Service quality optimization of online mental health communities: A KANO-IPA model approach

Abstract

Objectives

As an important extension of mental health services, online mental health communities (OMHC) have become vital to access psychological support, share experiences and seek professional help. To enhance OMHC user satisfaction and promote the sustainable development of the industry, this study investigated the key elements of service quality within OMHCs and develops optimization strategies to inform practice.

Methods

This study collected user comments from the “One Psychology” online community. Based on this data, Latent Dirichlet Allocation (LDA) topic modeling was conducted to establish a framework of service quality elements. Furthermore, the Kano- Importance Performance Analysis (Kano-IPA) model is employed for the classification and priority evaluation of these elements.

Results

This research proposes a framework for service quality in OMHCs by identifying its core elements. The findings reveal that users place high expectations on attractive quality elements such as content diversity and personalization, while emphasizing one-dimensional quality elements including privacy security and interactivity. Additionally, professionalism, effectiveness and system stability are identified as must-be quality elements. The priority sequence for service quality optimization is effectiveness, responsiveness, peer support, interactivity, stability, professionalism and simplicity.

Conclusion

To optimize the service quality of OMHCs, it is essential to respond to users’ demands in a hierarchical manner: prioritize ensuring the stability and reliability of must-be elements, strive to enhance the performance of one-dimensional quality elements, and explore potential attractive elements in a timely manner to create pleasant surprises for users. Meanwhile, efforts should be made to strengthen technology empowerment, resource integration, responsive support and emotional mutual assistance. This study enriches the research related to the service quality of OMHCs and provides theoretical foundations and practical guidelines for service quality optimization.

Keywords

online mental health communities (OMHC)service quality user comments LDA topic model Kano-IPA model

Introduction

In recent years, Online Health Communities (OHCs) have emerged as a vital platform for health information exchange and support.¹ These virtual communities enable users to share knowledge, consult with experts, and interact with peers regarding health-related issues or treatment options.² With increasing public awareness of mental health, OMHCs have become a crucial avenue for accessing psychological support, sharing experiences, and seeking professional assistance.³ As a specialized type of OHCs, OMHCs focus primarily on mental health and emotional well-being, encompassing online counselling platforms, virtual therapy centres, and mental health apps that utilize remote technologies such as telephones, video conferencing, and text-based communication to deliver services.⁴

Compared to traditional face-to-face interventions, OMHCs offer unique advantages including accessibility, convenience, reduced stigma, and personalized user experiences.⁵ Research indicates that online counselling interventions are nearly as effective as in-person counselling interventions,⁶ while also effectively addressing the persistent limitations of time, space, and stigma inherent in traditional models. By providing an accessible, anonymous, and user controlled environment, this transformation expands support for traditionally underserved populations, thereby strengthening the overall coverage and system resilience of public mental health services.⁷ However, as the online mental health service market continues to expand, OMHCs are gradually revealing shortcomings in areas such as service content, user experience, and technical support, making it difficult to meet users’ evolving needs.⁸ Furthermore, the neglect of critical individual factors such as privacy concerns and empathy can negatively impact service quality, ultimately affecting user retention and platform sustainability.⁹ This raises pressing questions: What factors influence the service quality of OMHCs, and how can these platforms enhance their offerings? Addressing these issues is essential for advancing OMHC development.

Existing studies on OMHCs service quality have largely focused on analyzing discrete dimensions such as interaction, content, usability, cost, privacy and technical features,^10–14 rather than incorporating them into a unified hierarchical framework. This fragmented approach hinders the identification of which basic services users perceive as essential, which key improvements could substantially enhance satisfaction, and which value-added features might deliver exceptional user experience. Furthermore, it limits the ability to systematically prioritize service optimization strategies based on the relative importance and actual performance of each dimension. Consequently, investigating the determinants of service quality in OMHC and implementing targeted optimizations are essential for enhancing user satisfaction and ensuring the sustainable growth of this emerging field.

With the popularization of big data and text mining technology, extracting service quality elements and optimizing services by analysing users’ online comments has become a research hotspot.¹⁵ Comment-based analysis can enhance research efficiency while providing actionable recommendations for platform improvement.¹⁶ This study investigates “One Psychology”, a well-known OMHC in China. By employing the Kano-IPA model, we establish a framework of service quality elements and propose targeted recommendations. Our findings aim to enhance OMHC service quality, offering theoretical insights and practical guidance for platform development.

Literature review

In recent years, with the increasing attention of the society to mental health problems and the wide application of OMHCs, a growing number of scholars have begun to conduct research in this field and have made considerable progress. Existing studies mainly focus on OMHC user characteristics, information behaviorand participation model, community service management, the key role played by communities in providing peer-based social support, especially in depression and anxiety disorders.^17–22 The research methods mainly include questionnaire, text analysis and machine learning technology.

In terms of community service management, the current research mainly discusses the influence of isolated factors such as personalized content, peer interaction, system stability on service satisfaction.^{11,12,23–26} For example, Rickwood et al. (2019) emphasized the importance of consultation duration and user interaction on the eheadspace platform.²³ Hogsdal et al. (2023) pointed out that factors such as the quality of mental health information, text complexity and system stability are the key to adolescent satisfaction using the system availability scale.²⁴ Melcher et al. (2023) showed that cost and privacy transparency are basic and non-negotiable factors in application selection.²⁵ Users’ expectations also include the need for credibility, customizability and instant response support to meet urgent needs. Opie et al. (2024) found that peer interaction, application-based service provision, asynchronous support and personalized content help to improve service satisfaction.²⁶

Research within the Chinese context reflects similar priorities while highlighting culturally and structurally distinct concerns. For instance, studies note the critical role of anonymity in facilitating self-disclosure within stigma-laden environments,²⁷ and emphasize hybrid moderation models where peers and facilitators jointly guide support exchanges.²⁸ Culturally attuned interaction styles, such as the use of metaphor is also noted as salient in text-based mental health support.²¹ Nevertheless, like the international literature, Chinese scholarship remains fragmented, examining attributes such as trust, interface accessibility, or peer interaction^11,29,30 in isolation rather than as interrelated components of an integrated service system.

Beyond identifying isolated factors, scholars have proposed service quality improvements at the macro level or technical level. For example, Research on mobile applications for immigrant populations shows that OMHC applications are positively associated with immigrants’ level of social integration, overall health, and mental health, with AI-assisted applications showing even better outcomes.³¹ Similarly, in China, scholars such as Jin et al. (2022) advocated systematic improvements in regulation, employee training and remote application technology,⁴ and Zhou et al. (2025) developed a recommendation framework using knowledge mapping and large-scale language model to optimize the accuracy and professionalism of information transmission.³²

In summary, existing research on OMHC service quality—both globally and in China—has predominantly examined discrete dimensions such as interaction, content, usability, cost, privacy, and technical features^10–14 without integrating them into a coherent, hierarchical framework. This fragmentation makes it difficult for us to identify which basic services users consider essential, which are the key improvement points that can significantly improve satisfaction, and which are value-added functions that may bring surprises. Moreover, it is impossible to systematically determine the strategic priority of service optimization according to the importance and actual performance of each element. In addition, the current research relies more on subjective survey data and ignores the rich and active feedback contained in user generated content (such as online comments). This often limits the depth and objectivity of opinions.

Therefore, we go beyond subjective investigation and use LDA topic modeling technology to objectively extract key service elements from large-scale online user reviews and establish a service quality framework. On this basis, we classify and optimize these elements based on Kano-IPA model. This approach enables us to: (1) systematically classify quality elements, (2) diagnose their current performance level, and (3) provide data-driven strategic suggestions for service improvement. By doing so, the focus of this study has shifted from listing factors to establishing an operable overall OHMCs quality management framework.

Research design

By collecting and analyzing user comments from the platform “One Psychology”, which is one of the most prominent OMHCs in China, we construct a service quality element framework based on LDA top model analysis. Then, the Kano-IPA model was used to classify the service quality elements, evaluate the importance and satisfaction of each quality element, locate the key points of service optimization, and put forward suggestions for the optimization of service quality of OMHC, thus painting roadmap as Figure 1.

Figure 1.

Analytical framework for optimizing service quality of OMHC. Note. $α$ denotes Kano satisfaction indicator, and $r denotes$ Kano importance indicator, which are used to classify elements, ρ denotes priority index of strategies.

