted.

Audio recordings were transcribed verbatim and checked for accuracy. Transcripts were managed in NVivo 11. Two researchers (WD and TT) independently coded the transcripts using an inductive thematic approach. Coding results were compared, and disagreements were resolved through discussion; when needed, a third reviewer (DQ) made the final decision. Data saturation was assessed during data collection and coding and was considered reached when three consecutive interviews produced no new attribute-relevant codes or candidate attributes. Candidate attributes mentioned by at least the mean number of participants per attribute were carried forward to expert review, whereas those mentioned by fewer participants were not retained. Semantically overlapping attributes were consolidated where appropriate.²⁶

Expert Consultation and Ranking

To ensure the validity and objectivity of the attributes and levels for the DCE, we convened a multidisciplinary expert panel including Internet hospital clinicians, digital health/hospital operations managers, and researchers with expertise in preference elicitation and DCE design. Three female researchers with qualitative research experience (YG, DQ, XY) invited 12 experts to participate in online or in-person consultations via email or telephone. The interview outline (Supplementary File 4) consisted of four parts: first, experts provided demographic and professional information; second, researchers explained the candidate attributes and their corresponding levels in detail; third, experts reviewed the list and proposed additions or deletions of attributes and levels; and fourth, experts rated the importance of each attribute and suggested revisions to their definitions or level descriptions as needed.

Because 16 candidate attributes were reviewed, experts were asked to identify the five most important and five least important attributes. This format was chosen to encourage prioritization while limiting expert burden across a relatively large candidate list. To facilitate quantitative comparison, positive scores (from +5 to +1, in descending order) were assigned to the “most important” attributes, and negative scores (from −5 to −1, in ascending order) were assigned to the “least important” attributes.²⁷ Attributes not selected by an expert were coded as NA for score aggregation. Composite scores were calculated by summing the scores across experts. Expert agreement was assessed using tie-adjusted Kendall’s coefficient of concordance (W). Expert scores were converted to ordinal ranks, and unselected attributes were assigned tied middle ranks. Statistical significance was assessed at P<0.05.

Principles for Developing Attributes and Levels for DCEs

Although no single standardized procedure exists, DCE attributes and levels should reflect the key features of the service and meet several practical criteria.^20,21 Attributes should be relevant to the decision context, conceptually distinct, and collectively cover the main dimensions of interest. They should also be meaningful for service design, information provision, or policy. Levels should be clear, realistic, mutually exclusive, and span a plausible range. To maintain cognitive feasibility and design efficiency, DCEs commonly include 4–7 attributes with 2–4 levels each.

Ethical Approval and Informed Consent

This study was approved by the Ethics Committee of the Shanghai Health Development Research Center (Shanghai Medical Information Center) (Approval No. 2025003). The study was conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent before the interviews, and the consent process covered the publication of anonymized responses and direct quotations.

Results

Characteristics of the Included Studies

As of December 16, 2025, a total of 746 relevant studies were identified, including 158 from PubMed, 78 from Web of Science, 108 from Embase, 21 from CNKI, 274 from Wanfang, 101 from VIP, and 6 from other sources. After excluding 188 overlapping records, 558 studies remained. Subsequently, 469 studies were excluded after title and abstract screening, and a further 78 were excluded after full-text review. Ultimately, 11 studies were included in the final analysis (Table 1). The screening process and results are illustrated in Figure 2.

Table 1.

Characteristics of Included DCE Studies on Internet Hospital Preferences Among Patients with Chronic Diseases

