Background

With internet access and use pervading society worldwide, concerns have been raised about the potential damage of excessive internet use [1]. ‘Internet Addiction (Disorder)’, ‘Problematic Internet Use’, ‘Pathological Internet Use’, ‘Internet Addictive Behaviour’ and other terms have been used to describe the combination of addiction-like symptoms and social problems seen in individuals spending huge amounts of time using the internet [2, 3]. However, this proposed disorder is still under discussion concerning its definition, validity as a construct, proposed diagnostic criteria, preferred measuring instruments, and treatment [4, 5]. Proponents of the disorder have made considerable efforts to see it included in the American Psychiatric Association (APA)‘s Diagnostic and Statistical Manual of Mental Disorders (DSM), resulting in the inclusion of Internet Gaming Disorder in section III under the heading “Conditions for Further Study” [6].

Previous epidemiological studies used some measurements including the Young Diagnostic Questionnaire (YDQ), the Internet Addiction Test (IAT), and the Compulsive Internet Use Scale (CIUS) to assess excessive internet use. High prevalence of internet addiction among adolescents and students has been reported by many authors. Cheng et al. [7] conducted a meta-analysis including 89.281 participants from 31 nations and estimated prevalence of internet addiction defined as a YDQ score from 5 to 8 or an IAT score from 70 to 100 to be 6.0% among adolescents and youth. Tsitsika et al. [8] reported that prevalence of dysfunctional internet behavior defined as an IAT score from 40 to 100 among 14–17-year old students to be 13.9% in European countries (N = 13.284) while Mak et al. [9] reported its prevalence to range from 13.7 to 50.9% among 12–18-year-old students in Asian countries (N = 5.366).

In contrast, only a few studies investigated the prevalence among the adult population. Bakken et al. [10] reported the prevalence of internet addiction (YDQ score 5–8) and at-risk internet users (YDQ score 3–4) to be 1.0 and 5.2% respectively in 3.399 Norwegian subjects. Zadra et al. [11] reported that 685 out of 15.023 subjects (4.5%) showed 21 points and more on the CIUS. Thus, the problem of excessive internet use in adults should not be underestimated, even though its prevalence may be lower in adults than among students.

A number of studies have identified the correlation between problematic internet use (PIU) and psychiatric disorders [12,13,14,15,16], including a systematic review by Carli et al. [3] and most notably a meta-analysis by Ho et al. [2]. PIU was shown to be associated with alcohol abuse, attention deficit hyperactivity disorder (ADHD), depression, and anxiety. The correlation with depression has also been found in longitudinal studies [17,18,19], with individuals overly exposed to the internet having a higher risk of developing depression. Furthermore, PIU has also been found to correlate with sleep problems and these also have their effects on patients’ well-being [20,21,22,23]. Therefore, we deemed it necessary to assess the correlation between PIU and comorbid psychiatric symptoms among patients with psychiatric disorders as their psychiatric symptoms could affect PIU or vice versa.

As far as we know, however, there have been no studies as of yet investigating the prevalence and comorbidity of PIU in a psychiatric population. Psychiatric symptoms might either induce PIU in patients with psychiatric illnesses, or PIU might induce or aggravate psychiatric symptoms. Therefore, this study aims to investigate the prevalence of PIU among adult psychiatric patients, and to reveal the degree of co-morbidity between PIU and psychiatric symptoms.

We hypothesize that the prevalence of PIU among psychiatric patients to be higher than among the general population. PIU might be associated with sleep problems, depression, anxiety, ADHD, autism, obsessive-compulsive disorder (OCD), social anxiety disorder (SAD), alcohol abuse, and impulsivity as previous studies reported in adolescents and students.

Methods

Study design and population

A cross-sectional study was performed using online questionnaires. Participants were recruited through their treating physicians at the outpatient clinic of psychiatry, University Hospital, Kyoto Prefectural University of Medicine from January 2016 to April 2016, for which Kyoto Prefectural University of Medicine Research Ethics Committee approved all procedures (ERB-C-485). After obtaining both oral and written consent, participants were asked to fill in an electronic questionnaire either at home or on tablets provided by the researchers.

