Visualization of Patient-Generated Health Data: A Scoping Review of Dashboard Designs

Abstract

Background Patient-centered clinical decision support (PC CDS) aims to assist with tailoring decisions to an individual patient's needs. Patient-generated health data (PGHD), including physiologic measurements captured frequently by automated devices, provide important information for PC CDS. The volume and availability of such PGHD is increasing, but how PGHD should be presented to clinicians to best aid decision-making is unclear.

Objectives Identify best practices in visualizations of physiologic PGHD, for designing a software application as a PC CDS tool.

Methods We performed a scoping review of studies of PGHD dashboards that involved clinician users in design or evaluations. We included only studies that used physiologic PGHD from single patients for usage in decision-making.

Results We screened 468 titles and abstracts, 63 full-text papers, and identified 15 articles to include in our review. Some research primarily sought user input on PGHD presentation; other studies garnered feedback only as a side effort for other objectives (e.g., integration with electronic health records [EHRs]). Development efforts were often in the domains of chronic diseases and collected a mix of physiologic parameters (e.g., blood pressure and heart rate) and activity data. Users' preferences were for data to be presented with statistical summaries and clinical interpretations, alongside other non-PGHD data. Recurrent themes indicated that users desire longitudinal data display, aggregation of multiple data types on the same screen, actionability, and customization. Speed, simplicity, and availability of data for other purposes (e.g., documentation) were key to dashboard adoption. Evaluations were favorable for visualizations using common graphing or table formats, although best practices for implementation have not yet been established.

Conclusion Although the literature identified common themes on data display, measures, and usability, more research is needed as PGHD usage grows. Ensuring that care is tailored to individual needs will be important in future development of clinical decision support.

Keywords

clinical decision support - patient-centered care - shared decision-making - patient-generated health data - other clinical informatics applications - patient engagement

Background and Significance

As health care strives to become more patient centered,[1] better methods are needed to capture and use data from patients to inform care delivery. Patient-centered care emphasizes incorporating patient preferences, social factors, and physiologic variations to tailor clinical management to an individual's goals and needs. Information technology applications designed to facilitate such decision-making and incorporate patient-specific data have been described as patient-centered clinical decision support (PC CDS). PC CDS is defined as a set of tools that significantly incorporate patient-centered factors into their underlying knowledge and data, as well as ultimate delivery and use, to support clinical decision-making. The knowledge is based on comparative effectiveness research or patient-centered outcomes research that incorporates meaningful outcomes for patients. The data include, if available, that which is generated directly from patients, which can include patient-reported outcomes (PROs), patient preferences, and some social determinants of health data. The delivery entails directly engaging patients and caregivers via desktop or mobile applications or patient portals across a range of settings. The use supports direct patient and/or caregiver involvement and applying decision support as part of a shared decision-making process.[2]

The Office of the National Coordinator for Health Information Technology (ONC) defines patient-generated health data (PGHD) as “health-related data created, recorded, or gathered by or from patients (or family members or other caregivers) to help address a health concern.”[3] Thus, ONC distinguishes PGHD from other patient-related data generated in clinical settings, because the capture of the data does not involve a health care provider, but only the patient. With advances in mobile and wearable technologies, PGHD comprising physiologic measurements (e.g., blood pressure, heart rate, weight) can now be measured and reported passively, requiring minimal patient attention or effort for the data to be recorded and available for analysis. Devices that capture and transmit such data are now more available and are being adopted worldwide.[4]

PGHD has been noted to have many potential benefits in longitudinal clinical care.[5] It reveals the unique physiologic patterns of an individual patient, thereby informing patient care tailored to a specific individual's needs. It provides information about the patient's health between provider visits and increases patients' engagement in their own care. PGHD can also offer providers more accurate measurements and temporal insight into a particular disease,[6] enabling clinicians to use the data to guide discussions with patients and use clinic time more efficiently.[6] [7]

There are several barriers that prevent the full benefits of PGHD from being realized in clinical care. First, integration with current EHRs and clinical workflows is limited,[6] which is a barrier to development of PC CDS centering on patients' individual needs.[7] Second, large volumes of automatically captured, high-frequency data can be time-consuming to view, and cognitively difficult to interpret, which contributes to clinician burnout.[8] Third, while clinical guidelines offer recommendations based on physiologic measurements obtained in outpatient clinic visits,[9] similar evidence-based interpretation of PGHD is lacking, hindering PGHD use in patient care.[10] Finally, best practices on how to analyze and present large volumes of PGHD in ways that are actionable and support effective decision-making have not yet been established.

Objectives

We sought to identify requirements to inform development of an effective software application, that is, a dashboard that incorporates PGHD and supports shared decision-making between patients and clinicians. We focused specifically on physiologic measurements that could be passively recorded (biometric[3] PGHD), because of the novel challenges with visualization and integration in decision-making of large-volume, high-frequency data as described above.