LDA topic model

LDA topic model (Latent Dirichlet Allocation) is a model that can mine potential topic information from massive texts. Its concept was first proposed by Blei et al. in 2003.³³ Its basic idea is to define a document as a collection of multiple hidden topics and repeatedly simulate the generation process of the document, so as to identify the potential topic information in the text data.³⁴ LDA topic model can effectively extract implicit topics from large-scale unstructured text data and realize feature dimensionality reduction and information extraction.³⁵

Compared to other topic models, LDA model offers greater flexibility and higher accuracy in identifying latent topics, making it particularly effective for precisely categorizing themes within complex textual structures.³² As an unsupervised learning algorithm, LDA does not require annotated training datasets. Instead, it only requires the specification of the optimal number of topic clusters.³⁶

The data for this study consist of authentic, unsolicited user comments from OMHCs. The unsupervised nature of LDA allows for the inductive derivation of themes directly from this user-generated content, minimizing preconceived structural biases.³⁷ This provides a complementary perspective to traditional survey-based approaches, which are inherently shaped by a priori researcher-defined constructs and scales while valuable for testing specific hypotheses.³⁸ Consequently, LDA is particularly well-suited to uncover the nuanced, emergent, and user-driven factors that constitute service quality in this context.³⁹ This paper aims to employ LDA topic model to mine and analyze the comments in OMHCs with the goal of identifying key factors for improving the quality of OMHC services.

Kano model

Inspired by Herzberg’s two-factor theory, Noriaki Kano, a Japanese quality management expert, proposed the Kano model in 1984.⁴⁰ Through a specifically designed questionnaire survey method—using paired functional and dysfunctional questions—the model collects user perception data. According to the nonlinear relationship between perceived importance of product/service characteristics and customer satisfaction, Kano model then divides the quality elements into five categories based on the collected questionnaire responses: (1) attractive quality element (A): the element that will significantly increase customer satisfaction with the increase of service level but will not significantly decrease due to insufficient service; (2) one-dimensional quality element (O): the element that high service level will lead to customer satisfaction and low service level will lead to customer dissatisfaction; (3) must-be quality element(M): the element that user satisfaction will not significantly improve with the improvement of service level, but will decline significantly due to the decline of service level; (4) indifferent quality element (I): the element that customer satisfaction will not change regardless of changes in service levels; (5) Reverse quality element(R): the elements that user satisfaction will gradually decrease as service levels improve. However, Kano model is a qualitative analysis method. The classification criteria for the elements are relatively subjective, which limits its potential for decision-making support in relevant fields.⁴¹

To gain a more precise understanding of user needs and behavioral patterns, researchers have drawn inspiration from the concept of analytical CRM and proposed the Analytical Kano model.⁴² It begins with the traditional paired questionnaire and the five-category Kano evaluation table. By systematically calculating the Better coefficient and Worse coefficient for each requirement item, it ultimately forms an analytical decision-making system that comprehensively reflects requirement attributes, quantifies satisfaction impact, and supports priority ranking.⁴³ This shifts demand classification from qualitative judgment based on statistical frequency to using quantitative coefficients, thereby enabling more precise decision-making in prioritizing requirements across the two dimensions of “enhancing satisfaction” and “preventing dissatisfaction”. Its effectiveness has been widely confirmed in studies. By this method, Meng et al. constructed a synchronous multi-service planning model based on the classification and reorganization of customer service demand.⁴⁴ Lizarelli et al. proposed support improvement decisions in an entrepreneurial education service.⁴⁵ In this paper, we will use this quantitative Kano model to classify the service quality elements and then determine the priority of quality elements.

IPA model

In 1977, Martilla and James proposed the importance-performance analysis model (IPA model) in marketing research.⁴⁶ The model divides product/service attributes into four regional quadrants with satisfaction as the abscissa and importance as the ordinate. Quadrant I is the performance maintenance area, where the importance of attributes and satisfaction performance are both high, and enterprises should continue to maintain positive advantages. Quadrant II is the key improvement area, with higher attribute importance but lower satisfaction, which is the main disadvantage of the product and should be improved. Quadrant III is low value area, with lower importance and lower satisfaction performance, which belongs to the secondary disadvantage attribute and can be considered to develop when resources are sufficient. Quadrant IV represents the over performance area, with lower importance but higher satisfaction, which is considered as the secondary focus area requiring only the maintenance of the current status and no further action.

As a commonly used tool for identifying critical quality problems, IPA model can be effectively combined with Kano model to formulate targeted strategies for enhancing customer service quality. For example, Zhai et al. adopted the analytical Kano model combined with the improved IPA to realize the priority analysis of user demands. Finally, the optimal path for product improvement can be obtained.⁴³ Chen et al. applied the Kano model and Importance-Performance Analysis (IPA) to clarify the categories of quality elements and the priority order of improvement for the online training of intellectual property information services provided by subject librarians in university libraries.⁴⁷ This paper will employ the model to classify the service quality elements of OMHCs and develop targeted improvement strategies.

Construction of service quality element system based on topic identification

Data collection and processing

The data for this study were collected from publicly accessible user comment sections of the “One Psychology” platform (website and mobile applications—Tencent My App and the Apple App) using the Octoparse web crawler, in compliance with the platform’s published terms of service. The dataset consists solely of textual comments and was anonymized prior to analysis, with all personally identifiable information removed. No private user data were accessed, and no interaction with users occurred during the research. “One Psychology” community (https://www.xinli001.com/) is one of the largest and most advanced online psychological theme communities in China.⁴⁸ This platform positioned to provide professional psychological support, with almost 53 million registered users worldwide, adopting an operation mode integrating professional content popularization, psychological assessment, counseling services, and a community interaction ecosystem that connects users and qualified counsellors. This paper selects “One Psychology” as the research object, which has a high reputation in the field of OMHC in China and covers the main fields of OMHC services, thereby demonstrating strong representativeness.

The collection of comments spanned from the time “One Psychology” enters the application store to December 15, 2024. A total of 6,270 user reviews were collected. Then the data were cleaned through the following operations. Firstly, invalid comments primarily fell into two categories: the first category consists of extremely brief comments, such as “good”, “positive review”, “thank you”, etc., which often lack substance and cannot reflect the theme well. The second category of comments comprises lengthy comments that exhibit characteristics of review spamming. These comments contain a significant amount of invalid information. In order to ensure the validity of the subsequent analysis, this study removed comments with lengths exceeding 150 words and those that appeared repeatedly. Secondly, the redundant and noisy data were cleaned, including removing emoticons, web links, and useless characters from the text data. After the above steps, a total of 4,127 valid comments were obtained as the original corpus set for LDA topic modelling.

Topic modelling and analysis based on user comments

LDA model is an unsupervised topic recognition method, and its parameter settings will directly influence the effectiveness of comment text clustering. In LDA topic model, there are three parameters that need to be set, which are the number of topics K, the prior parameters of topic distribution α and the prior parameters of word distribution β.³³ Existing studies predominantly employ the index of perplexity to determine the optimal number of topics.⁴⁹ A lower perplexity value indicates stronger predictive performance of the model.⁵⁰ Here α was set to 0.05, β was set to 0.01 and the number of topics was set to 1-19. Based on the number of topics, the perplexity was calculated and the corresponding perplexity curve was plotted, as shown in Figure 2. By testing the perplexity and coherence of topic distributions under different parameter settings, LDA can accurately identify topic structures with practical business implications. These data-driven topics provide a direct and comprehensive source for the specific attribute items used in the subsequent Kano questionnaire, ensuring that the established attribute system is both grounded in authentic user feedback and systematically measurable. Thereby, it supports the logical coherence of the entire research process, from text mining to quantitative analysis.⁵¹

Figure 2.

Perplexity curve and clustering effect. Note. The number of topics depends on the minimum of perplexity, and the number of bubbles denotes the number of clusters.

It can be seen that when K is taken as 3, the perplexity of the model obtains a local minimum value. Then with the increase of the value of K, the perplexity gradually increases. The topic discrimination degree is more obvious and the clustering effect is no longer significantly improved. Considering the relationship between perplexity and the number of topics, as well as the performance of LDA model, the optimal number of topics K was selected as 3. Using the trained LDA topic model, we extract the topic from the collected comment text, and get three user comment topics and the top ten high-frequency keywords under each topic. Each topic is named according to the respective sets of keywords, and finally three dimensions are obtained which are service content, user-centricity and system performance, as shown in Table 1.

