We searched Scopus, Web of Science, MEDLINE, and Google Scholar. We used the search strings “quantified self,” “quantif*+self*,” and “quantif*AND self.” The search included studies published from January 2007 to October 2015. We included journals, conference proceedings, and papers written in English. The review was guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement for reporting systematic reviews [19] (Figure 1).

Twenty-six empirical studies met our inclusion criteria (see Multimedia Appendix 1). This sample was achieved through screening titles, abstracts, and keywords of 340 studies (Scopus 211, Web of Science 102, MEDLINE 20, and Google Scholar 7). A total of 71 studies were duplicated; of the remainder, we excluded 243 studies, in two rounds. In the first round, 136 studies were excluded because these studies either did not consider the use of SQ technologies for health self-management or did not investigate users in real-world contexts or daily life settings (rather than in controlled environment laboratories) where people individually collected, managed, and reflected on data. As a result of this round, 133 articles were eligible for the next round of full-text screening. In this round, exclusion criteria were as follows: studies provided conceptual knowledge but not empirical evidence, eliminating 78 articles; and studies focused on proposing a new solution and reporting only its technical specifications, eliminating 29 articles (see Multimedia Appendix 2) [20-122].

To assess the quality of the included studies, we followed the method recommended by Thomas and Harden [123].

Included studies in this review were coded by authors’ names, year of publication, study aims, study design, and study main outcomes or results (Multimedia Appendix 1). Other characteristics noted for each study were as follows: the sampling frame (ie, source of data, selection of participants, recruitment methods, and consent); data collection methods (eg, interviews, questionnaires, and so on); the strategies used to ensure the reliability and validity of data collection methods; data analysis methods (eg, inductive content analysis, statistical quantitative analysis, and so on); and the strategies used to ensure the reliability and validity of data analysis methods.

Included studies in this review were then assessed according to 12 criteria, covering three main quality issues. Five criteria are related to the quality of the reporting of a study's aims and objectives, sampling frame, and data collection and analysis methods (eg, was there an adequate description of the sample used and the methods for how the sample was selected and recruited?). A further 4 criteria are related to the sufficiency of the strategies employed to establish the reliability and validity of data collection and analysis methods, and hence the validity of the findings. The final 3 criteria are related to the appropriateness of the study design (ie, appropriateness of the included study design considering our research aims). The first 9 criteria are suggested by Thomas and Harden [123], whereas the last 3 criteria, related to the appropriateness of the study design, are left open and can be decided by the researcher. In this review, these criteria are assumed to be met by setting the inclusion and exclusion criteria as explained in the previous section, and including only empirical studies for the thematic analysis. We found that all the included studies have reported adequate descriptions of the research aims and objectives, sampling frame, and data collection and analysis methods as well as good or some attempt to establish the reliability and validity of the data collection and analysis methods; thus, they met most of these quality criteria (see Multimedia Appendix 1).

To extract and synthesize themes about health SQ work, first we inductively coded the included studies [124,125]. Then, to investigate the constructs of health SQ work in the light of AT, we deductively coded the included studies into the following themes [124,125]. Subject is a person engaged in an activity and using instruments in the course of this work [16]. Object (in the sense of “objective”) is held by the subject and motivates activity, giving it a specific direction [126]. Division of work relates to the extent to which an activity involves collaboration and the sharing of tasks with others [16] or with tools [127]. Community is a group of people with whom the subject shares similar objectives [14,16]. Plans and rules are the norms and specifications of an activity that is undertaken by subjects to fulfill their objectives [16]. Open coding of the content of each study was done by looking at the language it used to explain and describe SQ and associating terms and synonyms with these themes. Both stages of analysis were undertaken by author MA, and then the preliminary results were critiqued jointly by authors KG and FMS, in several rounds of review, until all authors reached agreement on the interpretation of the data.

We found that the literature provided a variety of perspectives on health SQ as data-driven work that users undertake to fulfill their health objectives [4,127,128]. It described how users interact with the SQ tools to define what aspects are relevant to their health conditions (eg, weight, sleep, blood pressure, and so on) [4,5]; set goals and track data about these health aspects for a period of time [7]; analyze the collected data to extract insights on health status [129,130]; adjust behaviors based on the insights and knowledge obtained from the analyzed data [130]; and control the adapted behaviors by sustaining the changes until the desired health outcomes are achieved [127]. We found that it was possible to categorize the overall content of the literature on health SQ activity or work in two ways: work on data and work with data. Table 1 maps these to the included studies. We found that the literature addressed working on data much more that working with data.

