Department of Computer and Information Science
Informationsteknologi
LIU-IDA/LITH-EX-A--13/001--SE
Development and piloting of a fully
automated, push based, extended session
alcohol intervention on university students – a
feasibility study
Marcus Bendtsen
Tutor
Johan Åberg
Upphovsrätt
Detta dokument hålls tillgängligt på Internet – eller dess framtida ersättare – under 25 år från publiceringsdatum under förutsättning att inga extraordinära omständigheter uppstår.
Tillgång till dokumentet innebär tillstånd för var och en att läsa, ladda ner, skriva ut enstaka kopior för enskilt bruk och att använda det oförändrat för ickekommersiell forskning och för undervisning. Överföring av upphovsrätten vid en senare tidpunkt kan inte upphäva detta tillstånd. All annan användning av dokumentet kräver upphovsmannens medgivande. För att garantera äktheten, säkerheten och tillgängligheten finns lösningar av teknisk och administrativ art.
Upphovsmannens ideella rätt innefattar rätt att bli nämnd som upphovsman i den omfattning som god sed kräver vid användning av dokumentet på ovan beskrivna sätt samt skydd mot att dokumentet ändras eller presenteras i sådan form eller i sådant sammanhang som är kränkande för upphovsmannens litterära eller konstnärliga anseende eller egenart.
För ytterligare information om Linköping University Electronic Press se förlagets hemsida http://www.ep.liu.se/.
Copyright
The publishers will keep this document online on the Internet – or its possible replacement – for a period of 25 years starting from the date of publication barring exceptional circumstances.
The online availability of the document implies permanent permission for anyone to read, to download, or to print out single copies for his/hers own use and to use it unchanged for non-‐commercial research and educational purpose. Subsequent transfers of copyright cannot revoke this permission. All other uses of the document are conditional upon the consent of the copyright owner. The publisher has taken technical and administrative measures to assure authenticity, security and accessibility.
According to intellectual property law the author has the right to be mentioned when his/her work is accessed as described above and to be protected against infringement.
For additional information about the Linköping University Electronic Press and its procedures for publication and for assurance of document integrity, please refer to its www home page: http://www.ep.liu.se/.
© Marcus Bendtsen
Abstract ... 5
1 Introduction ... 6
1.1 Single session interventions ... 6
1.2 Multi session interventions ... 7
1.3 Advantage of e-‐health interventions ... 8
1.4 Challenges for alcohol e-‐health interventions ... 8
2 Aims ... 10
2.1 Research aims ... 10
2.2 Development aims ... 11
3 Theoretical considerations ... 12
3.1 Definition of push based ... 12
3.2 Intervention design ... 13
3.3 Message content ... 13
3.3.1 Food for thought ... 13
3.3.2 Reflective ... 13
3.3.3 Tasks ... 14
3.3.4 Challenge ... 14
3.4 Existing studies ... 14
4 Technical considerations ... 16
4.1 Service Oriented Architecture (SOA) ... 16
4.2 Multi tier architecture ... 17
4.2.1 Data layer ... 17
4.2.2 Logic layer ... 17
4.2.3 Exposure ... 17
4.3 Short Message Service (SMS) ... 17
4.4 Mobile applications ... 18 4.5 Email ... 19 4.5.1 Content ... 19 4.5.2 Deliverer ... 19 4.5.3 Identification ... 19 5 Methods ... 21 5.1 Study design ... 21 5.2 Technical platform ... 22 5.2.1 Services ... 24 5.3 Data preparation ... 24 5.3.1 Dataset ... 24 5.3.2 General data ... 24 5.3.3 Risky drinkers ... 24
5.3.4 Estimation compared to peers and thoughts of change ... 25
5.3.5 Enrolment data ... 25 5.3.6 Follow-‐up data ... 25 5.4 Correlations ... 25 6 Results ... 26 6.1 Aim 1 ... 26 6.1.1 Enrolment ... 26 6.1.2 Delivery method ... 27 6.1.3 Activation ... 27
6.1.4 Ordinal logistic regression ... 27
6.2.1 Choice of length and frequency ... 29
6.2.2 Satisfaction of length and frequency ... 30
6.3 Aim 3 ... 31
6.3.1 Responses ... 31
6.3.2 Content in general ... 32
6.3.3 Motivational, fact and challenges ... 32
6.3.4 Recommend to a friend ... 33
7 Discussion ... 34
7.1 Technical outcomes ... 34
7.2 Research outcomes ... 34
7.2.1 Is it worth further development and studies ... 34
7.2.2 Same length and frequency for all future participants ... 34
7.2.3 Human-‐computer interaction ... 35 8 Conclusions ... 36 9 Tables ... 37 10 References ... 49 11 Enclosures ... 53 11.1 Enclosure 1 ... 53 11.2 Enclosure 2 ... 54 11.3 Enclosure 3 ... 55 11.4 Enclosure 4 ... 56
Abstract
Alcohol consumption amongst university students in Sweden has repeatedly been measured to be at risky levels for more than 50% of students. Internet based brief interventions aimed to intervene and prevent risky drinking have been developed with some success during recent years. Single session interventions have been implemented into routine practice in Sweden and other countries, however not all risky drinkers benefit from these single session brief interventions.
