This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
A growing number of investigators have commented on the lack of models to inform the design of behavioral intervention technologies (BITs). BITs, which include a subset of mHealth and eHealth interventions, employ a broad range of technologies, such as mobile phones, the Web, and sensors, to support users in changing behaviors and cognitions related to health, mental health, and wellness. We propose a model that conceptually defines BITs, from the clinical aim to the technological delivery framework. The BIT model defines both the conceptual and technological architecture of a BIT. Conceptually, a BIT model should answer the questions why, what, how (conceptual and technical), and when. While BITs generally have a larger treatment goal, such goals generally consist of smaller intervention aims (the "why") such as promotion or reduction of specific behaviors, and behavior change strategies (the conceptual "how"), such as education, goal setting, and monitoring. Behavior change strategies are instantiated with specific intervention components or “elements” (the "what"). The characteristics of intervention elements may be further defined or modified (the technical "how") to meet the needs, capabilities, and preferences of a user. Finally, many BITs require specification of a workflow that defines when an intervention component will be delivered. The BIT model includes a technological framework (BIT-Tech) that can integrate and implement the intervention elements, characteristics, and workflow to deliver the entire BIT to users over time. This implementation may be either predefined or include adaptive systems that can tailor the intervention based on data from the user and the user’s environment. The BIT model provides a step towards formalizing the translation of developer aims into intervention components, larger treatments, and methods of delivery in a manner that supports research and communication between investigators on how to design, develop, and deploy BITs.
A growing number of investigators have commented on the lack of models informing the design of behavioral intervention technologies (BITs) [
The purpose of this paper is to describe a BIT model that supports the translation of the clinical aims of a BIT treatment and its intervention components into BIT features. The BIT model proposed here is intended to provide a broad hybrid framework that combines behavioral principles with technological features that can help bridge the fields of behavioral science and technology. Experts from both fields contribute to the development of BITs, but the vastly different training and knowledge backgrounds have led to differences in conceptual models that guide development and evaluation. A framework that integrates behavioral science, design, and engineering can support the definition of systems in terms of testable hypotheses that could then be evaluated. This would help avoid the all-too-common process of developing BITs that ignore psychological or engineering principles or that rely entirely on developer intuition [
We review here three design models proposed by Ritterband [
Ritterband provided one of the first generalizable models depicting how a Web-based intervention contributes to symptom change [
The Ritterband model is useful, as it specifies the elements and characteristics to consider when designing an intervention website. Many elements of the model could also be applicable to other technologies, such as mobile devices. However, the Ritterband model does not articulate how technological components might be mapped onto more specific (and proximal) intervention goals, which is important in intervention design. Furthermore, while Ritterband emphasizes that the model is not necessarily linear (eg, components do not necessarily need to be deployed sequentially), the non-linear properties are not articulated. These non-linear properties are increasingly important as technologies are able to receive and react to data obtained from the user, the user’s environment, and third parties such as a health care system or coaches.
The Fogg Behavior Model [
Fogg’s model is elegantly simple and very useful within the constraints he outlines. However, the restricted focus does not fit the goals of many treatment interventions that attempt to address more complex problems such as reducing symptoms of depression or anxiety, treating insomnia, improving self-management of chronic illnesses, coping with addictions, or implementing healthy lifestyle programs. Users may not know what steps to take to attain their goals and may require some education. It may even be difficult for users to identify behavioral goals that are circumscribed enough to be attainable. Motivation may wax and wane and thus can be a focus of BITs. However, Fogg’s model may be very useful for small behaviors. As such, the model could serve as a useful tool in considering and designing individual components of larger intervention programs.
Oinas-Kukkonan has described comprehensive models, which he refers to as Persuasive System Design and the Behavior Change Support System, which, in spite of the name, also addresses cognitive change and adherence to a BIT [
A strength of Oinas-Kukkonen’s model is that it supports the transfer of design components into software functionality. Its clear articulation also allows the evaluation of the value of these components, as evidenced by a meta-analysis that evaluated both the frequency of the use of these components, as well as their impact on adherence [
The BIT model provides a framework for the translation of treatment and intervention aims into an implementable treatment model. For the purposes of clarity, we use the term “intervention” to refer to a single interaction with a single element and the term “treatment” to refer to multiple interactions that unfold over the entire course of interaction with the BIT.
