People now can measure for themselves their heart rate, sleep quality and quantity, mood, workouts, blood pressure, food consumed, quality of surrounding air – anything from mental, emotional, and physical to social and environmental aspects of daily life – due to computational advances in self-quantification systems (SQS) or technologies. SQS refer to the sensing devices and apps that can be used for collecting and processing data. Examples of SQS are Fitbit for counting steps and tracking sleep; Adidas miCoach for tracking physical activities such as swimming and running; and Lumoback for monitoring posture. There are two types of SQS: A primary SQS collects data directly from the users; health-related examples are MoodPanda for tracking mood and iBGStar for monitoring blood glucose. A secondary SQS can aggregate data delivered by multiple primary tools or apps, to facilitate data analysis; TicTrac, BodyTrack, and Argus are health-related examples.
Self-quantifiers are individual users who use SQS to collect data about themselves. Health self-quantification (SQ) refers to the process of incorporating SQS into people’s daily lives as a form of health self-management. SQ has the potential to induce changes in behaviours, attitudes, and approaches to health self-management, and the use of SQS is widespread and growing. For example, according to one survey about tracking for health that was conducted by Fox and Duggan in 2013, 69% of US adults track themselves; 21% of the study population were using dedicated devices and 46% stated that they had changed their behaviour based on the collected data.
However, there are many challenges that self-quantifiers may face when they use SQS for managing health. For example, manual data collection generates inconsistent datasets, subjective measurements lead to incomparable results, interpreting the generated data is confusing due to complex graphs, finding patterns and relationships between different kinds of personal data are difficult due to data integration limitations, accessing raw data is restricted raising data portability issue, etc. These challenges may be holding back the optimal use of health SQS to obtain and apply knowledge about health status. The aim of this paper is to characterise the experience of using SQS in self-managing health, particularly its impact on self-quantifiers’ time and data, and to draw implications for health self-management and health activation.
An international online survey was conducted from December 2013 to March 2014 to elicit information from adult self-quantifiers, about the methods they used to perform key activities related to health data self-management – data collection, data handling, data analysis and aggregation, and data sharing. Participants were recruited by using social media like Quantified Self meetup forums, Twitter, Facebook, and LinkedIn. People who used one or more selfquantification tools or apps as part of their health selfcare were included. 103 individuals provided sufficient information for analysis.
Respondents from USA were the highest proportion (60%), followed by Australia (11%). Nearly two thirds (62%) were aged between 20 and 39 years. 75% were male. Almost one third (32.3%) had completed high school or equivalent; the rest had at least a university undergraduate degree. 68% reported good to very good health status.
The top three motivations to use SQS were:
- to know if a certain health-related variable could affect another variable (64%);
- to find answers to specific questions related to health (62%);
- to proactively minimise possible future health problems (61%).
From the surveyed self-quantifiers, 89% were using at least one tangible tracking device for self-quantification whereas 11% were using only apps. The most popular primary tangible self-quantification devices were for tracking and quantifying physical activities, sleep, weight, calories burned, and diet:
- the Fitbit collection (e.g., Fitbit Ultra, Fitbit One, Fitbit Zip, etc.) with nearly 25% of the respondents;
- Withings tools (e.g., Withings scale, and Withings activity tracker) with 19%; and
- Jawbone Up with about 12%.
The most popular primary tracking apps were:
- 80Bites (for tracking food consumption) with 13% of respondents;
- 42Goals (for tracking personal life goals such as quit smoking, lose weight, and reduce coffee consumption) with 11%; and
- Moves (records walking, cycling, and running activities through step counter with 9%.
The most popular secondary self-quantification tool was BodyTrack (20% of respondents). In addition, two types of ancillary tools were used for collecting data and for handling the generated datasets. The most popular ancillary tools for data collection were diary apps such as Evernote, Microsoft One Note, and iPhone note apps (30% of respondents). Most popular tools for data handling were cloud storage services such as Dropbox, SkyDrive, and Google Drive, collectively used by nearly one quarter (24%) of respondents.
Combined use of primary, secondary, and ancillary tools
No survey respondents were found to be using only primary SQS for health data self-management. Nearly one third were using all types of tools; i.e., primary, secondary, and ancillary tools. In data collection, nearly 80% were using three primary tools to collect health related data; however, 51% stated that their primary tools did not offer the kind of data that they need; thus, they had to use ancillary tools to collect the needed data.