In- and exclusion criteria

Participants were included if they had been diagnosed with a psychiatric diagnosis from the tenth revision of the International Classification of Diseases and Related Health Problems (ICD-10) [24] categories F1 (Mental and behavioural disorders due to psychoactive substance use), F2 (Schizophrenia, schizotypal and delusional disorders), F3 (Mood [affective] disorders), F4 (Neurotic, stress-related and somatoform disorders), or F5 (Behavioural syndromes associated with physiological disturbances and physical factors).

Participants were excluded if they were below 20 years of age, did not have access to the internet, or were in such a clinically bad shape that their treating physician preferred not to ask them to fill out the questionnaire.

In order to prevent and assess possible selection bias, whether or not the patient met the inclusion criterion for diagnosis or had to be excluded due to age was judged beforehand by the researchers by accessing the medical files for each patient scheduled for an appointment at the psychiatry outpatient department. If any of the treating physicians were unable to recruit a patient due to the patient’s clinical condition being too severe (i.e. the patient was not able to fill out the questionnaire), the patient not agreeing to participate in the research, or not turning up for their appointment, physicians were asked to note this on a form provided to them by the researchers.

Measurements

To assess internet addiction, we administered both IAT developed by Young [25] and the CIUS developed by Meerkerk et al. [26], both translated into and validated in Japanese [27]. There seem to be different versions of the IAT in use. The validated Japanese version used in this research has 20 questions that can be answered on a Likert-scale of 1 to 5, with proposed cut-off points of 40 (problematic internet use), as also used in the large studies by Tsitsika et al. [8] and Mak et al. [9]. The cut-off for PIU on the CIUS was defined as 21 points, as suggested by Guertler et al. [28]. To measure sleep problems, we administered the validated Japanese version of the Athens Insomnia Scale [29, 30]. To measure the severity of other symptoms of mental disorders, we administered the validated Japanese versions of the Beck Depression Inventory II (BDI-II) [31, 32], the State-Trait Anxiety Inventory (STAI) [33, 34], the Adult ADHD Self-report Scale (ASRS) [35], the Autism Spectrum Quotient (AQ) [36, 37], the Obsessive-Compulsive Inventory (OCI) [38, 39], the Liebowitz Social Anxiety Scale (LSAS) [40, 41], the Alcohol Use Disorders Identification Test (AUDIT) [42, 43], the Barratt Impulsive Scale (BIS) [44, 45], and questions regarding the amount and nature of internet use and the patients’ background (Appendix).

The physicians diagnosed their patients based on ICD-10 criteria, and then assessed the severity using the Clinical Global Impression Severity (CGI-S) [46].

Data was only sent when participants completely finished the questionnaire. Any participants who might have completed only part of the questionnaire were thus counted as non-respondent, as we did not receive any of their data.

Statistical analysis

Continuous variables were shown as mean (± standard deviation) or median (interquartile range), and categorical variables were presented as numbers and percentages. To study statistically significant differences between problematic internet users and normal internet users, we divided our sample in two groups. Respondents who scored 40 or higher on the IAT [8, 9] or 21 or higher on the CIUS [28] were defined as problematic internet users, while those who scored below those values on both were defined as normal internet users. Categorical variables were compared between two groups using chi-square test while Fisher’s exact test was performed in tables larger than 2 × 2 using R version 3.3.1 (http://www.r-project.org/). As results apparently didn’t follow a Gaussian distribution, differences in ordinal and interval variables between the two groups were assessed using the Mann–Whitney U test. The software used was SPSS for Windows version 23. Bonferroni correction was applied for multiple comparisons.

Results

Demographics and internet use

As shown in Figs. 1, 1.220 patients visited the hospital during the 3 months of our recruitment period and 333 of them were included in this study. Two hundred thirty-one patients completed the survey, making the response rate at 69.4% (231/333). Fifty patients (21.6%, 50/231) scored above the 40 points thresholds on the IAT. Furthermore, 43 patients (18.6%, 43/231) scored 21 or higher on the CIUS. This brought the total number of patients we defined as “problematic internet users”, those who scored either 40 or higher on the IAT or 21 or higher on the CIUS, to 58 (25.1%, 58/231).