BITs are intended to assist users in achieving a goal related to health, mental health, or wellness. A single BIT intervention enables users to change their current state (the state at the moment of the BIT use) using one or more possible interventions to achieve the intervention aims (desired future states) (see
The BIT model displayed in
Summary of BIT model.
|
|
BIT component | Examples | ||
|
|||||
|
Why | Aims | Clinical aims: | ||
|
|
|
|
Weight reduction: | |
|
|
|
|
|
Decrease caloric intake |
|
|
|
|
|
Increase physical activity |
|
|
|
|
Promote sleep hygiene | |
|
|
|
|
Decrease depression: | |
|
|
|
|
|
Increase positive activities |
|
|
|
|
|
Decrease avoidance behaviors |
|
|
|
Usage aims: | ||
|
|
|
|
Use of Intervention tools | |
|
How (Conceptual) | Behavior change strategies | Education | ||
|
|
|
Goal setting | ||
|
|
|
Monitoring | ||
|
|
|
Feedback | ||
|
|
|
Motivation enhancement | ||
|
|||||
|
What | Elements | Information delivery | ||
|
|
|
Notifications | ||
|
|
|
Logs | ||
|
|
|
Passive data collection | ||
|
|
|
Messaging | ||
|
|
|
Reports | ||
|
How (Technical) | Characteristics | Medium | ||
|
|
|
Complexity | ||
|
|
|
Aesthetics | ||
|
When | Workflow | User defined | ||
|
|
|
Frequency | ||
|
|
|
Conditions: | ||
|
|
|
|
Time-based rules | |
|
|
|
|
Task completion rules | |
|
|
|
|
Event-based rules | |
|
|
|
Tunneling |
BITs facilitate reaching future changes (ie, intervention aims) through possible interventions.
The overall goals of a BIT treatment, as well as the aims of any given intervention component, reflect the intentions of the developer [
Usage aims focus on maintaining engagement with the BIT generally and/or with its specific intervention components. Usage aims are often thought to be related to clinical aims, although the relationship between use and outcome has been mixed [
Behavioral change strategies are the methods used to attain clinical and use aims. They are grounded in models and theories of how behavior change occurs and is maintained.
Education aims to increase the user’s understanding of their past and current state and of the steps required to achieve the future state (see
Goal setting involves future planning to achieve desired future states. This can include activity scheduling, setting tasks of progressively greater difficulty, anticipation of barriers, or goals with respect to application use.
Monitoring involves recording of past states or current states. Examples include recording of current or past behaviors, cognitions, or events, reviewing previously set goals and identifying barriers, or monitoring intervention and application use.
Feedback provides information on current and past states, or the likelihood of future states, with the goal of increasing insight and understanding regarding the user’s condition or actions. Feedback may also overlap with other behavior intervention components, such as motivation enhancement (eg, feedback on goal attainment may provide information about progress and may also increase or decrease motivation).
Motivation enhancements are interventions that increase the likelihood that the user will engage in specific behaviors related to treatment goals or use of the application in the future (motivation to change current state into future state through behavior change or BIT use). These include positive reinforcement, contingent rewards, behavioral contracts, incentives, and social support [
BIT elements are distinct components or objects of a BIT intended to implement the behavior change strategies, which in turn support the user in achieving the clinical and usage aims required to attain the treatment goal. By BIT elements, we mean the actual technical instantiations present in the BIT. For example, a data entry field created in a food logging application supports the behavior change strategy of monitoring. Thus, the BIT elements are the aspects of the BIT, with which the user actually interacts. Below is a list of commonly used elements of existing BITs, but this list could expand as aims, designs, and technologies continue to advance.
Information delivery typically involves one-way interactions in which the system provides content to the user when the user initiates access. These can include such things as text, video, images, audio, or a combination of media. They are distinct from other similar components in that they commonly remain available after release and are often used didactically.
Notifications are individual messages pushed to the user, such as text messages, emails, or within app notifications.
Logs are a form of data collection that require the user to enter data. Examples include free entry, selection menus, and using a rating scale.
Passive data collection refers to data collected without any user effort, such as phone sensor data collection, data from external devices such as pedometer, and data collected through application programming interfaces (APIs) from other available sources (eg, weather data or prescription refills).
Messaging elements link the user with other individuals including those supporting the interventions (both professionals and paraprofessionals), peers drawn from their social network, or peers using the system. Messaging refers to more than just one-to-one correspondence and can include discussion boards.
Reports are reflections of data collected by the BIT that are provided back to the user (eg, calendars, calorie counts, thought records).
Visualizations may be considered a subset of reports and convey specific information derived from previously collected data and assessments. Data may be aggregated across an individual user or across groups of users.
BIT element(s) are mapped onto behavior change strategies. A specific behavior change strategy can be targeted by more than one BIT element, which may be delivered sequentially or may be embedded in each other. For example, education is often achieved by delivering didactic tools that rely on text-based information. But such learning may be augmented by embedded reports (visualizations or text) derived from data and assessment, thereby providing feedback to illustrate a point and support learning.
BIT elements can be further defined and/or refined across a number of dimensions to better fit the user and/or optimize the element to achieve its aim and overall treatment goal of the BIT, commonly by improving the user’s comprehension, ability to complete tasks, and engagement. We describe four characteristics (medium, complexity, aesthetics, personalization) that have received attention in BIT research, however, these are intended as examples and are by no means an exhaustive list.