This feasibility study attempted to develop and pilot an extended session intervention, where participants received messages with motivating content several times a week for a few weeks. All students on semester 1, 3 and 5 at Linköpings Universisty were invited to join a brief single session intervention as part of routine practice, and those who completed the single session intervention were invited to join the new extended intervention.
Out of a total of 11,284 students that were invited to complete the single session
intervention 4916 (%=43.6) responded. Out of these 1216 (%=24.7) decided to enrol to the extended intervention and 898 (%=77.9) completed the follow up questionnaire after the extended intervention. Participants that enrolled to the extended intervention were automatically placed in a draw for one of two iPads.
Issues were found with participants that wanted to receive messages via SMS, as 28.3% didn’t activate their SMS intervention and hence didn’t enrol to the extended
intervention. Furthermore there was some indication that participants exposed to more messages were more positive towards the content, as were participants receiving SMS messages over email message. This might be an indication that email may not be up to par with SMS for delivering this type of intervention.
The study showed that this kind of extended intervention is worthwhile pursuing. Risky drinkers were more likely to find the intervention useful, and a majority of all
participants would possibly or definitely recommend the intervention to a friend that needed help with their alcohol consumption. Future studies should focus on decreasing the number of participants not activating their SMS intervention, experimenting with enrolment without any prize and possibly detached from single session intervention, measuring the effect on alcohol consumption of the intervention as well as identifying any differences between receiving the intervention via email or SMS.
The responsibility of expanding and enhancing the research of fully automated brief interventions lay upon researches from several fields. There is a need of refining the human-‐computer interaction as well as the content and design of the intervention. This cannot occur effectively from a single department but should be a joint venture in order to be cost effective and to utilize expertise.
1 Introduction
Alcohol continues to be a widely spread reason for a number of health problems. An estimated 3.8% of all global deaths and 4.6% global disability-‐adjusted life-‐years are attributed to alcohol. This makes alcohol one of the largest avoidable risk factors for health problems, and the extent to which this is a burden is even greater within younger groups [1]. A worldwide study of heavy drinking, including over 17,000 students in 21 countries, showed that alcohol consumption is common in many countries, but also varies with social and economic factors [2].
A survey of Swedish undergraduate university students (n = 4575) concluded that 96% of students had consumed alcohol in the preceding 12 months, and that 33% had consumed enough to classify as binge drinking twice a month or more often (55% had consumed this quantity at least once a month) [3]. A more recent study of university undergraduate students (n = 2858) showed that more than 50% of all students
consumed enough alcohol to be considered binge drinking once a month or more often [4,5].
Although face-‐to-‐face interventions (i.e. motivating conversation with a health care provider in various settings) have shown to have a reasonable effect, the
implementation of such interventions in routine health care has been hampered by a number of barriers such as time, money, professional skills and not least uninterest from staff to engage in life style interventions [7,8]. The expansion of access to electronic devices with network communication capabilities has prompted a new approach for reaching individuals, including those not offered life style interventions within the health care system. This has led to a rapid growth of Internet based lifestyle interventions during the past decades.
