Abstract
Objective
We analyze observed reductions in physician note length and documentation time, 2 contributors to electronic health record (EHR) burden and burnout.
Materials and Methods
We used EHR metadata from January to May, 2021 for 130 079 ambulatory physician Epic users. We identified cohorts of physicians who decreased note length and/or documentation time and analyzed changes in their note composition.
Results
37 857 physicians decreased either note length (n = 15 647), time in notes (n = 15 417), or both (n = 6793). Note length decreases were primarily attributable to reductions in copy/paste text (average relative change of –18.9%) and templated text (–17.2%). Note time decreases were primarily attributable to reductions in manual text (–27.3%) and increases in note content from other care team members (+21.1%).
Discussion
Organizations must consider priorities and tradeoffs in the distinct approaches needed to address different contributors to EHR burden.
Conclusion
Future research should explore scalable burden-reduction initiatives responsive to both note bloat and documentation time.
Keywords: electronic health records, physician burnout, health policy, documentation
INTRODUCTION
Electronic health record (EHR) usability issues and documentation burden are substantial,1–6 and considered a key contributor to increasing rates of physician burden and associated burnout.7 This burden, attributed in part to regulatory requirements for physician billing,1 is comprised of 2 related but distinct issues. First, US physicians spend substantial time in the EHR, with a third of that time spent completing documentation, dwarfing other domains of EHR use like clinical information review, order placement, and inbox management.8,9 Second, notes are often “bloated”: excessively long and often filled with information considered redundant and unnecessary for clinical decision-making. Use of copy/paste10,11 as well as shortcut features to drop in large blocks of templated text12,13 contribute to bloated notes that are hard to navigate, lack clinical value, and may contribute to safety risks and diagnostic errors.14–17
Health system leaders have a vested interest in addressing these burdensome aspects of EHR use. And yet, documentation behaviors are slow to change—even in response to guideline changes explicitly intended to reduce documentation requirements.18 This may be because health care organizations struggle with a “one size fits none” problem of note optimization. Physicians are given significant flexibility in their approach to note composition and use of various tools and resources to support note-writing, leading to highly variable documentation practices.19,20 Organizations need to be clear on the end goal (reduced time, reduced note length, or both), then align and promote strategies accordingly. However, it is unclear what the “right” strategies are, and in what combination to balance potentially competing priorities.
In this study, we leverage a unique longitudinal nationwide dataset to identify outpatient physicians who demonstrate a reduction in note length and/or time spent in the EHR on notes. We focus on reductions in the months immediately following the 2021 implementation for reduced documentation requirements for Evaluation and Management (E/M) visits, a major policy change architected by the Centers for Medicare and Medicaid Services (CMS), and the American Medical Association (AMA).21 Among other changes, this policy eliminated the required history and physical exam portion of notes in cases where the information was not germane to the visit, thereby aiming to reduce documentation burden and note bloat. We first examine the extent to which changes in these 2 dimensions co-occur for the same physicians. We then analyze how physicians achieve these reductions in note length and/or time in notes based on changed use of various note support tools and strategies. These findings can support policymakers and health system leaders in prioritizing documentation optimization approaches that target time in notes, note length, or both, given local efforts to improve physician satisfaction, efficiency, and quality.
MATERIALS AND METHODS
Data
Our dataset consisted of de-identified weekly EHR usage metadata from 130 079 US outpatient physician users of the Epic EHR. Data were extracted from the Epic Signal data warehouse, a frequently used source of EHR usage metadata described in further detail elsewhere.8,22,23 Epic has the largest national share of ambulatory practices, and our study represents one of the largest physician-level datasets of EHR metadata compiled to date.24,25 Our study period covered 21 weeks from January, 2021 (start of the E/M guideline change) through May, 2021. We excluded physicians with fewer than 17 weeks of documentation data and used linear interpolation to otherwise ensure data completeness. This study was deemed exempt as non-human subjects research by the University of Pennsylvania Institutional Review Board.