No.	First Author (Year)	Country	Study Population	Attributes	Methods for Developing Attributes and Levels
1	Woode et al (2025)28	Australia	Chronic knee pain patients	Choice of physiotherapist; Travel time one-way (min); Waiting time (min); Listening and discussion time (min); Security and privacy; Pain outcome; Consultation cost (AU$/45 min)	Literature review; expert consultation
2	Chen et al (2025)13	China	Chronic patients	Cost (CNY); Type of internet hospital; Type of online doctor; Waiting time for an appointment; Communication mode; Online payment method	Literature review; expert consultation; pilot testing
3	Tierney et al (2024)29	United States	Hypertensive patients	Ability to see a clinician with whom you have an established relationship; Profession of available clinician; Copayment; Appointment type; Time of available appointment; Earliest available appointment	Literature review; expert consultation
4	Hu et al (2024)14	China	Obesity	Doctor level; Hospital level; Out-of-pocket cost (RMB); Waiting time (min;); Consultation format; Consultation duration (min;)	Literature review; focus group discussions; expert consultation; pilot testing
5	Wu et al (2023)15	China	Chronic patients	Disease severity; One way trip distance to hospital (min); The price of SOOC (¥)(the price of specialist outpatient offline consultation); The increasing ratio of medical insurance payment for online services compared to offline; The convenience of applying SOOC services; Doctors’ recommendation	Literature review; focus group discussions; pilot testing
6	Wang et al (2023)16	China	Chronic patients	Waiting time; Doctor’s professional title; Doctor’s evaluation score; Grades of the hospital; Scale of consultation platform; Cost (¥)	Literature review; focus group discussions; qualitative interviews; expert consultation
7	Savira et al (2023)17	Australia	Chronic patients	How do you talk with the GP?; How far do you have to travel? (if relevant); How well do you know the GP?; Does the GP have full access to your complete medical history?; How long do you have to wait before your appointment?; Technical quality (if relevant); Opportunity to ask additional questions; Out-of-pocket cost to you	Literature review; focus group discussions; expert consultation; pilot testing
8	Li et al (2023)18	China	Chronic patients	Hospital grade; availability of human-assisted services; average physician rating percentage; initial physician response time after inquiry (min;); insurance reimbursement; page navigation process; drug delivery support	Literature review; qualitative interviews; expert consultation
9	Wei et al (2022)30	China	Chronic patients	Operating platform type; physician recommendation index; consultation mode; service cost	Literature review; expert consultation
10	Buchanan (2021)31	United Kingdom	Chronic patients	Consultation waiting time; GP reputation; consultation cost; the process by which antibiotics would be collected; and the similarity of the consultation to a traditional ‘face-to-face’ appointment	Literature review; qualitative interviews; expert consultation
11	Park et al(2011)32	South Korea	Diabetic patient	Monthly fee; Type of service platform; Type of service provider; Service scope; Personalization of consultation; Service hour; Reply time; Assurance of service; System failure; Confidentiality	Literature review; expert consultation

Figure 2.

Flow Diagram of Literature Screening for DCE Studies on Internet Hospital Preferences among Patients with Chronic Diseases.

Among the 11 included studies, all examined preferences among patients with chronic diseases, but the specificity of the target population varied. Four studies focused on clearly defined disease groups such as chronic knee pain, hypertension, obesity, and diabetes and were conducted in Australia,²⁸ United States,²⁹ China,¹⁴ and South Korea.³² The remaining seven studies recruited broader chronic-disease populations without restricting to a single condition and were predominantly conducted in China, with five studies,^{13,15,16,18,30} while Australia¹⁷ and the United Kingdom³¹ each contributed one study. The methods used for attribute and level development primarily included literature review, focus group discussions, qualitative interviews, expert consultation, and pilot testing, with literature review applied in all studies. The number of attributes identified in these studies ranged from 4 to 10, and each attribute was generally defined with 2 to 4 levels.

After consolidating semantically similar items, we identified 34 distinct attributes. The research team reviewed each attribute based on its direct relevance to Internet hospital choice, applicability to the Chinese service context, conceptual clarity and distinctness, and suitability for exploration in patient interviews. Attributes that were only indirectly related to the decision context, overlapped substantially with other attributes, or were difficult to present as clear and understandable service features were excluded. This process resulted in 20 attributes for the qualitative interview stage. The five most frequently reported attributes were physician response time, patient ratings, physician’s professional title, consultation cost, and mode of consultation.

Qualitative Interview Findings

Participants’ Characteristics

A total of 24 patients with chronic diseases were interviewed (Table 2). Fourteen participants (58.33%) had previously used Internet hospital services, whereas 10 (41.67%) had not. The group included 15 women and 9 men, and both younger adults and older adults were represented. Participants with short-cycle courses or undergraduate education or above accounted for 54.17%, while those with junior secondary education or below accounted for 33.33%. Most participants were employed (50.00%) or retired (37.50%). The disease categories included cardiovascular and cerebrovascular diseases (33.33%), metabolic diseases (16.67%), cancer (8.33%), and other chronic conditions (41.67%). Basic Medical Insurance, including Urban-Rural Resident Basic Medical Insurance (URRBMI) and Urban Employee Basic Medical Insurance (UEBMI), covered 91.67% of participants. Commercial health insurance coverage was low.

Table 2.