75% of respondents used another type of ancillary tool for handling the collected data from these primary SQS, for example, to organise the collected data into folders, import or export data, update data or files. 70% of respondents were performing most of the related tasks manually. Thus, over two thirds (68%) had lost files permanently, and over three quarters (77%) had failed to locate old files.
75% of respondents used secondary SQS to aggregate the data generated from primary SQS. In over one third (34%) of these cases the secondary tools were not able to connect automatically with the primary tools; therefore, they had to upload the datasets manually into the secondary SQS.
Time devoted to data collection, data handling, data aggregation and analysis, and data sharing
The duration of these SQ activities was calculated by counting the cases of taking more than 10 minutes by users in order to complete each of these SQ activities. For representation simplicity, we refer to this with the symbol CTM (Count of Taking more than 10 Minutes). To measure the impact of using these tools on CTM, we identified a set of factors (explanatory variables) that had a statistically significant effect on the CTM. Each activity had different factors. Examples of these factors were: the number of primary and ancillary tools used for data collection; the number of collected data types; the method used to collect these data; the ICT skills level, etc. In this blog, we will talk about one factor which is the impact of ICT skills level on the duration of completing each of these activities. Details about other factors can be found in the full paper that is available at: http://www.ncbi.nlm.nih.gov/pubmed/26262066
To classify respondents based on their ICT (information and communication technology) skills, we used the Eurostat’s ICT model. It classifies users of ICT technologies into high, medium, and low level of ICT skills. In our survey, 53%, 39%, and 8% were found to belong to these categories respectively.
In data collection, to find the mean difference in time taken to collect data between users with low to medium ICT skills and users with high ICT skills, we built a general linear model. We found that the mean difference between these two groups was statistically significant (mean difference=1.65). On average, people with high level ICT skills took less time (0.4 when collecting five types of data) relative to people with low to medium ICT skills (2.4 when collecting five types of data) (p-value=0.024, and 95% CI=0.017, 0.236)(see Figure 1).
To calculate the time devoted to data handling, we built a logistic regression model. We found that the effect of having low to medium ICT skills on the odds of the data handling duration was significant – 18 times higher compared to users with high ICT skills (p-value<0.001, and 95% CI=4.987, 64.971).
Similarly, a logistic regression model built to find out the time devoted to data aggregation and analysis. The result indicated that the effect of having low to medium ICT skills on the data analysis duration odds was significant 17.6 times higher compared to users with high ICT skills (p-value=0.001, and 95% CI=4.872, 63.575).
Likewise, in data sharing, the effect of having low to medium ICT skills on the odds of data sharing activity taking longer was 50 times higher compared to users with high ICT skills (p-value=0.001, and 95% CI=11.567, 221.936).
Using SQS for managing data is neither easy to use nor efficient. It requires high level of ICT (information and communication technology) skills. It forces the persons to use multiple SQS in order to acquire the needed data, and devote a huge amount of time to collect, handle, analyse, and share data. Even for those with strong technical and mathematical skills, the time increases significantly with each additional tool used.
Currently, few single primary self-quantification systems can support all of a user’s information needs. This limitation could generate a gap between the two critical stages of health self-quantification (see Figure 2). We argue that the wider the gap gets, the more a self-quantifier has to focus on managing data instead of managing health; thereby, they face information and time limitations on taking an active role to turn the collected data into knowledge, and ultimately take informed actions. This will be tested in our next user trial study.
An online survey was used to elicit information from self-quantifiers about the methods they used to undertake key activities related to health self-quantification. Current self-quantification systems (SQS) are limited in their ability to support the acquisition of health-related information essential for individuals to make informed decisions based on their health status. They do not offer services such as data handling and data aggregation in a single place, and using multiple types of tools for this purpose complicates data and health self-management for self-quantifiers. This study provides empirical evidence about self-quantifiers’ time spent using different data collection, data handling, data analysis, and data sharing tools and draws implications for health self-management activities.
Almalki M, Martin Sanchez F, Gray K. Quantifying the Activities of Self-quantifiers: Management of Data, Time and Health. Studies in health technology and informatics. 2015;216:333-7. PMID:26262066. http://www.ncbi.nlm.nih.gov/pubmed/26262066