The prevalence of mobile health apps and wearable devices has enabled patients to track a variety of patient-generated data (PGD) outside the clinic, ranging from biometrics (eg, blood pressure, heart rate) to their everyday activities (eg, exercise, sleep) [1]. Researchers have found that PGD are a valuable source of information for clinicians because they can use the data for diagnosis and treatment [2-6]. Data-driven consultation has the potential to optimize patient care by leveraging effective PGD sharing [7,8]. However, little is known about whether and to what extent patients who track their PGD actually share their data with clinicians in the clinic, which is critical to fully leveraging the potential of PGD to improve health outcomes.

In prior studies that have examined how patients leverage PGD, a key related concept that has often been neglected is health literacy—the cognitive and social skills to gain access to, understand, and use health information in ways that maintain good health [9]. As a multidimensional concept, health literacy characterizes not only an individual’s ability to read and understand health information but also the ability to manage one’s health and engage with health care providers [9-11]. Although health literacy has been studied in public health for decades, this concept has rarely been examined within the context of studies focusing on PGD tracking and sharing.

Previous studies have found that through self-tracking of their PGD, patients can develop the skills necessary to interpret health information [12] and to communicate with clinicians [13]. In this light, we aim to understand whether patients’ self-tracking practices, as a way to help them gain knowledge about their health [14-18], relate to their health literacy and whether patients’ health literacy, as an important skill to fully engage with health care providers [10,11,19], relates to patients’ practice of sharing their PGD with clinicians.

Although health literacy has not yet been explicitly considered in studies related to PGD, researchers have found that tracking PGD can increase people’s awareness of their health [14] and enable them to better answer their clinicians’ questions [15,44,45]. Therefore, we speculate that tracking PGD may relate to one’s perception regarding HSI and AE. In addition, we assume that patients who shared their PGD in the clinic had tracked their PGD before and were equipped with sufficient HSI and AE. For example, a high level of HSI could enable patients to share their health data with clinicians by equipping patients with sufficient knowledge about their health and a high level of AE could facilitate patients’ data sharing as a way of engaging with clinicians. Thus, we investigate the following research questions (RQs):

Although most PGD-related studies have examined patients’ experiences retrospectively after some time had passed since their clinic visit [1,13], we sought to gather the immediate perceptions of patients regarding their experiences of meeting with a clinician. We surveyed 109 patients who had just met with a clinician at a university health center (UHC), asking them about their practices of tracking and sharing PGD and assessing their health literacy (ie, HSI and AE). Conducting the survey onsite at the UHC enabled us to collect the experiences of participants from their current clinic visit, which helped reduce recall bias and enabled us to gather ecologically valid data.

Throughout the survey, we used the term personal health data instead of PGD to aid participants’ understanding. At the beginning of the survey, we defined personal health data as health-related data that people keep track of about themselves, such as exercise (eg, step count, miles run), sleep, diet, heart rate, and blood pressure. Table 1 and Textbox 1 describe our survey measures, including how we defined and assessed each construct.

In addition, we collected participants’ (1) demographic information (eg, age, gender, occupation, educational attainment, household income, first language), (2) access to technology (eg, internet, mobile phone), and (3) details of their meeting with the clinician (eg, reasons for visit, satisfaction). The survey consisted of 39 questions (Multimedia Appendix 1) and was administered using Qualtrics [46] on a tablet device we provided.

Data collection for this study was part of a larger project that involved onsite surveys and follow-up interviews at a UHC. This paper focuses on the survey results. Following the UHC’s guidance, 3 researchers set up a study space next to several clinics, including primary care, women’ s health, behavioral health, immunization or allergy, and alternative medicine. We approached patients as they exited clinic offices and asked whether they had just met with a clinician. If the patient answered yes, we invited them to participate in the survey. If they expressed an interest, we then led them to the study space, secured their informed consent, and instructed them to take the survey using a tablet paired with a keyboard. As we had 2 tablets set up in the study space, we were able to host 2 participants at the same time. During the study, 2 to 3 researchers were sitting behind the tablets, and each participant was sitting in front of each tablet to complete the survey. Participants would inform the researchers when they completed the survey. We did not recruit any patients who had visited the UHC for mental health reasons, as per the UHC’s request.

The survey took participants about 4 to 16 minutes to complete (mean 7.60, SD 2.62). Each participant received a US $5 campus dining gift card after completing the survey. The study was approved by the Institutional Review Board at the researchers’ university.