Methods

We conducted a scoping review of peer-reviewed literature to understand the current landscape for PGHD visualization.[11] We followed the five-stage framework as described by Arksey and O'Malley.[12] Our research questions were: (1) What types of PGHD visualization developed with user input, or evaluated by users, would be helpful for PC CDS? (2) What principles affecting usability and acceptance have been established that could guide PGHD dashboard development?

Literature Search and Screening

We searched PubMed for all relevant publications up to August 2022. Keywords and MeSH terms included “person generated” and “patient generated health data,” “visual*,” “data display,” “clinical decision support,” and “user-computer interface” (the complete query is available in the expanded methods description in [Supplementary Appendix Tables A1–A3], available in the online version).

Our inclusion criteria were as follows: PGHD of physiologic measurements that are or could be passively recorded (e.g., blood pressure, weight, blood glucose), clinician user input either through development or evaluation, and articles published in English. We excluded studies that used types of “PGHD” that are not physiologic (e.g., subjective reports of symptoms or PROs), studies that were designs or proposals with no evaluation or involvement of end users, visualizations of research study data or viewing population data, interfaces that were only for patient usage, review articles, studies of technical integration of PGHD with EHRs, or research on general perspectives on PGHD rather than about a specific dashboard or visualization.

The search retrieved 468 results. As per accepted scoping[11] and rapid review practices,[13] we had one clinical informatician screen all titles and abstracts, conferring with team members on inclusion and exclusion decisions. Sixty-three articles were identified as eligible for full-text review, and ultimately 15 studies met criteria for inclusion in our study ([Fig. 1]).

Fig. 1 Literature flow diagram.

Data Abstraction, Analysis, and Synthesis

We used a data abstraction form to document key characteristics of included studies, including study type, attributes of visualization and dashboard components, extent of clinical implementation, and type of evaluation or user input (see [Appendix Tables A1–A3], available in the online version for detail). We used qualitative thematic synthesis to analyze and summarize our findings.[14]

Results

The studies that met our criteria varied widely in study design, clinical domain, and visualization formats used ([Table 1]). Several were user-centered design studies[15] [16] [17] [18] [19] [20] [21] for the development of PGHD visualizations, which involves soliciting users' feedback on needs, preferences, and workflows in the design and development process. One study[17] primarily gathered user input on impressions and usefulness of visualizations designed to support patients with multiple chronic conditions engaging in cancer care. One was a user study on different visualizations of activity data in arthritis patients.[22] Other studies were at various stages of implementation or varied in their primary research questions. Two were pilot studies for integrating PGHD with an EHR[23] [24] and developed a dashboard for physicians as a side effort. Yet other studies performed evaluations as secondary research objectives, including a retrospective study examining patient factors associated with use of a PGHD patient portal,[25] and an observational study examining the impact of PGHD–EHR integration on diabetes control.[26] Two studies involved user-centered design of a hypertension dashboard,[16] with subsequent postimplementation comparison of the utilization of the dashboard versus the paper report during clinic visits.[27] One study focused on the development of a protocol for analyzing physiologic data capture[28] initiated by patients. Another was a randomized controlled trial that compared the practice of remote PGHD monitoring and virtual visits with traditional, in-person clinic visits, and the impact on diabetes.[29]

Table 1 Study characteristics

Study authors and year	Study type	Clinical domain	Extent of clinical implementation	Type of evaluation or user input
Karni et al 2022[15]	User-centered design and development	Parkinson disease	None, design only with simulated data	Interviews
Cassarino et al 2021[16]	User-centered design and development	Blood pressure as a marker of brain health	None, design and build only	Interviews
Alpert et al 2020[17]	Development and usability study	None specified; general chronic disease	None, design and build only	18-item Likert scale usability survey
Koopman et al 2020[18]	User-centered design	Hypertension	None, design and build only	Focus groups and survey
Backonja et al 2018[19]	User-centered design	Cancer and chronic disease	None, simulated data only	Transcribed interviews (n = 8)
Gardner et al 2017[20]	Development and usability study	Heart failure	None, user study with 6 days of real patient data	SUS (n = 14)
Paterson et al 2017[21]	Development and user study	Hypertension	Development and then implementation	Qualitative feedback reported
Keogh et al 2020[22]	User study	Arthritis	None, design and build only	Focus groups
Salvi et al 2019[23]	Development and pilot study	Diabetes	Pilot integration for 6 months on 27 patients and 3 physicians	System Usability Scale (SUS) evaluation (n = 3)
Kumar et al 2016[24]	Feasibility study for PGHD integration into an EHR	Pediatric type 1 diabetes	Pilot integration of 10 patients	Designed by a team including users
Jung et al 2019[25]	Retrospective analysis examining factors associated with patient usage of portal; physician dashboard development a side effort	General chronic disease	Fully implemented and integrated with EHR	Designed by a team including users
Weissman et al 2016[26]	Observational study, assessing improvement in blood glucose control	Diabetes	Fully implemented and integrated with EHR (965 patients)	Likert scale questionnaire developed by investigators
Cohen et al 2022[27]	Retrospective study	Hypertension	Fully implemented and integrated with EHR	Observational analysis of transcriptions of video recorded visits
Pevnick et al 2020[28]	Protocol development	Atrial fibrillation detection from heart rate and step counts	Fully implemented and integrated with EHR	Designed by a team including users
Hsu et al 2016[29]	Randomized controlled trial of a remote monitoring program compared with usual care, A1c reduction as primary outcome	Type 2 diabetes	Implemented for trial (n = 40 patients, 20 in each arm)	Time needed for clinic visits