Table 1.

LDA topic distribution of user comments.

	Topic 1 (service content)	Topic 2 (user-centricity)	Topic 3 (system performance)
High Frequency Keywords	service	analyze	apply
	professional	customize	stable
	effectiveness	knowledge	login
	analyze	recommend	simple
	test	function	complaints
	settle	reply	safe
	treat	support	interface
	content	steer	system
	utility	choice	update
	practice	growth	privacy

Note. Each column corresponds to an identified topic, with words positioned higher indicating larger relative proportion.

As can be seen from Table 1, in topic 1, users put forward more detailed requirements for the characteristics of service content. For example, the original text “You are such a dedicated teacher, who has a sharp insight into problems and always hits the nail on the head. You have been able to provide relatively effective solutions under the existing conditions. It is so relaxing. Thank you so much!” reflects users’ demands for professional, effective and friendly service. The original text “The reading column of One Psychology offers abundant and high-quality content. Its Q&A area is vibrant, with a diverse array of questions relating to psychology, emotions, relationships, work and study, each drawing active and excellent answers.” reflects users’ demand for diverse content. Through comprehensive analysis, the core needs and expectations of users are extracted and topic 1 is named as “service content”.

Topic 2 indicates that users hold much expectations for the user-centric design of OMHC. For example, the original text “One Psychology is a heaven for psychological healing and spiritual purification. Whenever I am stressed and need to manage my emotions, I will turn to One Psychology. That’s because this platform provides tailored guidance and help alleviate mental distress, and they formulate one suitable mental health service plan after another for you.” reflects the platform providing personalized, on-demand guidance that aligns with individual emotional needs is very important. The original text “One Psychology feels like a soothing comfort and a godsend: it lets you ask questions and get personalized service suggestions that fit your situation.” indicates that users expect to obtain personalized service experiences. After comprehensive analysis, these are closely related to the functions of the community. Therefore, the core needs and expectations of users are extracted and topic 2 is named as “User centricity”.

In topic 3, the original text “The interface of this app is so visually appealing! This free online counseling website is fully-featured—it works like a space for relieving stress. No need to worry about heavy work pressure anymore. Thumbs up!” reveals the necessity of friendly interface from the perspective of users. The original text “An amazing app! The long-awaited update brings multiple favorite features, enhanced stability and better user experience, making this great emotional counseling platform highly impressive.” indicates that users have high expectations for the platform’s stability and smoothness and continuous optimization, which are highly related to the improvement of system performance. Therefore, the core needs and expectations of users are extracted and topic 3 is named as “system performance”.

Construction of service quality element system

Based on LDA topic analysis above, we construct a system of service quality elements for OMHCs, including 3 topics and 12 service quality elements in total, as shown in Table 2. The three topics constitute the first-level indicators in the element system, in which “service content” reflects users’ expectations regarding the content characteristics of services provided by OMHC, “User-centricity” reflects that the core attraction of OMHC lies in users’ social interaction and demand guidance, and focusing on “system performance” is consistent with the development direction of the current technology-driven mental health service platform, which aims to enhance user experience and stickiness through technological advancements.

Table 2.

Elements of OMHC service quality based on LDA topic analysis.

First-level indicator	Second-level indicator	Indicator description
Service Content	Professionalism	$f_{1 - 1}$ : Using professional consulting techniques to provide services
	Effectiveness	$f_{1 - 2}$ : Effectively improve the psychological condition of users
	Interactivity	$f_{1 - 3}$ : Friendly interaction between the consultant and the user
	Diversity	$f_{1 - 4}$ : Diverse knowledge of mental health
User-centricity	Personalization	$f_{2 - 1}$ : Focus on user needs and provide personalized services
	Peer support	$f_{2 - 2}$ : Users interact and mutually support one another
	Autonomy	$f_{2 - 3}$ : Gently guides the user to utilize the platform without imposing restrictions
System performance	Privacy and security	$f_{3 - 1}$ : Attaching paramount importance to user information security and Safeguarding Against Privacy Breaches
	Simplicity	$f_{3 - 2}$ : The platform interface is simple and easy to operate
	Stability	$f_{3 - 3}$ : Maintain system stability and prevent program crashes
	Feedback	$f_{3 - 4}$ : Provide user feedback channels and respond in time
	Update and iteration	$f_{3 - 5}$ : Continuously update and optimize to meet the changing needs of users

The second-level indicators are derived from the mapping of high-frequency words associated with relevant topics and are further supplemented based on existing research.^27,52 Different from the conclusion of traditional research,^25,26 autonomy, update and iteration are the new factors found in the text mining of user comments. Autonomy emphasizes the gentle guidance of the community and the autonomous participation of users, which can enhance users’ sense of belonging and long-term willingness to participate. When users feel a sense of control over their own rehabilitation process, their psychological compliance and treatment cooperation will be significantly improved, so as to improve the actual effect of service quality. Update and iteration reflect the key value of community dynamic evolution ability for service sustainability. Mental health needs are dynamic. The continuous optimization of community content, rules and functions can adapt to the changing needs of users, reflecting the key role of “dynamic evolution” for service sustainability.

Analysis of service quality elements of OMHC based on Kano-IPA model

Questionnaire design and data collection

In order to formulate elaborate strategies for improving service quality, the questionnaire was designed according to the service quality element system constructed before. Firstly, the design of Kano model indicators is conducted.

Berger pointed out that one of the ways to quantify the Kano model is to assign a level of customer satisfaction/dissatisfaction regarding the performance of a product or service.⁵³ This is achieved by designing satisfaction scales for both positive and negative aspects of elements, as shown in Table 3. Since affirmative responses are more tendentious than negative responses, asymmetrical scales are designed to reduce the impact of negative evaluations.^46,54 The five-point scale and its corresponding measurement criteria are defined as follows: for the provision of a given quality attribute, the respective ratings for the perceptual responses of “Dislike, Tolerable, Indifferent, Expected, and Like” are assigned numerical scores of “-0.5, -0.25, 0, 0.5, and 1”. For the non-provision of the quality attribute, the corresponding numerical scores for the same perceptual responses are “1, 0.5, 0, -0.25, and -0.5”.

Table 3.

Satisfaction scale for positive and negative questions.

		Dislike	Tolerable	Indifferent	Expected	Like
Product/Service	Provide	-0.5	-0.25	0	0.5	1
Quality Elements	Not Provide	1	0.5	0	-0.25	-0.5

Note. This scale is used to calculate $x, y$ in formula (4.2).

Then the customers’ perceived importance levels of product or service quality elements were measured on a scale ranging from 0 to 1.0, where higher values indicate greater perceived importance. The survey results were categorized into five distinct intervals: “Unimportant, Moderately Important, Important, Very Important, and Extremely Important”. The correspondence between the values and these intervals is detailed in Table 4.^44,55

Table 4.

Importance scale.

Unimportant	Moderately important	Important	Very important	Extremely important
0-0.2	0.2-0.4	0.4-0.6	0.6-0.8	0.8-1.0

Note. This scale is used to calculate w in formula (4.2).

The product or service quality element is expressed as F = { $f_{i}$ | i= 1,2,…,I}, in which F is the set of attributes of the product or service, $f_{i}$ is the i-th quality element of the product or service, the total value of quality element is I, and the total number of customers surveyed is J. Then the customer’s comprehensive evaluation of the factor $f_{i}$ = ( $\forall_{i}$ = 1,2,…,I) can be calculated as follows (formula (4.1) to formula (4.5) are cited from Xu’s research⁴³).

e_{i j} = (x_{i j}, y_{i j}, w_{i j})

(4.1)

In formula (4.1), $x_{i j}$ represents the evaluation of the j-th customer regarding the positive aspects of product or service attribute $f_{i}$ , $y_{i j}$ represents the evaluation of the j-th customer regarding the negative aspects of product or service attribute $f_{i}$ , and $w_{i j}$ represents the evaluation of the j-th customer regarding the importance of product or service attribute $f_{i}$ .

