J Med Syst (2014) 38:1 DOI 10.1007/s10916-013-0001-1
ORIGINAL PAPER
The SAMS: Smartphone Addiction Management System and Verification Heyoung Lee & Heejune Ahn & Samwook Choi & Wanbok Choi
Received: 7 September 2013 / Accepted: 23 December 2013 / Published online: 7 January 2014 # Springer Science+Business Media New York 2014
Abstract While the popularity of smartphones has given enormous convenience to our lives, their pathological use has created a new mental health concern among the community. Hence, intensive research is being conducted on the etiology and treatment of the condition. However, the traditional clinical approach based surveys and interviews has serious limitations: health professionals cannot perform continual assessment and intervention for the affected group and the subjectivity of assessment is questionable. To cope with these limitations, a comprehensive ICT (Information and Communications Technology) system called SAMS (Smartphone Addiction Management System) is developed for objective assessment and intervention. The SAMS system consists of an Android smartphone application and a web application server. The SAMS client monitors the user’s application usage together with GPS location and Internet access location, and transmits the data to the SAMS server. The SAMS server stores the usage data and performs key statistical data analysis and usage intervention according to the clinicians’ decision. To verify the reliability and efficacy of the developed system, a comparison study with survey-based screening with the K-SAS (Korean Smartphone Addiction Scale) as well as self-field trials is performed. The comparison study is done using usage data from 14 users who are 19 to 50 year old adults that left at least 1 week usage logs and completed the survey questionnaires. The field trial fully verified the accuracy of the time, location, and Internet access information in the usage measurement and the reliability of the system operation over more than 2 weeks. The comparison study showed that daily use count has a strong H. Lee : H. Ahn (*) : W. Choi Seoul National University of Science and Technology, Seoul, Republic of Korea e-mail: [email protected] S. Choi Department of Psychiatry, Gangnam Eulji Hospital, Eulji University and Eulji Addiction Institute, Seoul, Republic of Korea
correlation with K-SAS scores, whereas daily use times do not strongly correlate for potentially addicted users. The correlation coefficients of count and times with total K-SAS score are CC=0.62 and CC =0.07, respectively, and the t-test analysis for the contrast group of potential addicts and the values for the non-addicts were p=0.047 and p=0.507, respectively. Keywords Objective assessment . Smartphone addiction . Statistical analysis . ICT system
Introduction During the last decade, smartphones have gained popularity all over the world. In July of 2013, over 50 % of mobile subscribers in the US and over 65 % of mobile subscribers in South Korea were using smartphones, and the percentage keeps increasing. The key differences between smartphones and previous mobile phones are full-featured Internet access and easy installation of new applications through modern OS platforms and app store. Hence, smartphones are now considered handheld computers rather than traditional phones [1]. Although smartphones have given enormous convenience to our lives, pathological use of smartphones has created a new mental health concern among the community [2–4]. Behavior addictions, including smartphone addiction, are often difficult to define since they are related not only to physical, but also to psychological and social aspects [5–7]. Based on the definition of Internet addiction in the literature, we define smartphone addiction as the excessive use of smartphones that interferes with the daily lives of the users. In addition, it has various clinical features, such as tolerance, withdrawal symptoms, salience, mood modification, craving, loss of control, etc. Although smartphone addiction is a new behavior addiction, much more research data is still needed [6]. Nevertheless, it is prevalent worldwide, and it is causing concern to the society.
1, Page 2 of 10
The traditional clinical approach for screening and assessment in behavior addictions has been surveys and interviews. However, there are a massively large number of subjects in smartphone addiction. Furthermore, continuing intervention would be essential for daily-life treatment of addicted people. This paper presents the SAMS (Smartphone Addiction Management System), a comprehensive ICT system for objective assessment-based diagnosis and intervention for smartphone addiction. We describe the system requirements, architectures, and typical use scenarios of the smartphone addiction management system. Then, we describe the system verification methods: operation checking by manual records, and through a comparison study with a pilot test. The correlation between the usage measure and paper-based addiction scale is also examined. Finally, we discuss the future improvement of the SAMS, together with comparison with related ICT systems.
The SAMS: Smartphone addiction management system
J Med Syst (2014) 38:1 Table 1 Functions of SAMS system Functions
Description
Monitoring
- Monitoring smartphone usage for 24 h, without interruption, unnoticed by smartphone users. - Recording items including app use time, duration, location, and relevant information, such a URL access location. - Usage data is recorded in both the smartphone (for a fixed duration) and the server database (for a long term). - Data should be stored in an efficient format for the data analysis. - The assessment data is stored in a secure fashion. - Providing individual and group data analysis. - Time and trend pattern visualization. - Using a system architecture that can easily extend new analysis and visualization. - Fostering self-control. - Manual intervention by clinicians and automatic rule based intervention.
Data Archive
Data Analysis
Intervention and treatment
System requirements The key goals of the SAMS are to provide a comprehensive framework for an objective data-driven study for addiction study and treatment, specifically, to itemize the assessment metrics, analysis algorithm, and treatment strategies. We developed the system requirements starting from Young’s clinical guideline of Internet addiction [8] and Gustafon et al’s relapse prevention model in A-CHESS [9]. Then the unique properties and capabilities of smartphones, such as their mobility and 24-h-companionship, are included for system requirements. The system requirements are reviewed by the coworking national project team from St. Mary Hospital, Ulji Hospital, and Seoul National University Hospital, South Korea. Table 1 summarizes the key functionality for each stage of the SAMS. The first function is to effectively monitor and store the key usage data for usage patterns. The second function is to provide analysis tools for evaluation of usage data to diagnose symptoms. The last is to deliver intervention and provide a self-control mechanism. System architecture Figure 1 illustrates the overall system architecture and workflow of the SAMS framework. On the client side, the SAMS application continuously monitors applications in use, and stores the usage records locally. Users may view the locally stored records for self-recognition and control. Periodically, the new records are transmitted to the SAMS server via the Internet. The usage records are archived in the SAMS server’s database, and statistical and data mining analysis are performed on the data for the diagnosis and treatment to be
conducted by clinicians. Clinicians can determine feedback actions, such as requesting current condition check survey or updating the usage limit time table for a specific application. While in the early stage of studying smartphone addiction, most studies focus on assessment and evaluation of the extent of addictive use of smartphone. We also aimed to develop treatment techniques and study the outcome of their efficiency. Young [7] suggested possible treatment techniques for Internet addicts: (a) practice the opposite time in Internet use (for example, to shift the Internet use time from evening to morning, or reverse), (b) employ external stoppers, (c) set goals, (d) abstain from a particular application, (e) use reminder cards, (f) develop a personal inventory, and (g) enter individual therapy or a support group. The usage-abstaining function and individual therapy are included in the SAMS application. For example, when a user is playing a special game application for more than the designed usage time, the SAMS client will pop up a warning message and block the user from the game application. The usage restriction function is especially designed for elementary to high-school students for preventing overuse in classrooms, and will only be used for the contracted users. Client system design The SAMS client application is developed on Google Android based smartphones. In the Android system, a component of an application has to be one of four Java classes [10]: ‘Activity’ is for user interaction and output display, ‘Service’ is for a long running background service, ‘ContentProvider’ is for persistent data storage, and ‘BroadcastReceiver’ is for waiting for