Abbreviations: EHR, electronic health records; PGHD, patient-generated health data; SUS, System Usability Scale.

The clinical domains were primarily chronic diseases, with two unspecified. Four studies focused on diabetes with blood glucose level as the primary measure.[21] [22] [24] [27] A scoping review of PGHD–EHR integration similarly found diabetes mellitus to be the most common condition addressed.[30] Other studies we reviewed also displayed blood glucose measurements, but they incorporated these with other PGHD (such as blood pressure, weight, step counts) for a broader focus on general chronic disease or cancer.[17] [23] Three studies that examined control of hypertension collected blood pressure as the sole data type,[16] [19] [25] and another used blood pressure as a marker of brain health.[14] One study focused on heart failure and collected multiple data types.[18] Another study used heart rates and step counts, not to manage a known condition, but to detect previously undiagnosed atrial fibrillation.[28] One study focused on Parkinson's disease,[15] collecting data from wrist and bed sensors, in addition to multiple PROs, with the goal of improving patient self-management at home.

Data Collection

[Table 2] shows the types of data collected by the reviewed studies. As expected, physiologic parameters that change frequently were collected more often than other types of data. Activity data were not collected as instantaneous measurements but rather as tallies, durations, or proportions.

Table 2 Types of patient-generated health data visualized and the frequency with which they were obtained from the patient

Data type	Frequency collected
Heart rate	Near constant monitoring (exact frequency unspecified)
Blood pressure	Daily to multiple times per day
Blood glucose	Range: every 5 min,[24] every 15 min,[23] few[26] to multiple[19] times per day, once daily[29]
Oxygen saturation	Near constant monitoring
Weight	Once daily
Activity data
Step counts	Constant tracking, for day total and frequencies
Activity by intensity category (inactive, light, moderate, vigorous)[16]	Constant tracking (for a 28-d study period)
Sleep data	Constant tracking
Wrist movement (for bradykinesia and dyskinesia)	Constant tracking, producing scores every 2 min[15]

Specific monitoring device types were not tracked for this review due to inconsistent reporting, but we did note patient-generated data collection and entry methods. Some studies only used one device to collect PGHD (e.g., a blood glucose monitor), but one study had patients use five different devices.[20] Most studies recorded data passively, although some used blood pressure measurements actively entered by patients.[18] [21] [27] One study required patient data entry into a portal.[27] Another required patients to send their blood pressure measurement by text message.[21]

Visualization and Dashboard Development

Not all the studies described the design of the specific visualization in the study. Some studies involved clinician users in the design.[15] [16] [20] One homegrown dashboard was described as being “refined over several meetings” with end users.[28] A few studies cited their own prior research about health care professionals' perceptions and preferences regarding PGHD, developing a graphical interface based on the highest ranked factors. No visualizations were explicitly developed to be actual CDS tools. One study did say their dashboard was developed for doctors “to provide recommendations based on their results,”[25] although the paper did not illustrate the specific recommendations for clinical management. One system developed a screen view intended for use by both patients and their health care providers—along with some recommended insulin regimen adjustments based on a simple algorithm (a 3-unit increase or decrease every 3 days, based on the relation to a target blood glucose).[29]

Visualization Formats and Interpretations

Attributes of visualization as extracted from studies are summarized in [Table 3]. Almost all the visualization dashboards included line graphs plotting quantitative data points with trend lines over time. This is consistent with a recurrent principle of clinicians' need to see data longitudinally.[31] [32]

Table 3 Summary findings on patient-generated health data visualization graphical attributes