In order to make a clear distinction between the various new, so-‐called e-‐health interventions, the CONSORT-‐EHEALTH statement seeks to clarify the difference between all computerized interventions. The CONSORT-‐EHEALTH statement defines “Fully automated interventions” as: e-‐health interventions delivered without any human
interaction in the intervention method. This makes a clear distinction between
interventions that are fully automated and those who might use the Internet for recruitment to more hands on interventions, such as counselling over the phone. Research in the area of fully automated interventions has evolved around two main strains: single-‐ and multi-‐session interventions [9].
1.1 Single session interventions
Fully automated single session interventions are usually web-‐based self-‐assessment forms with personal feedback, sometimes including normative and safe drinking limits. Participants are lead to the website (either via email, advertisement etc.) and fills out the questionnaire. Once complete the participant is showed his or her personal feedback based on the responses. The effect of the method has been documented to be fairly on par with brief face-‐to-‐face interventions [10,11,12].
Some single session interventions are designed to be proactive [5,10,13,14], where participants are recruited to the intervention through email [10,14]. This allows
healthcare professionals to take an active role in large-‐scale prevention work. This type of system is in use today at 90% of student healthcare services across Sweden. Here a single staff member at the student healthcare centre can, within minutes, send several thousands of emails paired with an invitation to students, thus exposing them to an intervention they didn’t actively seek with a comparable effect to brief face-‐to-‐face conversations [15].
Other single session interventions take a more passive approach [16,17], creating an open website which participants can stumble upon, either through search engines, advertisements or referral from other websites. This passive approach broadens the target group, as there is no single healthcare professional that offers the intervention to his or her clients. However there is some indication that it’s harder to entice participants to start the intervention once they’ve found the website [18].
Whichever method is used, active or passive, fully automated single session
interventions show a lot of promise, and has matured to a level where systems can be implemented into routine practice of healthcare professionals [14].
Furthermore, recent research has opened up for the possibility that self-‐assessment alone, or with minimal feedback, could give a positive effect on participant’s
consumption [12,19]. This could lead to the effect of single session interventions being underestimated, and actually surpassing those of brief face-‐to-‐face interventions.
1.2 Multi session interventions
Fully automated multi session interventions are designed in a manner where
participants are asked to complete several steps on different occasions. This could be a website where participants are asked to login each week to answer questions or fill in a diary.
Studies of multi session interventions aren’t as easily generalized as single session interventions. These type of interventions can include several methods of delivery, ranging from TV, email, SMS, web etc. [20,21]. Studies conducted using multi session websites have shown varying effect [19,22]. A review of both single and multi session interventions show that methods with an extended nature might outperform single sessions [12], however the number of studies are still too few to make any definitive conclusions and also the enclosed studies (n = 9) where very different in their nature and execution.
One issue that multi session interventions suffer from is high attrition rates. Trials where participants are aware of the trial settings, give consent and self-‐report that they are motivated to change still have low adherence. One study showed that only 45% of participants ever accessed the website [17]. The same system was translated to the real world [18], and very few (5%) returned to complete more steps in the program other than the first self-‐assessment.
Using a more active approach could possibly increase adherence, as participants are reminded to complete their intervention or take part in necessary steps [23]. Greater adherence has been observed using SMS technology [20]. Here a motivational and diary keeping SMS service succeeded in keeping nearly 75% of participants throughout the trial, however it should be noted that the adherence rate probably was boosted by the fact that participants knew they were being observed, were given incentives and a limited sampling size. However the study takes steps towards a more proactive type of multi session interventions. Participants were recruited via email and face-‐to-‐face. An interesting observation was that both groups showed similar levels of adherence, making the step of actually talking to participants before the intervention redundant concerning attrition levels.
1.3 Advantage of e-‐health interventions
Individual face-‐to-‐face and group alcohol interventions suffer from similar issues such as participants not adhering to the program as well as high-‐risk groups not being reached [24]. These two issues might be related to phenomenon such as anonymity, geographical, religious or cultural differences. Furthermore face-‐to-‐face and group interventions are more time consuming and therefore naturally more expensive per individual reached. Fully automated interventions have the potential of reaching more people, without the aforementioned phenomenon becoming an obstacle.