For each product or service attribute $f_{i}$ , the average customer satisfaction level for negative questions is defined as $\bar{X_{i}}$ , and the average satisfaction level for positive questions is defined as $\bar{Y_{i}}$ , so we can get:

\bar{Y_{i}} = \frac{1}{J} \sum_{j = 1}^{J} w_{i j} y_{i j}, \bar{X_{i}} = \frac{1}{J} \sum_{j = 1}^{J} w_{i j} x_{i j}

(4.2)

The values of ( $\bar{X_{i}}, \bar{Y_{i}}$ ) can be plotted on a two-dimensional coordinate graph, where the abscissa represents the customer’s dissatisfaction level with attribute $f_{i}$ and the ordinate represents the satisfaction level. The majority of ( $\bar{X_{i}}, \bar{Y_{i}}$ ) values are expected to fall within the 0-1 range. Negative values indicate either inverse quality factors or problematic responses and should be excluded from the mean calculation. Thus, a product or service attribute is described as a vector: $f_{i}$ ∼ $\vec{r_{i}}$ ≡( $r_{i}$ , $α_{i}$ ),where the distance of $\vec{r_{i}}$ is:

| \vec{r_{i}} | = \sqrt{{\bar{X_{i}}}^{2} + {\bar{Y_{i}}}^{2}}

(4.3)

α_{i} = {t a n}^{- 1} (\bar{Y_{i}} / \bar{X_{i}})

(4.4)

The distance of the vector $| \vec{r_{i}} |$ is called Kano importance indicator where 0≤ $| \vec{r_{i}} |$ ≤2. The angle $α_{i}$ is called Kano satisfaction indicator where 0 ≤ $α_{i}$ ≤ π/2.

Due to the limited granularity of the traditional Kano model in classifying different attributes, certain elements may exhibit identical priority levels. Therefore, we will combine IPA and priority index $ρ_{i}$ to propose the order of improvement and maintenance strategies. The calculation of $ρ_{i}$ is as formula (4.5):

ρ_{i} = \frac{2 \sqrt{2}}{3} (1 - \frac{α_{i}}{π}) r_{i}

(4.5)

r_{i}

is the importance indicator which is located on the horizontal axis.

α_{i}

is the satisfaction indicator which is located on the vertical axis. When

α_{i}

is held constant,

ρ_{i}

is directly proportional to

r_{i}

, indicating that the service quality elements with higher importance should be prioritized for improvement. When

r_{i}

is held constant,

ρ_{i}

is inversely proportional to the value of

α_{i}

, indicating that the service quality elements with lower satisfaction should be prioritized for improvement. On this basis, the questionnaire was designed to calculate indicators

r_{i}

α_{i}

in the Analytical Kano and

ρ_{i}

in IPA, which mainly included three parts. The first part is basic information of users including gender, age, and the use of OMHC. In the second part, users were asked to rate their satisfaction of the community with or without various attributes. (e.g. How would you evaluate if OMHC provides counselor’s professional competence? How would you evaluate if OMHC doesn’t provide counselor’s professional competence? The formulation of items for all other variables follows the same approach.) In the third part, users were asked to rate the importance of different attributes by 12 items (e.g. Please rate the importance of stability provided by OMHC on a continuous scale from 0 to 1.0, where 0 indicates “unimportant” and 1.0 indicates “extremely important.”). Thus, the user sample data as well as their satisfaction levels and perceived importance regarding various community attributes could be collected.

During the period from November 22, 2024 to January 1, 2025, we distributed questionnaires through both online and offline channels. Online platforms included email, WeChat, QQ, Weibo and Xiaohongshu, while offline data collection primarily covered schools and corporate settings. Prior to the survey, respondents were informed of the research objectives, the voluntary nature of participation, and the principles of data confidentiality. No personally identifiable information (PII), such as names and contact details, was collected during the research process. All data were subjected to anonymization and de-identification processes, and data access permissions were strictly restricted to ensure the privacy and security of respondents. A total of 512 questionnaires were distributed to users and 342 questionnaires were recovered, of which 312 questionnaires were valid with an effective rate of 91.23%. The sample information is as Table 5. According to the collected data in Table 5, 66.2% were female. 72.6% were under the age of 40, and 94.5% of the respondents had prior experience using online mental health communities. OHMC’s users are characterized as young, female and inclined to accept online psychological consultation.^56,57 Therefore, this sample is representative to a certain extent and can be used to verify the validity of the indicator system.

Table 5.

Demographic characteristics of survey participants.

Characteristic	N (%)
Gender
Male	173 (33.8)
female	339 (66.2)
Use times
0	28 (5.5)
1-2	78 (15.2)
3-5	95 (18.6)
6-8	114 (22.3)
9-10	107 (20.9)
11 or more	90 (17.6)
Age
under 18	98 (19.1)
18-25	106 (20.7)
26-30	92 (18.0)
31-40	76 (14.8)
41-50	61 (11.9)
51-60	47 (15.2)
61 or older	32 (6.3)

Note. n=512.

The reliability and validity of the questionnaire was tested through SPSS 25.0. The Cronbach’s α coefficient was 0.873 and the KMO value obtained by Bartlett spherical test was 0.754, which are all above 0.7, indicating that the questionnaire had good reliability and validity and met the requirements for further analysis.

Classification of OMHC service quality elements based on Kano model

Referring to the study of Xu,⁴² the importance indicator of Kano $r_{0}$ = 0.5 was used as the dividing line, which has been adopted by relevant studies.^43,47,54 If $r_{i}$ < 0.5, the element $f_{i}$ was considered tobe an indifference quality element. Taking the satisfaction indicator $α_{0}$ =π/6 and $α_{0} =$ π/3 as the boundary respectively. Assuming that the lower limit of element satisfaction $α_{L}$ = π/6 and the upper limit $α_{H}$ =π/3, when $r_{i}$ > $r_{0}$ and $α_{i}$ < $α_{L}$ , the element $f_{i}$ is considered to be a must-be quality element. When $r_{i}$ > $r_{0}$ and $α_{i}$ > $α_{H}$ , the element is considered to be an attractive quality element. When $r_{i}$ > $r_{0}$ and $α_{L}$ < $α_{i}$ < $α_{H}$ , the element is considered to be a one-dimensional quality element.

The questionnaire data were aggregated and calculated in accordance with formula (4.1) to formula (4.5) and the results are shown in Table 6. According to the classification criteria above, the distribution map of service quality elements of OMHC was obtained as shown in Figure 3.

Table 6.

Summary of relevant data of quality elements.

	$r_{i}$	$α_{i}$	$ρ_{i}$
$f_{1 - 1}$	0.508	0.522	0.399
$f_{1 - 2}$	0.585	0.520	0.460
$f_{1 - 3}$	0.536	0.682	0.396
$f_{1 - 4}$	0.518	1.228	0.298
$f_{2 - 1}$	0.520	1.111	0.317
$f_{2 - 2}$	0.567	0.722	0.412
$f_{2 - 3}$	0.540	0.773	0.384
$f_{3 - 1}$	0.571	1.033	0.361
$f_{3 - 2}$	0.471	0.747	0.338
$f_{3 - 3}$	0.506	0.478	0.405
$f_{3 - 4}$	0.578	0.700	0.424
$f_{3 - 5}$	0.504	1.055	0.316

Note. The values of $α_{i}$ , $r_{i} and$ $ρ_{i}$ are respectively calculated with formula (4.3), (4.4), (4.5), and π=3.14.

Figure 3.

Distribution of service quality elements in OMHC. Note. Classification rule: if $r_{i} <$ 0.5: indifferent quality element; if $r_{i}$ >0.5 and $α_{i}$ < $π / 6 :$ must-be quality element; if $r_{i}$ > $0.5$ and $α_{i}$ > $π / 3 :$ attractive quality element; if $r_{i}$ >0.5 and $π / 6$ < $α_{i}$ < $π / 3 :$ one-dimensional quality element.

As can be seen from Figure 3 and Table 6, content diversity ( $f_{1 - 4}$ ), personalization ( $f_{2 - 1}$ ), update and iteration ( $f_{3 - 5}$ ) are classified as attractive quality elements, which will significantly increase customer satisfaction with the increase of service level but will not significantly decrease due to insufficient service. Among the three elements, content diversity ( $f_{1 - 4}$ ) accommodates users’ exploratory needs in different psychological domains, providing a beyond-expectation experience. Personalization ( $f_{2 - 1}$ ) can customize the content according to user’s interests and behaviors and thus improve user experience. The findings for both variables were consistent with those of existing studies.^5,13 By contrast, update and iteration ( $f_{3 - 5}$ ) was identified as a novel variable in this study, which will continue to provide users with freshness and further enhance user engagement. This finding may stem from the distinctive nature of online mental health services, whose effectiveness relies heavily on the ongoing optimization of content and the dynamic evolution of features. Such continuous adaptation is essential to align with users’ evolving psychological needs, sustain long-term engagement, mitigate digital fatigue, and ultimately transform the usage process into a progressively developing experience. When attractive quality elements are adequately fulfilled, user’s satisfaction with OMHCs will be substantially improved. Therefore, decision-makers should build a diversified content system, meet the personalized needs of users and continuously optimize the functions and services of the platform.

