Consumer health information needs in China – a case study of depression based on a Social Q&A community

With the rapid development of China’s economy and the implement of “Healthy China” strategy [1], people have paid more and more attention to their own health. In health care, the rapid increase in health information on the Internet has led more patients to turn to the Internet as their first source of health information and to gain knowledge about their health before seeking a professional diagnosis [2‐4]. In the 2013 Pew Internet survey, searching for health information was the third most popular online activity; 72% of respondents reported they had used the internet to search for health information in the past year [5]. In China, health and medical care ranks as the most popular scientific topic among Chinese netizens, with a query rate of 63.16% [6].

Different from Baidu, Google and other traditional search engines, the question-and-answer web service provides users with a more convenient and specific platform for expressing complex needs. Users can not only look for basic information, but also seek social and emotional support from their unique social attributes [7]. With its interactive nature, the online community has quickly made the social question and answer service a platform for users to search for relevant information [8]. Therefore, research on network health community information, discover user information needs, and study the characteristics of user information needs, can better promote the operation and information services of the network healthy community.

Depression is a state of low mood and aversion to activities. It can affect a person’s thoughts, behavior, motivation, feelings, and sense of well-being. Its main features include sadness, difficulty in thinking and concentration and a significant increase/decrease in appetite and time spent sleeping. People experiencing depression may have feelings of dejection, hopelessness and, sometimes, suicidal thoughts. It can either be short term or long term [9]. Depression is a common mental disorder. Globally, more than 300 million people of all ages suffer from depression [10]. It is different from other types of non-mental diseases. Which can be divided into subgroups based on different levels of social barriers. Patients are more inclined to seek health information through the online platforms. The social Q&A platform is one of the main channels for people with depression to exchange information and seek help because of their social attributes.

In order to obtain the healthy information different type needs considering the healthy information description types, healthy information needs type and the correlation with the attention, this article focuses on the social Q&A community commonly used by health consumers to obtain healthy relevant information. Taking the largest social Q&A community in China as an example, we selected 10,862 questions and related records of depression on the platform of Zhihu. Using latent Dirichlet allocation (LDA) to obtain the type of health information needs, refines its demand types, realizes the measurement of user information need characteristics which can helps them better organize information resources, and enhances the efficiency of users’ access to health information.

According to the definition of the National Network of Libraries of Medicine, consumer health information is health and medical information related to the general public, patients and their families [11]. Network health information plays a huge role in consumer health services. Health information research based on the social media focuses on four aspects: health information behavior research, health information needs research, health information assessment and health information user research [12]. In terms of network health needs, the main focus is on the research on the content and characteristics of health information needs of different types of users. Valero-Aguilera B et al. found that women with breast cancer are interested in information on cure rates, survival rates, and the effects of disease on shape. While the urological cancer patients pay more attention on the information linked to their sex life, keeping healthy, and exercise [13]. Tsuya et al. analyzed the profiles and posts of users who were concerned about nine cancer-related accounts on Twitter to understand the similarities and differences between social media usage and health information needs of different cancer patients [14]. Hyun Jung Oh et al. surveyed and analyzed 291 respondents who used Facebook, and found that emotional support for people looking for health information on Facebook is more frequent than other support [15]. Ramo et al. used a combination of questionnaires and interviews to explore how young people are interested in quitting smoking through Facebook and how to motivate users to participate more [16].

Research on the information needs of the network health community helps to understand consumer needs and improve the service level of the network healthy community. At present, the research on health information needs in the question and answer community is mainly based on Yahoo! Answers and the method mainly about content analysis. There are few studies on the health needs of the social Q&A community in China. Sun et al. examined different content, styles and emotion expression by using manual statistical methods and got the needs topics by taking eating disorder as an example [17]. Sanghee Oh et al. used a content analysis method to analyze the cancers and sexually transmitted diseases in Yahoo! Answers, and proposed a coding model for disease-related problems in social Q&A websites [18]. They also compared the college students in the USA and Korea using convenience sample, in order to understand the proportion and characteristics of college students looking for health information through social media [19]. Kim et al. also analyzed 1000 posts through content analysis methods to clarify the topic of question and the question time of married Korean women living in the USA who sought help for health and medical issues [20]. Yong Jeong Yi collected the Q&A of sexual health information on a social Q &A site to identify their needs and the factors with the best answers [21]. There is less research on other platforms and methods. Donghua Chen et al. proposed a knowledge representation framework based on authoritative knowledge sources in online health communities to meet information needs of online patients [22].