Data collected during fully automated interventions show that a great number of risky-‐ drinkers are reached using these methods; giving them access to an intervention they didn’t actively seek. Although there are costs involved in developing and maintaining fully automated systems, these are often justified in the light of the reach and effect the intervention might have. Even passive interventions, where one could argue that the approach might reach less people due to advertisement costs or other recruitment fees, show economic viability even with a modest number of participants [22].
1.4 Challenges for alcohol e-‐health interventions
Thus reaching larger proportions of individuals with a fully automated intervention has been proven effective with both single and multi session interventions. When using an active approach, for instance by emailing a link to the intervention group, one can target specific individuals. On the other hand using a more passive approach, for instance a public website, broadens the intervention group.
Naturally not all individuals (e.g. students) change their behaviour when completing an alcohol intervention. There is a lack of research exploring how fully automated
interventions can be improved or extended to increase their effectiveness.
1.5 Extending the session
Rather than placing the burden of action upon the recipient of the intervention, as is common in multi session interventions e.g. asking somebody to come back to a website to complete more steps at a later time, this study proposes the term extended session. Here the participant is kept in session without any (or minimal) need for action, removing the need for the user to remember any kind of password or other method to
proposed extended session, however the difference should be found in the delivery and burden of the intervention. In extended intervention the session is never broken, the participant is constantly in session over a given time period. The intervention content is delivered to the participant, rather than the participant fetching the content at will.
2 Aims
This study takes the first steps in creating a pairing of a single session intervention with an extended intervention. The extended session intervention will be offered as part of the individual feedback from a single session intervention, giving those who decide they want more help the opportunity of extending their intervention for a specified time period (e.g. a few weeks), furthermore the extended session intervention will be using
push technologies to help participants stick to the program rather than simply asking
them to return at a later date.
In short, the extended session intervention will consist of a number of messages delivered to participants several times a week during a specified time period. The messages are intended to increase motivation for change as well as prompting actual change. By doing this participants are encouraged to “stay in session” for an extended time period, and also bring the intervention with them into their daily routine.
2.1 Research aims
The research aims of this study is to identify participant’s ability and attitude in relation to joining and completing a push based extended session intervention after completing a single session intervention.
• Aim 1: Signing up for the extended intervention requires several steps. First the baseline questionnaire needs to be completed and the signup form, as part of the individual feedback, needs to be filled out. Getting started with SMS and Android requires activating phone numbers or downloading apps. The first aim of the study attempts to identify any issues signing up for the extended intervention.
o Who enrols to the extended intervention? o What predicts choice of delivery method? o How many activate their SMS intervention?
o How many download and activate the Android app?
• Aim 2: Participants are free to choose extended intervention length (3,4,5 or 6 weeks), frequency of messages (3, 5 or 7 messages per week). The second aim is to identify satisfaction of these choices.
o What predicts choice of length and frequency?
o What predicts satisfaction of the intervention length and frequency?
• Aim 3: The third aim of the study is to analyse how students responded to the content, and if they thought that the messages might be useful for a student that needs to decrease their consumption.
o What is the satisfaction of the message content? What predicts satisfaction? o Would the participant recommend the intervention to a friend, and what
predicts this willingness?
It’s important to note that this study will invite all participants regardless of alcohol consumption, and there will be a prize draw between those participating to boost participation rates. Thus the aim is not to assess if there is enough curiosity or motivation after a single session intervention, but rather to assess the attitudes in regards to the intervention and the completion of it.
2.2 Development aims
Besides the research aims the study also attempts to develop and pilot the necessary software components. These components are to be used during this study but also for future studies. Furthermore there is an interest in implementing the extended
intervention into routine practice and thus these software components need to be developed, modified and tested before they can be used live.
3 Theoretical considerations
3.1 Definition of push basedAs previously discussed multi session interventions have issues with adherence, and as such needs to become more engaging in their approach. One way of doing this is to continually activate the participants by prompting them to do something, such as an email reminder or an SMS. This approach will be referred to as push based interventions. In this setting participants don’t actively need to seek or remember to complete some task, but rather are prompted when a task or step needs completing.
To further the terminology, electronic means of communicating with individuals such as email, SMS, automated phone, Android app, iPhone app, etc. are referred to as push
based delivery methods if they don’t require the user to remember to interact with them.