Peer support ( $f_{2 - 2}$ ), autonomy ( $f_{2 - 3}$ ), interactivity ( $f_{1 - 3}$ ), privacy and security ( $f_{3 - 1}$ ) and feedback ( $f_{3 - 4}$ ) are classified as one-dimensional quality elements, which directly affect user’s experience and perceived value. For example, peer support ( $f_{2 - 2}$ ) plays a pivotal role in strengthening community belongingness. Autonomy ( $f_{2 - 3}$ ) empowers users with greater decision-making capacity and perceived control. Interactivity ( $f_{1 - 3}$ ) improves the user’s experience and effectiveness. Privacy and security ( $f_{3 - 1}$ ) is the basis of the user’s trust in the community, and feedback ( $f_{3 - 4}$ ) is the bridge between users and the community. Among these findings, autonomy ( $f_{2 - 3}$ ) is different from existing research focusing on community-led service,^25,58 which emphasizes the importance of user subjectivity in decision-making. This may stem from a fundamental shift in service paradigms. Traditional mental health services are typically guided by professionals, placing users in a relatively passive role. In contrast, the anonymity and user-led nature of online mental health communities transform this dynamic. Within such environments, users gain autonomous control over information acquisition, interaction pacing, and the degree of self-disclosure. This sense of agency itself constitutes a core dimension of perceived service value. These elements show a significant linear correlation with user satisfaction. Decision-makers should ensure the fulfilment of these elements under resource-constrained conditions.

Professionalism ( $f_{1 - 1}$ ), effectiveness ( $f_{1 - 2}$ ) and system stability ( $f_{3 - 3}$ ) are classified as must-be quality elements. They reflect the basic needs of users in the process of using OMHC that are consistent with prior studies.^23,26 They represent core user expectations concerning community functionality and reliability. Decision-makers should prioritize these aspects in resource allocation and implementation.

Simplicity ( $f_{3 - 2}$ ) is an indifferent quality element. Service practicality is the primary concern for users, while community interface design receives minimal attention, with changes in these aspects having no significant impact on user satisfaction. Therefore, decision-makers can prioritize resources for other more important elements.

Optimization strategy of OMHC service quality based on IPA model

In order to further clarify the improvement strategy of OMHC service quality, we constructed an IPA matrix to conduct in-depth analysis of the elements, in which the horizontal axis represents satisfaction $α_{i}$ and the vertical axis represents importance $r_{i}$ . The specific positions of each service quality element within the four quadrants are determined according to the mean of the two indicators and the results are shown in Figure 4.

Figure 4.

IPA matrix of service quality elements in OMHC. Note. IPA area types: Key improvement area; Performance maintenance area; Low value area; Over performance area.

It can be seen that the elements are distributed across all four quadrant regions. The level of user satisfaction in performance maintenance area (Quadrant I) and over performance area (Quadrant IV) is relatively high, so maintenance strategies could be implemented to maintain high user satisfaction. The level of user satisfaction in key improvement area (Quadrant II) and low value area (Quadrant III) is low, hence it is necessary to implement improvement strategies for the quality elements of these two areas. Based on the above analysis, combined with the classification of quality elements by the Kano model, the decision rules can be obtained as follows:

Prioritization of improvement strategies

Peer support ( $f_{2 - 2}$ ), effectiveness ( $f_{1 - 2}$ ), interactivity ( $f_{1 - 3}$ ) and feedback ( $f_{3 - 4}$ ) falls within the key improvement area, characterized by low satisfaction levels coupled with high importance. Among the four elements, effectiveness ( $f_{1 - 2}$ ) is a must-be quality element which has the greatest impact on user satisfaction. The poor performance of this element will lead to user attrition and negative evaluations, so it should be improved first. The other elements including peer support ( $f_{2 - 2}$ ), interactivity ( $f_{1 - 3}$ ) and feedback ( $f_{3 - 4}$ ) are all one-dimensional quality elements, and the order of improvement is determined according to the priority index $ρ_{i}$ . Therefore, the prioritization sequence for improvement strategies in descending order of precedence is as follows: $f_{1 - 2} > f_{3 - 4} > f_{2 - 2} > f_{1 - 3}$ .

Professionalism ( $f_{1 - 1}$ ), simplicity ( $f_{3 - 2}$ ) and stability ( $f_{3 - 3}$ ) falls within the low value area, characterized by low satisfaction levels coupled with low importance. Professionalism ( $f_{1 - 1}$ ) and stability ( $f_{3 - 3}$ ) are must-be quality elements, which should be improved first. And $ρ_{3 - 3} > ρ_{1 - 1},$ so stability ( $f_{3 - 3}$ ) should be improved first before professionalism ( $f_{1 - 1}$ ). Simplicity ( $f_{3 - 2}$ ) is the indifferent quality element which has no significant impact on satisfaction, thus positioned at the lowest priority for improvement. Therefore, the prioritization sequence for improvement strategies in descending order of precedence is as follows: $f_{3 - 3}$ > > $f_{1 - 1} > f_{3 - 2}$ .

Based on the above analysis, the prioritization of improvement strategies for individual elements is: effectiveness ( $f_{1 - 2}$ ), feedback ( $f_{3 - 4}$ ), peer support ( $f_{2 - 2}$ ), interactivity ( $f_{1 - 3}$ ), stability ( $f_{3 - 3}$ ), professionalism ( $f_{1 - 1}$ ) and simplicity ( $f_{3 - 2}$ ).

Prioritization of maintenance strategies

Autonomy ( $f_{2 - 3}$ ), privacy and security ( $f_{3 - 1}$ ) falls within the performance maintenance area, characterized by high satisfaction levels coupled with high importance. The good performance of elements in this area is conducive to maintaining user stickiness and high satisfaction of OMHCs. Specifically, the two elements are both one-dimensional quality elements with priority indexes $ρ_{2 - 3} > ρ_{3 - 1}$ . Therefore, the prioritization sequence for maintenance strategies is: autonomy ( $f_{2 - 3}$ ) takes precedence over privacy and security ( $f_{3 - 1}$ ).

Content diversity ( $f_{1 - 4}$ ), personalization ( $f_{2 - 1}$ ), update and iteration ( $f_{3 - 5}$ ) are located in over performance area, characterized by high satisfaction levels coupled with low importance. The importance of elements in this area is relatively low, but user’s satisfaction is high. So these elements can be addressed through a strategy of maintenance. The three elements are all attractive quality elements with priority indexes $ρ_{2 - 1} > ρ_{3 - 5} > ρ_{1 - 4}$ .

Based on the above analysis, the prioritization of maintenance strategies for individual elements is: autonomy ( $f_{2 - 3}$ ), privacy and security ( $f_{3 - 1}$ ), personalization ( $f_{2 - 1}$ ), update and iteration ( $f_{3 - 5}$ ), content diversity ( $f_{1 - 4}$ ).

According to the Importance-Performance Analysis, all attractive quality elements fall within over performance area. Most one-dimensional quality elements and must-be quality elements fall within the key improvement area, and the indifferent quality elements are in the low value area, which are in line with the psychological characteristics of OMHC users. Consequently, the maintenance strategy should be implemented for the attractive quality elements and certain one-dimensional quality elements, while the improvement strategy should be implemented for the must-be quality elements, other one-dimensional quality elements and indifferent quality elements.

Suggestions for optimizing service quality of OMHC

Based on the research conclusions above, we put forward suggestions for optimizing service quality of OMHC as follows.

