Chatbot-Based Dialogue for Early Psychosis Detection: Leveraging

Global Assessment of Functioning (GAF)

Astria Hijriani

1,2,* a

, Lailatus Salma

, Sulastri

, Yohana Tri Utami

and Yunda Heningtyas

Ulsan National Institute of Science and Technology, Unist-Gil 50, Ulsan, South Korea

Department of Computer Science, University of Lampung, Bandar Lampung, Indonesia

Nursing Department, Politeknik Kesehatan Negeri Tanjungkarang, Bandar Lampung, Indonesia

Keywords: Psychosis, Schizophrenia, Global Assessment of Functional, Chatbot, Mental Disorder.

Abstract: Mental disorders patients often make people restless because they have uncontrollable behavior and need

special rehabilitation. Not every patient's family knows how to handle them, and health workers find it

difficult to monitor their condition. Early detection of psychosis, specifically schizophrenia, is critical for

effective treatment and support. Existing studies have explored expert systems and mobile applications for

diagnosing mental disorders and depression. Still, none have concentrated on developing a system or

application for the early detection of psychosis based on its severity. This research seeks to significantly

contribute to the mental health field by improving the early detection of psychosis, ultimately enhancing the

lives of individuals affected by this condition using chatbot-based dialog. The Early detection of psychosis

features was developed using the Global Assessment of Functional (GAF) and integrated with the Jiwamuku

mobile application. This GAF was categorized into four classifications: not indicated, mild, medium, and

serious. This feature was tested on 15 respondents. The application shows that two people are not indicated:

11 are mild, one is medium, and one is severe. Of all the data taken, 14 tests are accurate. The result shows

that early detection of psychosis features with Global Assessment of Functioning can be detected with 93%

accuracy. This research underscores the critical importance of early detection of mental disorders, particularly

psychosis, by harnessing the potential of chatbot-based dialogue and the Global Assessment of Functioning

(GAF).

1 INTRODUCTION

Mentally healthy individuals do not display signs of

mental disorders and effectively manage life

challenges by handling stress and recognizing the

importance of mental health on par with physical

well-being. Mental health problems can stem not only

from heredity but also from overwhelming stress due

to life demands (Saputra, et al., 2018).

In Indonesia, approximately 27.3 million people

experienced mental health disorders in 2017, with 6.7

per 1000 households dealing with schizophrenia/

psychosis (Kemenkes, 2018). Schizophrenia, a

lifelong disorder affecting daily activities, disrupts

https://orcid.org/0000-0002-6073-6517

https://orcid.org/0000-0001-7363-7211

https://orcid.org/0000-0003-1082-1622

https://orcid.org/0000-0001-5401-9979

thought and perception patterns, impacting factors

like age, memory, and intellect (Sumner et al., 2018).

People with mental health disorders often exhibit

uncontrollable behavior, causing societal distress.

They may experience relapses and require specialized

rehabilitation. However, post-treatment relapses are

influenced by a lack of supervision and limited

knowledge among family members regarding patient

care.

Information system technology can assist in

monitoring post-treatment mental health patients.

Studies, such as the development of web-based

systems and the JiwaMuKu Android app, focus on

monitoring patients' health conditions via mobile

applications used by their families. Although these

Hijriani, A., Salma, L., Sulastri, , Utami, Y. T. and Heningtyas, Y.

Chatbot-Based Dialogue for Early Psychosis Detection: Leveraging Global Assessment of Functioning (GAF).

DOI: 10.5220/0013671300003873

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Medical Science and Health (ICOMESH 2023), pages 271-277

ISBN: 978-989-758-740-5

271

systems can screen and provide progress graphs,

reminders, and reports, they have not progressed to

diagnosing disorders like psychosis.

In addition to post-medical care processes, early

detection of schizophrenia and other symptoms is

equally crucial. Early education and symptom

recognition processes are vital for personal

assessment and those close to the individual. If

identified early on, patients can seek assistance for the

mental health issues they are facing. Typically,

patients or their families may feel reluctant to disclose

the early symptoms to others or mental health

professionals. Other research explores using expert

systems for psychosis detection, schizophrenia, or

student depression. Nevertheless, there is ample room

for improvement in identifying the severity of early-

stage psychosis.

This research uses the Global Assessment of

Functioning (GAF) Scale to determine severity. The

study focuses on developing a feature for early

psychosis detection using the GAF scale through

chatbot-based dialogues. Choosing a chatbot dialog

aims to make the application more user-friendly, with

different constructions for self-assessment or

assessments by others.