This definition needs to be interpreted from every individuals point of view, for instance some people might not read their emails every day or even every week and some wont read them unless they know they have one waiting for them. For these individuals email isn’t a push based delivery method. Therefore it is important to find a balance between what method delivers the intervention in an optimal way with what method can be considered push based for participants.
Different types of delivery methods all carry with them their own unique set of limitations and possibilities.
Emails have the capability of carrying longer body text and can optionally include images or links to questionnaires and websites allowing for two-‐way communication. However emails are often overlooked or forgotten in the inbox, as individuals might receive a great number of emails every day and prioritizes what emails to deal with first. Also email requires the participant to have Internet access. How individuals check
emails can range from some checking them every five minutes to others who actually plan ahead when to check them.
SMS messages are not delivered via the Internet but rather mobile networks that have a greater worldwide reach. This allows individuals to receive messages to their phone even when there is no Internet access. Messages are limited in size, but most modern phones concatenate multiple messages from the same sender so that it looks like one message. Two-‐way communication is done by allowing participants to respond with their own SMS back to the system using a predefined syntax. This quickly gets very complex and participants need to be thought what syntax is correct and what options are available. There is also a cost involved in sending and sometimes receiving SMS messages.
Mobile applications (Android or iPhone) are very free in their construction. It’s possible to send messages, prompt the user for answers to questions, visualize with graphs etc. Applications can also be designed to not constantly be in need of Internet connection. One of the main limitations is the fact that applications are distributed through either
Play (Android) or App Store (iPhone), this means that participants need to be prompted
Not everybody owns an Android device or an iPhone, which limits the reach of this method.
3.2 Intervention design
The main idea behind the intervention is to allow participants to sign up for a push based intervention that delivers a predefined number of messages during a predefined number of weeks. The intervention will not allow for any two-‐way communication. The idea is to keep participants in session during their normal day activities, hence creating an extended session from a single session intervention.
3.3 Message content
Numerous theories have arisen from years of research in the field of behaviour change. These theories have a common approach to change, allowing the subject to go through three distinctive phases: becoming motivated, translating motivation to change and maintain the new behaviour [25].
Messages are divided into four categories: “Food for Thought”, “Task”, “Challenge” and “Reflective”.
3.3.1 Food for thought
“Food for thought” messages are thought to increase intrinsic motivation to reduce drinking and self-‐efficacy, lending its theoretical base from Social Cognition Models such as Theory of Planned Behaviour and Health Belief Models that emphasize the importance of attitudes, subjective social norms, perceived risk and personal control or
effectiveness. The messages will also be created with inspiration from Self-‐determination
theory that emphasizes the need for finding an intrinsic motivation to change and
maintain a changed behaviour as opposed to external motivation, such as being told to do something.
The “Food for thought” messages contained fact like content as well as questions that respondents were meant to think about during the week.
3.3.1.1 Examples
Consuming alcohol increases the risk for negative consequences, such as falling over and hurting yourself.
Have you ever had to skip something that you enjoy due to being hung-‐over, such as a soccer game or similar?
3.3.2 Reflective
“Reflective” messages are more or less the same as “Food for thought”, however the wordings were a bit different as they were sent on Sundays and were targeted towards reflecting over the past couple of days.
3.3.2.1 Examples
Hope you had a nice evening. When you have time – now or maybe later – take a few
minutes to reflect over how you normally feel the day after you’ve drunken a fair amount of alcohol? Are you dissatisfied with any of your emotions?
3.3.3 Tasks
“Tasks” are inspired by Social Cognitive Theory that focuses on behaviour change
intention as an important factor for future change. Also, a great emphasize is placed on
self-‐efficacy as a learning process in which people select, react and learn from experiences. Having good examples in ones surrounding that have succeeded with similar change is also valuable, as is appraisal and support from others.
3.3.3.1 Examples
Write down 3 good things about your alcohol consumption and 3 not so good. Does the positive outweigh the negatives? Amongst the negatives, what is most important? Think about why that is.
If you want to decrease your consumption then tell the people in your surroundings that you are doing so. You will surely feel an unexpected amount of respect for your decision.
3.3.4 Challenge
“Challenge” will rely on the Model of Action Phases that emphasizes the need to formulate specific goals i.e. “when, where and how to act” by thinking ahead and formulating “if-‐then” plans for various situations.