Graphic category	Findings
Line graph variations	Line connecting data points, 7-day moving average,[16] smoothed line,[18] [27] lines of each day's maximum and minimum value[23] [24] Multiple variables on same plot with different y-axis labels[28] Different markers by source (home vs. clinic blood pressure)[27] Each day over same plot to show blood glucose trends by time of day[24]
Visualization of summary statistics	Box plots of day's range of values[26] Circle plots of percent time spent in states (e.g., activity types[23]; high, low, or target blood glucose[24]) Bar plots of tallies of abnormal values each day[28] Stacked and clustered bar charts indicating activity intensity[22]
Calendar views	Four-week overview highlighting days with abnormal measurements[19] Shaded blocks by day of sleep quality (measured by duration and interruptions)[15]
Statistical interpretations	Descriptive statistics: e.g., maximum, minimum, range, average, median, and standard deviation Percent of data within predetermined categories (e.g., within target blood glucose range)
Clinical interpretations	Binary designation of normal/abnormal Comparison to standard reference ranges, an individual patient's goal,[16] or defined targets based on a guideline incorporating other comorbidities[21]
Usage of color for clinical interpretation	Abnormal values in red,[21] orange,[19] or yellow[40] Green, yellow, and red “stoplight” coding Increasing severity in yellow, orange, red, and maroon for blood pressure by the American Heart Association categories[16] Blood glucose values in yellow for hyperglycemia, green for the target range, and red for hypoglycemia[24]
Other contextual data in view (e.g., from EHR)	Current vasoactive medications, with daily dose and drug class (e.g., “Diuretic” and “Calcium Channel Blocker”), listed in a table, for a blood pressure study[16] Timeline bar plot of antihypertensives[18] or all medications[27] Subjective PRO scores on symptoms[15] [20] Adherence to medication regimen (when collected)[15] [19] [29]
Clinical inferences	Estimated hemoglobin A1c based on the average glucose readings[24] Plot of estimated pharmacokinetics of diabetes medications, to highlight times when medication levels might be subtherapeutic from missed doses[29]

Abbreviations: EHR, electronic health record; PRO, patient-reported outcome.

A few studies noted that familiar types of data presentations are helpful, which influenced their decisions to use line graphs—with the authors inferring that a new or unusual type of visualization would not help clinician understanding.[19] [22] One medical doctor was quoted as saying: “you don't want to have something too novel where you have some bizarre bar graph or some kind of odd, interesting pattern that's in 3D…that people haven't seen [before].”[19]

A user study with calendar overviews noted that some interfaces depicted regions without data present[19]; user feedback indicated that it was unclear whether empty fields meant data were missing or normal. Users wanted values and descriptions of normal ranges, even for common data points like blood pressure and blood glucose.

Stratification of the severity of abnormality was commonly done with green, yellow, and red “stoplight” coding. One legend describes green as “Ok,” yellow “keep an eye,” and red “needs attention.”[17] This is consistent with other literature visualizing PROs, in which users were described as appreciating the use of color for interpretation and that the tricolor designation was described as standard practice.[6] [30]

Color was also used for purposes other than interpretation. In some circle graphs, for example, different colors were used to indicate different regions (e.g., sleep and exercise as different activities[23]). Similarly, color was commonly used to distinguish data series (for example, in one graph,[20] heart rate was plotted in red, activity in green, and oxygen saturation in blue). In one visualization,[20] black blocks were plotted on a timeline when some abnormal parameter occurred, which turned to red when the user selected a particular one to view in further detail. Another study, rather than using the common “stoplight” color coding, explained that “blue, orange, and yellow colors were used because they can be distinguished by individuals with color blindness.”[19] The study noted earlier as plotting multiple days' blood glucose values on the same 24-hour graph used seven different colors for each day of the week.[24]

With respect to other contextual data (i.e., information coming from either the EHR or other data source), studies in which other data appeared with PGHD showed the contextual data viewable in various ways, as shown in [Table 3].[17] [21] [22] [23] [24] [26] [28] Several studies had no other contextual data in the provider's view.

Interactivity with Dashboard

The extent to which the clinician view was interactive was also highly varied. Static reports had no interactivity. One was only a PDF report displaying a table and graph of blood pressures, with summary statistics and sometimes a warning message.[21] Some visualizations had minimal interactivity, such as allowing adjustment of the data lookback period,[15] [17] [18] selection of which data points to view (e.g., home or office blood pressures),[18] or selection of a different screen or view. To identify “nocturnal” hypoglycemia in one visualization, a user could adjust what time of day to consider “night.”[24] Most interactivity was optional, but for some visualizations of specific data it was required. For example, one with black blocks on a timeline necessitated selection of a particular block to view what type of abnormality was captured.[20] Some allowed hovering within a region to get additional data. Pop-ups providing details were described[19] as aligning with the “Shneiderman Visual Information Seeking Mantra” of “Overview first, zoom and filter, then details-on-demand.”[33]

Clinical Decision Support Capabilities

Use of CDS to provide specific recommended actions to take based on the PGHD was limited. A table of blood pressures, for example, gave a “consider change in regimen” message[21] if measurements were outside the target range. One exception was a diabetes system,[29] in which a 3-unit adjustment in insulin dosing would be suggested based on a specific diabetes algorithm at the institution. No other dashboards included specific management recommendations.