2 RELATED WORKS

Mental health disorders are conditions in which an

individual has trouble adapting to their surroundings.

The inability to solve problems can lead to excessive

stress, making the mental health of the individual more

vulnerable and eventually diagnosed as a mental

health disorder (Lubis et al., 2014). Psychotic

disorders, on the other hand, are characterized by a

disturbance in assessing reality. Psychosis includes

various types, such as schizophrenia, schizoaffective

disorder, persistent delusional disorder, bipolar with

psychotic features, and depression with psychotic

features (Idaini et al., 2018).

Schizophrenia is a group of psychotic disorders

with distinctive distortions in thought processes.

Individuals with schizophrenia may sometimes feel

controlled by external forces, experience strange

delusions, perceptual disturbances, abnormal affect

integrated with real or actual situations, and autism.

Schizophrenia is the most common psychotic

disorder, affecting nearly 1% of the world's population

during their lifetime (Zahnia & Sumekar, 2016).

People with schizophrenia generally require

assistance from those around them, known as

caregivers. Caregivers help with the daily activities of

individuals needing care, such as the sick or children.

Caregivers can include family members or close

individuals who assist patients in their daily activities

(Farkhah & Suryani, 2017). The level of assistance

provided may vary for each patient.

Several studies have explored the use of

information systems in primary healthcare

institutions, including the development of web-based

systems for monitoring the recovery process of mental

health disorder survivors (Waskito, 2020),

(Adiyaksatama, 2020) and the JiwaMuKu Android-

based application (Syahputra, 2020). These systems

allow for tracking patients' health conditions,

integrated with a mobile application used by their

families.

Screenings can be performed in the mobile

application, and healthcare professionals can view the

results through a web application. The web application

can display patient progress graphs, provide

reminders, and generate patient reports. However,

these studies have not progressed to diagnosing

mental health disorders such as psychosis

(schizophrenia).

In 2018, Landung Sudarman and Febty Lestari

developed an expert system for diagnosing

schizophrenia (Sudarmana & Lestari, 2018) using a

forward chaining method and Dempster-Shafer

theory. (Widodo and Jaya, 2018) and (Anisa, 2018)

created an expert system for analyzing the level of

depression in students and schizophrenia detection

utilizing the certainty factor method. Another study

focused on early detection of psychosis using a mobile

application based on fingerprints. However, no

research has been conducted on developing a system

or application to detect early-stage psychosis based on

its severity.

This research utilized the Global Assessment of

Functioning (GAF) Scale to determine the severity

level based on weighted symptoms.

The GAF is used to assess the seriousness of

mental health disorders and measures the impact of

symptoms on daily life on a scale of 0-100 (Smith,

2021). GAF scores can be assessed through

interviews, questionnaires, medical records,

information from doctors, caregivers, or close

relatives, and police or court records documenting

violent or illegal behavior.

3 METHODS

3.1 Design

This study was to develop an early psychosis

detection feature on the JiwaMuKu (Jiwa Munyai

ICOMESH 2023 - INTERNATIONAL CONFERENCE ON MEDICAL SCIENCE AND HEALTH

272

Jiwa Strong) platform in the form of an Android

application that general users can use.

3.2 Population and Sampling

The psychosis early detection feature is carried out by

several community groups, including five people

from the Sukarame neighborhood, five people who

are experiencing mental disorders, and five people

mental health professionals.

Testing with people with psychiatric disorders

was conducted in two different places. The first place

was at Wisma Ataraxis, while the second was at the

Yayasan Sinar Jati Kemiling. During the testing, they

were directly accompanied by mental health experts.

3.3 Procedure

Application development begins with a discussion

with the Ikatan Perawat Kesehatan Jiwa Indonesia

(IPKJI) to analyze the needs and definitions of the

application during development. After the discussion,

the application design was carried out by designing

the application's functional features, workflows, data

structures, and appearance. After that, the coding

stage is carried out based on the design that has been

made previously.

The next stage is black box testing to ensure

that all functions run smoothly before being used by

users. If the black-box test has met the expected

results, then testing with users is carried out. From

testing with users, it can be concluded how accurately

the application works.

4 RESULTS

The early detection of psychosis is a new feature that

will be added to the JiwaMuKu application.