To answer RQ1 (“Do patients’ self-tracking practices relate to their HSI and AE?”), we used multivariate multiple regression to examine whether the number of data types participants tracked and their tracking group (ie, nontracker, memory tracker, or tool tracker) were related to their HSI and AE levels. As patients’ health status has been found to highly correlate with their health literacy [34], we controlled their health status as a covariate. In addition, we dummy coded the tracking group using nontrackers as the reference group.

To answer RQ2 (“Do patients’ self-tracking practices, along with their HSI and AE, relate to their practices of sharing data with doctors?”), we used multiple logistic regression to predict whether participants had shared their PGD during a particular visit based on their HSI, AE, the number of data types they tracked, and their tracking group. Using this same set of predictors in addition to whether participants had shared their data, we used multiple linear regression to predict participants’ willingness to share their data with clinicians in the future.

To answer RQ3 (“What benefits and barriers do patients encounter with regard to sharing PGD with doctors?”), we first analyzed participants’ responses to the questions regarding their perceived benefits and barriers of sharing PGD with clinicians, and then categorized these responses using a bottom-up (inductive) approach. In this manner, we developed a coding dictionary that reflected different sharing-related benefits and barriers. We later reorganized the initial codes into potential themes, which complemented our findings for RQ2 by explaining why participants had or had not shared their health data with their clinician during the particular clinic visit.

In total, 112 patients participated in our study; however, 3 patients were subsequently excluded from our analysis as they had visited the UHC simply to receive a flu shot, fill out a form, or get a massage. Among the remaining 109 patients, 86 (78.9%) were female and 73 (67.0%) were native English speakers, and their ages ranged from 18 to 37 years (mean 21.0, SD 3.82). Most of the participants were college students (77/109, 70.6%), some (18/109, 16.5%) were full-time employees, and the rest (14/109, 12.8%) were unemployed. Except for 2, all participants (107/109, 98.2%) had access to the internet and a mobile phone. Participants reported visiting the clinic for an inquiry or examination for a specific injury, illness, or condition (65/109, 59.6%), a follow-up visit (22/109, 20.2%), a regular checkup (18/109, 16.5%), or something else (4/109, 3.7%).

Of the total, 87.2% (95/109) of participants reported tracking at least one type of PGD. Of these 95 participants, 38 (35%) were memory trackers who tracked their health data relying on their memory only and 57 (52%) were tool trackers who tracked their data using at least one tool (eg, paper journal, mobile app, website). The remaining 12.8% (14/109) of participants were nontrackers, who did not track any PGD. Table 2 provides descriptive statistics for each of the 3 tracking groups.

The regression model that predicted HSI based on the participants’ health status, the number of data types they tracked, and their tracking group was significant. However, there were no significant effects of the number of data types tracked or the participant tracking group (Table 3). Using the same independent variables, the regression model that predicted AE was found to be significant, and there was a significant main effect of the number of data types participants tracked. This result indicates that when controlling for participants’ health status and their tracking group, those who tracked more data types tended to have higher AE.

When asked to list up to 3 types of PGD that the participants felt would have been helpful if they had tracked and shared with their clinicians, 72.5% (79/109) of our participants provided valid responses. These responses included body measures (eg, blood pressure, glucose level, heart rate, body temperature; 47/79, 60%), food (36/79, 46%), sleep (27/79, 34%), water (25/79, 32%), exercise (19/79, 24%), stress (7/79, 9%), menstrual cycle (6/79, 8%), symptoms (6/79, 8%), mood (3/79, 4%), and others (5/79, 6%). In the subsequent open-ended question, participants explained why these data would have been helpful if they had tracked and shared them: (1) to better inform clinicians in detail about their health condition; (2) to receive more personalized care; and (3) to better articulate their illness and health concerns (Textboxes 2 and 3).

Our results show that neither the number of data types that participants tracked nor how they tracked their data relates to their HSI; however, when more types of PGD participants were tracked, their AE tended to be higher. We suspect that as participants tracked more types of PGD, they had more information about their health in mind. Armed with this information, participants could more actively engage with their clinicians, articulate their health concerns, and ask and answer questions regarding their health. However, tracking more types of data did not necessarily transfer to useful knowledge that participants felt helped them to manage their health. Many participants mentioned that they were not sure about the relevance and usefulness of their data.