J Med Syst (2014) 38:1
Page 3 of 10, 1
Messaging Server (GCM)
Intervention
Messaging
Messaging
Usage Monitoring
Data Storing
Data Transmission
Data Receive
feeback
Decision (Clinicians)
Clinician
Smartphone User
Data Visualization
SAMS client (Google Android)
Data Receive Data Archive
Data Analysis
SAMS server
Fig. 1 System architecture and workflow of SAMS
system wide events. The proposed SAMS client application is composed of eight activities, two services, one content provider, and three broadcast receivers. Due to space limitations, we describe only the operation of the usage monitoring mechanism, which is the key function for our study. First, the foreground application in the Google Android system can be obtained from the system service ‘android.app. ActivityManager’. The monitoring service periodically checks the changes in the foreground application and records its start time, end time, GPS location, and URL if the application is a browser. The GPS location can be obtained from the Android service, ‘android.location. LocationManager’, and the current URL can be obtained from t h e A n d r o i d b r o w s e r h i s t o r y c o n t e n t p r o v i d e r, ‘android.provider. Browser’. Then the usage data is first stored in local ‘SQLite’ database, and then transmitted to the server whenever Internet access, either WiFi or cellular data, is available. Taking into account the LCD display device’s status, the SAMS client does not include the duration of the time the LCD display is off. Special consideration has been taken to maintain the periodic checking process in the Android system without interruption due to its memory limitation and process kill mechanism. In the current implementation, the system sets up the Android system alarm service to send ‘PendingIntent’ periodically every 3 s. We compared this ‘Alarm service plus PendingIntent’ method with an alternative, ‘Service with TimerTask’ method, and finally chose the proposed one because of its reliability (not unintentionally terminated) and low power-consumption (Fig. 2). Figure 3 shows the UI scenarios for users. The SAMS application runs in the background without the user being aware of it. On the first launch after installation (Fig. 3a), the SAMS application gets its GCM (Google Cloud Messaging) registration ID from the Google GCM server, and then registers the ID to the SAMS server for use in messaging service later. When the usage button is pressed, the usage profiles for
applications are presented (Fig. 3b). The user can examine the daily usage pattern of an application during the most recent week (Fig. 3c). A survey view pops up when a clinician asks the current status of the user (Fig. 3d). A blocking application view intercepts the application when the time is in the range forbidden by clinicians (Fig. 3e). Figure 3e is the setting view of the SAMS application which is only for developers. Server system design Figure 4 illustrates the architecture of the SAMS server system. The server receives and stores the data (user profile and client system information) from SAMS clients, performs data analysis/mining, and finally provides graphical and numerical information for the clinicians for analysis, diagnosis and treatment. For basic treatment strategies, the SAMS client is
Fig. 2 Component Structure of the SAMS Android application
1, Page 4 of 10
J Med Syst (2014) 38:1
Fig. 3 SAMS client application user interfaces: a is the launch view, i.e., main view of SAMS app. b shows the app list in the decreasing order of total usage time. c shows the daily usage time graph during 7 most recent days for a clicked app in (b). d shows the survey view pop up on demands or periodically. e shows a blocking view pop-up for preventing from using an app. f is the configuration view for SAMS server’s URL and monitoring period
a
Main View
b app usage time summary
d short status survey view requested through GCM messaging to update the limitation time table or prompt the current status survey of the user. The system architecture of the SAMS server follows the MVC (model-view-control) pattern, and uses the express.js framework of the Node.js web architecture [11]. First, the database schema is designed for reflecting the user profile, data from users (including use records), the installed packages for each user, the social relationship between users, and forbidden application data for each user. Second, the view and transport part is implemented following RESTful [12] methods, i.e., using HTTP and JSON. The data visualization is done using a D3 [13] Javascript- based framework. Lastly, because NoSQL DB [15] is used for the SAMS server, the control-data analysis engine implements the statistical analysis function, built on map-reduce flows [14]. When the SAMS server is requested for analysis results from the browser (clinicians), the data analysis engine processes the data records and transmits the available data. The browser draws the result data using the D3 framework. The current SAMS server provides two groups of data analysis functions, individual analysis and group statistics analysis. The individual analysis functions are used to
e app blocking view
c daily usage history for an app
f
app setting view
examine and diagnose the individual use patterns. Figure 5 shows the individual usage analysis and control pages. The daily usage histogram page (Fig. 5a) shows which application was used and how long it was used on average in a given period. The location information is visualized in the tracking information page using a Google map overlay (Fig 5b). When required (for example, to see what time and which app was used, which is not included in our analysis page yet), the clinician further examines the usage log records (Fig. 5c); each record contains the start time, end time, location, and optional URL information. Even though we did not use it in this study, the clinician can use the application control page to block an application’s usage on a weekly basis (Fig. 5d). For example, in the Fig. 5d, the specific application is marked as orange sections in the weekly schedule table. The update event is sent through the GCM server to the user’s smartphone. Then the smartphone retrieves the updated blocking schedule from the SAMS server. Whenever a new application is launched, the SAMS client will check the permission and block it if it is not allowed. The group statistics analysis is illustrated in Fig. 6, where the correlation between use time/use count and survey
J Med Syst (2014) 38:1 messaging interface
Page 5 of 10, 1 client interface
clinician interface
data visualization (d3)
GCM
data analysis engines
Restful (Json)
webservice (express.js)
DB API (mongoose) server framework (node.js)
DB users usages (mongoDB) Apps (other data)
Fig. 4 System Architecture of SAMS Server: SAMS severs is based on V8, the recently popular javascript engine. Node.js and express.js frameworks are used for building services, control and data analsysis. The interface with the client is defined with RESTful (REpresentation State Transfer) methods, i.e., using HTTP and JSON (JavaScript Object Notation). The data visualization is done using a D3 (Data-Driven Documents). NoSQL DB is used for data storage due to its rich features for data mining and data aggregation
meters from the real locations due to the properties of GPS sensors in commercial smartphones. We believe this much inaccuracy makes the location information useless for our purpose. The extra power consumption is also practically important since the users would not keep our application if they feel that the time between battery recharges is shortened due to the SAMS application. Considering the trade-off between the power consumption and measurement accuracy, a measurement period of 3 s was chosen for deployment. Using ‘GO battery saver’ utility app, the power consumption of SAMS client is measured to be less than 2 % in general user patterns. For the compatibility issue, we tested SAMS clients on as many Android phones as available, which included major models in South Korea and Android versions 2.2, 2.3, 3.0, 4.0 and 4.1 as in Table 2. We updated some implementations for better compatibility. We verified that all functions worked correctly for our specifications, except for the graphic chart package (‘achartengine’) which did not work due to the Android version compatibility problem.