Fig. 1
figure 1

Flowchart of the inclusion of patients

Full size image

The proportion of problematic internet users is remarkably higher in the younger age groups and unemployed group (Table 1). Sex, education level, living alone or with others, and marital status do not seem to significantly affect internet addiction scores. Furthermore, PIU did not seem to differ greatly between the ICD-10 categories of the respondent’s main diagnosis (as reported by their treating physicians). CGI-S scores of patients with PIU assessed by their treating physicians were significantly higher than those of normal internet users.

Table 1 Demographics of the respondents
Full size table

Psychiatry scale scores

Problematic internet users scored higher on all scales measuring co-morbid psychiatry showing a significantly uneven distribution of scores between the problematic and normal internet users, with the exception of the state subscale of STAI and AUDIT (Table 2).

Table 2 Average scores (and standard deviation) for the different scales measuring psychopathological co-morbidity of normal internet users compared with problematic internet users
Full size table

Sleep habits

Some of the questions regarding to internet use (Appendix) revolved around the amount of internet use before sleep. The question “After you’ve gone to bed but before you sleep, how much on average per day do you use the smartphone or tablet?” (median < 10 min vs 30 min-1 h, Mann–Whitney U test, p < 0.001) and “When you wake up during your sleep, how much do you use your smartphone or tabled in bed?” (median “I don’t use it at all” vs “I hardly use it”, Fisher’s exact test, p < 0.001) got significantly higher answers from problematic internet users than from normal internet users (Additional file 1: Tables S1 and S2).

Internet activities

While problematic internet users do not use the internet for work significantly more than normal internet users, they do use it more for private purposes (median 1 h–3 h vs >3 h, Mann–Whitney U test, p < 0.001) (Table 3). Several private activities on the internet seemed more closely associated with PIU (Additional file 1: Table S3). Problematic internet users seem to spend more private time on the internet for the purposes of communicating, looking up information, reading blogs/forums or making blog/forum posts, surfing social networking sites, and watching videos or listening to music.

Table 3 Reported time spent for work and private related activities on the internet among problematic internet users compared with normal internet users
Full size table

Other results

We also asked about the different devices participants used to access the internet (Appendix). Of these devices, only smartphone use was reported significantly more often by problematic internet users than normal internet users in all samples (median “I often use it” vs “I use it really often”, Fisher’s exact test, p < 0.001) (Additional file 1: Table S4). When the data was analyzed by age, only the twenties showed the significant difference in smartphone use frequency between the two groups (Table 4). However, this finding was not significant after Bonferroni correction.

Table 4 Comparison of smartphone use frequency between normal internet users and problematic internet users
Full size table

Discussion

To our knowledge, this is the first study that investigated the prevalence of PIU, and degree of co-morbidity between PIU and psychiatric symptoms among adult psychiatric patients.

We identified 58 problematic internet users who scored either high IAT (40 and more) or CIUS (21 and more) out of 231 patients (25.1%). Although their demographic background was different, Osaki et al. [47] reported only 4.0% of people scored 40 and more on the IAT in the Japanese general population (N = 4.153). Bakken et al. [10] in Norway found percentages for at-risk internet users (YDQ score 3–4) and internet addiction (YDQ score 5–8) of 5.2% (177/3.393) and 1.0% (35/3.393) respectively. Zadra et al. [11] reported that 685 out of 15.023 German general population (4.5%) showed 21 points and more on the CIUS. Thus, the prevalence of problematic internet users among the general population might be estimated to be 4.0 to 6.2%. Bakken et al. [10] reported that 13.0% of their respondents did not use the internet at all, which means 6.0% of their internet using population displayed at-risk internet users, and 1.2% internet addiction. Considering these results, the prevalence of PIU among psychiatric patients could be much higher compared to that of general population.