Demographic and Health Characteristics of Participants in the Attribute-Development Study for a Future DCE on Internet Hospital Preferences Among Patients with Chronic Diseases

Category	Number (%)	Category	Number (%)
Gender		Educational Attainment
Male	9 (37.50%)	Junior Secondary Education or Below	8 (33.33%)
Female	15 (62.50%)	Senior Secondary Education/Secondary Vocational Education	3 (12.50%)
Age (years)		Short-cycle Courses/Undergraduate Degree	10 (41.67%)
18–30	8 (33.33%)	Postgraduates (Master’s Degree or Above)	3 (12.50%)
31–40	2 (8.33%)	Occupation Type
41–50	3 (12.50%)	Employed	12 (50.00%)
51–60	1 (4.17%)	Retired	9 (37.50%)
>60	10 (41.67%)	Unemployed	3 (12.50%)
Type of Chronic Disease		Type of Medical Insurance
Cardiovascular and Cerebrovascular Diseases	8 (33.33%)	Urban–Rural Resident Basic Medical Insurance (URRBMI)	12 (50.00%)
Metabolic Diseases	4 (16.67%)	Urban Employee Basic Medical Insurance (UEBMI)	10 (41.67%)
Cancer	2 (8.33%)	Commercial Health Insurance	2 (8.33%)
Others	10 (41.67%)	Prior Use of Internet Hospitals
		Yes	14 (58.33%)
		No	10 (41.67%)

Identification of Potential Attributes

During the interview phase, patients mentioned a total of 29 candidate attributes relevant to choosing an Internet hospital, including nine attributes that were newly identified from patient narratives (Table 3). These potential attributes were further categorized into four dimensions: platform and institutional characteristics, physician service quality, specific services, and drug delivery and related costs.

Table 3.

Frequency of Candidate Attributes Mentioned by Patients with Chronic Diseases When Choosing Internet Hospitals

Dimension	Potential Attributes	n*
Platform and Institutional Characteristics	Hospital Grade	20
	Operating Platform Type	16
	Information Provision	13
	Distance to Medical Facility	9
	Data Security and Privacy Protection	7
Physician Service Quality	Physician’s Professional Title	20
	Physician Response Time	12
	Patient Ratings	13
	Physician Consultation Time	4
	Physician Continuity	4
Specific Services	Ease of Use	22
	Mode of Consultation	17
	Medical Record and Health Data Management	16
	Upload of Self-Monitoring Data and Automatic Alerts for Abnormal Readings	15
	Follow-up Reminder Function	14
	Support Services	13
	Psychological Counseling	12
	Personalized Health Education	9
	Appointment Flexibility	9
	Family Coordination Function	7
	Referral Convenience	4
	Access to On-site Resources	2
	Authentication Method	3
	Prescription Availability (for Long-term Medication)	1
Drug Delivery and Related Costs	Out-of-Pocket Cost	18
	Drug Delivery Method	16
	Drug Delivery Time	14
	Consultation Cost	14
	Medical Reimbursement	7

Notes: *n: The number of respondents who mentioned this attribute.

Specifically, for platform and institutional characteristics, hospital grade and operating platform type were frequently raised, suggesting that participants used institutional cues to judge platform credibility and perceived reliability (eg, “I feel more reassured when it’s a tertiary hospital platform rather than an unknown site”, P08).

Within physician service quality, participants were most attentive to physicians’ professional title, viewing it as a salient signal of expertise and clinical trustworthiness (eg, “I would rather choose a senior doctor—even online—because the title reflects experience”, P03; “If it’s a chief physician, I’m more confident about the advice”, P19).

In terms of specific services, ease of use (page operation process) was mentioned most often, followed by mode of consultation and functions such as access to electronic medical records and health data management (eg, “If the steps are complicated, I may give up halfway”, P11; “Text is sometimes not enough—I prefer video when the condition is complex”, P06).

Moreover, regarding drug delivery and related costs, participants primarily focused on economic burden and medication accessibility, paying particular attention to out-of-pocket cost and drug delivery method (eg, “Cost is the first thing I look at—if it’s too expensive, I won’t use it”, P21;

“If the delivery service is convenient and reliable, I would prefer using an Internet hospital because I can get my medicines without making an extra trip to the hospital”, P09).

The mean number of participants mentioning each candidate attribute was 12. Semantically overlapping attributes were merged where appropriate; for example, “consultation cost” was incorporated into “out-of-pocket cost” to reduce redundancy and respondent burden in subsequent choice tasks. Following this process, 16 candidate attributes were retained for expert consultation and ranking.