Experiment and data validation Benchmark method
addiction scores (to be defined in the next section) is shown. In the plots, the vertical axis denotes the total K-SAS score, and the horizontal axis denotes the daily use time and count. The lines are linear regressions fitted to data: red lines for females, blue ones for males, and black ones for both. The data analysis itself will be discussed in the following section. System verification We considered three aspects in verifying the SAMS system. First, the correctness of the monitoring data and analysis results. Second, the reliability of the system operation and power consumption. Last, the proper operation of core function on various Android versions in the market. The verification of the accuracy and reliability of input data is crucial for diagnosis and statistical analysis. Note that the SAMS client application has to run 24 h and maintenance efforts have to be minimized. For these purposes, the SAMS application should restart automatically after the smartphone is re-booted or process-killed in a low memory condition. The broadcast receiver and service mechanisms were carefully selected and implemented. For testing data correctness and reliability, the developers and trial users use the apps for a couple of weeks, making notes for app usages and location, and then comparing the usage records with the notes. In this test period, we optimized system parameters such as the sampling period and reporting periods. We observed that even in a rare case, the GPS data was deviated up to several hundred
In order to evaluate the reliability of usage data from the SAMS, we performed a comparison study with a surveybased smartphone addiction test. The survey-based scales are not always reliable and need to be confirmed by clinicians’ close interview. However, we assume the validity of survey results for evaluating the efficacy of the obtained data from the SAMS system. We used the K-SAS (Korean Smartphone Addiction Scale) [2] because it reflects the Korean society very well. A recently published smartphone addiction scale is also based on the KSAS [3]. The K-SAS is composed of 15 question items in Table 3. The 15 questions are organized into four groups based on the factors. Group 1 questions (1, 5, 9, and 12) are for ‘daily-life disturbance’, group 2 questions (2, 6, and 8) are for ‘virtual life orientation’, group 3 questions (3, 7, 10, and 14) are for ‘withdrawal’, and group 4 questions (4, 8, 11, and 15) are for ‘tolerance’. As answer choices for each question, 4 level Likert items, ‘strongly disagree’,’ disagree’, ‘agree’ and ‘strongly agree’ were used. For confirming the validity of answers, we used the reverse-question 8, 10, and 13. Answers to the non-reverse questions are their values in the Likert scale, while answers to reverse questions are 1 minus their Likert scale values. The sum value of all questions is called the total K-SAS scale, and sum values of each subgroup are called group N’s subscales (i.e., G1, G2, G3, and G4). The diagnosis criteria is illustrated by the decision flow chart in Fig. 7. The total scale and subscales are used to determine the
1, Page 6 of 10
a
b
J Med Syst (2014) 38:1
c
daily usage profiles for a user (in the decreasing order of use time)
geographical usage pattern of a user (the numbers are the usage counts for the locations)
d
detail usage log of a user (app package name, start time, duration, GPS location, url)
week time table for an app of a user (orange is the allowed time; grey is for the unallowed time)
Fig. 5 SAMS’s iindividual analysis server clinician interface (the daily usage histogram, usage location graph, usage record details, and usage control page)
highly addicted, the potentially addicted, and the normal (i.e., not-addicted). The subject is determined to be highly addicted when the total scale is over 45, subscale of daily life disturbance is over 16, subscale of withdrawal is over 13, or subscale of tolerance is over 14. The subject is determined to be potentially addicted when total scale is between 42 and 45, subscale of daily life disturbance is over 14, subscale of withdrawal is over 12, or subscale of tolerance is over 13. In the other cases, the subject is determined to be a non-addict. Participant recruit and data acquisition The SAMS client application was uploaded to the SAMS server and SMS messages to request install SAMS app were sent to about 120 anonymous users that were from 4 persons (lab members, including the SAMS developers) phone number list. All recipients were requested to respond to the K-SAS test online, and to download and install the SAMS client application. No benefits were given for joining this test. Among 120 recipients, about 40 installed the SAMS application. However, for data analysis, we used data from 14 users
that had more than 1 week usage log and finished the K-SAS survey faithfully (we checked the consistency of reverse questions). Reasons for unfaithful answers or cessation of application use were not studied carefully. The 14 users are 7 males and 7 females, and the age ranges from 18 to 51 (average of 26.57, standard deviation of 8.25). The main focus of the paper is to introduce a new ICT system, not medical interpretation of the experimental data. We took into account the possible ethical and privacy problems. An IRB approval was not obtained for this verification test. However, the study participants provided informed consents, and all the assessed data are protected on a secured database. We will apply the IRB approval for real scale medial analysis. Statistical analysis The main goal of this trial test was to check and understand the relation between the basic usage statistics and the K-SAS score regarding the K-SAS scale as the true addiction level. The group analysis in Fig. 6 showed the correlation graphs of the usage statistics and the K-SAS score. The correlation
J Med Syst (2014) 38:1
Page 7 of 10, 1 Table 3 K-SAS survey questions Factor Group
Daily life disturbance (G1)
Fig. 6 SAMS’s group analysis functions (top to bottom: plots of (daily use time, K-SAS score comparison) and (daily use counts, K-SAS value))
analysis with the total scale showed no correlation (CC=0.07) for daily use time, but a significant correlation (CC=0.62) for daily use count and the total scale. However, the t-test analysis after applying the sub-group criterion revealed p=0.507 for daily use time and p=0.047 for daily use count. The result implies that the sub-condition in addiction classification is important in distinguishing the potential addicts from the
Table 2 Tested Systems for the SAMS client Android version
Models (Release Year)
2.2 2.3
HTC Desire (2010), SKY Vega S (2011) HTC Desire (2010), LG Optimus One (2010), SKY Vega S (2011) Samsung Galaxy Tab 10.1 (2011) Samsung Galaxy S3, LG optimus G (2012)
3.1 4.0
Question
Having performance degradation in class or office due to smartphone use. Using smartphone when I am not supposed to (in classroom, during meeting, etc). The people around me tell me that I use the smartphone too much. Having difficulties in concentrating on job/study due to the smartphone use. Smartphone use has nothing to do with my performance in class or office. Virtual world Feeling empty when not using my orientation smartphone. (G2) Feeling more pleasant or excited while using smartphones than while being with friends or family. Withdrawal Won’t be able to stand not having a (G3) smartphone. Feeling depressed, anxious, or oversensitive when I am not able to use my smartphone. Not feeling impatient when not holding my smartphone. Not being able to work/study without my smartphone. Tolerance (G4) Having tried to shorten smartphone use time but fails all the time. Keeps on using smartphone while thinking about stopping it. Using my smartphone longer than I had intended. Not spending much time using smartphone.
Question num 1
5
9 12
15
reverse
2 6
8 14
4
reverse
11 7 3 13 10
reverse
non-addicts. From the fact that social network and chatting applications have frequent but short usage time, these applications can be considered as key addiction applications. This suggests that application categorization should be considered for smartphone addiction. To examine the details of application categories, we merged the categories of apps in GooglePlay, the Android app store into 5 ones, social network, games, entertainment, web-browsing, and others. The data shows that the potentially addicts use more social network applications, such as Kakao talk and Facebook than the non-addicts. Kakao talk is the most popular messaging and chatting applications and we categorized it into ‘social network’ in spite of its original category of communication in Google app store which is ‘communications’Tables 4 and 5. We also examined the location of smartphone use. The potentially addicted were asked about most frequent use
1, Page 8 of 10
J Med Syst (2014) 38:1
Fig. 7 K-SAS diagnosis criteria flow charts: highly addict when total K-SAS score>45, G1>16, G3 >13, or G4 >14; potentially addict when total K-SAS score> 42, G1>14, G3>12, or G4>13; otherwise non addict
Total score, G1, G3, G4
next
yes
total score > 42
yes
total score > 45 no
no yes G1> 16
yes G1> 14
no
no yes
yes
G3> 13
G3> 12
no G4> 14
next
locations, and in this study, all three cases were their homes. This implies that detailed statistical study is required to reveal the relation between family environment and smartphone addiction level.
Table 4 Contrast group comparison: daily usage count and use time Contrast group
total K-SAS score (subscales)
daily use time (mins)
daily use count
The potential addicts
52 36 (G1=14, G3=13) 32 (G3=14) Average 37 31 29
273 464 369 369 381 193 148
318 244 137 233 98 95 94
28 27 25 34 26 17 24 20 Average
159 277 317 442 101 179 469 626 299
93 215 212 253 98 86 151 113 137
The non addicts
no yes
yes G4> 13
addict
non- addict
potentially addict
Discussion and conclusion The traditional clinical approach for screening and assessment in behavior addictions has been surveys and interviews, and this also applied to smartphone addiction. However, the survey approach has serious drawbacks and needs to be complemented by objective assessment analysis. First, in interview-based methods, it is difficult to follow the change of subjects’ status, especially when there is a massively large number of subjects as in smartphone addiction. Furthermore, smartphone addiction exhibits a chronically relapsing nature as do other behavioral addictions. Continuing intervention is essential for daily life treatment of addicted people. Second, the survey questions are often subjective and the replies from those surveys depend upon their seriousness, mood, and attitude to the survey. For example, the frequently used Likert score of “strongly agree”, “agree”, “disagree”, and “strongly disagree”, is apt to be subjective, and people who overuse their smartphones may underestimate their own use. Hence, objective and quantitative usage needs to be recorded to perform diagnosis and treatment based on concrete and accurate data. To address these problems, we developed the SAMS (Smartphone Addiction Management System), a comprehensive ICT system for objective assessment-based diagnosis and intervention for smartphone addiction. It has already been established that continuing ICT-based intervention helps in
J Med Syst (2014) 38:1
Page 9 of 10, 1
Table 5 The usage statistics of application categrories: potentially addicts and non addicts Category
Social Network
Game
Entertainment
Browsing
Others
App name
Rank by usage count
Rank by usage time
Daily usage count
Daily usage time (min)
Addicts
Non
Addicts
Non
Addicts
Addicts
Non
KakaoTalk
1
1
1
1
54.6
25.16
75.12
50.37
Coms.