There may be several reasons explaining the high prevalence of PIU among psychiatric patients. First, our study found that problematic internet users scored higher on scales measuring sleep problems, depression, trait anxiety, ADHD, autism, OCD, SAD, and impulsivity. This is in line with previous research in non-psychiatric students and adolescents [2, 3, 12]. The severity of these psychiatric symptoms in our samples may very well be higher than that of previous studies’ subjects because all of our subjects were diagnosed with at least one of primary psychiatric disorders. Therefore, it is reasonable to assume that the more severe the psychiatric illness of the studied population, the higher the prevalence of PIU. Then, some measurements used in our study such as STAI, AQ, and BIS assess the inherent characteristics of each subject, and these measurements also showed a significant difference between the two groups. Due to its cross-sectional nature, this study is unable to identify any causal relationships between PIU and psychiatric disorders [3], but intrinsic characteristics that predispose towards anxiety, autism, and impulsiveness may be the same characteristics that cause people to develop PIU. Further study to reveal the causal relationship between psychiatric symptoms and PIU is needed. In contrast to previous research [2], we did not find PIU to be associated with alcohol abuse. This is likely due to the small number of patients with alcohol abuse attending our psychiatric outpatient clinic, as these patients are often referred to other hospitals where specialized treatment is available for alcohol addicts and abusers. Further studies that include a larger number of alcohol addiction patients are necessary.

The study also showed a strong inverse association of PIU and age, which might be caused by the high amount of smartphone use among the younger population. There is also a number of studies done in student and adolescent populations which also show an inverse association between PIU and age [12]. An investigation done by the Japanese Ministry of Internal Affairs and Communications among Japanese high schoolers showed figures of 4.6% of high school students scoring 70 or higher on the IAT, and 55.2% scoring between 40 and 70 (N = 14.071) [48]. The group closest in age to these studies was our group of respondents between the ages of 20 and 29, which showed a percentage of 46.0 of problematic internet users. Due to the strong apparent inverse correlation between age and PIU, it is very well possible that psychiatric patients of high school age would score higher than older psychiatric patients or healthy high school students. Further studies should clarify the prevalence of PIU among adolescent patients with psychiatric illnesses.

There was no significant difference in time spent for work-related activities on the internet between the two groups, while problematic internet users spent significantly longer private time compared to normal internet users. Regarding time spent for various private activities on the internet, amount spent for activities such as blogging, social networking, and watching movies would be associated with PIU. These findings offer an important suggestion in clinical settings, as it shows that psychiatric patients need not avoid using the internet for work. Furthermore, physicians could suggest their patients to restrict their internet use to specific activities in order to prevent PIU. Lastly, patients with PIU seem to use their smartphone a lot after they have gone to bed, or when they wake up during sleep. Physicians could help their patients by discouraging this bad sleep hygiene.

This investigation is limited by the relatively small sample size. Only patients who were examined at the University Hospital, Kyoto Prefectural University of Medicine’s Psychiatry outpatient clinic were recruited. Due to the small population size, we were unable to divide our population into groups with a specific psychiatric diagnosis and analyse correlations with scores on the questionnaires. Another limitation is the cross-sectional nature of the study. Because we did not measure our population repeatedly over time, there can be no conclusions drawn regarding causality. Furthermore, this study using an online questionnaire is likely to have greatly improved response rates (69.4%) among people with familiarity to the internet, but prevented us from collecting data on patients who were not used to answering questionnaires online and who did not have internet connection. In our study, we also didn’t employ more advanced statistical methods such as bootstrapping to address the deviations from normality in our data. Finally, the questionnaire relied on self-report. One of the strengths of this study is that many different scales were filled in by each respondent, providing a clear and broad overview of any overlapping or co-morbid illness in every patient.

Future research could aim at finding causal relationships between PIU and various co-morbidities. Future research could also investigate the effectiveness of specialized treatment for psychiatric patients with PIU, due to the high degree of co-morbidity.

Conclusion

This study in adult psychiatric patients suggest that the prevalence of PIU is higher (25%) than that of general population (6%). In addition, PIU in psychiatric patients is associated with higher scores on questionnaires investigating sleep problems, depression, trait anxiety, ADHD, autism, OCD, SAD, and impulsivity. More research is needed to determine the causal relation between PIU and psychiatric symptoms.