Medium refers to media employed, such as text, video, audio. Variation in the medium can be varied for many of the intervention elements, including information delivery, social networking, or data collection. In considering the medium, it can be useful to apply a framework, such as media richness theory, which can provide information on how media may vary in their suitability for communicating different types of information effectively [
Complexity can be varied depending on the user, target population, and the task (eg, providing didactic information, a notification, or data collection). For example, some users may prefer more elaborate content, while others may prefer leaner content [
Aesthetics may vary depending on the user characteristics and tastes [
Personalization refers to altering the characteristics or content of a BIT to increase the relevance for an individual user. For example, the content of information may be tailored to fit the user’s needs and capabilities by altering language or providing examples that are more likely to be relevant to the user [
The characteristics in this model are intended to reflect the need to modify BIT elements. Conceptually, the elements could be considered objects, and the characteristics could be considered the potential attributes of those objects. It is beyond the scope of this paper to provide guidance on the methods one might use to decide which attributes best meet the needs of users, as these questions are the subject of entire fields of study such as human factors engineering and human computer interaction (HCI).
Most BITs are designed for repeated interactions over an extended period of time. That is, within our terminology, most BITs are intended as a treatment consisting of a series of interventions. The workflow defines when and under what conditions BIT interventions are delivered and can take into account changes in the aims, elements, and/or characteristics that occur over the course of a treatment. The workflow identifies when an intervention is delivered and potentially also the sequence of interventions. Below we describe common examples of workflows (user defined, frequency, conditions, tunneling).
User defined workflows allow the user access to all intervention elements and content, permitting the user to decide the sequence and timing of their use.
Frequency refers to the frequency with which any intervention is deployed. Some interventions have expectations of the frequency of use.
Conditions use data to determine when an intervention will be delivered. A variety of types of conditions can be employed. (1) Time-based rules define the release of an intervention element based on the passage of time. For example, Web-based treatments modeled on standard face-to-face treatments sometimes release new content on a weekly basis [
Tunneling uses data to determine which interventions are most like to meet the needs or preferences of an individual at a given time. For example, an intervention for anxiety can use information on comorbidities to provide specific interventions targeting those problems to improve efficacy [
Workflows may use and integrate a number of these elements, for example, providing core interventions in a predetermined sequence with a mixture of time-based and task completion rules and then allowing the user to select from a variety of additional interventions that the user believes are most useful [
To further explain the BIT model, we provide an example of a portion of a popular fitness app (MyFitnessPal). MyFitnessPal is an Internet website and mobile application designed to help people lose weight. The MyFitnessPal mobile application is freely available for the Android, BlackBerry, iOS, and Windows platforms. The overall clinical aim of MyFitnessPal is to promote weight loss. Two of the sub-aims of the application are to reduce caloric intake and increase physical activity. Although MyFitnessPal makes use of several behavior change strategies (education, feedback, goal setting, motivation enhancement), the major behavior change principle used is monitoring. That is, weight loss is promoted by helping people track what they eat and how much they exercise. Reviewing every feature of MyFitnessPal is beyond the scope of this paper; however, we present aspects of the core functionality of entering food into one’s diary to illustrate the BIT model.
The use of the food diary requires the user to initiate the interaction, requiring the user to remember and be sufficiently motivated to engage in the task. To mitigate the effects of forgetfulness or low motivation, MyFitnessPal makes use of another behavior change strategy, motivational enhancement, to support the usage aim of entering consumed food into the application. One technological manifestation of a motivation enhancer is through the BIT element of a notification. In MyFitnessPal, these notifications are delivered via text push notifications from the application. These notifications are created from the system using task completion rules. As
The MyFitnessPal example is an illustration of how the BIT model maps onto an existing application. High quality applications often contain these elements; however, there is no shortage of poorly designed applications available that do not effectively engage users to accomplish the intended actions or achieve the intended aims. The BIT model is intended to support developers and designers by providing a clear model of how to move from a general clinical aim to a clearly defined and effective application. We now move to a discussion of translating these conceptual design decisions into technological implementations.
BIT model example using MyFitnessPal calorie intake monitoring features.
The instantiation of a design based on the BIT model requires technological implementation in a system that can actually deliver the BIT to users. In this section, we provide an example of a hybrid model that integrates a general technological framework using the BIT model. We refer to this as BIT Technological, or BIT-Tech. BIT-Tech is an example that can be used by system designers and developers as a conceptual guideline. It shows the relationship among (1) software components developed for supporting BITs, (2) the user, and (3) the environment.