Facebook Vintage Red Story Subtotal Marble of All Cookie Run Gray City Subtotal YouTube African TV for LGU + Genie Music Subtotal Android Browser Seoul Bus NAVER Subtotal Go Locker Camera Gallery 3D
3 23
3 10
4 23
3 10
50 167 158
29 62 69
50 167 158
42 65 32
15 34 140
42 65 32
15 34 140
4 33 9
2 71 6
4 33 9
2 71 6
2 15 10
174 33 13
2 15 10
174 33 13
8.75 2.50 62.1 0.32 0.01 0.02 7.2 1.80 0.65 0.04 12.0 13.59 0.19 6.42 33.7 0.01 0.69 0.30
29.05 1.20 111.8 7.50 0.19 0.08 22.6 5.38 0.16 0.59 88.2 43.03 0.19 13.67 77.5 5.88 0.76 3.13
10.26 3.09 86.0 3.92 0.02 0.16 46.3 15.15 31.41 0.80 89.3 22.54 0.14 14.84 61.9 0.01 0.27 0.14
Social Social
29 67 74
27.34 0.94 98.2 0.71 0.07 0.03 3.7 0.33 0.05 0.62 21.0 20.05 0.61 5.71 37.8 29.05 1.55 3.74 44.0
15.4
17.9
15.7
Subtotal
Non
Original Google category
Casual Game Casual Game Puzzle Media and Video Media and Video Music and Audio Comms Travel Books and Reference Decoration Photo Photo
treatment of patients in the domain of long-term treatment for disorders such as addictions, diabetes, and obesity [8, 9]. Thus, we believe that an ICT system will be very useful in treating smartphone addiction since the smartphone itself is the very target of monitoring and controlling. In this paper, we will demonstrate this belief using the SAMS program from both IT technical and clinical aspects.
time is not very clear in smartphone usage. For example, when a user used a smartphone 30 times in an hour and each time for 10 s, do we calculate it as 300 s or 1 h? Though in our study, the use count predicted the addiction score relatively well, better measures should be devised.
Principal results
We can find various addiction-related applications in smartphone app stores. A comprehensive survey on ICTbased health applications was presented in [10, 15]. In addition, many researchers and clinicians have developed and tried various mobile-assisted health care systems, which have similarities in system architecture and enabling technologies. Most systems have mobile-side-only systems, which provide only the self control function, and a few provide more comprehensive service together with network service. Tw o s p e c i f i c a p p l i c a t i o n s , ‘ A - C H E S S ’ a n d ‘Netaddiction’[16], need to be mentioned here because we used them as references for designing the SAMS. A-CHESS is a smartphone-based self-management system for preventing the relapse of alcohol addicts. A-CHESS follows Witkiewitz and Marlatt’s model [17] and provides methods to prevent
Through the system operation verification and the pilot data study, we fully verified the reliability of the SAMS and also showed examples of its efficacy. Daily use time, daily use count, and location data were all correctly measured for SAMS clients, and archived and analyzed for clinicians in the SAMS system. From our pilot trial and its statistical analysis, we obtained two research questions. First, the subscale score in the K-SAS test played a crucial role, especially for classifying the subjects in the potentially addicted range. The correlation between the total K-SAS score and use patterns was less distinct than when the sub condition was applied. Second, the pilot statistical result on correlation of the use time, use count and K-SAS sores left us with a significant question. The definition of use
Comparison with prior work
1, Page 10 of 10
relapses in each relapsing stage. We adopted the stage model concept and features unique in smartphones. Since 1995, Young has developed and maintained an online consultation service for intervention for Internet addicts, (http://www. netaddiction.com). The website provides education, support, and treatment to people concerned about Internet addiction. Individual clients requesting treatment are screened for Internet addiction using the IAT (Internet Addiction Test). Those classified for inclusion in the study completed an intake counseling form administered during the initial session that evaluated information related to compulsive use of the Internet as part of this study. Once an appointment had been scheduled, sessions were conducted between the client and the principal investigator. The SAMS is designed based on the treatment model of the two prior systems, and includes monitoring, transmission, archiving, analysis, and feedback procedures. Due to little requirement for the user’s interaction in assessment and treatment, most subjects have been using the SAMS even after the experiment periods. Limitations and further works The data presented in this study is not yet enough for the purpose of medical diagnosis analysis. The sample is very small and not bias-controlled. There is a lack of a control group or comparison condition. Data acquisition and analysis is to be conducted with a larger randomized control. The current system includes only limited data analysis functions and intervention methods. Advanced diagnosis techniques using statistical analysis and data mining, and carefully developed treatment mechanisms from prior research are to be included. Moreover, the investigation should conduct outcome studies of the included intervention methods for effective treatment management. Despite of all these limitations, the pilot study shows the accuracy of assessed data and efficacy of an ICT based system in smartphone addiction. Though the objective measures, using SAMS can importantly contribute to the assessment and diagnosis of addictive disorders, but it cannot be a goal to eliminate questionnaires. The diagnosis should be carefully conducted by clinicians, considering all the available measurement from questionnaires and objective measurement.
J Med Syst (2014) 38:1 Acknowledgments This study was supported by a grant of the Korea Healthcare Technology R&D Project, Ministry for Health Welfare, the Republic of Korea (A120157). Conflicts of interest None declared.
References 1. Zheng, P., Lionel, N., Smart phone and next generation mobile computing, Morgan Kaufmann. 2010. 2. Porter, G. Alleviating the dark side of smart phone use. In: Technology and Society (ISTAS), 2010 I.E. International Symposium; June 7–9, Rutgers. Conference Publications 435–440, 2010. 3. Kim, D. I., Chung, Y. J., Lee, J. Y., et al., Development of smartphone addiction proneness scale for adults: Self-report. Korean J. Couns. 13(2):629–644, 2012. 4. Kwon, M., Lee, J. Y., Won, W. Y., et al., Development and validation of a smartphone addiction scale (SAS). PLoS One 8(2):e56936, 2013. 5. Douglas, A. C., Mills, J. E., Niang, M., et al., Internet addiction: meta-synthesis of 1996–2006 quantitative research. Comput. Hum. Behav. 24(6):3027–3044, 2009. 6. Griffiths, M., Does Internet and computer addiction exists? some case study evidence. Cyber-Psychol. Behav. 3(2):211–218, 2000. 7. Young, K. S., Internet addiction: symptoms, evaluation, and treatment. In VandeCreek, L., and Jackson, T. (eds.), Innovations in Clinical Practice: A Source Book, 17:19–31. 8. Berg, M., Arts, J., and Lei, J., ICT in health care: Socio-technical approaches. Methods Inf. Med. 42(4):297–301, 2003. 9. Gustafson, D. H., Boyle, M. G., Shaw, B. R., et al., An e-health solution for people with alcohol problems. Alcohol Res. Health 33(4): 327–337, 2011. 10. Brady, P., Android anatomy and physiology. In Google IO developer conference. 2008. 11. Stefan, T., and Vinoski, S., Node. js: Using JavaScript to build highperformance network programs. Internet Comput. IEEE 14(6):80–83, 2010. 12. Pautasso, C., Zimmermann, O., Leymann, F., Restful web services vs. big web services: Making the right architectural decision. In Proc. of ACM the 17th Int’l conference on World Wide Web: 805–814. 2008. 13. Cattell, R., Scalable SQL and NoSQL data stores. ACM SIGMOD Record39, no. 4, 12–27, 2011. 14. Dean, J., and Ghemawat, S., MapReduce: simplified data processing on large clusters. Commun. ACM 51(1):107–113, 2008. 15. Bostock, M., Ogievetsky, V., and Heer, J., D3 data-driven documents. IEEE Trans. Vis. Comput. Graph. 17(12):2301–2309, 2011. 16. Dennison, L., Morrison, L., Conway, G., Yardley, L. Opportunities and challenges for smartphone applications in supporting health behaviors change: qualitative study. J. Med. Internet Res. 15(4) 2013. 17. Witkiewitz, K., and Marlart, G. A., Relapse prevention for alcohol and drug problems: that was Zen, this is tao. Am. Psychol. 59(4):224– 235, 2004.