Data aggregation and visualization did assist with diagnosis and individual adjustments to treatments. Overlying activity data with heart rate provided insight about the etiology of tachycardia,[28] and calendar views of medications enabled users in one study to observe that missed doses were always on the same weekday.[19] Color-coded weekday plots of blood glucose enabled clinicians managing diabetes to observe that recurring nocturnal hypoglycemia was only happening on two specific weeknights[24]; further inquiry revealed that this was after a patient's sports practice, prompting the decision to decrease the insulin dose only on the specific days, with subsequent resolution of the problem.

Types of Evaluations

Not all the studies that incorporated user input into design reported any evaluation, but those that did used a variety of methods. These included standard usability assessments (often Likert-type scales), informal subjective feedback reported to the authors, transcribed and coded interviews, or other objective metrics such as measurement of the clinical time spent using the implemented systems.

Likert scale usability assessments included the commonly used System Usability Scale (SUS).[34] One study[20] reported a SUS average of 92 with 14 clinician users (scores over 68 are considered to be above average).[35] It is worth noting, however, that—although such a usability assessment measures ease of use and learnability of a technological system—it elicits little information about whether the system was an improvement on what professionals were previously using, or how it met their needed work objectives. A questionnaire developed by the International Organization for Standardization (ISO) to evaluate human–computer interactions (called the ISO 9241/110-S[36]) was used in another study[17]; this instrument uses an 8-point Likert scale for its 18 items about interface clarity, appropriateness, and ease of use. Another study used a 5-point scale ranging from worst to best change on the impact on practice.[26] This study indicated that 79% of physicians who responded “reported willingness to continue using [the system] in the future.”

A few studies reported only informal subjective feedback in their research findings. Example statements include “clinicians found the system easy to use,” “feedback on the layout of the report was favorable,” and “the reports…were found to support and enhance clinical decision making.”[21] In a user-oriented design study, interviews were performed, recorded, and transcribed.[19]

Some studies reported on time spent doing clinical work. One involving self-monitoring of glucose[26] said that 61% of users reported the new system required less time than their previous workflow; 10% said it required more. That study also reported that the subjective quality of clinical decision-making improved for 79% of patients, the quality of discussion improved for 76% and medication adherence improved for 61%. This is consistent with the findings of another pilot study that did not describe data visualization but found that remote monitoring in type 1 diabetes mellitus saved time during clinic visits by allowing providers to focus on behaviors, disease trends, and management.[37] Use of an EHR-integrated dashboard of home blood pressure data resulted in similar or less time needed for discussion in clinic visits compared with viewing paper reports, although the difference was small.[27] A randomized controlled trial comparing remote monitoring to usual clinical care found that health care providers spent approximately one-third the amount of time on virtual visits compared with usual care, with greater improvements in hemoglobin A1c by the end of the 14-week study.[29] This last study was the only example in our review that compared clinical outcomes.

Evaluation Findings

Several recurrent themes emerged of what features clinicians appreciate most about PGHD integration. One study cited speed of use, data comprehensibility and clarity, and simplicity.[26] Aggregation of multiple data types and displaying data trends over time is consistent with research done primarily with PROs.[30]

Specific improvements requested by users, if solicited by the researchers, were along similar lines. A user study received many suggestions for additional information to be included in the data views—including subjective symptoms, patient-reported reasons for particular data points, and additional clinical data such as laboratory results, treatments, and goals of care.[19] Another study found that users desired the ability to annotate visualizations with similar types of information themselves (e.g., reason for stopping treatment, date of starting an exercise program).[18]

Discussion

Overall, the current body of research incorporating clinician user insights or feedback into PGHD visualization for CDS is nascent and heterogeneous, with no best practices for viewing PGHD clearly established. Few studies gave the specific research purpose of evaluating PGHD dashboards. No studies directly addressed which visualizations best supported specific clinical decision-making. Even though the studies varied in design, clinical domain, and visualization formats, themes did emerge regarding longitudinal display, interpretation, and aggregation.

Although not explicitly discussed in the literature or spelled out into any type of taxonomy of PGHD usage, two distinct categories emerge in how different PGHD types factor into clinical decision-making. First, specific physiologic measurements, like blood pressure and blood glucose, are themselves the main targets for clinical intervention and outcome measures. Second, contextual factors are viewed as causing or influencing the primary data of interest, with some studies collecting activity, but only to glean insight into blood glucose[23] or heart rate.[28] For example, in one study the authors state that “daily step count can sometimes aid in understanding whether abnormal heart rates might be due to vigorous exercise.”[28] In another study, the “dawn effect” in blood glucose, defined as a normal blood glucose at night followed by hyperglycemia at wake up time, is a pattern inferred by collecting sleep schedules.[23] It is possible that, for some clinical purposes, activity data could be the primary outcome of interest (e.g., for diagnosing sleep disorders, meeting exercise, or functional goals).

