Background: The psychiatric treatment gap is substantial in Korea, implying barriers in seeking help. Objectives: This study aims to explore barriers of seeing psychiatrists, expressed on the internet by age groups. Methods: A corpus of data was garnered extensively from internet communities, blogs and social network services from 1 January 2016 to 31 July 2019. Among the texts collected, texts containing words linked to psychiatry were selected. Then the corpus was dismantled into words by using natural language processing. Words linked to barriers to seeking help were identified and classified. Then the words from web communities that we were able to identify the age groups were additionally organized by age groups. Results: 97,730,360 articles were identified and 6,097,369 were included in the analysis. Words implying the barriers were selected and classified into four groups of structural discrimination, public prejudice, low accessibility, and adverse drug effects. Structural discrimination was the greatest barrier occupying 34%, followed by public prejudice (27.8%), adverse drug effects (18.6%), and cost/low accessibility (16.1%). In the analysis by age groups, structural discrimination caused teenagers (51%), job seekers (64%) and mothers with children (43%) the most concern. In contrast, the public prejudice (49%) was the greatest barriers in the senior group. Conclusions: Although structural discrimination may most contribute to barriers to visiting psychiatrists in Korea, variation by generations may exist. Along with the general attempt to tackle the discrimination, customized approach might be needed.
Keywords: Mental health; Service; Barrier; Discrimination; Stigma; South Korea
The global burden of mental illness is considerable. Recent research suggests that mental illness is on a par with cardiovascular disease in terms of the level of disability caused [[
Although effective pharmacological and psychological interventions have been developed, globally, the proportion of people receiving treatment for a mental illness is low [[
Although Korea is economically developed, the psychiatry treatment gap remains huge. An epidemiologic survey in Korea reported that only 22% of people with a mental illness seek professional help during their lifetime [[
Text mining (TM) is a novel technique that enables researchers to process an unprecedented amount of textual data by subdividing and extracting necessary information. Given the tremendous quantity of communications available on the internet, careful exploration of web-based information has the potential to provide valuable insight on specific issues. TM has proved to be a powerful tool in health research, particularly when combined with natural language processing (NLP) [[
Korea has an international reputation for its strong internet-access infrastructure [[
In this study, we used text available on the internet (
Figure 1 shows the outline of the method. We obtained social media data from VAIV (Seoul, South Korea), one of the leading companies for social media analyses in South Korea. Data were collected from 1 January 2016 to 31 July 2019. Text was collected from web communities, social network services (SNS), and personal blogs because internet users commonly express their opinions using such sites. In Korea, web communities called 'cafés' are served by internet platform service providers such as Naver (
Graph: Fig. 1 Flow chart showing the texts included in this study
Original text had to be analyzed automatically. Furthermore, many articles published online have errors in spelling and grammar. NLP allows computers to analyze language. The SOCIALmetrics™ engine, provided by VAIV, subdivided the original texts that were identified into sentences and morphemes after the external links and stop words were removed. Stop words include prefixes and suffixes. In a Korean sentence, prefixes and suffixes determine the meanings and grammatical functions of the words they are attached to.
For each word, the most appropriate combination of morphemes was tagged as parts of speech, such as noun or adjective. If expressions were difficult to analyze, they were paraphrased as simple words based on a normalized dictionary. Noun phrases were processed individually. For example, 'big hospital' was processed as a single item rather than 'big' and 'hospital.'
After NLP, additional TM techniques were applied. Synonyms were transformed into representative words using the dictionaries installed in the analytical software. To simplify the data and the analysis procedure, noun groups and predicates were used as keywords and phrases, and levels of linkage among the extracted keywords were calculated. Keywords and phrases were categorized as shown in Table 1. The analysis procedure used in this study was developed from a dictionary containing 8,940 words, categorized into 9 major categories and 26 subcategories. For example, the keyword 'schizophrenia' falls into the major category 'health' and the minor category 'illness.'