Psychosis detection diagnoses its severity based on

the symptoms experienced using the Global

Assessment of Functioning (GAF) scale. The public

can access this feature, and it does not require login

beforehand. The questions asked are based on the

individual's conditions and experiences. As for the

answers provided, they indicate the level of

confidence in those events or conditions.

Like an expert system, this feature also utilizes

rules to conclude. Table 1 presents a list of symptoms

and rules for the early detection of psychosis features.

Meanwhile, figure 1 shows the main menu of the

Jiwamuku application.

Figure 1: Home menu for the user.

The "Check Me" menu is where users can

diagnose psychosis. To begin the test, options will be

presented for self-detection or detection of others.

The difference between these two options lies in the

questions' wording, which is based on the target being

detected. If conducting the test for oneself, the

questions will use the word 'kamu' (you), while for

testing someone else, the word 'dia' (he/she) will be

used. The answer and the calculation will also differ

based on the viewpoint.

Figure 2: Display Self-Doubt Questions for Oneself (Left)

and Self-Doubt Questions for Others (Right).

Figure 2 displays examples of questions related

to a lack of confidence. Feeling lacking in confidence

is something everyone may experience, and it's a

normal occurrence. However, if one consistently feels

Chatbot-Based Dialogue for Early Psychosis Detection: Leveraging Global Assessment of Functioning (GAF)

273

low in self-confidence and believes that others always

harbor negative thoughts about them, it falls under the

symptoms of psychosis.

After answering various displayed questions, a

summary page of the provided answers will appear.

Figure 3 shows the display of the resume answer. If

the answers seem appropriate, the user can view the

conclusion of the detection by pressing the Next

button. The displayed conclusion will be based on the

obtained score. There are four categories:

• The first category indicates no mental health

issues if the score is >91. Individuals in this

category show no symptoms, function optimally,

and have no unresolved problems.

• The second conclusion, with a score between 61-

90, indicates a mild condition. In this category,

individuals with mental health issues experience

temporary and manageable symptoms, like

everyday problems.

• The conclusion with a score between 51-60

indicates a moderate condition, where the

experienced symptoms are at a moderate level.

• The last conclusion, in the severe category, with

a score between 0-50, suggests that individuals

with mental health issues are in a persistent and

more serious state. This may include some

disabilities in communication and cognitive

functions, a risk of self-harm or harm to others,

and severe impairments in various functions.

Figure 3: Display of the summary page of the application.

When the application is ready, a system test is

conducted to ensure that all functions within the

application and system operate correctly using the

black box testing technique. Figure 4 shows the

documentation of black box testing.

Figure 4: Black box testing at the Sinar Jati Foundation in

Kemiling.

The system developers and healthcare personnel

perform black box testing on the JiwaMuKu

information system. The testing activities occurred at

the Kedaton Inpatient Health Center, the Rajabasa

Indah Health Center, the Sinar Jati Kemiling

Foundation, and the Way Halim Health Center.

The second testing involves accuracy testing to

determine the accuracy of early psychosis detection

and conclusion, as documented in Figure 5. Various

community members were involved in this testing.

There were five individuals in the Sukarame

community, five individuals experiencing moderate

mental disorders, and five mental health

professionals.

Figure 5: Testing activities for early detection of psychosis

conducted by Mr. E at the Sinar Jati Kemiling Foundation.

ICOMESH 2023 - INTERNATIONAL CONFERENCE ON MEDICAL SCIENCE AND HEALTH

274

Table 2: Symptoms and Rules.

Code Explanation Not

Indicated

(

100-91

)

Mild

(91-61)

Moderate

(60-51)

Severe

(50-0)

G1 Men

asin

kan diri dari lin

kun

an sekita

oran

lain x x x

G2 Perilaku ketika makan tidak bai

x x

G3 Perubahan emosi secara signifikan x x x

G4 Perubahan pada pola tidu

xxx x

G5 Berteriak tiba-tiba dan marah tanpa alasan x x x

G6 Sulit berkonsentrasi (terlebih jika sedang membaca ataupun

menonton)