Measuring note length and time in notes
We operationalized 2 measures of physician documentation burden: clinical note length and time spent in the EHR documenting (“time in notes”). Our measure of overall note length was constructed as the average total number of characters per visit note, aggregated at the physician-week level. To measure the time physicians spent in notes, we computed a volume-adjusted measure of “active time per note” by dividing total active time in notes by the total number of notes a physician signed in each week.22
Mechanisms underlying documentation burden reductions
We constructed 5 measures capturing potential mechanisms that may underlie changes to both note length and time in notes. First, we constructed 3 measures of note composition as the number of characters per note contributed by each of manual text entry, copy and paste, and SmartTools (Epic’s templated text and form shortcuts). On average, these 3 text sources contribute more than 80% of note text per physician-week in our sample. Second, we constructed 2 measures of note contribution source: characters per note authored by the focal physician and characters per note authored by other care team members.
Statistical analysis
To identify physicians who reduced note length over time, we calculated correlation coefficients between note length (characters per note) and study week (ie, numeric measure of weeks 1 through 21) for each physician. Physicians with a correlation coefficient at least one negative standard deviation from the overall mean were considered “note length decreasers”. We repeated this procedure for time in notes, using the same criteria of one-negative standard deviation or more from the mean correlation between time in notes and study week to identify physicians who were “note time decreasers”.
Having identified physicians who decreased either note length or time in notes over the study period, we categorized physicians into 4 mutually exclusive groups: (1) those who decreased only note length; (2) those who decreased only time in notes; (3) those who decreased both note length and time in notes; and (4) those who decreased neither. We first calculated descriptive statistics for the full physician sample as well as each of the 4 categories of physicians in our sample. Second, we visualized study period trends in average note length and time in notes across the 4 physician categories to establish face validity for our definitions of “decreasers”. Finally, we examine how the underlying note composition mechanisms (text type and text contributors) changed within each physician group across the study period. We did this by calculating within-group relative changes (percent decrease or increase) for each of the 5 mechanism measures between the first and final week of the study period.
RESULTS
Our study sample consisted of 130 079 physicians across 309 organizations. The sample was 43.3% primary care physicians (n = 56 364); 48.2% from medical specialties (n = 62 644); and 8.5% from surgical specialties (n = 11 071). Specialties were similarly distributed within each of the 4 “note decreaser” categories. For physicians who only decreased note length (n = 15 647), average note length dropped from 4996 to 3996 characters per note over the study period, a 20.0% decline (Figure 1). Physicians who decreased only time in notes (n = 15 417) had an average drop from 7.1 min per note to 4.4 min per note by the end of the period, a 38.0% decline. Physicians who decreased both note length and time in notes (n = 6793) had a starting average note length of 4943 characters per note and average time in notes of 7.1 min per note; these declined to 3858 characters per note and 4.5 min per note, respectively, over the study period. Of note, among physicians who only decreased note length, time in notes stayed relatively flat; among physicians who only decreased note time, note length increased by almost 200 characters per note (Figure 1). The remaining 92 222 physicians were classified as “non-decreasers”; these physicians demonstrated slight increases in both note length and time in notes (Figure 1 and Table 1).
Figure 1.
Trends in note length and time in notes across physician groups. Lines depict smoothed trends to adjust for week-to-week noise related to holidays and other seasonal trends. The left axis measures time in notes per note (the solid line in all panels), while the right axis measures note length per note (the dashed line in all panels).
Table 1.