The first type, work on data, refers to activities that users carry out to manage their health SQ data. We found that the description of health SQ activities was not consistent across the reviewed literature, even when studies were based on the same human-computer interaction concepts. One study [4] stated that there are five key activities: preparation (eg, determining what information to collect and what collection tool to use), data collection, data integration (ie, combining data from multiple tools into one place), reflection (ie, analyzing data and looking at or exploring visualized information), and action-taking by the tools (eg, sending alerts to the user). However, another study [7] asserted that there are three key activities: goal setting and collecting data via using the tools (eg, BodyMedia—an armband for weight management—allows users to set goals toward more physical activity, lose weight, or stay the same); interpretation or reflection on the data (eg, analyzing data and calculating the time necessary to reach goals by the software); and providing feedback and coaching by the software (eg, Larklife—a wristband that tracks steps taken—will glow in a certain color when the users have not been moving for a while). Furthermore, we found that data collection and analysis activities received high attention, whereas hardly any study examined data handling activity (eg, data organizing, storing, and so on).

The second type, work with data, refers to activities that users carry out in using their data as the basis for actively managing their health status [8,10]. Some studies used the term “reflection” to imply this type of activities; to reflect upon data, users are assumed to look at their analyzed data, attempt to understand them, and act upon this understanding to improve or maintain health status (as in studies [4,7]). Other studies described this kind of interaction from the perspective of users’ cognition in relation to one or more factors: building belief [12,131]; intentions for collecting data [131]; perception of the usefulness of using such tools in tracking health [132]; perception of the usefulness of objective data in making informed decisions [133], building knowledge or awareness about health and function status [5,7,130,134], in deciding what action should be taken on the generated data [8], the ability to act [132], and maintaining behavior or keeping track to cope with growing health problems or conditions [135]; and perception of themselves as good or bad self-quantifiers [8]. However, the definition of these factors is vague and questions remain open about how these factors are interrelated. We found that there is a need for further conceptualization to define and describe this kind of interaction between users and their health SQ data.

Applying the concepts of AT to our analysis of included studies, we found that one study [8] mentioned all 6 AT constructs, although it did not examine their structure or the relationships among them. The remaining studies mentioned only 3 constructs—user, tools, and objectives. Table 1 maps the AT constructs mentioned in the included studies, and the following paragraphs give detailed examples of how we found these occurring in the literature.

In the AT system, the subject has a goal, tools, colleagues or friends, and rules when he or she is working on the goal to transform it into the intended outcome [15,16,148]. This goal is transformed into outcomes through a process of doing or activities [16]. This process is called the transformation process [16,148]. In this paper, this transformation process represents the work on data and with data that health self-quantifiers perform to manage their own data and health, respectively. In the light of AT, we can say that self-quantifiers have health objectives (eg, being an active or fit person). To attain such objectives, they use SQ tools and set SQ plans to track data in a certain way (eg, directive toward a goal), for example, to walk 10,000 steps per day. They interact with the SQ tools in order to collect, analyze, store, and share data with others as well as to gain knowledge on health, take actions on the generated data to protect or maintain health status, and so on. Thus, AT appears to provide an appropriate conceptual basis to describe health SQ work. Therefore, it is used as a foundation to develop the Health Self-Quantification Activity Framework (see Figure 2).

To illustrate how this framework fits with the AT conceptualization of work, the discussion next presents the reconciliation between the AT’s 5 principles and the nature of health SQ practice. These principles constitute the general conceptual system of an activity, which are as follows: object-orientedness, tool mediation, the historical development of activity, the hierarchical structure of activity, and internalization/externalization [149]. Each of these principles will be discussed in turn.