Expert Consultation Results

Characteristics of the Experts

Twelve experts participated in the consultation. Consultation duration ranged from 32 to 98 minutes, with a mean of 51 minutes (Table 4). Five experts were from higher education institutions and had experience in DCE design and application. Six were from medical institutions and had experience in chronic disease diagnosis, treatment, and management. One was from a provincial center for chronic disease prevention and health management and focused on policy development and health service system construction. Eight experts (66.67%) had 20 years or more of professional experience, and ten (83.33%) held senior professional titles or managerial positions, such as Professor, Chief Physician, Associate Chief Physician, or Attending Physician.

Table 4.

Characteristics of Experts Participating in the Attribute-Development Consultation for a Future DCE on Internet Hospital Preferences Among Patients with Chronic Diseases

Expert No.	Gender	Institution	Title/Position	Years of Experience	Consultation Duration (Min)
N1	Male	Provincial Center for Health Management	Division Director	30	47
N2	Male	Tertiary Hospital	Department Director	30	72
N3	Female	University	Professor	30	98
N4	Male	Community Health Service Center	Director	20	46
N5	Male	University	Professor	30	65
N6	Male	University	Associate Professor	10	41
N7	Male	University	Professor	30	55
N8	Female	University	Lecturer	10	39
N9	Male	Tertiary Hospital	Associate Director	20	38
N10	Female	Tertiary Hospital	Supervisor	15	39
N11	Female	Tertiary Hospital	Deputy Section Chief	10	32
N12	Male	Tertiary Hospital	Department Director	30	36

Expert Evaluation and Finalization of Attributes and Levels

The 12 experts evaluated and ranked the 16 candidate attributes, and the results are presented in Table 5. The tie-adjusted Kendall’s W was 0.31 (P<0.001), indicating statistically significant but moderate agreement among the experts. We further reviewed expert comments together with evidence from the literature review and patient interviews, and retained attributes that were clearly interpretable, minimally overlapping, actionable in Internet hospital services, and feasible for a concise DCE task.

Table 5.

Attribute Scores from the Expert Consultation on DCE-Based Preferences for Internet Hospital Selection Among Patients with Chronic Diseases

Rank	Attribute	Expert No.												Total
		N1	N2	N3	N4	N5	N6	N7	N8	N9	N10	N11	N12
1	Out-of-pocket cost	5	NA	NA	4	NA	NA	3	2	3	5	NA	NA	22
2	Ease of use	NA	2	2	NA	2	3	1	NA	2	1	3	5	21
3	Physician’s professional title	3	4	−5	5	NA	NA	NA	NA	5	3	4	NA	19
4	Mode of consultation	4	3	1	−2	NA	−1	4	4	NA	NA	1	4	18
5	Upload of self-monitoring data and automatic alerts for abnormal readings	2	−3	3	2	4	5	−3	5	−1	NA	NA	NA	14
6	Access to electronic medical records and test results	−2	NA	4	3	5	4	−4	NA	NA	4	NA	−2	12
7	Patient ratings	NA	1	NA	NA	1	2	NA	−1	−2	2	NA	3	6
8	Follow-up reminder function	−1	NA	5	−3	3	NA	2	3	NA	−1	−2	NA	6
9	Drug delivery method	1	NA	NA	1	NA	−3	NA	NA	NA	NA	2	NA	1
10	Hospital grade	NA	5	−3	NA	−1	1	NA	−4	NA	−3	NA	2	−3
11	Drug delivery time	−3	NA	−1	−1	−3	NA	NA	NA	4	NA	−1	NA	−5
12	Physician response time	NA	−4	−2	NA	NA	NA	NA	−5	NA	−2	5	−3	−11
13	Information provision	NA	−2	NA	NA	−4	−2	−2	−2	1	NA	−3	1	−13
14	Support services	−5	−5	NA	−5	NA	NA	−5	1	−3	NA	NA	−1	−23
15	Psychological counseling	NA	−1	NA	NA	−5	−5	−1	NA	−5	−5	−4	−5	−31
16	Operating platform type	−4	NA	−4	−4	−2	−4	5	−3	−4	−4	−5	−4	−33

Notes: *NA = not selected among the five most or five least important attributes; treated as 0 when calculating the composite score.

Ultimately, seven candidate attributes influencing patients’ preferences for Internet hospital services were retained (Table 6). The aggregated ranking scores of the seven retained candidate attributes are Figure 3, and the detailed per-expert scoring distributions across all attributes are provided in Supplementary File 2 (Figures S1–S12). An attribute tracking table was used to document the generation, merging, exclusion, and refinement of attributes across all study stages (Supplementary File 5).