ORIGINAL PAPER
The SAMS: Smartphone Addiction Management System and Verification Heyoung Lee & Heejune Ahn & Samwook Choi & Wanbok Choi
Received: 7 September 2013 / Accepted: 23 December 2013 / Published online: 7 January 2014 # Springer Science+Business Media New York 2014
Abstract While the popularity of smartphones has given enormous convenience to our lives, their pathological use has created a new mental health concern among the community. Hence, intensive research is being conducted on the etiology and treatment of the condition. However, the traditional clinical approach based surveys and interviews has serious limitations: health professionals cannot perform continual assessment and intervention for the affected group and the subjectivity of assessment is questionable. To cope with these limitations, a comprehensive ICT (Information and Communications Technology) system called SAMS (Smartphone Addiction Management System) is developed for objective assessment and intervention. The SAMS system consists of an Android smartphone application and a web application server. The SAMS client monitors the user’s application usage together with GPS location and Internet access location, and transmits the data to the SAMS server. The SAMS server stores the usage data and performs key statistical data analysis and usage intervention according to the clinicians’ decision. To verify the reliability and efficacy of the developed system, a comparison study with survey-based screening with the K-SAS (Korean Smartphone Addiction Scale) as well as self-field trials is performed. The comparison study is done using usage data from 14 users who are 19 to 50 year old adults that left at least 1 week usage logs and completed the survey questionnaires. The field trial fully verified the accuracy of the time, location, and Internet access information in the usage measurement and the reliability of the system operation over more than 2 weeks. The comparison study showed that daily use count has a strong H. Lee : H. Ahn (*) : W. Choi Seoul National University of Science and Technology, Seoul, Republic of Korea e-mail: [email protected] S. Choi Department of Psychiatry, Gangnam Eulji Hospital, Eulji University and Eulji Addiction Institute, Seoul, Republic of Korea
correlation with K-SAS scores, whereas daily use times do not strongly correlate for potentially addicted users. The correlation coefficients of count and times with total K-SAS score are CC=0.62 and CC =0.07, respectively, and the t-test analysis for the contrast group of potential addicts and the values for the non-addicts were p=0.047 and p=0.507, respectively. Keywords Objective assessment . Smartphone addiction . Statistical analysis . ICT system
Introduction During the last decade, smartphones have gained popularity all over the world. In July of 2013, over 50 % of mobile subscribers in the US and over 65 % of mobile subscribers in South Korea were using smartphones, and the percentage keeps increasing. The key differences between smartphones and previous mobile phones are full-featured Internet access and easy installation of new applications through modern OS platforms and app store. Hence, smartphones are now considered handheld computers rather than traditional phones [1]. Although smartphones have given enormous convenience to our lives, pathological use of smartphones has created a new mental health concern among the community [2–4]. Behavior addictions, including smartphone addiction, are often difficult to define since they are related not only to physical, but also to psychological and social aspects [5–7]. Based on the definition of Internet addiction in the literature, we define smartphone addiction as the excessive use of smartphones that interferes with the daily lives of the users. In addition, it has various clinical features, such as tolerance, withdrawal symptoms, salience, mood modification, craving, loss of control, etc. Although smartphone addiction is a new behavior addiction, much more research data is still needed [6]. Nevertheless, it is prevalent worldwide, and it is causing concern to the society.
1, Page 2 of 10
The traditional clinical approach for screening and assessment in behavior addictions has been surveys and interviews. However, there are a massively large number of subjects in smartphone addiction. Furthermore, continuing intervention would be essential for daily-life treatment of addicted people. This paper presents the SAMS (Smartphone Addiction Management System), a comprehensive ICT system for objective assessment-based diagnosis and intervention for smartphone addiction. We describe the system requirements, architectures, and typical use scenarios of the smartphone addiction management system. Then, we describe the system verification methods: operation checking by manual records, and through a comparison study with a pilot test. The correlation between the usage measure and paper-based addiction scale is also examined. Finally, we discuss the future improvement of the SAMS, together with comparison with related ICT systems.
The SAMS: Smartphone addiction management system
J Med Syst (2014) 38:1 Table 1 Functions of SAMS system Functions
Description
Monitoring
- Monitoring smartphone usage for 24 h, without interruption, unnoticed by smartphone users. - Recording items including app use time, duration, location, and relevant information, such a URL access location. - Usage data is recorded in both the smartphone (for a fixed duration) and the server database (for a long term). - Data should be stored in an efficient format for the data analysis. - The assessment data is stored in a secure fashion. - Providing individual and group data analysis. - Time and trend pattern visualization. - Using a system architecture that can easily extend new analysis and visualization. - Fostering self-control. - Manual intervention by clinicians and automatic rule based intervention.
Data Archive
Data Analysis
Intervention and treatment
System requirements The key goals of the SAMS are to provide a comprehensive framework for an objective data-driven study for addiction study and treatment, specifically, to itemize the assessment metrics, analysis algorithm, and treatment strategies. We developed the system requirements starting from Young’s clinical guideline of Internet addiction [8] and Gustafon et al’s relapse prevention model in A-CHESS [9]. Then the unique properties and capabilities of smartphones, such as their mobility and 24-h-companionship, are included for system requirements. The system requirements are reviewed by the coworking national project team from St. Mary Hospital, Ulji Hospital, and Seoul National University Hospital, South Korea. Table 1 summarizes the key functionality for each stage of the SAMS. The first function is to effectively monitor and store the key usage data for usage patterns. The second function is to provide analysis tools for evaluation of usage data to diagnose symptoms. The last is to deliver intervention and provide a self-control mechanism. System architecture Figure 1 illustrates the overall system architecture and workflow of the SAMS framework. On the client side, the SAMS application continuously monitors applications in use, and stores the usage records locally. Users may view the locally stored records for self-recognition and control. Periodically, the new records are transmitted to the SAMS server via the Internet. The usage records are archived in the SAMS server’s database, and statistical and data mining analysis are performed on the data for the diagnosis and treatment to be
conducted by clinicians. Clinicians can determine feedback actions, such as requesting current condition check survey or updating the usage limit time table for a specific application. While in the early stage of studying smartphone addiction, most studies focus on assessment and evaluation of the extent of addictive use of smartphone. We also aimed to develop treatment techniques and study the outcome of their efficiency. Young [7] suggested possible treatment techniques for Internet addicts: (a) practice the opposite time in Internet use (for example, to shift the Internet use time from evening to morning, or reverse), (b) employ external stoppers, (c) set goals, (d) abstain from a particular application, (e) use reminder cards, (f) develop a personal inventory, and (g) enter individual therapy or a support group. The usage-abstaining function and individual therapy are included in the SAMS application. For example, when a user is playing a special game application for more than the designed usage time, the SAMS client will pop up a warning message and block the user from the game application. The usage restriction function is especially designed for elementary to high-school students for preventing overuse in classrooms, and will only be used for the contracted users. Client system design The SAMS client application is developed on Google Android based smartphones. In the Android system, a component of an application has to be one of four Java classes [10]: ‘Activity’ is for user interaction and output display, ‘Service’ is for a long running background service, ‘ContentProvider’ is for persistent data storage, and ‘BroadcastReceiver’ is for waiting for