Few studies described systems that provide recommendations on actions to take in response to PGHD, although visualization did enable individualized management changes. Notably, many of the evidence-based clinical guidelines were not developed with the expectation that massive volumes of PGHD would be available to incorporate into the clinician–patient decision-making process. It may be that new clinical management guidelines will need to be developed before the full benefits of PGHD can be realized in PC CDS.

Key Lessons Learned

Recurrent themes throughout individual studies and reviews emphasized several key attributes for development of new PGHD visualizations.

1. Longitudinal data display: the capability to show data over time, commonly in the form of line graphs, was almost universally desired or utilized.

2. Aggregation of multiple data types: having multiple data types, including from different sources, all simultaneously viewable, improves comprehensibility of data context and allows for the user to spot correlations.[8] Clinicians frequently request additional data types to be added to the display. Providers want to see PGHD integrated with other EHR data,[30] as PGHD are not sufficient in themselves for clinical decision-making.[22] [31] According to one framework developed for how to address new PGHD, a key principle is considering how other EHR data can be leveraged as context to understand what the PGHD means.[28] However, there is a need to distinguish between PGHD and health care-generated data,[18] [38] and some clinicians have concerns about the legal liability of viewing data outside their scope of practice.[39]

3. Interpretation and actionability: visualization dashboards need to provide interpretation and actionable suggestions for clinicians,[40] as they have sometimes not referred to or used PGHD because of the lack of actionability.[31] Clinician users request interpretation, meaning, and guidance,[37] even for common data types. Some call for artificial intelligence and machine learning to aid with interpretation.[41]

4. Customization of visualization and more options: it was noted that familiar types of graphs (e.g., line and bar graphs) are better than providing a type of graph not commonly seen by clinicians.[19] [22] However, views that deviate from classic linear time graphs could allow for noticing unusual patterns. It may be that standard visualizations are adequate for most patients, but additional options may prove their greatest benefit with unusual cases. Individual users also will have their own preferences, and usability heuristics indicate that flexibility should be allowed for users to customize data display for their specific needs.[17] [19] [27] [31]

5. Availability for other purposes (e.g., documentation): in some studies, even if PGHD was integrated with an EHR, it was only saved to flowsheets and not permitted to be included in notes.[30] Physicians want to be able to automatically insert PGHD summaries or interpretations into a clinical note. While integration with EHRs is likely underreported and best practices are not yet established, information needs to be where it will be used if it is to be actionable.[30]

6. Speed: clinicians want data visualization options to be simple, compatible with existing EHR systems, and data flow to closely resemble, or align with, existing processes.[21] For example, one dashboard was developed with the goal of being viewed in less than 30 seconds,[16] and quick activation or opening of the dashboard was one feature clinicians reviewed positively.[17]

Our review has certain limitations. The language to describe PGHD in the reviewed literature is varied, and some studies did not use the phrase “patient-generated” at all, meaning that our review may not have captured all salient publications. This yields inconsistent indexing by keywords and MeSH terms, making any comprehensive search difficult. Even so, our findings were consistent with several related reviews, giving us some confidence that our main conclusions are likely generalizable. Our review did not attempt to address how PGHD improves patient engagement or involvement in decision-making, and only in the few studies mentioned were effects on outcomes measured; these were not the focus of this review.

Conclusion

Use of PGHD in clinician-facing dashboards for decision support is still a nascent effort, and no best practices exist based on our literature review. Our findings indicate that longitudinal visualization, data aggregation, speed, actionability, and customization are desired attributes for PGHD visualizations for clinicians. Further effort toward better accessibility of PGHD within clinical workflows is crucial as health care moves toward more patient-centered care.

Clinical Relevance Statement

As PGHD increases in volume and availability, there will be increasing need to incorporate it into shared clinical decision-making in many clinical contexts. This review provides insights on what is currently known on how best to display and aggregate such data in ways that are clinically useful.

Multiple-Choice Questions

1. Which of the following principles is a recurring theme among clinician user perspectives on dashboards using PGHD?

   • Users want to see only one type of data on the screen at a time

   • Users want to have data interpretation and summaries

   • Users want only one visualization option determined for them without need for customization

   • The speed of use is irrelevant in workflows if visualization is clear

   Correct Answer: The correct answer is option b. Clinician users prefer to have multiple types of data on the same screen, and options for customization of visualizations, as well as having data interpretations and summaries. Speed of use for smooth integration into clinical workflows is extremely important for adoption.

2. Current limitations with usage of PGHD in patient-centered clinical decision-making include all of the following except:

   • Lack of guidelines on how to analyze PGHD

   • Poor integration of PGHD data and EHRs

   • Decreasing availability of devices that capture PGHD

   • Difficulty in visualizing and interpret large volumes of PGHD

   Correct Answer: The correct answer is option c. Devices that capture PGHD are increasing in availability, but there are currently limitations for the usage of such data due to lack of guidance on visualization and interpretation, as well as poor integration with existing EHRs.