Table 6.

Final Candidate Attributes and Levels Included in the DCE on Internet Hospital Preferences Among Patients with Chronic Diseases

No.	Attribute	Levels	Description
1	Out-of-pocket cost	10 CNY	The portion of the medical expenses paid by patients at Internet hospitals that is not reimbursed by medical insurance.
		70 CNY
		200 CNY
		1000 CNY
2	Ease of use	Simple	The operational convenience experienced by patients when using Internet hospital services.
		Complex
3	Physician’s professional title	Chief Physician	The physician’s professional rank, reflecting their qualification and level of clinical experience.
		Associate Chief Physician
		Attending Physician
4	Mode of consultation	Video consultation	The communication mode patients choose for online consultations.
		Voice consultation
		Text consultation
5	Access to electronic medical records and test results	Available	Whether the Internet hospital allows access to integrated medical records and test results to support chronic disease management.
		Unavailable
6	Upload of self-monitoring data and automatic alerts for abnormal readings	Available	Whether the Internet hospital supports remote upload of self-monitoring data and provides automatic alerts for abnormal readings (eg, blood glucose spikes or blood pressure fluctuations).
		Unavailable
7	Patient ratings	Good	Patient feedback on physicians’ service quality, communication, and treatment outcomes.
		Poor

Figure 3.

Aggregated expert ranking scores for the seven retained candidate attributes reviewed by twelve experts.

Out-of-Pocket Cost

Out-of-pocket cost was retained as a final attribute because it received the highest expert ranking score and was also prominent in the earlier stages of attribute development. It was defined as the portion of Internet hospital medical expenses that patients pay themselves and is not reimbursed by medical insurance. Experts considered this attribute to reflect patients’ direct financial burden and to be relevant to reimbursement scenarios.^20,33 They also advised using out-of-pocket cost as an umbrella concept instead of separating several fee components, such as consultation fees and service fees, because separate fee attributes could overlap and increase task complexity.

Cost has to be in there—it directly determines whether people will use the service, and it also matters for how reimbursement and compensation should be designed. (N1)

Patients care about what they end up paying out of pocket, not whether it’s labeled as a consultation fee or a service fee—splitting it just creates overlap and makes the survey harder to interpret. (N10)

Experts advised that out-of-pocket cost levels should be anchored in real-world pricing and should span a sufficiently wide range to capture potential gradient and threshold effects, consistent with ISPOR good research-practice recommendations.³⁴ Existing evidence also indicates substantial variation in online consultation fees, with a reported median around 49 CNY³⁵ and an average of approximately 85 CNY.³⁶ More broadly, a national cross-sectional survey of Internet hospitals reported consultation fees ranging from 0 to 1,000 CNY per visit, with the median at 20 CNY and the 75th percentile at 200 CNY.³⁷

To further contextualize the price landscape, we reviewed publicly available pricing and reimbursement contexts in four representative provinces/municipalities (Shanghai, Yunnan, Guangdong, and Shandong).^38–41 Across these settings, routine online consultations are commonly priced in the tens of yuan, while charges for specialist or higher-intensity services can extend into the hundreds of yuan, depending on clinician seniority and consultation mode; public reporting has described a typical online consultation range of roughly 10–500 CNY. In addition, our pricing scan found that Internet hospital consultations were priced up to 600 CNY (USD 86.96) in Yunnan, Guangdong, and Shandong, and approximately 1,000–3,000 CNY (USD 144.93–434.78) in Shanghai.

Synthesizing expert input with the above evidence, we finalized four out-of-pocket cost levels: 10 CNY (USD 1.45), 70 CNY (USD 10.14), 200 CNY (USD 28.99), and 1,000 CNY (USD 144.93). Among these levels, 10 CNY reflects minimal patient payment under high reimbursement; 70 CNY aligns with common mainstream online consultation prices; 200 CNY represents a higher but still commonly observed payment level; and 1,000 CNY was included to reflect high-cost online service options observed in real-world Internet hospital practice.