Table 1 Analyses of keywords
Major Category Subcategory Keywords Lifestyle Health, life, exercise... Crime Gangnam station crime... Season and weather Spring, Winter... Time Now, today, everyday... Place Hospital, house, outside... Normal Lecture, arrangement, birth... Special Trauma, sexual abuse... Age Teens, twenties... Object Alone, kids, friends... Relationship Divorce, break-up... Life cycle Adolescent, childhood... Personal Service fee, side effect... Social Record, discrimination... Physical Skin, face, hair... Disease Disease, early, cancer... Mental disorder Stress, depression... Symptom Headache, self-harm... Media Program, internet... Institution Health minister, prosecutor... General Counseling, mental health... Positive Good, helpful, grateful Negative Bad, sick, difficult... Adjective be, far, much... Verb Receive, go, come... Noun Person, task, idea... Group word Mental health
Prior to focusing on psychiatry specifically, we identified words that were associated with 'psychiatry,' 'internal medicine,' and 'surgery.' To cover texts relevant to psychiatry as widely as possible, 'mental health' and 'mental illness' were settled as main words for their conceptual comprehensiveness. Then, words associated with the two words were selected based on the frequency. Texts containing selected words were included for further analyses. These words included 'psychiatry,' 'mental health,' 'mental illness,' 'mental hospital,' 'psychological therapy,' 'distress,' 'treatment,' 'prescription,' 'depression,' 'suicide,' and 'antidepressant.' We excluded specific disorders as they could bias the result. To enhance the specificity of the corpus, texts containing approximately 500 words that were associated with advertisements, such as 'second-hand car,' were excluded.
Among the words associated with 'psychiatry,' we focused on those that described barriers to seeking psychiatric treatment. We categorized these words and read as many articles as possible to understand the contexts in which the words were used. The words were categorized at a structural and individual level by a TM professional, based on a theoretical multilevel conceptual framework suggested by Megan et al. [[
Although the writers' demographic data could not be accessed, a part of corpus could be classified generally by age group, based on their web community. Some communities were considered to represent specific age groups with similar interests. Also, these web communities have cultural characters shared by the age group in question. For example, writers in the community for sharing information on university entrance exam were deemed teenagers. Texts from the web community 'people searching for jobs seriously' were likely written by 20- to 30-year-old job seekers as the government official employment was the main topic in the community. Texts from the web community 'pregnancy and bringing up children café' were likely written by 30- to 40-year-old mothers. Most members of café 'elegant menopause' may be 50-year-olds. In the long list of web communities, such communities that could be identified were selected for the further analysis. In addition, some message boards and internet portals classify their members according to age group. Therefore, these data were also analyzed separately for each age group.
A total of 97,730,360 articles were identified for the period in question. Table 2 shows the 20 keywords most frequently associated with 'psychiatry,' 'internal medicine,' and 'surgery' in this corpus. The word most frequently associated with 'psychiatry' was 'information' (7.8%; Table 2). The word most frequently associated with both 'internal medicine' and 'surgery' was 'symptom.'
Table 2 The 20 words that were most frequently associated with psychiatry, internal medicine, and surgery
No 1 Symptom Symptom 2 Symptom 7.70% Dysfunction 8.20% Dysfunction 8.20% 3 Mind 7.40% Medicine 6.90% Condition 7.80% 4 Psychology 6.90% Dysfunction 6.70% Problem 6.60% 5 Idea 6.90% Problem 6.40% Effect 5.80% 6 Condition 6.80% Condition 6.10% Management 5.40% 7 Appliance 6.40% 5.60% Result 4.80% 8 Way 6.40% Function 4.80% 4.50% 9 Problem 5.80% Food 4.60% Drug 4.40% 10 Medicine 5.70% Result 3.70% Disorder 4.10% 11 Disorder 5.50% Management 3.70% Function 3.90% 12 Dysfunction 4.10% Appliance 3.70% Experience 3.90% 13 Behavior 3.50% Usage 3.60% Hobby 3.70% 14 Instance 3.10% Effect 3.60% Everyday life 3.70% 15 Hobby 3.50% Idea 3.70% 16 Anxiety 2.80% Usual 3.40% Usual 3.60% 17 Relationship 2.60% Idea 3.10% Mind 3.40% 18 Emotion 2.60% Feeling 3.10% Recent 3.30% 19 Result 2.60% Mind 3.00% Recovery 3.20% 20 Society 2.50% Panic 2.90% Personal 3.20%
Among the texts, 6,097,369 contained keywords associated with psychiatry, including 2,323,303 texts from web communities (36.4%), 1,896,239 texts from blogs (31.1%), and 1,963,827 texts from Twitter (32.2%).
We identified approximately 3,000 words associated with 'psychiatry' according to their frequency. Many of these words were linked to the topic of discrimination. A recurring theme was concern that disclosure of having a mental illness would be disadvantageous due to government policy or social stigma. Consequently, these words were classified in the 'structural discrimination' word group. These terms/words included 'medical record,' 'public official employment,' 'buying insurance,' 'disadvantage,' and 'non-insurance' (Table 3). Another recurring theme was stereotypes and prejudice associated with mental illness. This word group was labeled the 'public prejudice group' and included the terms/words 'mad person,' 'negative attitude,' 'prejudice,' 'stigma,' and 'sympathy.' Words expressing concern regarding adverse drug effects were classified in the 'adverse drug effect' word group and included the terms/words 'adverse effect,' 'tolerance,' 'withdrawal symptom,' 'addiction,' and 'dependence.' Words expressing concern regarding medical costs were classified in the 'low accessibility' word group and included the terms/words 'medical fee,' 'medication fee,' 'expensive,' 'burdensome,' and 'counseling fee.' Although these terms/words could have been classified in the 'structural discrimination' word group, we created the 'low accessibility' word group to highlight a distinct barrier to psychiatric treatment.