Descriptive statistics, stratified by group
| Physician Group | Variable | Start of period mean (SD) | End of period mean (SD) | % Change |
|---|---|---|---|---|
| Decreased both note length and time in notes n = 6793 | Note length (char. per note) | 4942.6 (2799.4) | 3858.2 (2259.5) | –21.9% |
| Time in notes (min per note) | 7.1 (9.2) | 4.5 (5.8) | –36.3% | |
| Manual text (char. per note) | 407.7 (521.6) | 280.9 (357) | –31.1% | |
| Copy/paste text (char. per note) | 590.6 (1301) | 461.3 (994.6) | –21.9% | |
| SmartTools text (char. per note) | 2080 (1662.2) | 1590.8 (1317.2) | –23.5% | |
| physician characters per note | 3656.2 (2620.7) | 2807.7 (2057.6) | –23.2% | |
| Non-physician characters per note | 466.8 (1284.4) | 372.7 (1040.4) | –20.2% | |
| Decreased note length n = 15,647 | Note length | 4995.8 (3061) | 3996.5 (2488.1) | –20.0% |
| Time in notes | 5.6 (8.2) | 5.6 (8.3) | –1.2% | |
| Manual text | 322.7 (425.7) | 288 (342.4) | –10.8% | |
| Copy/paste text | 608.1 (1523) | 493.3 (1199.7) | –18.9% | |
| SmartTools text | 1862.9 (1578.6) | 1542.9 (1312.7) | –17.2% | |
| Physician characters per note | 3288.7 (2653.6) | 2754.6 (2171.5) | –16.2% | |
| Non-physician Characters per note | 623.3 (1643.9) | 446.2 (1213.6) | –28.4% | |
| Decreased time in notes n = 15 417 | Note length | 4690.6 (3018.8) | 4837.8 (3012.7) | 3.1% |
| Time in notes | 7.1 (12.2) | 4.4 (4.2) | –38.5% | |
| Manual text | 363.2 (447) | 264 (331.9) | –27.3% | |
| Copy/paste text | 622.3 (1439) | 624.8 (1365.2) | 0.4% | |
| SmartTools text | 1863.6 (1581.8) | 1795.4 (1584.4) | –3.7% | |
| Physician characters per note | 3343.7 (2676.9) | 3180.9 (2621.1) | –4.9% | |
| Non-physician characters per note | 475.3 (1358.2) | 575.6 (1513) | 21.1% | |
| Non-decreasers n = 92 222 | Note length | 4608.9 (3120.6) | 4846.7 (3183.2) | 5.2% |
| Time in notes | 5.9 (8.2) | 6.3 (8.9) | 5.9% | |
| Manual text | 332.2 (411.5) | 329.7 (412.4) | –0.8% | |
| Copy/paste text | 712 (1651.9) | 754.2 (1647.3) | 5.9% | |
| SmartTools text | 1778.2 (1565.6) | 1865.7 (1648.9) | 4.9% | |
| Physician characters per note | 3281.5 (2725.8) | 3430.4 (2800.5) | 4.5% | |
| Non-physician characters per note | 470.1 (1429.3) | 497.6 (1482.4) | 5.9% |
Changes in note composition
Each of the 4 physician groups demonstrated distinct patterns of changes in note composition over the study period. Physicians who decreased both note length and time in notes demonstrated relative decreases of over 20% across all 5 mechanism measures. Of the 3 measures of note composition, manual text decreased the most for these physicians by 31.1% on average (Figure 2). Physicians who only decreased note length also saw decreases in all 5 composition measures, with text from other team members decreasing the most (28.4%). Physicians who only decreased time in notes had large decreases in manual text (27.3%) paired with minimal changes to copy/paste and SmartTools text. These physicians also had large increases in the amount of text contributed by other team members (21.1%). This was the only substantial increase in any dimension of note composition in any of the 3 groups of “decreaser” physicians.
Figure 2.
Changes in note composition and team contribution across physician groups, January 1, 2021 to May 31, 2021. Figure depicts average relative changes in each documentation composition mechanism from the beginning of the study period to the end of the study period.
DISCUSSION
Using a novel dataset of physician EHR use measures that capture note length, time in notes, and note composition, we found that decreasing note bloat and decreasing physician time in notes are distinct goals supported by different strategies of note composition. Of the 37 857 physicians identified as either note length or note time decreasers, only 6793 (17.9%) achieved decreases in both measures. Our exploration of mechanisms further supports this interpretation, as we show that decreases in these 2 measures are functions of distinct mechanisms of note composition and team-based contribution. These distinct—and in the case of team-based note text, divergent—underlying forces suggest that most physicians may be pursuing one of these goals at the expense of the other. As a result, the endpoint chosen matters greatly for measuring the success of burden reduction initiatives, as reductions in note length may reduce bloat without saving time (and vice versa). Physicians and organizations should determine the most pertinent endpoint for their own current state and operational goals; our results provide insight into the strategies that may prove successful in attaining those goals.