x x x

G7 Merasa cemas berlebihan x x x

G8 Kesulitan untuk membaca dan menulis x

G9 Sulit merasa senan

atau

uas x x

G10 Kehilan

an minat dan motivasi

ada berba

ai aktivitas x x x x

G11 Tidak peduli pada penampilan dan kebersihan diri x x

G12 Curiga dan tidak percaya dengan orang lain x x

G13 Sering merasakan/mendengar adanya suara yang mengancam

atau memberi

erintah

x x

G14 Meyakini bahwa dirinya menjadi sosok tertentu x x

G15 Ekspresi wajah dan nada bicara tidak sesuai dengan situasi x x

G16 Gerakan tubuh tidak teratur dan gerak tubuh yang tidak

normal atau sulit di

rediksi

x x

G17 Berkeinginan atau melakukan untuk mencederai diri sendiri

mau

un oran

lain

G18 Tidak

erca

a diri, takut dikritik atau dihakimi oran

lain x x

G19 Putus asa dan pesimis x x

G20 Merasa dirinya dikendalikan oleh orang lain x x

G21 Mendengar suara halusinasi yang berkomentar secara terus-

menerus

x x

G22 Merasa bersalah, tidak beguna, rasa terbebani, tidak berdaya

dan tidak ber

una

G23 Ber

ikir in

in mati x

5 DISCUSSIONS

The challenge in this research is how to transform the

language of symptoms into an early detection

language easily understood by users. The developed

chatbot dialog should also be able to view two

different sides, including self-assessment and

assessment by others.

Testing conducted at Wisma Ataraxis involved

two individuals facing mental health issues. The first

respondent, R, experienced mental health problems

with symptoms of anxiety, difficulty sleeping,

concentration issues, and withdrawal from social

environments. The conclusion from R's early

psychosis detection test indicated mild mental health

issues due to R's improved condition and ability to

communicate. The second respondent, Mr. M, had

mental health problems with symptoms of loss of

interest in hobbies, hallucinations, and suicidal risk.

Like R, Mr. M's results also showed mild mental

health issues, as his condition improved during the

test. The third and fourth respondents, both staff at

Wisma Ataraxis, showed no indications of mental

health issues.

The next testing location was conducted at

Yayasan Sinar Jati Kemiling, where early psychosis

detection tests were performed on three patients and

one staff member. The first respondent at Yayasan

Sinar Jati, a healthcare worker, received a mild result.

The second respondent, Mr. E, experiencing

symptoms of anxiety, emotional changes,

unexplained anger, and neglecting personal hygiene,

showed mild results. Mr. E had been mistreated by his

family and remained unaccepted by them despite two

years of treatment. Despite his tendency to raise his

voice, Mr. E was genuinely concerned about his

surroundings, leading to a mild psychosis detection

result.

The next respondent was Mr. R, with symptoms

of anxiety, emotional changes, difficulty

concentrating, neglect of personal hygiene,

suspicious feelings, hallucinations, and suicidal risk.

Chatbot-Based Dialogue for Early Psychosis Detection: Leveraging Global Assessment of Functioning (GAF)

275

He also faced family rejection and had been treated at

Yayasan Sinar Jati for four years. During the early

psychosis detection test, Mr. R scored 79, falling into

the mild category.

The final respondent from Yayasan Sinar Jati

Kemiling, identified as S, had been under treatment

for four years, experiencing symptoms of anxiety,

hallucinations, and suicidal risk. S's result indicated a

moderate level. Since direct interaction with the

patient was not allowed during data collection, there

was no documentation during the testing activities.

From the conducted tests, the application's

conclusions were compared with those of mental

health experts conducting direct psychosis detection.

Out of 15 test data, two individuals showed healthy

results, 11 showed mild results, 1 showed moderate

results, and 1 showed severe results.

Of the 14 accurate test data, the accuracy rate was

93%. Therefore, this feature can conclude with an

accuracy rate of 93%. The results from the collected

data do not represent the overall since the sampling

size was small. Hence, these results are used as a

reference, indicating that the application's displayed

results have an accuracy rate of 93%.

6 CONCLUSIONS

In conclusion, this research emphasizes the

paramount significance of early detection in

addressing mental disorders, specifically psychosis.

Leveraging chatbot-based dialogue and the Global

Assessment of Functioning (GAF), our study has

demonstrated a promising approach to enhance the

detection and monitoring of mental health conditions.

Moving forward, there are several avenues for

future research in this domain. Firstly, developing

early psychosis detection systems could benefit from

applying expert system development methods such as

certainty factor, fuzzy logic, or other advanced

techniques. These methodologies can potentially

enhance the accuracy and efficiency of early

detection systems.

Furthermore, exploring chatbot-based solutions

for early psychosis detection presents an intriguing

avenue for future work. Integrating chatbots into

mental health diagnostics offers a user-friendly and

accessible approach, allowing individuals to engage

in conversations with the bot to diagnose psychosis.