BIT-Tech is defined with respect to the previously defined BIT model concepts, that is, intervention
We use the superscript notation
Inspired by the robotics paradigm, we will describe our model in terms of
The BIT-Tech aspect of the model is composed of several components (
Profiler: The profiler is responsible for collecting data to define the user and environment at any given point in time. The profiler passes data
Intervention Planner: The intervention-planner is responsible for planning interventions at current time
Intervention Repository: The intervention repository stores all the intervention elements developed for the use with the BIT and can be implemented in terms of a database. Once the intervention repository receives the specification of the current intervention step from the intervention-planner at time
User Interface: This delivers an intervention
The unfolding of these interventions over time is specified by the workflow
Note that the selection of aims and elements can be predefined by the BIT based on the developer’s expertise, or alternatively can be chosen by the user or may be determined adaptively based on information received during the intervention. A treatment as a sequence of intervention steps is defined in terms: (1) elements
More specifically, the workflow
The transition among intervention steps is defined by function
The transition function might also be designed using partial contextual information. For example, a transition might be triggered simply if a certain amount of time has elapsed (eg, 1 week) or by the completion of specified tasks, regardless of the contextual information about previous interventions, aims, and collected data. However, it is also possible to begin developing adaptive BITs that employ artificial intelligence techniques to adapt the workflow to the user’s preferences and/or needs over time. That is, the workflow structure can be modified over time using collective and individual data to provide and sequence specific intervention elements with specific modifications to the characteristics to increase the likelihood of achieving the treatment and intervention aims.
Data gathered through the profiler may be initialized with specific profile data, such as demographic information and clinical status (user data), or specific time and location (environment data), which may determine the BIT tools delivered, any tailoring or refinement of the elements, characteristics, and workflow. The profiler may also gather additional types of data over the course of the treatment, such as updated data on clinical status, information on the patient’s use of the application elements, or environmental data such as location or weather (see
After the system is developed and deployed, the system performance and effectiveness can be evaluated using different computational metrics. For example, the interaction aspects of the BIT can be evaluated using HCI measures such as usability, ease of use, and usefulness [
Three paradigms: Reactive, Deliberative, and Hybrid.
BIT-Tech framework: required environment and user data is collected by the Profiler component; collected data is passed to the Intervention Planner, which is responsible for planning intervention at time t; the Intervention Repository component stores all the interventions and passes specific details of the selected intervention to the User interface component, which then delivers the intervention.
Example workflow generated by the workflow-planner specifying the elements (rectangular nodes), element’s characteristics (elliptical nodes), as well as order of transitions among elements.
Workflow for MyFitnessPal specifying the elements (rectangular nodes), element’s characteristics (elliptical nodes), as well as order of transitions among elements.
We have described the BIT model, which includes both a framework for articulating the relationship between intervention aims, elements, characteristics, and workflow and its technological counterpart (BIT-Tech). This model has a number of potential uses and implications for BIT research.
The BIT model can help developers formalize their intentions with respect to each design consideration, as well as assist in the clarification of how the intervention aims will be implemented in terms of intervention elements. This formalization can be assisted through the development of checklists and flow diagrams that allow the developer to utilize the general model in the development of a specific intervention. In this way, the BIT model can guide development (particularly those who are new to the area) to think through the various decisions that are critical to the design of a BIT and promote the integration of behavioral and psychological theory with BIT design.
Much of the development of BITs to date has been informed primarily through the application of behavioral and psychological theories and by developer intuition [
The formalization of the design and development process also supports the translation of the developer’s intentions into testable hypotheses regarding the effects of specific intervention elements, characteristics, and workflow decisions, as well as determining the required data outputs to test those hypotheses. Much of the evaluation of BITs has focused on efficacy, which limits the growth in our knowledge regarding the mechanisms by which BITs achieve both their proximal intervention aims as well as the ultimate treatment goals [
The proposed conceptual framework could be further refined through the development of more detailed ontologies that further define BIT elements, characteristics, and workflows. An ontology is a formal language used to create a map of a domain, which can provide the conceptual framework to facilitate the rapid or automated construction of BIT applications [
There are several caveats and limitations of the present work that should be mentioned. First, the BIT model we present is intended to be generalizable and is therefore a simplification. It is intended as a general framework that should be modified and elaborated to fit the needs of a specific BIT treatment protocol. Second, the proposed model is intended to respond to calls for a framework that integrates developers’ intentions, behavioral and psychological theory, the design of BIT treatment protocols, and the implementation in a technology framework. We fully expect and encourage the modification of this framework to take into account the ideas of other investigators, new technological developments, the needs and intentions of other stakeholders such as purveyors and care systems [
The BIT model builds on existing models. Our BIT model extends the Ritterband model [
application programming interface
behavioral intervention technology
Behavioral Intervention Technology—Technological Instantiation
finite state machine
human computer interaction
This paper was funded by National Institutes of Health (NIH) Grant Nos. P20MH090318, R01MH100482, R01MH095753, and R34MH095907.
None declared.