3.3.4.1 Examples
Tonight, or the next time you will consume alcohol, try not drinking any alcohol until you get to the party and meet your friends. Notice how the first positive effect of alcohol is noticeable whilst among friends rather than at home getting ready. Good evening.
Tonight, or the next time you will consume alcohol, decide to drink a glass of water between each alcoholic beverage. You will feel a lot better the following day, guaranteed. You will save a fair amount of money as well. Good evening.
3.4 Existing studies
Using simple push messages delivered via SMS or email has been used successfully several times in trials where consideration has been taken to smoking habits, physical activity and weight loss [38,39,40]. Although slightly different in their nature the main component of the interventions have been messages delivered continuously during a set time period. This method has show promise, e.g. one of the more recent ones, txt2stop [38], was found to significantly improve smoking cessation rates at 6 months.
Using simple push messages as part of alcohol intervention trials has been used to try and improve upon diary keeping methods, where SMS was used to try and improve self-‐ assessment [20]. Similarly SMS messages have been used to assess and give feedback to
participants in an attempt to reduce heavy drinking [41]. In preparation there is a study trying to improve adherence to pharmacotherapy [42].
The method of having simple push messages central to the intervention has been trialled to some extent [43] where messages were used to improve outcome for patients with depression and co-‐morbid Alcohol Use Disorder.
There is however still a lack of research in the area of using push based interventions within high-‐risk alcohol treatment. One limitation of existing studies has been sample size and recruitment methods. An interesting approach is to increase sample size and use a more proactive approach that reaches not only those already curious or motivated to change.
4 Technical considerations
4.1 Service Oriented Architecture (SOA)
Service oriented architecture can be described as a set of services communicating to create a larger system. These services can be within the same enterprise or
communications can also be done to external services. A common example is the use of pre existing services to handle credit card payments on an e-‐commerce website.
However SOA shouldn’t only be viewed as a means of abstracting services externally, but as a set of guidelines and principles for building entire systems and by doing so abstracting complexity, knowledge, workload, stability etc.
SOA is a set of guidelines and principles, although software companies might have an interest in developing and marketing frameworks for creating SOA services enabling standardizations for development and deployment. This set of guidelines and principles are agnostic and do not force upon the architect any kind of framework or programming environment. The guidelines and principles themselves also differ somewhat between architects and companies, however the basic idea of splitting a big system into several independent services is always at the core of SOA.
SOA has been used for a long time within enterprise application development. Over time drastic changes have been made to the technologies used to implement SOA, especially the method of exposing the services going from complex binary-‐based communication such as CORBA or DCOM to more lightweight methods such as REST.
Four commonly used SOA principles are:
• Contract – There should be some kind of written documentation that helps others to understand the externalized functions of the service. At least there should be information regarding endpoint, expected parameters as well as return values and possible errors.
• Abstraction – The service developer should keep all parts of the implementation hidden from the client. The client shouldn’t have to understand the
implementation to be able to use the service.
• Autonomy – The service developer shouldn’t be governed by external forces as to how the service is developed. The implementation is completely done by the service developer and as such the correctness and dependability of the service is the developer’s responsibility.
• Stateless – The service itself shouldn’t remember state using local memory. Agnostic scalability requires that the service sees every request as completely new and has no local history as to what this particular client has requested before. This is not so say that history can’t exist in a distributed memory fashion, such as a database etc.
4.2 Multi tier architecture
Once in discussion around SOA the topic usually pivots into discussion around
implementation and fine-‐grained architecture. A common place to start is with the multi tier architecture. Design patterns such as Smalltalk’s MVC are usually confused with multi tier architecture, and rightfully so as they look very common, however multi tier architecture is not a full blown design pattern with rigorous implementation, but instead a principle of viewing a service as a stack, where each layer of the stack has a very clear and distinct principle.
These layers that make up the multi tier architecture can be referred to as: data, logic and exposure. These layers are written in the service-‐preferred language, such as Java, Python, JavaScript etc. Staying true to SOA principles the preferred language should be the developer’s choice. However as multi tier architecture isn’t a rigorous design pattern, some implementations use more layers to get finer control of certain parts of the architecture.