Time in notes is likely to be a more salient aspect of burden to physicians than note length. Note length has not demonstrated a consistent relationship with burnout measures,5,26 but excess time in the EHR has been correlated with several measures of physician burnout.5,27–29 Furthermore, relative to other EHR use domains, active time in notes is the largest single share of physician EHR time, occupying between one-third and one-half of all active EHR time.8,9 Our findings illustrate that decreasing time in notes is primarily a function of manual text reduction and teamwork. Neither copy/paste text nor templated text increased meaningfully among note time decreasers, contrary to the branding of these “efficiency” tools. Efforts to optimize documentation to reduce burden should identify opportunities to deploy automated text in more value-added ways. For example, organizations could focus on support of concise shortcuts or macros (eg, Epic’s SmartPhrases) explicitly intended to replace manually authored text to describe care plans for commonly seen clinical problems (eg, upper respiratory tract infections). Use cases like this have the potential to save clinical time while preserving meaningful physician input to the note and valuable note content. EHR vendors can support these efforts by developing and disseminating sample templates that organizations can adapt to their local environments.
Time in notes also appears responsive to increased collaboration in note writing, implying that scribes and multi-collaborator note-authoring tools may play an important role in reducing the physician time burden tied to documentation. Our analysis shows that physicians who only reduced their time in notes counterintuitively increased note length on average. To the extent that support such as scribe-assisted documentation can increase prevalence of bloated notes that may be even harder to navigate and/or lack the distilled information crucial for clinical decision-making, organizations need proactive strategies to mitigate this unintended consequence of removing physicians from direct documentation. Additional studies are required to establish efficacy and more precisely estimate the full causal impact of documentation support, not only on time spent and the length of notes but also on other aspects of EHR burden like time after hours and on off days, both of which are antecedents to physician burnout.
Finally, any efforts to reduce documentation burden should prioritize preserving (or distilling) the clinical value of notes, rather than exacerbating the existing “catch-all” role that notes sometimes play. Helping physicians to produce more concise notes with less time spent in the process is important for advancing efficiency and productivity, but organizations cannot risk overcorrecting at the expense of note quality. While we cannot observe note quality directly in our data, substantial variation across providers in documentation practices, time in notes, and note length19,23 suggest that there is ample room for organizations to optimize documentation without negative impacting patient care.
Limitations
Our study has several limitations. First, our study design samples on the dependent variable; specifically, we only look at physicians with statistically identified decreases in note length and time in notes. Because our aim was to better understand how physicians who shorten notes and reduce documentation time do so, sampling to look only at this population is appropriate. Second, we are unable to observe in our data the specific roles of team members who contributed to note text; we are similarly unable to observe the nature of the manual text that was reduced, as the data contain only aggregated character counts by text source rather than actual note text. This second aspect precludes us from drawing conclusions regarding note quality. Third, our sample includes only outpatient physicians. Results from this analysis may not translate to the distinct care delivery environment and associated documentation behaviors in the inpatient setting. Finally, our sample includes only organizations using Epic’s ambulatory EHR software. While Epic has the largest ambulatory market share, its client base is not representative of the mix of ambulatory care organizations in the United States and is less commonly adopted among independent, small, and rural practices.
CONCLUSION
In this study, we analyzed mechanisms underlying decreases in note length and time in notes for a national sample of physicians during the first 5 months of 2021. We found that these 2 aspects of documentation burden are distinct, as a minority of physicians achieved reductions in both domains. Decreased time in notes is associated with team-based support, but not use of efficiency tools that leverage copied or templated text. More judicious, concise, and value-oriented use of these tools may contribute to decreased time as well as note length. Future research should explore scalable strategies that help organizations design and implement burden-reduction initiatives responsive to both note bloat and documentation time.