Activity Theory asserts that the subject’s work is an objective-oriented activity [126,150]. Objectives are sense-forming motives [126] where subjects’ consciousness of the world around them is formed by their acting upon it [15]. In addition, an objective can be a motive-stimulus that stimulates a subject to engage in the activity until the desired goals are reached [126]. In health SQ, self-quantifiers track themselves for various purposes or objectives such as to improve health (eg, track blood glucose to hit the target range); to improve other aspects of well-being (eg, track time spent on doing things to be mindful); and to find new life experiences (eg, track heart rate for as long as possible to see what can be learned from it) [5]. Setting goals by using SQ tools in relation to these objectives helps users to collect the data necessary to attain their targets [5,8,13,137]. For example, in case of using Jawbone UP—a wearable wristband for tracking physical activities such as walking—when the users have not been moving a lot, Jawbone UP will start vibrating after being still for a certain time [7]. This helps the users to make sense of their health status (eg, I have not been active today), make a real-time decision (eg, do more walking now), or perform medium-term self-assessment (eg, if a person with diabetes is not feeling well or is inactive today, they may need to check their blood sugar) [8]. It also helps to illustrate one’s progress toward reaching what they want; hence, it stimulates or motivates users to keep on working until their goals are achieved [4-7,13].

Activity Theory asserts that the subjects’ work is mediated by tools [14-16]. Thus, tools could be at the same time both enabling and limiting. They may empower the users during their work to attain their goals, or they may restrict such interaction [15]. Either effect may be observed in health SQ. Self-quantification activities are mediated by the users’ tools [9-13]. As an example of mixed effects, a study about using SQ tools to measure progress by athletes [9] found that if the generated data or scores are below what was expected (ie, because of limitation in data accuracy that users are not aware of), their confidence and how they perceive themselves as athletic individuals may be distorted; consequently, their performance could decrease and eventually they might stop analyzing or even collecting such data.

Activity Theory states that the objective-oriented activity has a history of its own from which the human mind develops an understanding [150]. The historical analysis of the activity is often needed to understand the recent situation [15]. In health SQ, the advances in computational analysis of SQ tools make building a history of work possible, which is a major facilitator to understand current health status and obtain self-knowledge [129,130,134,142]. Data generated from using SQ tools are also beneficial for people to evaluate their future health status. Providing a history of the collected data and detecting trends over time can help users to not only quantify the current health and function status, but also calibrate expectations for upcoming activities based on previous experience. Through evaluating current performance against past performance, they may compare their health status at different times in the year [9,141].

Activity Theory differentiates between activity, actions, and operations to offer a hierarchical structure of activity. Achieving the subjects’ objectives requires them to go through a series of activities [15,16,18]. Each of these activities is composed of actions, and actions are composed of operations [16]. In health SQ, users undertake various DM and HM activities in pursuit of achieving their health objectives. These activities are composed of actions and operations [1]. We can take as an example the data collection activity, which is one of the DM activities: in the case of quantifying the walking habit, it could be composed of several actions such as setting goals, wearing the tools, going for a walk, and so on that are required for generating data or measuring performance [1]. However, the conceptualization of activities and actions in the examined literature is vague; thus, the boundary between them is not clear. This could be one of the reasons for inconsistencies in describing the DM activities, as discussed previously. For example, goal setting was considered a key activity by De Maeyer and Jacobs [7], whereas data collection is an action that constitutes this activity. In contrast, Li [4] considered data collection as a key activity, whereas goal setting is an action that may be a part of this activity.

Activity Theory asserts that any activity has a recursive structure in which feedback that evaluates performance is decisive [150]. On the basis of this feedback, the subject’s internal mental or cognitive actions are formed and this is called internalization. The opposite of the internalization process is the externalization process where internal cognitive actions are transformed into external actions. These two processes demonstrate that cognitive activity is tightly interconnected with external objective-practical activity [150]. In health SQ, data generated from SQ tools may provide decisive feedback about the person’s current health status because they are neutral and not intuitive or emotional [3,8,11,133]. Such data offer an opportunity for users to obtain self-knowledge about health status, and based on this knowledge they decide on what actions they need to take [4,7,9,13]. However, our analysis of the literature shows that these internalization/externalization processes have not been fully investigated yet, and the transition between these two processes remains open for examination. Therefore, a more detailed discussion of these processes cannot be provided in this paper and will be taken up in future work.