4.2.1 Data layer
Consisting of routines to communicate with some kind of persistent storage (SQL, LDAP, tape etc). The key to successfully creating a data layer is to ensure that all
communication to the persistent storage is done via it alone, no other parts of the multi tier architecture should be allowed to created connections to storage. The data layer exposes routines that allow for retrieval and modification of data in storage.
4.2.2 Logic layer
The only layer that is allowed to communicate with the data layer, hence creating a stack. The name logic comes from the longer business logic, and refers to the use of the layer. This is the layer that should implement routines that allow success for the service (and in length the enterprise). Typical routines combine calls to other services (external and internal) with the data layer. An example would be a routine that first uses an external service to debit a credit card then uses another routine to send a written notification via mail and finally uses the data layer to write this information to storage. 4.2.3 Exposure
A lot has happened to this layer through time, since moving from complex exposure such as CORBA and DCOM to lighter REST based exposure. The layer’s only responsibility is to accept incoming request, routing them to the correct routines in the logic layer and returning correct responses to requests. Today it’s very popular to do this using the JSON format and large systems (Facebook, Twitter, etc) are heavily reliant of this today.
4.3 Short Message Service (SMS)
Short Message Service gives users of GSM based mobile phones the ability to send and receive short text messages over the GSM network. SMS messages have a limit of 140 characters, however modern mobile phones often simulate the possibility of sending and receiving longer SMS messages by concatenating them. The SMS specification is maintained by 3GPP [26]. If the recipient is not available (mobile phone turned off or out of network range) messages are saved at the service carrier and retried later.
When using SMS as part of software there are two viable options. There are online services that allow easy access to their REST API’s, giving programmers access to
resources for sending and receiving SMS messages [27,28,29]. The entire feature set for these services are not yet available in Sweden, and an example price is approx. $0.036 per SMS [27]. Assuming that we had 1000 participants for 3 weeks receiving 1 message per day, that gives us 1000 * 3 * 7 = 21000 messages, priced at $0.036 gives us a cost of $756 (approx. 5400 SEK, USD/SEK rate at October 2012).
For this study we might send more or less messages than this, however even at this modest estimation the cost is high. Considering external validity of the intervention this price wouldn’t make the intervention useable in a real world setting.
The second option is to invest in a GSM modem (approx. 3000 SEK) and a SIM card from a carrier. Currently there are SIM cards available in Sweden at 699 SEK per month that will allow unlimited SMS messages to be sent. Although obvious upside in cost one has to be aware that the GSM modem needs to be integrated into the software by writing code and maintained by being connected to a server at all times. The cost of
development and maintenance could exceed the gateway option, especially if the software is only used once.
4.4 Mobile applications
Mobile applications are written in different programming languages and different work environments depending on what operative systems are to be targeted, these could include iOS (iPhone/iPad), Android, Windows Mobile, Windows 8, etc. There are also considerations to be made regarding what versions of these operative systems should be supported. Newer versions of the operative systems will allow for a more advanced feature set available for programmers whilst at the same time limiting the market, as users with older devices won’t be able to use the application.
Applications written in the preferred language of the operative system are usually
(when written well) more advanced and responsive. This includes advanced 3D graphics as well as use of cameras etc. However if several operative systems are to be supported costs of developing using these native language approach could be costly and difficult to maintain.
To avoid this cost the option of writing HTML5 applications that can be deployed to all operative systems is sometimes used. A small program still needs to be written for each operative system that includes this HTML5 application, however it’s usually a very straightforward matter that doesn’t take a significant amount of time and shouldn’t incur great cost. This type of hybrid application however is usually not as responsive and should only be used if there are simple interactions with the application.
To be able to communicate with the user of an application mobile operative systems have developed the ability to “push” messages. This requires additional setup depending on operative system. For Android, the mobile application needs to be modified to send a notice of registration when the application is started for the first time. Furthermore there is need of a server that can receive device registrations and send messages [30]. These features may or may not be available depending of operative system versions, and the use of them might exclude potential users that don’t have an updated phone.
Once a “push” system has been setup the application developer can at any time send a message (via Internet not GSM) to all devices registered. This could be used to inform users of new content that is available, or remind them to complete some task. As there are no costs involved in the actual sending of messages (there might be cost involved in setting it up and maintaining it), this is an interesting possibility to replace SMS costs.