Table 3 The 10 words that were most frequently associated with each barrier
1 Medical record 14,690 Mad person 7710 Adverse effect 11,535 Medical fee 6361 2 Public official employment 4202 Negative attitude 6340 Tolerance 1841 Medication fee 1981 3 Buying insurance 2968 Prejudice 5587 Withdrawal symptom 1021 Expensive 1341 4 Disadvantage 2885 Stigma 935 Addiction 331 Burdensome 1149 5 Non-insurance 828 Sympathy 735 Dependence 176 Counseling fee 784 6 Disadvantage on job 699 Abnormal 549 Potent 158 Treatment cost 655 7 University 666 Finger pointing 337 Taking duration 97 Copay 394 8 Browse 465 Tag 245 Discontinuation 78 High cost 190 9 F code 435 Other's view 226 Sequela 75 Subsidy 172 10 Entrance examination 218 Loser 184 Overdose 67 Poor 69
Figure 2 shows the frequency of keywords in each of the following categories: structural discrimination, public prejudice, adverse drug effects, and low accessibility. Structural discrimination was the greatest barrier to receiving psychiatric treatment, accounting for 34% of all of the keywords. Public prejudice was the next biggest barrier (27.8%), followed by adverse drug effects (18.6%), and low accessibility (16.1%). The 10 words that were most frequently associated with each barrier are shown in Table 3. In the structural discrimination section, 'medical record' was overwhelmingly the most frequently used term, followed by 'public official employment,' 'insurance,' 'disadvantage,' 'non-insurance,' 'disadvantage in seeking jobs,' 'university,' 'browsing records,' 'F-code diagnosis,' and 'entrance examination.' 'Mad person' was the most frequently used term in the public prejudice section, followed by 'negative perception,' 'prejudice,' 'stigma,' 'sympathy,' 'abnormal,' 'finger-pointing,' 'tag,' 'other's view,' and 'loser.' Concerns about medication in the adverse drug effects section were clear from words/terms such as 'adverse effect,' followed by 'tolerance,' 'withdrawal symptom,' 'addiction,' 'dependence,' 'potent,' 'taking duration,' 'discontinuation,' 'sequela,' and 'overdose.' The low accessibility section included 'medical fee' followed by 'medication price,' 'expensive,' 'burdensome,' 'counseling fee,' 'treatment cost,' 'copay,' 'high cost,' 'subsidy,' and 'poor.'
Graph: Fig. 2 Barriers according to keywords
Data from selected web communities were also separated into the following four groups: teenagers, 20- to 30-year-old job seekers, 30- to 40-year-old mothers with children, and 50- to 60-year-old seniors. Figure 3 shows the barriers to psychiatric treatment arranged by age group. Structural discrimination was the greatest barrier to treatment in the teenagers (51%), young job seekers (64%), and mothers with children (43%), followed by public prejudice, low accessibility, and adverse drug effects. However, in the seniors group, public prejudice (49%) was the greatest barrier to treatment, followed by adverse drug effects and structural discrimination. The seniors group was least concerned about low accessibility, whereas accessibility was of greater concern to the other groups than adverse drug effects. 'Record' was the word most frequently associated with barriers to treatment in all groups except the seniors group, whereas 'university' was only linked to the teenagers group, 'public official' was only linked to the jobseekers group, and 'buying insurance' was only linked to the mothers with children group. 'Prejudice' and 'negative perception' were only linked to the seniors group (Table 4).
Graph: Fig. 3 Barriers according to keywords, arranged by age group
Table 4 The five words that were most frequently associated with barriers, arranged by age group
Rank Teenagers Share (%) 20–30 years Job seekers Share (%) 30–40 years Mothers with children Share (%) 50–60 years Seniors Share (%) 1 Record 25.9 Record 22.4 Record 22.1 Mad person 19.0 2 Mad person 14.4 Public official 15.6 Mad person 14.8 Adverse effect 16.7 3 University 10.6 Disadvantage 13.6 Adverse effect 10.9 Prejudice 13.1 4 Disadvantage 8.0 Mad person 7.1 Medical fee 9.3 Record 10.7 5 Medical fee 6.8 Medical fee 4.5 Buying insurance 5.1 Negative attitude 9.5
This study investigated the barriers to seeking help from psychiatrists in Korea. To the best of our knowledge, this is the first study to analyze these barriers using TM of internet big data within Korea. The results show that concerns about structural discrimination are the greatest barrier to seeking help, followed by public prejudice, adverse drug effects, and low accessibility. Interestingly, the greatest barrier is contrasting by age groups. While younger Koreans were more concerned about structural discrimination, older Koreans regarded public prejudice as the greatest barrier to seeking help.