J Med Syst (2014) 38:1
Page 3 of 10, 1
Messaging Server (GCM)
Intervention
Messaging
Messaging
Usage Monitoring
Data Storing
Data Transmission
Data Receive
feeback
Decision (Clinicians)
Clinician
Smartphone User
Data Visualization
SAMS client (Google Android)
Data Receive Data Archive
Data Analysis
SAMS server
Fig. 1 System architecture and workflow of SAMS
system wide events. The proposed SAMS client application is composed of eight activities, two services, one content provider, and three broadcast receivers. Due to space limitations, we describe only the operation of the usage monitoring mechanism, which is the key function for our study. First, the foreground application in the Google Android system can be obtained from the system service ‘android.app. ActivityManager’. The monitoring service periodically checks the changes in the foreground application and records its start time, end time, GPS location, and URL if the application is a browser. The GPS location can be obtained from the Android service, ‘android.location. LocationManager’, and the current URL can be obtained from t h e A n d r o i d b r o w s e r h i s t o r y c o n t e n t p r o v i d e r, ‘android.provider. Browser’. Then the usage data is first stored in local ‘SQLite’ database, and then transmitted to the server whenever Internet access, either WiFi or cellular data, is available. Taking into account the LCD display device’s status, the SAMS client does not include the duration of the time the LCD display is off. Special consideration has been taken to maintain the periodic checking process in the Android system without interruption due to its memory limitation and process kill mechanism. In the current implementation, the system sets up the Android system alarm service to send ‘PendingIntent’ periodically every 3 s. We compared this ‘Alarm service plus PendingIntent’ method with an alternative, ‘Service with TimerTask’ method, and finally chose the proposed one because of its reliability (not unintentionally terminated) and low power-consumption (Fig. 2). Figure 3 shows the UI scenarios for users. The SAMS application runs in the background without the user being aware of it. On the first launch after installation (Fig. 3a), the SAMS application gets its GCM (Google Cloud Messaging) registration ID from the Google GCM server, and then registers the ID to the SAMS server for use in messaging service later. When the usage button is pressed, the usage profiles for
applications are presented (Fig. 3b). The user can examine the daily usage pattern of an application during the most recent week (Fig. 3c). A survey view pops up when a clinician asks the current status of the user (Fig. 3d). A blocking application view intercepts the application when the time is in the range forbidden by clinicians (Fig. 3e). Figure 3e is the setting view of the SAMS application which is only for developers. Server system design Figure 4 illustrates the architecture of the SAMS server system. The server receives and stores the data (user profile and client system information) from SAMS clients, performs data analysis/mining, and finally provides graphical and numerical information for the clinicians for analysis, diagnosis and treatment. For basic treatment strategies, the SAMS client is
Fig. 2 Component Structure of the SAMS Android application
1, Page 4 of 10
J Med Syst (2014) 38:1
Fig. 3 SAMS client application user interfaces: a is the launch view, i.e., main view of SAMS app. b shows the app list in the decreasing order of total usage time. c shows the daily usage time graph during 7 most recent days for a clicked app in (b). d shows the survey view pop up on demands or periodically. e shows a blocking view pop-up for preventing from using an app. f is the configuration view for SAMS server’s URL and monitoring period
a
Main View
b app usage time summary
d short status survey view requested through GCM messaging to update the limitation time table or prompt the current status survey of the user. The system architecture of the SAMS server follows the MVC (model-view-control) pattern, and uses the express.js framework of the Node.js web architecture [11]. First, the database schema is designed for reflecting the user profile, data from users (including use records), the installed packages for each user, the social relationship between users, and forbidden application data for each user. Second, the view and transport part is implemented following RESTful [12] methods, i.e., using HTTP and JSON. The data visualization is done using a D3 [13] Javascript- based framework. Lastly, because NoSQL DB [15] is used for the SAMS server, the control-data analysis engine implements the statistical analysis function, built on map-reduce flows [14]. When the SAMS server is requested for analysis results from the browser (clinicians), the data analysis engine processes the data records and transmits the available data. The browser draws the result data using the D3 framework. The current SAMS server provides two groups of data analysis functions, individual analysis and group statistics analysis. The individual analysis functions are used to
e app blocking view
c daily usage history for an app
f
app setting view
examine and diagnose the individual use patterns. Figure 5 shows the individual usage analysis and control pages. The daily usage histogram page (Fig. 5a) shows which application was used and how long it was used on average in a given period. The location information is visualized in the tracking information page using a Google map overlay (Fig 5b). When required (for example, to see what time and which app was used, which is not included in our analysis page yet), the clinician further examines the usage log records (Fig. 5c); each record contains the start time, end time, location, and optional URL information. Even though we did not use it in this study, the clinician can use the application control page to block an application’s usage on a weekly basis (Fig. 5d). For example, in the Fig. 5d, the specific application is marked as orange sections in the weekly schedule table. The update event is sent through the GCM server to the user’s smartphone. Then the smartphone retrieves the updated blocking schedule from the SAMS server. Whenever a new application is launched, the SAMS client will check the permission and block it if it is not allowed. The group statistics analysis is illustrated in Fig. 6, where the correlation between use time/use count and survey
J Med Syst (2014) 38:1 messaging interface
Page 5 of 10, 1 client interface
clinician interface
data visualization (d3)
GCM
data analysis engines
Restful (Json)
webservice (express.js)
DB API (mongoose) server framework (node.js)
DB users usages (mongoDB) Apps (other data)
Fig. 4 System Architecture of SAMS Server: SAMS severs is based on V8, the recently popular javascript engine. Node.js and express.js frameworks are used for building services, control and data analsysis. The interface with the client is defined with RESTful (REpresentation State Transfer) methods, i.e., using HTTP and JSON (JavaScript Object Notation). The data visualization is done using a D3 (Data-Driven Documents). NoSQL DB is used for data storage due to its rich features for data mining and data aggregation
meters from the real locations due to the properties of GPS sensors in commercial smartphones. We believe this much inaccuracy makes the location information useless for our purpose. The extra power consumption is also practically important since the users would not keep our application if they feel that the time between battery recharges is shortened due to the SAMS application. Considering the trade-off between the power consumption and measurement accuracy, a measurement period of 3 s was chosen for deployment. Using ‘GO battery saver’ utility app, the power consumption of SAMS client is measured to be less than 2 % in general user patterns. For the compatibility issue, we tested SAMS clients on as many Android phones as available, which included major models in South Korea and Android versions 2.2, 2.3, 3.0, 4.0 and 4.1 as in Table 2. We updated some implementations for better compatibility. We verified that all functions worked correctly for our specifications, except for the graphic chart package (‘achartengine’) which did not work due to the Android version compatibility problem.