4.5 Email
Email is cheap but suffers from deliverability issues. Legitimate messages are often not received as they are filtered into SPAM folders or even stopped by the mail server. Although a common problem, there are steps that can be taken to ensure increased deliverability of emails.
4.5.1 Content
Emails can contain HTML elements so that content is formatted prettier than just plain text. This also opens up for images to be included in the email content. However if sending HTML emails one should always append a plain text version for those who are not able to read HTML in their email clients, this can be done using multipart emails. Clients able to read HTML emails will not see the plain text email.
Emails containing only HTML and no plain text will have a higher SPAM score. If images are used it’s important to have a good ratio between images and text. If there is to many images compared to text characters then SPAM scores will rise. Furthermore it’s
important to avoid certain sentences, such as “free drugs”, “viagra”, etc. Email servers recognize these key words and SPAM scores are increased [31].
4.5.2 Deliverer
Simply installing email server software on a computer and sending emails from a static or dynamic IP is going to be perceived as SPAM. When sending emails from a mail server one should make sure that the mail server is listed in so called whitelists. A common whitelist [32] classifies senders in trust categories (high, mid, low, none), being listed as a trusted sender helps emails avoiding SPAM classification at receiving mail servers. 4.5.3 Identification
It’s possible for any mail server to send email from any sender, e.g. a server could send an email from name@example.com without having any kind of ownership of this domain name. This kind of spoofing infers great risk, as it’s easy for someone to
impersonate somebody else. By introducing SPF records [33] domain owners are able to list authorized senders in their DNS configuration. SPAM filters are then able to verify that an email comes from an authorized sender.
Although SPF ensures that the SMTP header MAIL FROM is authorized it does not consider the actual content of the email. By using DKIM [34] the integrity of the headers in the email (From, To, etc) and the body can be checked. DKIM uses a set of keys (public and private) to create a signature of these fields. The DNS contains a field with the public key, and as such any receiver can use the public key and the signature in the email to ensure that the sender actually had access to the private key [35].
To further minimizing spoofing attempts Message-‐ID was introduced as a way of
creating a unique id for every email [36]. This means that two emails should never have the same Message-‐ID. This is an attempt to avoid replay attacks where a malicious user
could re-‐use DKIM signatures or resend the exact same email. Exactly how these Message-‐IDs are created slightly differ depending on the senders email client, however there are recommendations [37].
5 Methods
5.1 Study designThe study is a pilot and feasibility study. All students on semester 1,3 and 5 (n = 11283) at Linköping’s University were, in mid October 2012, invited via email to complete an alcohol single session intervention. This is part of the student healthcare centre’s routine practice.
The contents of the email sent to students was the same used as during routine practice with the addition of information about the possibility to participate in a research project at the end of the questionnaire. See enclosure 1 for the email contents.
After answering the questionnaire each student was presented with his or her personal and normative feedback, this is part of the routine practice. However for the sake of this study, information about the research project was prepended to the feedback. All
students, regardless of alcohol consumption, were offered to join the extended intervention. No other means of registering for the intervention was offered to the students.
During registration students were made aware that they are giving informed consent to participate. Participants were also made aware that they entered a draw for one of two iPad’s when signing up for the extended intervention. For the prepended information please see enclosure 2.
Participants that were willing to join the extended intervention were given the option of selecting delivery method (email, SMS or Android), how many weeks they would like to get messages delivered (3,4,5 or 6) and how often they would like messages (once a day, five a week or three a week). If selecting email as delivery method they would have completed the signup (as we already had validated their email address from the initial invitation to the single session intervention), if selecting SMS they would have to send an SMS with an activation code to a displayed phone number and if selecting Android they were given instructions on how to download the app and activate it. Those who selected Android but weren’t able to find the app on Google Play were allowed to go back and choose another delivery method. This courtesy was not given to the other delivery methods. Once signup was complete messages were delivered at the length and frequency requested. See figure 1 for a schematic overview of message delivery.
Figure 1 – Schematic overview of message delivery Messages
per week Mon Tue Wed Thu Fri Sat Sun
7 FFT Task FFT Task Challenge Challenge Reflection
5 FFT FFT Challenge Challenge Reflection
3 FFT Challenge Reflection