Enhancing technological empowerment to improve community service efficacy

Given the digital characteristics of OMHC, it is imperative for community operators to utilize advanced technologies such as big data, cloud computing and artificial intelligence to analyse user behaviours, build accurate user portraits and provide users with personalized psychological resources recommendations, intervention programs and counselling services.⁵⁹ Furthermore, community operators could explore the development of AI-driven psychological counselling assistant and provide instant mood analysis and initial psychological support to help users alleviate emotional problems such as anxiety and depression, and at the same time provide auxiliary decision-making support for professional counsellors.⁶⁰

Community operators could adopt modern system architecture such as federated learning and differential privacy to improve system performance. For example, Kumari and Kaur integrated ensemble federated learning (EFL) and transfer learning (TL), combined with differential privacy (DP) to design a robust and privacy-preserving framework for detecting abusive language on social media, which enables collaborative training across platforms without exposing user data, while improving both generalization and detection accuracy.⁶¹ OMHC operators can improve the service quality of the core function modules of the system by promoting user centered improvement. For example, regularly publish mental health science articles, videos or podcasts to help users understand mental health knowledge and eliminate misunderstandings about psychological problems. Integrate online booking, interactive feedback, case query and other functions to make services more accessible and more timely.⁶² Operators can continuously optimize the system through iterative development based on user data and behaviour analysis, so as to bridge the gap between technical design and practical application.

Additionally, community operators can provide multi-terminal support. In addition to mobile and web, WeChat mini programs could also be developed to allow users to access community services anytime and anywhere, enhancing the convenience and accessibility of services.⁶³ For example, based on ReproSchema and cross platform framework, operators can use visualization tools to independently configure content and interaction processes, and rapidly deploy them.⁶⁴ While reducing the technical threshold, they can realize the agile iteration and scenario adaptation of service modules, so as to transform the usability design into the improvement of actual user experience.

Optimizing professional resource integration to improve the effect of user service

Community regulators should strictly review the qualifications of online psychological counselors to ensure that they have legal practice certificates and relevant qualifications.⁴ At the same time, experience assessment and skill test of counselors could be carried out to examine their core professional competencies such as communication skills, problem solving and crisis intervention, ensuring that the counsellors are able to provide users with high-quality psychological support services.⁶⁵

In order to provide comprehensive mental health services, community operators could establish a multidisciplinary professional team by integrating expertise from psychiatrists, clinical psychologists and social workers. This approach can be implemented by combining professional expertise with structured intervention programs. For instance, following the model developed by Przybylko et al., a structured online program centered on “lifestyle medicine and positive psychology” can be designed to promote behavioral activation through theoretical guidance, practical tasks, and interactive community features. The effectiveness of such programs in improving mental health, enhancing vitality, and alleviating symptoms of depression and anxiety has been validated through multidimensional assessments.⁶⁶ In practice, operators can develop standardized service modules based on similar frameworks, supported by expert supervision and digital tools, thereby establishing a closed-loop service process that spans from interdisciplinary collaboration to outcome tracking.

Moreover, professional scientific content on mental health and accessible knowledge should be provided by professional counsellors, enabling users to better understand mental health issues, identify problems in a timely manner and seek professional assistance. Typical psychological cases should be organized and provided to users with reference and learning resources. To assist users in self-assessment and adjustment, community developers should develop more practical mental health tools such as psychological assessment tools, emotion monitoring tools and integrate self-help intervention tools, therefore further enhancing the practicability and convenience of services.

Establishing user feedback mechanisms and optimizing community response support

The development of OMHCs is fundamentally supported by user engagement. It is essential for community operators to establish a user feedback and response mechanism to promptly identify user needs and service pain points, thereby fostering user trust and enhancing user stickiness. Simultaneously, strengthening data privacy protection measures is crucial. Community operators should adhere to relevant laws and regulations to foster a safer and more reliable environment for mental health communication.⁶⁷

In order to improve the service quality, the OMHC platform should support multiple feedback channels such as surveys, open comment sections and anonymous messaging.⁶⁸ At the same time, a dedicated feedback processing team or an intelligent system should be established to ensure that users’ questions and suggestions receive prompt responses, and the processing progress is updated openly and transparently. For example, the intelligent feedback processing framework developed by Huang et al. may be adopted, which leverages GPT-4o to enable systematic and empathetic user interactions through role definition, structured processes, and ethical safeguards.⁶⁹

Community operators could establish a “User Suggestion Adoption List” and put the excellent suggestions contributed by users into practice and provide rewards. To address urgent mental health-related feedback (e.g. self-harm, suicidal ideation), it is recommended to develop an intelligent crisis early warning system, which would utilize keyword recognition and real-time response mechanisms, integrating with professional rescue resources to provide timely and effective interventions for users.⁷⁰

In order to ensure service continuity, the community operators also needs to establish user profiles, conduct regular follow-ups and track the service effect to adjust service plans promptly, thereby providing users with sustained support.⁷¹

Fostering emotional communication and mutual support for a positive community atmosphere

OMHCs offer emotional support to users, with their core function centered on alleviating psychological stress and fostering resilience through shared experiences, peer support and empathetic engagement. Community operators can facilitate peer-to-peer communication among individuals facing similar challenges by implementing grouping mechanisms that categorize users based on shared interests, specific conditions, or therapeutic goals. For example, Akar summarized 12 common themes such as suicidal tendencies, depression and anxiety, and interpersonal relationships,⁷² which can be used as a realistic basis for grouping. In specific implementation, the operator can dynamically classify users based on the above topic tags and establish corresponding topic groups or support communities. At the same time, continue to optimize the grouping strategy in combination with user behaviour data, such as pushing mindfulness training content for high anxiety groups, and building peer communication boards for parenting pressure users. In order to improve the pertinence of support, and make the service continuously adjust with the change of user needs, so as to systematically implement the service goal of “personalized support”.

Community operators can also design incentives to encourage users to share mental health stories and coping experiences, and enhance the community atmosphere through interactive features such as likes, comments or badge rewards, thereby reinforcing community cohesion and participation.⁷³ Simultaneously, to ensure the quality and safety of communication, experienced users can be trained to serve as “community facilitators” or “peer mentors”, guiding discussions and promoting positive interactions within the group.⁷⁴ To avoid the spread of misinformation, professional counsellors can regularly review and give feedback on the content shared by users in OMHC and provide scientific supplementary suggestions. Thus, the community will not only provide emotional support but also serve as a platform for mutual empowerment, thereby further enhancing its functionality and social value.

Conclusions

The service quality of OMHCs plays a crucial role in enhancing user satisfaction and promoting the sustainable development of the industry. This study focuses on “One Psychology” community as a case, collecting user comments and conducting LDA topic analysis to establish a framework of service quality elements for OMHCs.

Different from the conclusion of traditional research, this study found that autonomy is an important factor influencing the service quality of online mental health community, users prefer to make counselling plans independently rather than accept fixed routine. In addition, this study discovered that update and iteration become notable merits, with most studies just focusing on effectiveness, personalization, react, privacy and so on.^10,12–14 This discovery provides a new perspective for optimizing the quality of community service, emphasizing the key value of community gentle guidance, user autonomous participation, and community dynamic evolution ability for service sustainability.

Using the Kano model, the service quality elements are systematically categorized, in which content diversity, personalization, update and iteration are attractive quality elements; peer support, autonomy, interactivity, privacy and security and feedback are one-dimensional quality elements; professionalism, effectiveness and system stability are must-be quality elements; and simplicity is an indifferent quality element. On this basis, utilizing the IPA matrix analysis, the prioritization of improvement strategies for individual elements is proposed and practical recommendations are provided from the perspectives of technological empowerment, resource integration, responsive support and emotional mutual assistance. This study enriches the research on service quality of OMHCs, providing a theoretical foundation and practical guidance for optimizing service quality. It contributes to enhancing the service level of OMHCs, thereby improving user experience and promoting the sustainable development of online mental health industry.

Footnotes

Acknowledgments

The authors like to acknowledge and thank all those who participated in the interviews and questionnaire surveys.

ORCID iDs

Hua Yang

Wanting Li

Ethical considerations

This study utilized publicly available social media data and voluntary user surveys, involving no patients, clinical data, or identifiable personal information. The research adhered to platform terms and relevant ethical guidelines for internet-based research. All data were anonymized; no personally identifiable information (PII) was retained.

Consent to participate

Survey participation was voluntary and based on informed consent. Analysis uses aggregated data only, ensuring participant anonymity and adherence to ethical research standards.