Many self-quantifiers have been frustrated in accomplishing their health objective [4-10]. They gradually lose their initial motivation and may stop tracking, and thus may not achieve the desired health outcomes [1,4-10,128]. To examine what could cause such a situation in a more systematic and structured way, the Health Self-Quantification Activity Framework puts SQ tool use into context, in relation to the goals, plans, and competence of the user. In the following paragraphs, examples of how a person could fail to achieve her or his goals are provided in the light of the health SQ constructs as follows.

“Users” need to have high information and communication technology (ICT) skills [6] and mathematical skills [9] in order to successfully use the SQ tools to track health. However, not everyone is so skilled. A study shows that few people in Europe have high ICT skills (only 29% of the European population from 28 countries) [151]. Therefore, factoring in the users’ technical and mathematical competence levels (eg, low, moderate, or high) could lead to better understanding their health outcomes from health SQ [152,153].

“SQ tools” have a major influence over users. A study found that “users were emotionally involved with their tools, and saw this more as their external conscious to see the details on their physical activity, food and sleep patterns and act on it where necessary” [7]. However, any issues that arise when these tools are in use may increase the time and cognitive effort needed to resolve these issues and to gain self-awareness [4,5,7,154]. For example, if many types of health indicators are of interest, finding a single tool that can support acquisition of most of these data is difficult [4-7,12,138], thus, the users may decide not to collect all the data that would be useful to successfully manage health [4]; if they proceed, they may find themselves using multiple tools, which increases the person’s time and efforts to integrate, analyze, and learn from the collected data [4,5]; and if the same data are tracked by different tools, when combined, data may be inconsistent, giving rise to concerns about data accuracy [139]. Therefore, users may lose their trust in these measurements or may stop tracking [4,10,139,141].

“The division of work”—for example, dividing data collection work between the user and tools in a form of manual or automatic data collection [4,6,10]—may be cumbersome, and this can impede the user’s motivation to collect data of interest or pursue the SQ practice [7]. For example, one study [4] found that self-quantifiers were eager to log their food intake manually at the beginning, but this decreased because as time passed they found it very time consuming.

“Community” is a very real experience that is enabled by health SQ: some users go online and share their data or results about their health status with others who could be friends, family, or strangers, in order to observe, compare, compete, and so on [143]. However, sharing this kind of information has its own set of problems. For example, it could expose the users to privacy breaches and different kinds of discrimination, including discrimination in insurance and other financial dealings [143]. In addition, if the users’ results appeared lower when compared with others, they might consequently develop a low opinion of themselves and thus they might stop tracking their performance [9,137].

“SQ plans and rules” are different for each person, and most SQ systems do not consider this variation. For example, using SQ tools for conducting self-experimentation raises many issues [5,10,136]. When the individuals have personal hypotheses (eg, if I drink sugar water over a period of 8 weeks, then I will lose weight), they want to design their tracking plan to either prove or disprove their hypotheses. However, the lack of integration of single-case experimental design standards into health SQ systems leads to insufficient scientific rigor in data collection and analysis [5,136]. Thus, with current SQ tools, testing one’s thinking about personal weight loss would be challenging and potentially misleading.

To summarize, health SQ is data-driven and objective-oriented work that is mediated by tools. This work is composed of two types of activities: DM activities (ie, work on data) and HM activities (ie, work with data). These activities comprise 6 constructs (ie, users, SQ tools, health objectives and goals, division of work, community, and SQ plans and rules). Understanding health SQ work, activities, and constructs is important because it is the first step to operationalize health SQ fully. This may in turn help to achieve the aims of health professionals and researchers who seek to make or study changes in the self-quantifiers’ health systematically. This review makes two significant contributions to research in this field: it explores health SQ work and its constructs thoroughly and it adapts AT to describe health SQ activity systematically.

However, many gaps exist in the literature (eg, inconsistency in describing tools, lack of a comprehensive view of health objectives, and vagueness in describing DM and HM activities) that need further investigation. To fill these gaps, our ongoing work is leading toward developing a taxonomy that accounts for various aspects of SQ tools’ functionality and characterizations [1]; a conceptual model and classification schema for explaining data or health aspects [155]; and a model that describes the key DM and HM activities that are necessary for health self-management [1,154]. These are critical contributions to establishing a holistic and rigorous theoretical framework within which to understand and improve health SQ activity.