Our results differ from those of previous studies. A study of barriers to mental health treatment based on the World Health Organization's mental health surveys reported that attitudinal barriers to initiating and maintaining treatment are more important than structural barriers [[
Stigma is a controversial concept, and it is difficult to define [[
In Korea, medical costs are covered by the national insurance system. Diagnoses and treatment records are monitored to calculate the financial cost of this service. A major benefit of this system is that it increases accessibility by decreasing the cost of treatments. However, some patients worry that their medical records may be disclosed inappropriately. The Korean standard classification of diseases records mental and behavioral diagnoses in section 'F' and the so-called 'F code' is sometimes considered a 'Scarlet Letter' [[
Corrigan et al. suggested that structural discrimination against people with mental illnesses may occur intentionally or unintentionally [[
Discrimination against people applying for private insurance exemplifies unintentional discrimination because this may be justified commercially. Insurance companies frequently refuse to underwrite people with mental disorders. One reason for this may be that mental disorders are heterogeneous and difficult to diagnose precisely [[
Public prejudice may include both internal stereotyping and prejudice against people with mental illnesses. Our results showed that younger people were more concerned about structural discrimination, whereas older people feared public prejudice. This reflects the different attitudes exhibited by each of these generations. Interestingly, some studies have reported a change in public attitudes toward mental illness. A meta-analysis of national surveys, conducted mainly in western countries, showed that attitudes toward mental illness have remained unchanged or worsened over the last few decades, despite considerable improvements in mental health literacy [[
This study had several limitations. First, although the TM method can scan a huge quantity of information, its results resemble an aerial photograph rather than a detailed map. Previously defined barriers to using mental health services could not be applied to our data. Therefore, it is difficult to compare our results with those from other studies. Second, barriers were categorized based only on words, and the same words were interpreted as having the same meaning, despite the possibility of different uses in different contexts. Third, although the internet is used widely in Korea, the population willing to communicate in cyberspace may differ from the general population, resulting in sampling bias. Fourth, although we have sought to cover the texts relevant to psychiatry as widely as possible, the selection process could be arbitrary to some extent. Although psychiatrically related, if the words were rarely mentioned they could be missed in the selection of words relevant to psychiatry. Finally, the analysis by age groups lacks quantitative robustness despite its insightful implication. As only a part of corpus was included, the representativeness of the selected could be limited. Also, communities were loosely classified into age groups based on their interest-group topics. Therefore, some texts in a community could be written by writers from other age groups, making the difference by age groups more ambiguous.
Structural discrimination is the greatest barrier to receiving psychiatric help in Korea. Difference in the weight of barriers, however, is among age groups. As well as addressing structural issues for all, more tailored approaches may be required by generations to lower the gap. Further studies are needed to validate the factors associated with barriers to psychiatric service use.
The authors would like to thank Dr. Sun Kyun Kim, a president and CEO of Signature Healthcare services, for funding support.
Hwo Yeon Seo wrote the article, Gil Young Song and Jee Won Ku analyzed the data as data specialists, Hee Jung Kim and Hye Yoon Park classified the words, Woojae Myung made the tables and figures, Chang Hyeon Baek named the groups, Hee Jung Yoo and Nami Lee proofread the draft and fleshed the discussion part out, Jee Hoon Sohn outlined the research and methodology, and Jee Eun Park coordinated the research and wrote this article with Hwo Yeon Seo. The author(s) read and approved the final manuscript.
Sun Kyun Kim Research Funding.
The corpus used for the analysis is not accessible.
The whole corpus included in this study is anonymous and accessible to all. The corpus from communities is collected Naver(https://section.cafe.naver.com/ca-fe/) and Daum (https://top.cafe.daum.net/%5fc21%5f/home). The corpus from blogs are mainly hosted by Naver (
We consent the publication of this article.
The authors certify that they have no affiliations with or involvement in any organization with any financial or non-financial interest in the subject matter or materials discussed in this manuscript.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
By Hwo Yeon Seo; Gil Young Song; Jee Won Ku; Hye Yoon Park; Woojae Myung; Hee Jung Kim; Chang Hyeon Baek; Nami Lee; Jee Hoon Sohn; Hee Jeong Yoo and Jee Eun Park
Reported by Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author