Author contributions

Conceptualization: Yang Hua and Li Wanting; Methodology: Yang Hua and Li Wanting; statistical analysis: Li Wanting; Data collection: Li Wanting; writing – Li Wanting; writing – review and editing, Yang Hua. All authors have read and agreed to the published version of the manuscript.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Research project the Humanity and Social Science Youth Foundation of Ministry of Education of China under Grant Number 23YJC630214.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to ethical restrictions.*

Author biographies

Hua Yang (1980-), female, associate professor, research interests: information management, quality management, supply chain management.

Li Wanting (2002-), female, master candidate, research interests: information systems and business intelligence.

References

Chen

Guo

, et al. Exploring the online doctor-patient interaction on patient satisfaction based on text mining and empirical analysis. Information Processing & Management 2020; 57(5): 102253. https://doi.org/10.1016/j.ipm.2020.102253

Lin

Kishore

. Social media-enabled healthcare: A conceptual model of social media affordances, online social support, and health behaviors and outcomes. Technological Forecasting and Social Change 2021; 166: 120574. https://doi.org/10.1016/j.techfore.2021.120574

Chen

. Exploring the effect of social support and empathy on user engagement in online mental health communities. International Journal of Environmental Research and Public Health 2021; 18(13): 6855. https://doi.org/10.3390/ijerph18136855

Jin

Zhang

Zheng

, et al. Telepsychology: applications, advantages, and challenges. Advances in Psychological Science 2022; 30(1): 141–156. https://doi.org/10.3724/sp.j.1042.2022.00141

Zainudin

Hassan

Ahmad

, et al. A comparison of a client’s satisfaction between online and face-to-face counselling in a school setting. Pertanika Journal of Social Sciences & Humanities 2021; 29(S1): 135–153. https://doi.org/10.47836/pjssh.29.s1.08

Ierardi

Bottini

Riva

. Effectiveness of an online versus face-to-face psychodynamic counselling intervention for university students before and during the COVID-19 period. BMC psychology 2022; 10(1): 35. https://doi.org/10.1186/s40359-022-00742-7

Leah

Riewpaiboon

Sirirungruang

, et al. Feasibility of online psychosocial interventions to promote mental health recovery and well-being. International Journal of Social Psychiatry 2024; 70(7): 1311–1324. https://doi.org/10.1177/00207640241264656

Yao

Tang

. The engagement behaviors and treatment barriers for depressed patients in an online health community: a pre-post-treatment comparison. Acta Psychologica 2026; 262: 106051. https://doi.org/10.1016/j.actpsy.2025.106051

Albikawi

Abuadas

Rayani

. Nursing Students’ Perceptions of AI-Driven Mental Health Support and Its Relationship with Anxiety, Depression, and Seeking Professional Psychological Help: Transitioning from Traditional Counseling to Digital Support. Healthcare 2025; 13(9): 1089. https://doi.org/10.3390/healthcare13091089

10.

Cheng

Kuo

Chang

, et al. Integrating the Kano model and IPA to measure quality of museum interpretation service: a comparison of visitors from Taiwan and Mainland China. Asia Pacific Journal of Tourism Research 2019; 24(6): 483–500. https://doi.org/10.1080/10941665.2019.1591474

11.

Zhao

Xie

Bao

. Research on university patent information service from the perspective of service encounter: elements and promotion strategies. Journal of Modern Information 2020; 40(11): 107–115.

12.

Martinez-Martin

Greely

Cho

. Ethical development of digital phenotyping tools for mental health applications: Delphi study. JMIR mHealth and uHealth 2021; 9(7): e27343. https://doi.org/10.2196/27343

13.

Saleem

Kühne

De Santis

, et al. Understanding engagement strategies in digital interventions for mental health promotion: scoping review. JMIR mental health 2021; 8(12): e30000. https://doi.org/10.2196/30000

14.

Fouyaxis

Bidargaddi

, et al. Critical design decisions and user demographics in enhancing real-time digital mental health interventions: A systematic review. Digital Health 2024; 10: 20552076241306782. https://doi.org/10.1177/20552076241306782

15.

Peng

Liu

. Research on user satisfaction of museum service based on fine-grained comment mining of large language model. Library and Information Service 2024; 68(17): 54–67.

16.

Feng

Cao

. A Study on Users’ Usage Evaluation of Online Health Communities Based on Topic Modeling and Sentiment Analysis. Researches in Library Science 2025; 8: 112–125.

17.

Shiner

Millard

, et al. Chronic health conditions and disability are prevalent among community users of a digital mental health service: a scoping survey. Disability and Rehabilitation: Assistive Technology 2024; 20(3): 562–571. https://doi.org/10.1080/17483107.2024.2389208

18.

Perich

Andriessen

. Predictors of digital technology-based mental health programs in young adults for mental health support. Health Promotion International 2023; 38(3): daad024. https://doi.org/10.1093/heapro/daad024

19.

Sequeira

. Determinants of e-mental health use during COVID-19: cross-sectional Canadian study. Journal of Medical Internet Research 2022; 24(11): e39662. https://doi.org/10.2196/39662

20.

Sun

Kang

. Health communication from a social support perspective: an empirical study of patient online communities. Journal of Shanxi University (Philosophy and Social Sciences) 2022; 45(1): 66–73.

21.

Jing

Jiang

. ‘No man is an island’: how Chinese netizens use deliberate metaphors to provide ‘depression sufferers’ with social support. Digital Health 2024; 10: 20552076241228521. https://doi.org/10.1177/20552076241228521

22.

Hawke

Sheikhan

MacCon

, et al. Going virtual: youth attitudes toward and experiences of virtual mental health and substance use services during the COVID-19 pandemic. BMC Health Services Research 2021; 21(1): 340. https://doi.org/10.1186/s12913-021-06321-7

23.

Rickwood

Wallace

Kennedy

, et al. Young people’s satisfaction with the online mental health service eheadspace: development and implementation of a service satisfaction measure. JMIR Mental Health 2019; 6(4): e12169. https://doi.org/10.2196/12169

24.

Høgsdal

Kaiser

Kyrrestad

. Adolescents’ assessment of two mental health–promoting mobile apps: results of two user surveys. JMIR Formative Research 2023; 7(1): e40773. https://doi.org/10.2196/40773

25.

Melcher

Camacho

Lagan

, et al. College student engagement with mental health apps: analysis of barriers to sustained use. Journal of American College Health 2022; 70(6): 1819–1825. https://doi.org/10.1080/07448481.2020.1825225

26.

Opie

Vuong

Welsh

, et al. Outcomes of best-practice guided digital mental health interventions for youth and young adults with emerging symptoms: part I. A systematic review of socioemotional outcomes and recommendations. Clinical Child and Family Psychology Review 2024; 27(2): 424–475. https://doi.org/10.1007/s10567-024-00469-4

27.

Jiang

. Digital Healing and Scene Adaptation: A Case Study Based on the Social Skills Rehabilitation Group on Douban. Journalism & Communication Review 2025; 78(6): 88–99.

28.

. Digital Mental Health Intervention: A New Paradigm for Reshaping Social Stability in the VUCA Era. Academic Exploration 2026; 2: 1–14.

29.

Wong

. The Impacts of Help-Seeking Stigmatization, Ethical Concerns about Online Counseling and Interpersonal Trust in Cyberspace on College Students’ Attitudes Toward Online Counseling. Psychological Research 2022; 15(5): 454–461.

30.

Zhu

. Challenges and Reconstruction of Psychological Resilience Among Generation Z Youth. Youth Exploration 2026; 1: 37–51.

31.

Drydakis

. Mobile applications aiming to facilitate immigrants’ societal integration and overall level of integration, health and mental health. Does artificial intelligence enhance outcomes? Computers in Human Behavior 2021; 117: 106661. https://doi.org/10.1016/j.chb.2020.106661

32.

Zhou

Wang

. Multi coupling resource portrait of public cultural data open platform in crisis situation. Information Science 2023; 41(8): 71–80+88.

33.

Blei

Jordan

. Latent Dirichlet allocation. Journal of Machine Learning Research 2003; 3(4-5): 993–1022.

34.

Liu

Zhao

. A Study on the Complexity Measurement of Crowdsourcing Tasks Based on Topic Modeling. Management Review 2023; 35(9): 142–154.

35.

Zhao

Sun

Wang

. The impact of analysts’ forward-looking information on stock investment returns: a quantitative study based on textual analysis. Systems Engineering – Theory & Practice 2024; 44(12): 3851–3861.