Ease of Use

Ease of use was ranked as the second-highest priority attribute by experts and was consistently identified as an important concept throughout the earlier stages of attribute development. Previous DCE studies had already included similar concepts, such as the page operation process.^18,42 Patients also frequently reported difficulties related to navigation, service entry, function searching, and completing online care procedures. Experts emphasized that usability is a key determinant of real-world uptake, particularly for chronic disease populations that include many older adults and patients requiring repeated service use.⁴³ They highlighted that information overload and redundant navigation can substantially reduce acceptance of Internet-based care.²³

A lot of patients with chronic disease are older—if the interface is even slightly complicated, people get discouraged and drop off, so usability is what makes the service actually workable. (N11)

We want this attribute to capture whether it ‘feels easy to use’—fewer steps, clearer pathways, and friendly prompts matter more than adding another set of features. (N12)

Experts recommended two levels, “simple” and “complex”, because patients generally described usability in broad, contrasting terms. More granular measures, such as the number of clicks or pages, were difficult to apply consistently across platforms. The binary structure reduced cognitive burden while capturing navigation clarity, operational steps, page layout, and supportive prompts. “Simple” referred to a clear, intuitive interface that allowed key functions to be completed with minimal effort, whereas “complex” referred to a less intuitive interface requiring more steps, repeated page switching, or additional effort.

Physician’s Professional Title

Physician’s professional title was ranked as the third-highest priority attribute by experts. They viewed professional title as one of the most salient and credible quality cues in Internet hospital settings, where patients often have limited information to judge clinical competence directly.^16,44 Experts also noted that while hospital grade signals institutional capacity, physician title more directly reflects individual expertise and is therefore likely to carry greater weight in remote consultations.^16,44

In online care, patients can’t easily judge a doctor’s capability—professional title is one of the most straightforward and trustworthy quality signals, so it should stay. (N4)

Hospital grade says something about the institution, but what really determines consultation quality is the physician’s own credentials—especially in Internet hospitals. (N9)

Accordingly, experts recommended three levels—chief physician, associate chief physician, and attending physician—to represent clear gradients in clinical experience and perceived credibility. Resident physicians were not included, as experts considered them unlikely to provide independent consultations on Internet hospital platforms and less preferred by patients.⁴⁵

Mode of Consultation

Mode of consultation was ranked as the fourth-highest priority attribute by experts. They emphasized that text, voice, and video consultations offer different levels of interaction and clinical value, and that modality choice can meaningfully affect decision quality and patient experience—particularly for chronic disease follow-up and more complex consultations.^46,47 Patients also highlighted that consultation mode influenced their sense of communication quality and confidence in online care.

Text, voice, and video aren’t interchangeable—especially for chronic follow-up or complicated cases, the communication mode can change both clinical judgment and the patient’s experience. (N7)

Keeping the modes separate helps us estimate patients’ trade-offs across communication formats and cost. (N8)

Accordingly, experts recommended three levels reflecting mainstream Internet hospital service formats: text consultation, voice consultation, and video consultation. These levels capture differences in clinician-patient interaction and are easy to present in choice tasks.

Upload of Self-Monitoring Data and Automatic Alerts for Abnormal Readings

Upload of self-monitoring data and automatic alerts for abnormal readings was ranked fifth by experts. This attribute refers to whether Internet hospitals allow patients to upload home measurements, such as blood glucose, blood pressure, or ECG data, and receive automated alerts when readings are abnormal.⁴⁸ Experts considered this function relevant to chronic disease management because it can reduce information gaps between patients and clinicians and support earlier detection of risk.⁴⁹ Evidence also suggests that remote monitoring and alert-enabled follow-up can improve outcomes and self-management adherence in conditions such as diabetes and heart failure.^48,50

If patients can upload blood glucose or blood pressure and trigger alerts, Internet hospitals start to look like real long-term chronic care—not just a one-off consultation. (N3)

This feature supports early warning and makes it easier to adjust treatment in time. (N5)

Experts also noted that, although many platforms already support online consultations and basic access to medical records,⁵¹ real-time data upload and automated alerts remain less widely available, and monitoring data from disease-specific applications are often not integrated into Internet hospital systems or clinical workflows.^52,53 Given the multiple functions involved—such as data upload, integration, visualization, and automated alerts—further subdivision could create overlapping levels and increase task complexity. This attribute was therefore defined using two levels, “available” and “unavailable”, according to whether patients could upload self-monitoring data and receive automated feedback or alerts for abnormal readings.