Experiment and data validation Benchmark method
addiction scores (to be defined in the next section) is shown. In the plots, the vertical axis denotes the total K-SAS score, and the horizontal axis denotes the daily use time and count. The lines are linear regressions fitted to data: red lines for females, blue ones for males, and black ones for both. The data analysis itself will be discussed in the following section. System verification We considered three aspects in verifying the SAMS system. First, the correctness of the monitoring data and analysis results. Second, the reliability of the system operation and power consumption. Last, the proper operation of core function on various Android versions in the market. The verification of the accuracy and reliability of input data is crucial for diagnosis and statistical analysis. Note that the SAMS client application has to run 24 h and maintenance efforts have to be minimized. For these purposes, the SAMS application should restart automatically after the smartphone is re-booted or process-killed in a low memory condition. The broadcast receiver and service mechanisms were carefully selected and implemented. For testing data correctness and reliability, the developers and trial users use the apps for a couple of weeks, making notes for app usages and location, and then comparing the usage records with the notes. In this test period, we optimized system parameters such as the sampling period and reporting periods. We observed that even in a rare case, the GPS data was deviated up to several hundred
In order to evaluate the reliability of usage data from the SAMS, we performed a comparison study with a surveybased smartphone addiction test. The survey-based scales are not always reliable and need to be confirmed by clinicians’ close interview. However, we assume the validity of survey results for evaluating the efficacy of the obtained data from the SAMS system. We used the K-SAS (Korean Smartphone Addiction Scale) [2] because it reflects the Korean society very well. A recently published smartphone addiction scale is also based on the KSAS [3]. The K-SAS is composed of 15 question items in Table 3. The 15 questions are organized into four groups based on the factors. Group 1 questions (1, 5, 9, and 12) are for ‘daily-life disturbance’, group 2 questions (2, 6, and 8) are for ‘virtual life orientation’, group 3 questions (3, 7, 10, and 14) are for ‘withdrawal’, and group 4 questions (4, 8, 11, and 15) are for ‘tolerance’. As answer choices for each question, 4 level Likert items, ‘strongly disagree’,’ disagree’, ‘agree’ and ‘strongly agree’ were used. For confirming the validity of answers, we used the reverse-question 8, 10, and 13. Answers to the non-reverse questions are their values in the Likert scale, while answers to reverse questions are 1 minus their Likert scale values. The sum value of all questions is called the total K-SAS scale, and sum values of each subgroup are called group N’s subscales (i.e., G1, G2, G3, and G4). The diagnosis criteria is illustrated by the decision flow chart in Fig. 7. The total scale and subscales are used to determine the
1, Page 6 of 10
a
b
J Med Syst (2014) 38:1
c
daily usage profiles for a user (in the decreasing order of use time)
geographical usage pattern of a user (the numbers are the usage counts for the locations)
d
detail usage log of a user (app package name, start time, duration, GPS location, url)
week time table for an app of a user (orange is the allowed time; grey is for the unallowed time)
Fig. 5 SAMS’s iindividual analysis server clinician interface (the daily usage histogram, usage location graph, usage record details, and usage control page)
highly addicted, the potentially addicted, and the normal (i.e., not-addicted). The subject is determined to be highly addicted when the total scale is over 45, subscale of daily life disturbance is over 16, subscale of withdrawal is over 13, or subscale of tolerance is over 14. The subject is determined to be potentially addicted when total scale is between 42 and 45, subscale of daily life disturbance is over 14, subscale of withdrawal is over 12, or subscale of tolerance is over 13. In the other cases, the subject is determined to be a non-addict. Participant recruit and data acquisition The SAMS client application was uploaded to the SAMS server and SMS messages to request install SAMS app were sent to about 120 anonymous users that were from 4 persons (lab members, including the SAMS developers) phone number list. All recipients were requested to respond to the K-SAS test online, and to download and install the SAMS client application. No benefits were given for joining this test. Among 120 recipients, about 40 installed the SAMS application. However, for data analysis, we used data from 14 users
that had more than 1 week usage log and finished the K-SAS survey faithfully (we checked the consistency of reverse questions). Reasons for unfaithful answers or cessation of application use were not studied carefully. The 14 users are 7 males and 7 females, and the age ranges from 18 to 51 (average of 26.57, standard deviation of 8.25). The main focus of the paper is to introduce a new ICT system, not medical interpretation of the experimental data. We took into account the possible ethical and privacy problems. An IRB approval was not obtained for this verification test. However, the study participants provided informed consents, and all the assessed data are protected on a secured database. We will apply the IRB approval for real scale medial analysis. Statistical analysis The main goal of this trial test was to check and understand the relation between the basic usage statistics and the K-SAS score regarding the K-SAS scale as the true addiction level. The group analysis in Fig. 6 showed the correlation graphs of the usage statistics and the K-SAS score. The correlation
J Med Syst (2014) 38:1
Page 7 of 10, 1 Table 3 K-SAS survey questions Factor Group
Daily life disturbance (G1)
Fig. 6 SAMS’s group analysis functions (top to bottom: plots of (daily use time, K-SAS score comparison) and (daily use counts, K-SAS value))
analysis with the total scale showed no correlation (CC=0.07) for daily use time, but a significant correlation (CC=0.62) for daily use count and the total scale. However, the t-test analysis after applying the sub-group criterion revealed p=0.507 for daily use time and p=0.047 for daily use count. The result implies that the sub-condition in addiction classification is important in distinguishing the potential addicts from the
Table 2 Tested Systems for the SAMS client Android version
Models (Release Year)
2.2 2.3
HTC Desire (2010), SKY Vega S (2011) HTC Desire (2010), LG Optimus One (2010), SKY Vega S (2011) Samsung Galaxy Tab 10.1 (2011) Samsung Galaxy S3, LG optimus G (2012)
3.1 4.0
Question
Having performance degradation in class or office due to smartphone use. Using smartphone when I am not supposed to (in classroom, during meeting, etc). The people around me tell me that I use the smartphone too much. Having difficulties in concentrating on job/study due to the smartphone use. Smartphone use has nothing to do with my performance in class or office. Virtual world Feeling empty when not using my orientation smartphone. (G2) Feeling more pleasant or excited while using smartphones than while being with friends or family. Withdrawal Won’t be able to stand not having a (G3) smartphone. Feeling depressed, anxious, or oversensitive when I am not able to use my smartphone. Not feeling impatient when not holding my smartphone. Not being able to work/study without my smartphone. Tolerance (G4) Having tried to shorten smartphone use time but fails all the time. Keeps on using smartphone while thinking about stopping it. Using my smartphone longer than I had intended. Not spending much time using smartphone.
Question num 1
5
9 12
15
reverse
2 6
8 14
4
reverse
11 7 3 13 10
reverse
non-addicts. From the fact that social network and chatting applications have frequent but short usage time, these applications can be considered as key addiction applications. This suggests that application categorization should be considered for smartphone addiction. To examine the details of application categories, we merged the categories of apps in GooglePlay, the Android app store into 5 ones, social network, games, entertainment, web-browsing, and others. The data shows that the potentially addicts use more social network applications, such as Kakao talk and Facebook than the non-addicts. Kakao talk is the most popular messaging and chatting applications and we categorized it into ‘social network’ in spite of its original category of communication in Google app store which is ‘communications’Tables 4 and 5. We also examined the location of smartphone use. The potentially addicted were asked about most frequent use
1, Page 8 of 10
J Med Syst (2014) 38:1
Fig. 7 K-SAS diagnosis criteria flow charts: highly addict when total K-SAS score>45, G1>16, G3 >13, or G4 >14; potentially addict when total K-SAS score> 42, G1>14, G3>12, or G4>13; otherwise non addict
Total score, G1, G3, G4
next
yes
total score > 42
yes
total score > 45 no
no yes G1> 16
yes G1> 14
no
no yes
yes
G3> 13
G3> 12
no G4> 14
next
locations, and in this study, all three cases were their homes. This implies that detailed statistical study is required to reveal the relation between family environment and smartphone addiction level.
Table 4 Contrast group comparison: daily usage count and use time Contrast group
total K-SAS score (subscales)
daily use time (mins)
daily use count
The potential addicts
52 36 (G1=14, G3=13) 32 (G3=14) Average 37 31 29
273 464 369 369 381 193 148
318 244 137 233 98 95 94
28 27 25 34 26 17 24 20 Average
159 277 317 442 101 179 469 626 299
93 215 212 253 98 86 151 113 137
The non addicts
no yes
yes G4> 13
addict
non- addict
potentially addict
Discussion and conclusion The traditional clinical approach for screening and assessment in behavior addictions has been surveys and interviews, and this also applied to smartphone addiction. However, the survey approach has serious drawbacks and needs to be complemented by objective assessment analysis. First, in interview-based methods, it is difficult to follow the change of subjects’ status, especially when there is a massively large number of subjects as in smartphone addiction. Furthermore, smartphone addiction exhibits a chronically relapsing nature as do other behavioral addictions. Continuing intervention is essential for daily life treatment of addicted people. Second, the survey questions are often subjective and the replies from those surveys depend upon their seriousness, mood, and attitude to the survey. For example, the frequently used Likert score of “strongly agree”, “agree”, “disagree”, and “strongly disagree”, is apt to be subjective, and people who overuse their smartphones may underestimate their own use. Hence, objective and quantitative usage needs to be recorded to perform diagnosis and treatment based on concrete and accurate data. To address these problems, we developed the SAMS (Smartphone Addiction Management System), a comprehensive ICT system for objective assessment-based diagnosis and intervention for smartphone addiction. It has already been established that continuing ICT-based intervention helps in
J Med Syst (2014) 38:1
Page 9 of 10, 1
Table 5 The usage statistics of application categrories: potentially addicts and non addicts Category
Social Network
Game
Entertainment
Browsing
Others
App name
Rank by usage count
Rank by usage time
Daily usage count
Daily usage time (min)
Addicts
Non
Addicts
Non
Addicts
Addicts
Non
KakaoTalk
1
1
1
1
54.6
25.16
75.12
50.37
Coms.