Although relatively rare, some researchers have examined the role of health literacy in the ability of individuals to track and monitor their health behaviors [47,48]. For example, Porter et al [47] conducted a diary study focused on people’s recording of their daily physical activities and found that highly health literate individuals were able to achieve a higher accuracy in their diaries. Zoellner et al [48] significantly reduced the sugar-sweetened beverage intake of the participants using a health literacy–centric intervention. Therefore, health literacy is seen as an essential human tool to support individuals in taking advantage of self-tracking technologies [49]. Although these studies did not examine how to improve the health literacy of individuals, we see opportunities to improve health literacy in people through their engagement in the very act of self-tracking itself and through feedback generated based on one’s personal health data. One opportunity is to deliver health knowledge relevant to individuals’ PGD. For example, food tracking apps, such as MyFitnessPal [50], suggest whether one’s food intake meets their nutritional needs when they record an entry and provide recommended dietary guidelines with feedback on one’s food consumption. Unlike traditional health literacy programs that focus on delivering general health knowledge [39,51], specific information related to an individual’s own health data can be easier to understand because it is tied to their own health concerns. However, we must keep in mind that interpretation of automatically tracked data is sometimes unreliable or may even cause misunderstandings [52,53]. For example, sleep tracking apps using pressure sensors to estimate sleep hours can cause false-positive detections when people are performing other activities in bed [53]. Therefore, in delivering health information based on an individual’s data to promote health literacy, it is important to ensure the accuracy of the data and to provide relevant and actionable advice.

Our results show that patients who track their data do not always share those data with clinicians. Patients who shared their data during their clinic visits and those with a higher level of AE reported greater willingness to share their PGD with clinicians in the future. Although participants acknowledged several benefits of sharing their data with clinicians, over half of the participants reported one or more barriers to data sharing. First, participants were uncertain about the relevance of the data to their health or felt that their data were not relevant. Although not all types of PGD are necessarily related to patients’ immediate health concerns, these data can be important health indicators for preventative care [54-56]. For example, an individual’s activity level and caffeine consumption can affect their sleep quality [54]; similarly, changes in one’s heart rate may indicate a risk of cardiovascular disease [56]. To optimize the value of PGD, it is important to empower patients to share their data with clinicians when necessary. We believe that both people and technology play important roles in this process: clinicians can point out specific types of PGD that may be helpful for diagnosis and decision making; tracking tools can be designed to inform patients of how their PGD may or may not be relevant to their present health concerns. For example, by supporting self-experimentation, tracking tools can draw potential correlations between patients’ health indicators (eg, symptoms) and their daily behaviors [22]. These tools can also provide feedback based on the data and other contextual information of individuals to help them reflect on their behaviors and better assess the relevance of their data to their health [16].

Second, participants reported that sharing their PGD with clinicians was sometimes difficult because they did not have enough data. Some participants worried that their data were imprecise, whereas others lacked access to professional devices to measure important body metrics, such as heart rate and blood pressure. To encourage patients to share their PGD with clinicians, we argue that there is a need to better support people’s ability to capture sufficient and accurate PGD over time. Due to the high data capture burden [14] and lack of flexibility in existing tracking tools [13,14], many people cannot adhere to the tracking plans, and may even abandon tracking. To address these challenges, we call for flexible and low-burden tracking tools to better support the ability of patients to collect their health data over a long period, which can help both patients and clinicians. For example, semiautomated tracking tools can be used to lower data capturing burden by leveraging automatic tracking (eg, sleep data from Fitbit [57]) and enable flexible data capture through manual tracking (eg, mood) [58,59]. In addition, efforts need to be made to lower the barriers to tracking devices and to make them more accessible to the public (eg, setting up blood pressure kiosks).

Third, participants were reluctant to share their data because of the limited time they had to meet with their clinician. Participants explained that to make the best use of the time, they did not want to waste the time on sharing their PGD, especially when they considered that their data would not help clinicians. However, because patients are not medical experts, they could not always tell which data are relevant to their health, especially if they kept track of various types of data. Although our work did not examine clinicians’ perspectives on PGD sharing, previous work has found that clinicians are also concerned about the time they need to review PGD in the clinic [6,60] and may even consider these data to be distracting [52]. To resolve the dilemma of clinicians having limited time and patients feeling uncertain about whether their personal health data can be helpful, an approach should be developed to help patients share their data efficiently. For example, Fitbit recently rolled out a health and wellness report [61] that includes a visual or textual summary of a user’s data, which the individual can print out and take to the clinician’s office [4]. Although the relevance of these data to one’s health may be unclear to the patient, we believe that this is a promising step to support patients in sharing their data with their clinicians, leveraging consumer health technologies.