Access to Electronic Medical Records and Test Results

To improve conceptual clarity and patient comprehension, medical record and health data management was renamed “access to electronic medical records and test results.” Access to electronic medical records and test results was ranked sixth by experts. It was defined as whether an Internet hospital provides access to integrated medical records and test results, such as prior examinations, medication history, and clinician notes, to support chronic disease management.⁵¹ Experts noted that chronic care decisions often depend on longitudinal information. Without integrated access, care may become fragmented and may lead to repeated tests or inconsistent decisions across visits.⁵⁴ Record access can support patient self-management and clinician judgment, including through decision-support tools where available,^55,56 and may also support data-driven chronic disease management at the system level.^55,57

You can’t really manage chronic disease without past tests and medication history—being able to see the full record directly affects whether online decisions are safe and efficient. (N5)

Data integration isn’t just a convenience; it reduces repeat testing and improves consistency, and it’s what helps Internet hospitals move from one-off ‘consults’ to continuous care. (N6)

Experts similarly recommended using two levels, “available” and “unavailable”, for this attribute. Because access to electronic medical records and test results involves multiple related functions, including previous examinations, medication history, and clinical notes, further separating these components could lead to overlapping concepts. Therefore, this attribute focused on whether integrated access to relevant medical information was provided within the Internet hospital platform.

Patient Ratings

Patient ratings were ranked seventh by experts. This attribute refers to the platform-displayed feedback on physicians (eg, star ratings, satisfaction percentages, or brief reviews) and serves as a practical quality cue when patients have limited information about a provider, which can shape trust and choice in Internet hospital settings.⁵⁸ Experts noted that many users—especially older adults or those with lower health literacy—tend to rely on simple, glanceable indicators rather than lengthy text, and therefore recommended keeping this attribute intuitive and easy to process.⁵⁸

Platform ratings are a very common quality shortcut—especially when patients don’t know the doctor and there’s clear information asymmetry. (N10)

We’d keep ratings simple and visual so people can actually use them to compare options, instead of getting overwhelmed by details. (N12)

Accordingly, experts recommended dichotomizing ratings into two levels: “good” and “poor.” Although platforms may display ratings using different formats, including star scores, satisfaction percentages, or written reviews, these formats are not standardized across Internet hospitals and are difficult to translate into comparable levels. A binary classification was therefore considered more intuitive for patients and more feasible for a hypothetical choice task. The “good” level represented physicians receiving generally favorable feedback and positive patient evaluations, whereas the “poor” level represented relatively unfavorable feedback or lower patient satisfaction.

Discussion

This mixed-methods study used a literature review, patient interviews, and multidisciplinary expert consultation to develop candidate attributes and levels for a future DCE on Internet hospital preferences among Chinese patients with chronic diseases. The stepwise process included a systematic review of 11 studies, in-depth interviews with 24 patients with chronic diseases, and consultation with 12 experts. The resulting framework comprised seven candidate attributes and 18 corresponding levels: out-of-pocket cost, ease of use, physician’s professional title, mode of consultation, access to electronic medical records and test results, upload of self-monitoring data and automatic alerts for abnormal readings, and patient ratings.

Comparison with Previous Studies

The candidate set combined service attributes commonly used in previous DCEs with digital functions more closely related to the continuity of chronic disease care. Out-of-pocket cost, physician qualification or professional title, consultation mode, patient ratings, and usability or service-process convenience have appeared in previous DCEs of Internet hospitals, online consultations, and telehealth services.^{13–18,28,29} Access to longitudinal medical information has also been considered in telehealth choice research.¹⁷ The recurrence of these attributes across the literature review, patient interviews, and expert consultation supports their inclusion as candidates for further testing.

Compared with conventional service attributes, access to electronic medical records and test results and the upload of self-monitoring data with automatic alerts were less consistently represented in the reviewed DCEs. Their inclusion may reflect the repeated monitoring and information-continuity needs of patients with chronic diseases. These functions could extend Internet hospital services beyond episodic online consultation by linking information across visits and supporting ongoing disease management.^48–57 Nevertheless, they should be interpreted as potentially relevant, continuity-oriented candidate attributes rather than as proven determinants of Internet hospital use.

Several attributes identified in earlier stages were not retained in the candidate set. Follow-up reminders were considered relevant to continuity of care, but their scope partly overlapped with self-monitoring and automatic-alert functions and could include heterogeneous activities such as medication, examination, and appointment reminders. Drug delivery was also valued by patients, but it incorporated several distinct dimensions—including availability, delivery time, additional cost, medication eligibility, and cold-chain requirements—that could not be represented adequately by one generic attribute without mixing service scenarios. Hospital grade overlapped with physician professional title and patient ratings as a signal of perceived quality. Physician response time, although frequently used in previous studies, received a lower expert score and was considered partly dependent on consultation mode and broader service processes. Psychological counseling was mentioned by fewer participants and may be more relevant to particular diseases or patient groups than to a general chronic disease DCE. These exclusions were made to maintain a manageable and minimally overlapping seven-attribute design; they should not be interpreted as evidence that the excluded features are unimportant.