FUNDING
NCA was supported by a training grant from the Agency for Healthcare Research and Quality (T32-HS026116-04) at the University of Pennsylvania, Perelman School of Medicine during the production of the manuscript. DAC is supported by the National Institutes of Health’s National Center for Advancing Translational Sciences, grants KL2TR002492 and UL1TR002494 at the University of Minnesota. The content is solely the responsibility of the authors and does not necessarily represent the official views of AHRQ or the National Institutes of Health’s National Center for Advancing Translational Sciences. This study did not receive any other direct funding.
AUTHOR CONTRIBUTIONS
All authors conceptualized the study. NCA conducted analyses and drafted the initial manuscript. AJH and SF provided critical review and input on analysis and study design. DAC oversaw analysis and provided manuscript revisions.
ACKNOWLEDGMENTS
We thank Debra Laumer, Santana Briones, and Sarah LaMar from the PennMedicine EHR Transformation team for their assistance in acquiring pilot data for this study. We thank Dr Bill Hanson, Dr John McGreevey, Dr JT Howell, and Dr Rachel Werner for their support for the project. Additionally, we thank Chris Gates from Epic Systems for his assistance in extracting the data for this study.
CONFLICT OF INTEREST STATEMENT
None declared.
Contributor Information
Nate C Apathy, National Center for Human Factors in Healthcare, MedStar Health Research Institute, Washington, District of Columbia, USA; Center for Biomedical Informatics, Regenstrief Institute, Indianapolis, Indiana, USA.
Allison J Hare, Brigham and Women’s Hospital, Boston, Massachusetts, USA.
Sarah Fendrich, Emmett Interdisciplinary Program in Environment & Resources, Doerr School of Sustainability, Stanford University, Stanford, California, USA.
Dori A Cross, Division of Health Policy & Management, School of Public Health, University of Minnesota, Minneapolis, Minnesota, USA.
DATA AVAILABILITY
Data available upon request.
REFERENCES
- 1. Downing NL, Bates DW, Longhurst CA.. Physician burnout in the electronic health record era: are we ignoring the real cause? Ann Intern Med 2018; 169 (1): 50. [DOI] [PubMed] [Google Scholar]
- 2. Melnick ER, Fong A, Nath B, et al. Analysis of electronic health record use and clinical productivity and their association with physician turnover. JAMA Netw Open 2021; 4 (10): e2128790. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Kroth PJ, Morioka-Douglas N, Veres S, et al. The electronic elephant in the room: physicians and the electronic health record. JAMIA Open 2018; 1 (1): 49–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Akbar F, Mark G, Prausnitz S, et al. Physician stress during electronic health record inbox work: in situ measurement with wearable sensors. JMIR Med Inform 2021; 9 (4): e24014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Hilliard RW, Haskell J, Gardner RL.. Are specific elements of electronic health record use associated with clinician burnout more than others? J Am Med Inform Assoc 2020; 27 (9): 1401–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Gardner RL, Cooper E, Haskell J, et al. Physician stress and burnout: the impact of health information technology. J Am Med Inform Assoc 2019; 26 (2): 106–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Rotenstein LS, Torre M, Ramos MA, et al. Prevalence of burnout among physicians: a systematic review. JAMA 2018; 320 (11): 1131–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Holmgren AJ, Downing NL, Bates DW, et al. Assessment of electronic health record use between US and non-US health systems. JAMA Intern Med 2021; 181 (2): 251. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Tai-Seale M, Olson CW, Li J, et al. Electronic health record logs indicate that physicians split time evenly between seeing patients and desktop medicine. Health Aff (Millwood) 2017; 36 (4): 655–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Huang AE, Hribar MR, Goldstein IH, Henriksen B, Lin W-C, Chiang MF.. Clinical documentation in electronic health record systems: analysis of similarity in progress notes from consecutive outpatient ophthalmology encounters. AMIA Annu Symp Proc 2018; 2018: 1310–8. [PMC free article] [PubMed] [Google Scholar]