36.

Peng

Zhou

Shi

. Mining, analysis and application of citation features from Chinese books in library servicescape: a case study of G-class books. Journal of Modern Information 2023; 43(10): 107–119.

37.

Feldhege

Moessner

Bauer

. Who says what? Content and participation characteristics in an online depression community. Journal of Affective Disorders 2020; 263: 521–527. https://doi.org/10.1016/j.jad.2019.11.007

38.

Dai

Lyu

, et al. Medical service quality evaluation based on LDA and sentiment analysis: Examples of seven chronic diseases. Digital health 2024; 10: 20552076241233864. https://doi.org/10.1177/20552076241233864

39.

Zou

Zhao

Childs

, et al. User Experience Design for Online Sports Shoe Retail Platforms: An Empirical Analysis Based on Consumer Needs. Behavioral Sciences 2025; 15(3): 311. https://doi.org/10.3390/bs15030311

40.

Kano

Seraku

Takahashi

. Attractive quality and must-be quality. Journal of the Japanese Society for Quality Control 1984; 31(4): 147–156.

41.

Meng

Zou

Chen

, et al. An ‘individual agent’ evaluation method and its application in standard target system. Management Review 2009; 21(12): 86–93.

42.

Jiao

Yang

, et al. An analytical Kano model for customer need analysis. Design studies 2009; 30(1): 87–110. https://doi.org/10.1016/j.destud.2008.07.001

43.

Zhai

Wei

. Identification and analysis of product improvement needs considering online review user attention: based on KANO-IPA model. Science and Technology Management Research 2023; 43(9): 95–104.

44.

Meng

Han

Meng

, et al. Research on Synchronous Multi-Service Planning Method and Its Application Considering the Classification and Reorganization of Customer Service Demands. Journal of Technology Economics 2014; 33(8): 68–75.

45.

Lizarelli

Osiro

Ganga

GMD

, et al. Integration of SERVQUAL, Analytical Kano, and QFD using fuzzy approaches to support improvement decisions in an entrepreneurial education service. Applied Soft Computing 2021; 112: 107786. https://doi.org/10.1016/j.asoc.2021.107786

46.

Martilla

James

. Importance-performance analysis. Journal of Marketing 1977; 41(1): 77–79. https://doi.org/10.1177/002224297704100112

47.

Chen

Liu

. Research on online training quality of intellectual property information services for subject librarians in universities: based on data from FULink alliance librarians. Library and Information Service 2021; 65(21): 76–83.

48.

Cao

Zeng

Zhang

. An analysis of factors influencing the continuance intention of users of psychological apps: a case study of ‘One Psychology’. Statistics and Management 2020; 35(4): 83–88.

49.

Wang

. Activation and inscription of online forum memory: an analysis with the example of Tianya community. Journalism and Mass Communication 2025; 1: 74–86.

50.

Zhu

Sun

. Sentiment classification prediction and topic identification of reviews related to the ChatGPT large language model. Information Science 2025; 43: 1–35.

51.

Sun

. An Exploration and Implications of AI Competency Requirements for American Librarians Based on LDA Topic Modeling. Library and Information 2025; 6: 104–112.

52.

Kim

Park

. MOSS-6: a multi-label dataset and deep learning model for detecting diverse social support-seeking behaviours in online mental health communities. Information, Communication & Society 2025; 28(4): 688–722. https://doi.org/10.1080/1369118x.2025.2460558

53.

Berger

. Kano’s methods for understanding customer-defined quality. Center for Quality Management Journal 1993; 2(4): 3–36.

54.

Meng

Zhou

Jiang

. Service family planning for mass customization based on customer requirements classification and recombination. Journal of Industrial Engineering and Engineering Management 2015; 29(1): 82–88.

55.

Yang

. The refined Kano’s model and its application. Total Quality Management & Business Excellence 2005; 16(10): 1127–1137. https://doi.org/10.1080/14783360500235850

56.

Batterham

Calear

. Preferences for internet-based mental health interventions in an adult online sample: findings from an online community survey. JMIR Mental Health 2017; 4(2): e7722. https://doi.org/10.2196/mental.7722

57.

Naslund

Aschbrenner

Araya

, et al. Digital technology for treating and preventing mental disorders in low-income and middle-income countries: a narrative review of the literature. The Lancet Psychiatry 2017; 4(6): 486–500. https://doi.org/10.1016/S2215-0366(17)30096-2

58.

Alvarez-Jimenez

Rice

D'Alfonso

, et al. A novel multimodal digital service (moderated online social therapy+) for help-seeking young people experiencing mental ill-health: pilot evaluation within a national youth e-mental health service. Journal of medical Internet research 2020; 22(8): e17155. https://doi.org/10.2196/17155

59.

Nian

, et al. “Internet Plus” Psychological Platforms: A New Approach to Evidence-Based Practice in Mental Health in the Big Data Era. Chinese Journal of Clinical Psychology 2019; 27(1): 210–214+205.

60.

D'alfonso

Santesteban-Echarri

Rice

, et al. Artificial intelligence-assisted online social therapy for youth mental health. Frontiers in psychology 2017; 8: 796. https://doi.org/10.3389/fpsyg.2017.00796

61.

Kumari

Kaur

. Towards Enhanced Cyberbullying Detection: A Unified Framework with Transfer and Federated Learning. Systems 2025; 13(9): 818. https://doi.org/10.3390/systems13090818

62.

Peng

Zuo

. The Current Situation, Problems, and Prospects of Mental Health Service Apps in the Context of “Internet +”. Chinese Journal of Clinical Psychology 2017; 25(2): 333–336+309.

63.

Yom-Tov

Lekkas

Heinz

, et al. Digitally filling the access gap in mental health care: an investigation of the association between rurality and online engagement with validated self-report screens across the United States. Journal of psychiatric research 2023; 157: 112–118. https://doi.org/10.1016/j.jpsychires.2022.11.024

64.

Klein

Clucas

Krishnakumar

, et al. Remote digital psychiatry for mobile mental health assessment and therapy: MindLogger platform development study. Journal of medical Internet research 2021; 23(11): e22369. https://doi.org/10.2196/22369

65.

. The Fragmentation of Mental Health Theories and Practices and Corresponding Countermeasures. Journal of Soochow University Educational Science Edition 2025; 13(6): 37–46.

66.

Przybylko

Morton

Kent

, et al. The effectiveness of an online interdisciplinary intervention for mental health promotion: a randomized controlled trial. BMC psychology 2021; 9(1): 77. https://doi.org/10.1186/s40359-021-00577-8

67.

Oti

Pitt

. Online mental health interventions designed for students in higher education: A user-centered perspective. Internet interventions 2021; 26: 100468. https://doi.org/10.1016/j.invent.2021.100468

68.

Jones

Ashurst

. Online anonymous discussion between service users and health professionals to ascertain stakeholder concerns in using e-health services in mental health. Health Informatics Journal 2013; 19(4): 281–299. https://doi.org/10.1177/1460458212474908

69.

Huang

Ding

, et al. Construction and Effect Evaluation of a Self-help AI Psychological Counseling System Based on Large Language Models. Acta Psychologica Sinica 2025; 57(11): 2022–2042. https://doi.org/10.3724/sp.j.1041.2025.2022

70.

School Psychology Professional Committee . Professional Competence Standards for School Mental Health Education Teachers of the Chinese Psychological Society. Acta Psychologica Sinica 2026; 58(1): 185–190.

71.

Liu

Zhou

, et al. The Mechanism of Mental Health Problems Influencing Professional Psychological Help-Seeking Behaviors Among College Students: A Longitudinal Study. Chinese Journal of Special Education 2021; 7: 89–96.

72.

Akar

. Connecting for well-being: a role-based network analysis of online mental health communities. Behaviour & Information Technology 2025; 44: 1–15. https://doi.org/10.1080/0144929x.2025.2484393

73.

Liu

Wang

. Users’ intention to continue using online mental health communities: empowerment theory perspective. International journal of environmental research and public health 2021; 18(18): 9427. https://doi.org/10.3390/ijerph18189427

74.

Giles

Newbold

. Self-and other-diagnosis in user-led mental health online communities. Qualitative health research 2011; 21(3): 419–428. https://doi.org/10.1177/1049732310381388