In summary, the current research on the information needs of the social question and answer community mainly focuses on the theme and the influence of the needs. Few studies focus on the systematic classification of the types of needs. At the same time, the research methods are mostly based on statistical analysis methods (such as manual annotation and content analysis). LDA (latent Dirichlet allocation) is widely used in text mining, including text topic recognition, categorization and similarity calculation in recent years [23]. There are very few studies on the application of LDA to find the health information topic on the social Q&A community. Therefore, this paper uses the combination of LDA and manual labeling to systematically encode and classify consumers’ health information needs with depression as an example.

According to the 41st “Statistical Report on the Development of China’s Internet Network” released by the China Internet Network Information Center (CNNIC), until December 2017, the number of users in Zhihu had increase most in social applications, from 7.6 to 14.6%. Ranked fourth in social application [24]. As the largest social Q&A platform in China, the latest operational data has been released by Zhihu recently: by the end of August 2018, the number of registered users has exceeded 200 million. In 2018, there were more than 80 million new users [25]. The number of followers on the topic of “health” and relative questions has exceeded 8.12 million and 690,000 respectively. In Zhihu, users are able to ask questions, discuss and interact with others on the site. There’re a variety of operations in Zhihu for an answer. In addition to responding, you can also agree, share, thank, collect, and consider it as no help. A question in Zhihu can be “followed” by users so that they would receive notifications about this question, e.g. when the question is answered.

The data coding rules of the traditional network health community are mostly determined by manual methods. The researchers combine the existing thesaurus to adjust the coding rules according to the actual situation of the collected data, and finally form the topic division strategy, and then perform manual coding. This method is easy, but there are problems such as subjective standard differences between different coding personnel, time consuming operations and unobjective classification. This paper combined LDA and manual methods to determine the subject coding rules.

LDA (latent Dirichlet allocation) is an example of a topic model in natural language processing that was rediscovered independently by David Blei, Andrew Ng in 2003 [26]. LDA is a three-layer Bayesian probability model that contains documents, words and topics. The specific model is shown as Fig. 1 and Table 1.

Each Q&A record is used as a document, and the LDA model is used to train. Finally, the keywords corresponding to the topics and themes are obtained. The training process is as follows:

Derived from the data we got from the Zhihu, the mean, median, mode and standard deviation of the essential features of the data were computed; see Table 4. Which contains the number of topics, answers, followers, the length of question, description and the question time.

At least one topic tag should be added to each question. The average number of tags for the topic of depression is 2.63 (SD = 1.48). The problems collected were located in topic of depressive so the tag “Depressive disorder” (N = 2743) and “Depression” (N = 742) are most. Depression is a mental disorder, so psychosocial topics are also popular, “mental illness” (N = 541) “psychology” (N = 523) and “psychology” (N = 427). After removing these five topics, the top 50 hot topics are shown in Fig. 2.

After coding the data and subject analysis, it was found that the health information needs were mainly focused on describing or asking about the symptoms of depression (24.93%) and social life (24.86%), see Fig. 3. Symptoms question are focused on asking if certain symptoms are suffering from depression such as low sleeping quality or suicidal tendency. Social life mainly includes boyfriends or girlfriends (Boyfriend is in severe depression, how can I help him?), relatives (Mother has major depression, how can I deal with her relationship?) and classmates/friends (friends have depression, often loses temper, what should I do?). Depression is a type of mental illness that requires more social attention than non-psychiatric diseases and so they pay more attention on the social life. The problems involving complications (0.69%) and prevention (0.47%) were all below 1%. The most frequently mentioned complications are anxiety, two-way affective disorder and obsessive-compulsive disorder.