This interactive and conversational model has the

potential to reach a broader audience, providing

timely and personalized assistance in mental health

assessments.

The research lays the foundation for innovative

approaches to mental health diagnostics, emphasizing

the importance of proactive measures in identifying

and addressing mental health issues. Integrating

technology, particularly chatbot-based systems, holds

promise in revolutionizing early detection methods

and facilitating more widespread access to mental

health support.

ACKNOWLEDGEMENTS

The researcher would like to extend gratitude to

Puskesmas Kedaton, Puskesmas Rajabasa Indah,

Puskesmas Way Halim, Wisma Ataraxis, and

Yayasan Sinar Jati Kemiling for their assistance and

collaboration in this research.

REFERENCES

Adiyaksatama, M. Y., 2020. Desain Antarmuka Dan

Implementasi User-Centered Design Pada Aplikasi dan

Sistem Informasi Jiwamuku (Jiwa Munyai Jiwa Kuat),

Minithesis, Bandar Lampung: Universitas Lampung.

Annisa, R., 2018. Sistem Pakar Metode Certainty Factor

Untuk Mendiagnosa Skizofrenia. IJCIT (Indonesian

Journal on Computer and Information Technology),

3(1), pp. 40–46.

Farkhah, L. & Suryani, S., 2017. Faktor Caregiver dan

Kekambuhan Klien Skizofrenia. Jurnal Keperawatan

Padjadjaran, 5(1), p. 37–46.

Idaini, S. et al., 2018. Prevalensi Psikosis di Indonesia

berdasarkan Riset Kesehatan Dasar. Jurnal Penelitian

dan Pengembangan Pelayanan Kesehatan, 3(1), pp. 9-

16.

Kementrian Kesehatan RI, "Laporan Riskesdas,"

Kementrian Kesehatan RI, Jakarta, 2018.

Lubis, N., Krisnani, H. & Fedryansyah, M., 2014.

Pemahaman Masyarakat Mengenai Gangguan Jiwa

Dan Keterbelakangan Mental. Share: Social Work

Journal, 4(2), pp. 137-144.

Saputra, F. A., Ranimpi, Y. Y. & Pilakoannu, T., 2018.

Kesehatan Mental dan Strategi Koping Studi

Sosiodemografi di Kudangan Kecamatan Delang

Kabupaten Lamandau Kalimantan Tengah. Jurnal

Keperawatan Jiwa., 2(1), pp. 63-74.

Smith, M., 2021. What Is the Global Assessment of

Functioning (GAF) Scale? [Online] Available at:

https://www.webmd.com/mental-health/gaf-scale-

facts#091e9c5e815efdfe-1-2

Sudarmana, L. & Lestari, F., 2018. Aplikasi Sistem Pakar

Untuk mendiagnosis Gangguan Jiwa Schizophrenia.

Jurnal Informatika: Jurnal Pengembangan IT (JPIT).,

3(1), pp. 40-44.

Sumner, C. B. et al., 2018. Process Evaluation of a Pilot

Intervention for Psychosocial Rehabilitation for Service

Users with Schizophrenia in Northwest Province, South

ICOMESH 2023 - INTERNATIONAL CONFERENCE ON MEDICAL SCIENCE AND HEALTH

276

Africa. Community Mental Health Journal, Volume 54,

p. 1089–1096.

Syahputra, A., 2020. Pengembangan Aplikasi Jiwamuku

Berbasis Android Untuk Monitoring Proses Recovery

Survivor Gangguan Jiwa Pasca Perawatan, Minithesis,

Bandar Lampung: Universitas Lampung.

Widodo, E. & Jaya, S., 2018. Implementasi Sistem Pakar

Untuk Mendiagnosa Tingkat Depresi Pada Mahasiswa

Tingkat Akhir dengan Metode Certainty Factor. Jurnal

Teknologi Pelita Bangsa, 8(2), pp. 233-240.

Waskito, S. T., 2020. Pengembangan Sistem Informasi

Berbasis Web Untuk Monitoring Proses Recovery

Survivor Gangguan Jiwa Pasca Perawatan, Minithesis,

Bandar Lampung: Universitas Lampung.

Zahnia, S. & Sumekar, D. W., 2016. Kajian Epidemiologis

Skizofrenia. Medical Journal of Lampung University.,

4(4), pp. 160-166.

Chatbot-Based Dialogue for Early Psychosis Detection: Leveraging Global Assessment of Functioning (GAF)

277