- 11. Brown PJ, Marquard JL, Amster B, et al. What do physicians read (and ignore) in electronic progress notes? Appl Clin Inform 2014; 5 (2): 430–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Cimino JJ. Putting the “why” in “EHR”: capturing and coding clinical cognition. J Am Med Inform Assoc 2019; 26 (11): 1379–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Johnson KB, Neuss MJ, Detmer DE.. Electronic health records and clinician burnout: a story of three eras. J Am Med Inform Assoc 2020; 28 (5): 967–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Singh H, Giardina TD, Meyer AND, et al. Types and origins of diagnostic errors in primary care settings. JAMA Intern Med 2013; 173 (6): 418–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Cohen R, Elhadad M, Elhadad N.. Redundancy in electronic health record corpora: analysis, impact on text mining performance and mitigation strategies. BMC Bioinformatics 2013; 14: 10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Gantzer HE, Block BL, Hobgood LC, Tufte J.. Restoring the story and creating a valuable clinical note. Ann Intern Med 2020; 173 (5): 380–2. [DOI] [PubMed] [Google Scholar]
- 17. Rule A, Bedrick S, Chiang MF, et al. Length and redundancy of outpatient progress notes across a decade at an academic medical center. JAMA Netw Open 2021; 4 (7): e2115334. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Apathy NC, Hare AJ, Fendrich S, Cross D.. Early changes in billing and notes after Evaluation and Management guideline change. Ann Intern Med 2022; 175 (4): 499–504. [DOI] [PubMed] [Google Scholar]
- 19. Cohen GR, Friedman CP, Ryan AM, et al. Variation in physicians’ electronic health record documentation and potential patient harm from that variation. J Gen Intern MED 2019; 34 (11): 2355–67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Rule A, Hribar MR.. Frequent but fragmented: use of note templates to document outpatient visits at an academic health center. J Am Med Inform Assoc 2021; 29 (1): 137–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. American Medical Association. Implementing CPT Evaluation and Management (E/M) Revisions. The American Medical Association; 2021. https://www.ama-assn.org/practice-management/cpt/implementing-cpt-evaluation-and-management-em-revisions. Accessed August 4, 2021. [Google Scholar]
- 22. Baxter SL, Apathy NC, Cross DA, Sinsky C, Hribar MR.. Measures of electronic health record use in outpatient settings across vendors. J Am Med Inform Assoc 2021; 28 (5): 955–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Rotenstein LS, Jay Holmgren A, Downing NL, Bates DW.. Differences in total and after-hours electronic health record time across ambulatory specialties. JAMA Intern Med 2021; 181 (6): 863. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Overhage JM, Johnson KB.. Pediatrician electronic health record time use for outpatient encounters. Pediatrics 2020; 146 (6): e20194017. [DOI] [PubMed] [Google Scholar]
- 25. Overhage JM, McCallie D Jr. Physician time spent using the electronic health record during outpatient encounters: a descriptive study. Ann Intern Med 2020; 172 (3): 169–74. [DOI] [PubMed] [Google Scholar]
- 26. Nguyen OT, Turner K, Apathy NC, et al. Primary care physicians’ electronic health record proficiency and efficiency behaviors and time interacting with electronic health records: a quantile regression analysis. J Am Med Inform Assoc 2022; 29 (3): 461–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Yan Q, Jiang Z, Harbin Z, Tolbert PH, Davies MG.. Exploring the relationship between electronic health records and provider burnout: a systematic review. J Am Med Inform Assoc 2021; 28 (5): 1009–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Micek MA, Arndt B, Tuan W-J, et al. Physician burnout and timing of electronic health record use. ACI Open 2020; 4 (1): e1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Adler-Milstein J, Zhao W, Willard-Grace R, Knox M, Grumbach K.. Electronic health records and burnout: time spent on the electronic health record after hours and message volume associated with exhaustion but not with cynicism among primary care clinicians. J Am Med Inform Assoc 2020; 27 (4): 531–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Data available upon request.