Conflict of Interest

A.B. is a stockholder in Elimu Informatics.

Protection of Human and Animal Subjects

No human or animal subjects were included in this project.

Authors' Contributions

All authors made substantial contributions to the conception, design, and execution of this research. All authors participated in drafting the manuscript or revising it critically for important intellectual content and gave final approval of the version published.

• References

• 1 Gerteis M. Ed. Through the Patient's Eyes: Understanding and Promoting Patient-Centered-Care. 1st ed.. Jossey-BassTM An Imprint of Wiley; 1993
• 2 Dullabh P, Sandberg SF, Heaney-Huls K. et al. Challenges and opportunities for advancing patient-centered clinical decision support: findings from a horizon scan. J Am Med Inform Assoc 2022; 29 (07) 1233-1243
• 3 What are patient-generated health data? | HealthIT.gov. Accessed June 24, 2022 at: https://www.healthit.gov/topic/otherhot-topics/what-are-patient-generated-health-data
• 4 Loncar-Turukalo T, Zdravevski E, Machado da Silva J, Chouvarda I, Trajkovik V. Literature on wearable technology for connected health: scoping review of research trends, advances, and barriers. J Med Internet Res 2019; 21 (09) e14017
• 5 Boston D, Cohen D, Stone J. et al. Integrating patient-generated health data into electronic health records in ambulatory care settings: a practical guide. Prepared for: Agency of Healthcare Research and Quality Contract No. 75Q80120D00019 AHRQ Publication No. AHRQ 22-0013; 2021
• 6 Cohen DJ, Keller SR, Hayes GR, Dorr DA, Ash JS, Sittig DF. Integrating patient-generated health data into clinical care settings or clinical decision-making: lessons learned from project HealthDesign. JMIR Human Factors 2016; 3 (02) e26
• 7 Dullabh P, Heaney-Huls K, Lobach DF. et al. The technical landscape for patient-centered CDS: progress, gaps, and challenges. J Am Med Inform Assoc 2022; 29 (06) 1101-1105
• 8 Ye J. The impact of electronic health record-integrated patient-generated health data on clinician burnout. J Am Med Inform Assoc 2021; 28 (05) 1051-1056
• 9 Unger T, Borghi C, Charchar F. et al. 2020 International Society of Hypertension global hypertension practice guidelines. Hypertension 2020; 75 (06) 1334-1357
• 10 Lavallee DC, Lee JR, Austin E. et al. mHealth and patient generated health data: stakeholder perspectives on opportunities and barriers for transforming healthcare. mHealth 2020; 6: 8
• 11 Pham MT, Rajić A, Greig JD, Sargeant JM, Papadopoulos A, McEwen SA. A scoping review of scoping reviews: advancing the approach and enhancing the consistency. Res Synth Methods 2014; 5 (04) 371-385
• 12 Arksey H, O'Malley L. Scoping studies: towards a methodological framework. Int J Soc Res Methodol 2005; 8: 19-32
• 13 Garritty C, Gartlehner G, Nussbaumer-Streit B. et al; Interim Guidance from the Cochrane Rapid Reviews Methods Group. Cochrane Rapid Reviews Methods Group offers evidence-informed guidance to conduct rapid reviews. J Clin Epidemiol 2021; 130: 13-22
• 14 Barnett-Page E, Thomas J. Methods for the synthesis of qualitative research: a critical review. BMC Med Res Methodol 2009; 9 (01) 59
• 15 Karni L, Jusufi I, Nyholm D, Klein GO, Memedi M. Toward improved treatment and empowerment of individuals with Parkinson disease: design and evaluation of an internet of things system. JMIR Form Res 2022; 6 (06) e31485
• 16 Cassarino N, Bergstrom B, Johannes C, Gualtieri L. Monitoring older adult blood pressure trends at home as a proxy for brain health. Online J Public Health Inform 2021; 13 (03) e16
• 17 Alpert JM, Kota NSP, Ranka S. et al. A simulated graphical interface for integrating patient-generated health data from smartwatches with electronic health records: usability study. JMIR Human Factors 2020; 7 (04) e19769
• 18 Koopman RJ, Canfield SM, Belden JL. et al. Home blood pressure data visualization for the management of hypertension: designing for patient and physician information needs. BMC Med Inform Decis Mak 2020; 20 (01) 195
• 19 Backonja U, Haynes SC, Kim KK. Data visualizations to support health practitioners' provision of personalized care for patients with cancer and multiple chronic conditions: user-centered design study. JMIR Human Factors 2018; 5 (04) e11826
• 20 Gardner CL, Liu F, Fontelo P, Flanagan MC, Hoang A, Burke HB. Assessing the usability by clinicians of VISION: a hierarchical display of patient-collected physiological information to clinicians. BMC Med Inform Decis Mak 2017; 17 (01) 41