The number of answers, followers, and views represent how much attention is paid to a question. In order to study whether there is correlation between nine different types of needs description and the user’s attention, we did a correlation analysis between different types of needs description and degree of attention in order to study their relations. The degree of attention is expressed by the sum of the number of answers, followers and views.

First, a descriptive statistical analysis of the attention of different needs descriptions is shown in Table 5. The question (topic) with highest level of attention was still symptomatic (average = 3493.60), and the standard deviation of the symptom class was also the largest (SD = 48,489.67), indicating that the participation of health information consumers participating in the symptom category was quite different. Prevention is the only type of requirement description with less than 100 attention (Mean = 32.46). Among the nine types of requirement descriptions, only etiology and pathology and Symptom showed a very weak positive correlation, while others showed very weak negative correlation. There is no significant correlation between the type of requirement description and the level of attention.

Defining the type of health information needs helps to better understand the behavior of health information, and the type of demand for depression information is shown in the Fig. 4. Among them, the cognitive needs are the most dominant, accounting for 60.32%. The physical level has the least demand, accounting for only 2.24%. Most of them are asking about drugs and dietary information, such as: “What are the drugs that are effective for depression”, “What food is good when you take medicine in depression?”

Similarly, in order to study the correlation between the three types of health information needs and the degree of attention, we did a correlation analysis, as shown in the Fig. 5. Similarly, cognitive needs are the most concerned, with an average number of concerns of 4033.834 (SD = 33,138.66), as shown in the Table 6. The correlation between the three needs and the concerns of the questions is very low, and the cognitive need is the only one of the three needs that are positively related to the degree of attention (corr = 3.07E-04).

In order to pay more attention to the emotional needs of patients with depression, we have subdivided 482 emotional needs into positive, neutral and negative I, negative II, and negative III. Among 482 questions, only 6 are positive emotions. For example: Are there any patients who have overcome depression together and move forward together? Negative emotions account for the largest proportion (68.54%), in which mild negative emotions accounts for the most. (for example: no energy now, and no interest in most of things, what should I do?). This is related to the persistent, hopeless, pessimistic and unsatisfactory characteristics of depressed patients. In order to explore whether higher concern is correlated with more negative emotions, the higher the concern in the community? We labeled positive, neutral, negative I, negative II, and negative III as − 1, 0, 1, 2, 3 and then did correlation analysis with attention, see Fig. 5. Neutral average attention is the highest (Mean = 475.90), negative III has the lowest attention (Mean = 112.95), which proves that intense emotions cannot lead to wider attention on the socialized question and answer platform.

There were 49.16% health information consumers chose to ask question at night. This is in line with the fact that people are more leisurely and depressed in the evening. In order to study the relationship between different post time and emotion, we labeled positive, neutral, negative I, negative II, and negative III as − 1, 0, 1, 2, 3 and then did correlation analysis with the time of ask question. There was weak correlation (corr = − 0.02882877) indicate there are more health information consumer tendency to ask negative question at night.

Our study identified the cognitive needs are the most dominant, accounting for 60.32%. There are two reasons for this result. First, the social Q&A community serve as a platform for knowledge sharing for depression persons with social barriers. Second, the cognitive needs not limited by identify and professional knowledge compare with the specialized physical needs and the emotion needs which limited by the identify such as lover relative. At the same time, there were low correlation and high SD between the needs type and the concerns of the question. These results reflect the health information consumers have different needs for online health information and People have low trust in online health information, which makes it difficult to build close contact between information consumers and the online community.

Our study found that the health information needs were mainly focused on describing or asking about the symptoms of depression (24.93%) and social life (24.86%). Depression is a type of mental illness that requires more social attention than non-psychiatric diseases and so they pay more attention on the social life. In addition, the symptomatic need also has the most attention and SD indicating that the participation of health information consumers participating in the symptom category was quite different.