Finally, some participants reported that they were reluctant to share their PGD because of privacy concerns and fear of being judged, especially when asked to share data relating to unhealthy behaviors, such as smoking. This fear may stem from how tracking tools often pass judgment on people’s data, as researchers have noticed many self-tracking tools (eg, MyFitnessPal [50], Lose It! [62]) posit weight gain and extra calorie intake as negative outcomes (eg, using red to highlight these data points). Although such designs serve as an alert for individuals to maintain a healthy lifestyle, they can increase the anxiety of those who have weight concerns or dietary problems [63]. To help patients overcome their fear of being judged, tracking tools should avoid creating an environment that perpetuates stigma and that passes explicit or implicit judgments on users.

This study has limitations that should be considered when interpreting the results and implications. First, although significant, the adjusted R-square statistics in our regression models were relatively low, with a medium effect size (Table 3), indicating that the models explained only a small proportion of the variances in our outcome variables (ie, HSI, AE, willingness to share one’s data with clinicians in the future). This could be partly because of either (1) omitting other confounding factors that could influence the health literacy of participants (eg, socioeconomic status, health insurance) [64,65] and data sharing practices (eg, the questions clinicians asked) or (2) sampling bias resulting from recruiting mostly young, healthy, and tech-savvy college students with adequate technology access and higher HSI and AE than patients in general [66,67]. Second, we characterized the tracking practices of participants based on the number of data types they tracked and how they tracked their data, which did not capture all dimensions related to self-tracking. Third, some participants reported sharing PGD they had not tracked (eg, blood pressure taken in the clinic), possibly because (1) they interpreted PGD as encompassing health-related data captured in the clinic, despite our best efforts to define and communicate its meaning or (2) they were able to retrospectively recall or estimate some forms of PGD, even if they had not explicitly tracked the data (eg, weight). Fourth, to make the survey easy and quick to respond to, we omitted questions inquiring about how the PGD sharing happened during the clinic visit. For example, we did not specifically ask participants what triggered their data sharing (eg, whether participants were asked by the clinician, had proactively shared their data themselves, or had their body metrics assessed in the clinic). In addition, we did not ask participants how they shared their data (eg, verbal communication, screen sharing). Finally, we conducted the survey at a UHC, where the clinicians might not be the primary physicians of the participants, which may limit the generalizability of our findings.

Despite these limitations, we believe that, in optimizing health care for the next generation, investigating the interrelationships between PGD tracking and sharing practices, health literacy, and clinic experiences of tech-savvy college students is an important starting point. Going forward, we aim to further examine whether and how health literacy relates to different PGD tracking and sharing contexts—for example, replicating the study with different patient populations, such as those with more severe health conditions and those with low health literacy. Other dimensions of patients’ self-tracking practices, such as the frequency with which they track their data (eg, routine vs casual trackers), and triggers and methods of data sharing in the clinic warrant future research [68]. In addition, our study focused only on 2 health literacy constructs (HSI and AE). It is worth investigating how patients’ data tracking and sharing practices may relate to other health literacy constructs, such as one’s ability to navigate the health care system [11]. Finally, by surveying patients immediately after they met with a clinician, our study demonstrates the value of gathering immediate perspectives of patients on their experiences with PGD sharing, which allowed us to quantitatively explore the interrelationships between patients’ data tracking practices, data sharing practices, and their health literacy (Figure 1). In the future, similar approaches can be used to examine the immediate perspectives of clinicians on their experiences in responding to and using PGD shared by their patients.

We presented an onsite survey study of 109 patients who had just met with a clinician at a UHC. Our aim was to investigate the potential relationships between patients’ PGD tracking practice, sharing practices, and their health literacy. We found that neither the number of data types the participants tracked nor how they tracked their data related to participants’ perceptions that they had sufficient information to manage their health (HSI). However, participants who reported tracking more types of PGD provided higher ratings when asked about their ability to actively engage with their health care providers. Our results also highlighted that tracking PGD does not always lead to sharing these data with clinicians, as exemplified by the many barriers participants reported regarding data sharing. Reflecting on our findings, we discussed ways to close the gaps between patient tracking and sharing of their PGD. We also suggested design opportunities to improve patient health literacy by leveraging the value of PGD to support patients in assessing the relevance of their PGD to their health and to facilitate patient data sharing with their clinicians.