Facebook Vintage Red Story Subtotal Marble of All Cookie Run Gray City Subtotal YouTube African TV for LGU + Genie Music Subtotal Android Browser Seoul Bus NAVER Subtotal Go Locker Camera Gallery 3D
3 23
3 10
4 23
3 10
50 167 158
29 62 69
50 167 158
42 65 32
15 34 140
42 65 32
15 34 140
4 33 9
2 71 6
4 33 9
2 71 6
2 15 10
174 33 13
2 15 10
174 33 13
8.75 2.50 62.1 0.32 0.01 0.02 7.2 1.80 0.65 0.04 12.0 13.59 0.19 6.42 33.7 0.01 0.69 0.30
29.05 1.20 111.8 7.50 0.19 0.08 22.6 5.38 0.16 0.59 88.2 43.03 0.19 13.67 77.5 5.88 0.76 3.13
10.26 3.09 86.0 3.92 0.02 0.16 46.3 15.15 31.41 0.80 89.3 22.54 0.14 14.84 61.9 0.01 0.27 0.14
Social Social
29 67 74
27.34 0.94 98.2 0.71 0.07 0.03 3.7 0.33 0.05 0.62 21.0 20.05 0.61 5.71 37.8 29.05 1.55 3.74 44.0
15.4
17.9
15.7
Subtotal
Non
Original Google category
Casual Game Casual Game Puzzle Media and Video Media and Video Music and Audio Comms Travel Books and Reference Decoration Photo Photo
treatment of patients in the domain of long-term treatment for disorders such as addictions, diabetes, and obesity [8, 9]. Thus, we believe that an ICT system will be very useful in treating smartphone addiction since the smartphone itself is the very target of monitoring and controlling. In this paper, we will demonstrate this belief using the SAMS program from both IT technical and clinical aspects.
time is not very clear in smartphone usage. For example, when a user used a smartphone 30 times in an hour and each time for 10 s, do we calculate it as 300 s or 1 h? Though in our study, the use count predicted the addiction score relatively well, better measures should be devised.
Principal results
We can find various addiction-related applications in smartphone app stores. A comprehensive survey on ICTbased health applications was presented in [10, 15]. In addition, many researchers and clinicians have developed and tried various mobile-assisted health care systems, which have similarities in system architecture and enabling technologies. Most systems have mobile-side-only systems, which provide only the self control function, and a few provide more comprehensive service together with network service. Tw o s p e c i f i c a p p l i c a t i o n s , ‘ A - C H E S S ’ a n d ‘Netaddiction’[16], need to be mentioned here because we used them as references for designing the SAMS. A-CHESS is a smartphone-based self-management system for preventing the relapse of alcohol addicts. A-CHESS follows Witkiewitz and Marlatt’s model [17] and provides methods to prevent
Through the system operation verification and the pilot data study, we fully verified the reliability of the SAMS and also showed examples of its efficacy. Daily use time, daily use count, and location data were all correctly measured for SAMS clients, and archived and analyzed for clinicians in the SAMS system. From our pilot trial and its statistical analysis, we obtained two research questions. First, the subscale score in the K-SAS test played a crucial role, especially for classifying the subjects in the potentially addicted range. The correlation between the total K-SAS score and use patterns was less distinct than when the sub condition was applied. Second, the pilot statistical result on correlation of the use time, use count and K-SAS sores left us with a significant question. The definition of use
Comparison with prior work
1, Page 10 of 10
relapses in each relapsing stage. We adopted the stage model concept and features unique in smartphones. Since 1995, Young has developed and maintained an online consultation service for intervention for Internet addicts, (http://www. netaddiction.com). The website provides education, support, and treatment to people concerned about Internet addiction. Individual clients requesting treatment are screened for Internet addiction using the IAT (Internet Addiction Test). Those classified for inclusion in the study completed an intake counseling form administered during the initial session that evaluated information related to compulsive use of the Internet as part of this study. Once an appointment had been scheduled, sessions were conducted between the client and the principal investigator. The SAMS is designed based on the treatment model of the two prior systems, and includes monitoring, transmission, archiving, analysis, and feedback procedures. Due to little requirement for the user’s interaction in assessment and treatment, most subjects have been using the SAMS even after the experiment periods. Limitations and further works The data presented in this study is not yet enough for the purpose of medical diagnosis analysis. The sample is very small and not bias-controlled. There is a lack of a control group or comparison condition. Data acquisition and analysis is to be conducted with a larger randomized control. The current system includes only limited data analysis functions and intervention methods. Advanced diagnosis techniques using statistical analysis and data mining, and carefully developed treatment mechanisms from prior research are to be included. Moreover, the investigation should conduct outcome studies of the included intervention methods for effective treatment management. Despite of all these limitations, the pilot study shows the accuracy of assessed data and efficacy of an ICT based system in smartphone addiction. Though the objective measures, using SAMS can importantly contribute to the assessment and diagnosis of addictive disorders, but it cannot be a goal to eliminate questionnaires. The diagnosis should be carefully conducted by clinicians, considering all the available measurement from questionnaires and objective measurement.
J Med Syst (2014) 38:1 Acknowledgments This study was supported by a grant of the Korea Healthcare Technology R&D Project, Ministry for Health Welfare, the Republic of Korea (A120157). Conflicts of interest None declared.
References 1. Zheng, P., Lionel, N., Smart phone and next generation mobile computing, Morgan Kaufmann. 2010. 2. Porter, G. Alleviating the dark side of smart phone use. In: Technology and Society (ISTAS), 2010 I.E. International Symposium; June 7–9, Rutgers. Conference Publications 435–440, 2010. 3. Kim, D. I., Chung, Y. J., Lee, J. Y., et al., Development of smartphone addiction proneness scale for adults: Self-report. Korean J. Couns. 13(2):629–644, 2012. 4. Kwon, M., Lee, J. Y., Won, W. Y., et al., Development and validation of a smartphone addiction scale (SAS). PLoS One 8(2):e56936, 2013. 5. Douglas, A. C., Mills, J. E., Niang, M., et al., Internet addiction: meta-synthesis of 1996–2006 quantitative research. Comput. Hum. Behav. 24(6):3027–3044, 2009. 6. Griffiths, M., Does Internet and computer addiction exists? some case study evidence. Cyber-Psychol. Behav. 3(2):211–218, 2000. 7. Young, K. S., Internet addiction: symptoms, evaluation, and treatment. In VandeCreek, L., and Jackson, T. (eds.), Innovations in Clinical Practice: A Source Book, 17:19–31. 8. Berg, M., Arts, J., and Lei, J., ICT in health care: Socio-technical approaches. Methods Inf. Med. 42(4):297–301, 2003. 9. Gustafson, D. H., Boyle, M. G., Shaw, B. R., et al., An e-health solution for people with alcohol problems. Alcohol Res. Health 33(4): 327–337, 2011. 10. Brady, P., Android anatomy and physiology. In Google IO developer conference. 2008. 11. Stefan, T., and Vinoski, S., Node. js: Using JavaScript to build highperformance network programs. Internet Comput. IEEE 14(6):80–83, 2010. 12. Pautasso, C., Zimmermann, O., Leymann, F., Restful web services vs. big web services: Making the right architectural decision. In Proc. of ACM the 17th Int’l conference on World Wide Web: 805–814. 2008. 13. Cattell, R., Scalable SQL and NoSQL data stores. ACM SIGMOD Record39, no. 4, 12–27, 2011. 14. Dean, J., and Ghemawat, S., MapReduce: simplified data processing on large clusters. Commun. ACM 51(1):107–113, 2008. 15. Bostock, M., Ogievetsky, V., and Heer, J., D3 data-driven documents. IEEE Trans. Vis. Comput. Graph. 17(12):2301–2309, 2011. 16. Dennison, L., Morrison, L., Conway, G., Yardley, L. Opportunities and challenges for smartphone applications in supporting health behaviors change: qualitative study. J. Med. Internet Res. 15(4) 2013. 17. Witkiewitz, K., and Marlart, G. A., Relapse prevention for alcohol and drug problems: that was Zen, this is tao. Am. Psychol. 59(4):224– 235, 2004.