• 21 Paterson M, McAulay A, McKinstry B. Integrating third-party telehealth records with the general practice electronic medical record system: a solution. J Innov Health Inform 2017; 24 (04) 915
• 22 Keogh A, Johnston W, Ashton M. et al. “It's not as simple as just looking at one chart”: a qualitative study exploring clinician's opinions on various visualisation strategies to represent longitudinal actigraphy data. Digit Biomark 2020; 4 (Suppl. 01) 87-99
• 23 Salvi E, Bosoni P, Tibollo V. et al. Patient-generated health data integration and advanced analytics for diabetes management: the AID-GM platform. Sensors (Basel) 2019; 20 (01) E128
• 24 Kumar RB, Goren ND, Stark DE, Wall DP, Longhurst CA. Automated integration of continuous glucose monitor data in the electronic health record using consumer technology. J Am Med Inform Assoc 2016; 23 (03) 532-537
• 25 Jung SY, Kim JW, Hwang H. et al. Development of comprehensive personal health records integrating patient-generated health data directly from Samsung S-Health and Apple health apps: retrospective cross-sectional observational study. JMIR Mhealth Uhealth 2019; 7 (05) e12691
• 26 Weissmann J, Mueller A, Messinger D, Parkin CG, Amann-Zalan I. Improving the quality of outpatient diabetes care using an information management system: results from the observational VISION study. J Diabetes Sci Technol 2015; 10 (01) 76-84
• 27 Cohen DJ, Wyte-Lake T, Canfield SM. et al. Impact of home blood pressure data visualization on hypertension medical decision making in primary care. Ann Fam Med 2022; 20 (04) 305-311
• 28 Pevnick JM, Elad Y, Masson LM, Riggs RV, Duncan RG. Patient-initiated data: our experience with enabling patients to initiate incorporation of heart rate data into the electronic health record. Appl Clin Inform 2020; 11 (04) 671-679
• 29 Hsu WC, Lau KHK, Huang R. et al. Utilization of a cloud-based diabetes management program for insulin initiation and titration enables collaborative decision making between healthcare providers and patients. Diabetes Technol Ther 2016; 18 (02) 59-67
• 30 Tiase VL, Hull W, McFarland MM. et al. Patient-generated health data and electronic health record integration: a scoping review. JAMIA Open 2020; 3 (04) 619-627
• 31 Zhang R, Burgess ER, Reddy MC. et al. Provider perspectives on the integration of patient-reported outcomes in an electronic health record. JAMIA Open 2019; 2 (01) 73-80
• 32 Rotenstein LS, Agarwal A, O'Neil K. et al. Implementing patient-reported outcome surveys as part of routine care: lessons from an academic radiation oncology department. J Am Med Inform Assoc 2017; 24 (05) 964-968
• 33 Shneiderman B. The eyes have it: a task by data type taxonomy for information visualizations. In: VL '96 Proceedings of the 1996 IEEE Symposium on Visual Languages. IEEE Computer Society Press; 1996: 336 . Accessed July 29, 2022 at: https://www.cs.umd.edu/~ben/papers/Shneiderman1996eyes
• 34 Brooke J. SUS: a quick and dirty usability scale. Usability Eval Ind. 1995: 189 . Accessed July 29, 2022 at: https://www.researchgate.net/publication/228593520_SUS_A_quick_and_dirt
• 35 Affairs AS for P. System Usability Scale (SUS). September 6, 2013. Accessed July 7, 2022 at: https://www.usability.gov/how-to-and-tools/methods/system-usability-scale.html
• 36 Langer J, Zeiller M. Evaluation of the user experience of interactive infographics in online newspapers. Forum Media Technol 2017; x: 97-106
• 37 Crossen S, Romero C, Reggiardo A, Michel J, Glaser N. Feasibility and impact of remote glucose monitoring among patients with newly diagnosed type 1 diabetes: single-center pilot study. JMIR Diabetes 2022; 7 (01) e33639
• 38 Jeevanandan N, Nøhr C. Patient-generated health data in the clinic. Stud Health Technol Inform 2020; 270: 766-770
• 39 Reading MJ, Merrill JA. Converging and diverging needs between patients and providers who are collecting and using patient-generated health data: an integrative review. J Am Med Inform Assoc 2018; 25 (06) 759-771
• 40 Blackford AL, Wu AW, Snyder C. Interpreting and acting on PRO results in clinical practice: lessons learned from the patient viewpoint system and beyond. Med Care 2019; 57 (Suppl 5 1): S46-S51
• 41 Melstrom LG, Rodin AS, Rossi LA, Fu Jr P, Fong Y, Sun V. Patient generated health data and electronic health record integration in oncologic surgery: a call for artificial intelligence and machine learning. J Surg Oncol 2021; 123 (01) 52-60

Publication History

Received: 05 June 2023

Accepted: 11 September 2023

Accepted Manuscript online:
13 September 2023

Article published online:
22 November 2023