Implications of the Candidate Attributes and Further Validation

The candidate attributes differ in how they may inform service design and policy. Ease of use, consultation mode, access to electronic medical records and test results, and the upload of self-monitoring data with alerts are service features that can be modified through platform and workflow design. Internet hospitals could modify these features through interface design, information-system integration, clinical workflow development, and service configuration. Out-of-pocket cost is both patient-facing and policy-sensitive because it may vary with pricing, reimbursement, and payment arrangements. By contrast, physician professional title and patient ratings mainly serve as quality signals under information asymmetry. They may help patients assess unfamiliar providers, but they are not equivalent to functions that platforms can directly modify. Their inclusion may therefore be more relevant to the presentation and transparency of provider information than to service-function development itself.

These implications remain provisional until the attributes and levels are evaluated in actual choice tasks. Cognitive interviews should first examine whether participants interpret the attributes and levels as intended. Particular attention is needed for the broad out-of-pocket cost range and the binary specifications of ease of use, record access, self-monitoring functions, and patient ratings. Pilot testing should assess whether the levels are realistic, sufficiently distinguishable, and free from unintended dominance, and whether respondents interpret all cost levels within the same Internet hospital service context. A subsequent full DCE can then estimate preference weights, trade-offs, and willingness to pay. Future analyses should also explore heterogeneity by age, digital literacy, education, medical insurance, disease type, region, and prior experience with Internet hospitals, as these characteristics may influence how patients evaluate digital services.

Strengths and Limitations

This study has several strengths. First, the staged mixed-methods design combined existing DCE evidence with patient perspectives and multidisciplinary expert input, rather than relying on a single source to develop attributes. Second, the attribute-tracking process documented how attributes were generated, merged, excluded, and refined across each stage, improving the transparency of the transition from 34 initial attributes to the final seven. Third, the study focused on patients with chronic diseases and considered both conventional healthcare-service characteristics and digital functions relevant to Internet hospital use. The resulting framework provides a structured basis for future DCE research and may also inform the design and evaluation of Internet hospital services.

Several limitations should be acknowledged. First, this study developed candidate attributes and levels but did not conduct a full DCE; therefore, patient preference weights, trade-offs, willingness to pay, and service demand were not estimated. The performance of the proposed framework in actual choice tasks requires further cognitive testing and pilot validation. Second, participants were recruited from a tertiary hospital and a community health service center in Shanghai. The single-city sample may limit geographic representativeness, particularly for rural areas and regions with different levels of digital health development. Therefore, the findings should be generalized with caution. Third, the formal literature review focused on previous DCE studies. Although patient interviews and expert consultation broadened the candidate pool, factors reported only in qualitative, implementation, or technology-adoption studies may not have been fully captured.

Conclusion

This study developed a preliminary attribute-level framework for a future DCE on Internet hospital service preferences among Chinese patients with chronic diseases. Through a literature review, patient interviews, and multidisciplinary expert consultation, seven candidate attributes and 18 corresponding levels were identified and refined. The framework provides a structured basis for subsequent preference research but does not quantify the relative importance of these attributes or determine how they influence patients’ choices. Further cognitive testing and pilot studies are needed to assess the clarity, realism, and feasibility of the proposed attributes and levels. A full DCE should then be conducted to evaluate and refine the framework and estimate patients’ preferences, trade-offs, and willingness to pay.

Funding Statement

This study was supported by the Policy Research Project of the Shanghai Municipal Center for Disease Control and Prevention (2025JZ23).

Data Sharing Statement

The data from this study cannot be publicly shared because it contains personal privacy information. De-identified datasets that were used and analyzed in this study can be obtained from Da He upon request, provided that the request complies with the project’s ethics approval requirements.

Ethics Approval

This study was approved by the Ethics Committee of the Shanghai Health Development Research Center (Shanghai Medical Information Center) (Approval No. 2025003).

Informed Consent

All participants provided written informed consent prior to the interviews. The consent process included permission to publish anonymized responses and direct quotations.

Author Contributions

Co-first authors: Yichun Gu, Dan Qin. All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. All authors have read and approved the final version of the manuscript and consent to its publication.

Disclosure

All authors declare that they have no conflicts of interest.