Demographic Disparities in Video Visit Telemetry: Understanding Telemedicine Utilization

ABSTRACT

Objectives: To investigate demographic disparities in failed episodes of telemedicine utilization. The primary hypothesis was that certain demographic groups, including older adults and specific racial or ethnic groups, would experience disparate amounts of failed video visits.

Study Design: A retrospective review was conducted using electronic health record–integrated scheduled telehealth video visit telemetry data gathered for all video visits at a California academic health center from September 1, 2020, to November 30, 2020. For each visit, we collected demographics including age, sex, ethnicity, primary language, and race.

Methods: Outcomes were categorized as successful or failed based on review of telemetry data. Successful visits were defined as simultaneous connections and completion of video visit, whereas failed visits were defined as provider-reported failure or lack of simultaneous connections for the telemedicine visit. Binomial generalized logistic regression using a generalized estimating equation approach was used to assess the impact of demographic factors on video visit success. Of 47,065 scheduled telemedicine video visits, telemetry data were available for 30,996; the 16,069 visits excluded from the study were due to no-shows, cancellations, or a nonintegrated solution being utilized.

Results: Of 30,996 visits included in the study, 27,273 were successfully completed. Analysis of the 3723 failed visits revealed that older adults and African American/Black patients were more likely to experience failed video visits, with ORs of 2.02 and 1.56, respectively.

Conclusions: This study highlights the significant demographic disparities in failed video visit occurrence caused by technical failure as demonstrated by telemetry data. These findings highlight the need for targeted interventions and opportunity for improved outcomes.

Am J Manag Care. 2025;31(3):e69-e73. https://doi.org/10.37765/ajmc.2025.89699

Takeaway Points

A stratified analysis of telehealth video visit telemetry data reveals higher failure rates among older adults and African American/Black individuals. Enhanced technical support, patient preparation, and risk assessment tools are recommended to mitigate these disparities and improve digital health equity.

• Older adults were 102% more likely than adults younger than 65 years and African American/Black individuals were 56% more likely than individuals of other races to experience telehealth video visit failures.
• An individual’s sex or primary language did not significantly affect video visit outcomes.
• Enhanced technical support and patient preparation can improve visit success rates.
• Developing a risk assessment tool can help identify and support patients at higher risk for technical difficulties.

Telehealth visits are now an accepted and rapidly growing method of health care delivery. Although in-person visits will remain an essential part of health care, some study findings have suggested that one-third of current medical encounters potentially could be delivered through telehealth.1 Increased telehealth adoption has been driven by advances in policy, tools, and external factors such as the COVID-19 pandemic.2 Many published studies indicate no difference in telehealth visit outcomes compared with traditional clinical settings,3 although the majority of studies that provide evidence for outcome parity were narrowly focused on specific well-defined conditions, clinical situations, and patient populations.4

Results of recent studies have shown disparities in telehealth utilization, finding that telehealth overall is less likely to be used by certain populations. Moreover, video telehealth is also less likely to be used compared with audio-only telehealth among certain populations.5,6 Digital health equity issues can therefore present as both an underutilization of telehealth visits overall and an overreliance on audio-only visits when video visits would be possible if proper preparation and support were provided. The clinical and demographic features associated with telehealth visit failure or connection difficulty were not known previously.

Objectives

Recognizing this emerging disparity in the completion of telehealth visits, UC Davis Health in the Sacramento area of California launched an electronic health record (EHR)–integrated scheduled telehealth video visit (STVV) platform, ExtendedCare from ExtendedCare Solutions. This software offers functionality to collect and analyze video telemetry, allowing for precise identification of connection failure cause. Additionally, to maximize clinical quality, workflows were designed to offer video telehealth first, rather than audio-only, as the primary option for patients wanting a remote visit. Supporting this policy, the information technology department established patient preparation instructions and a patient self-test environment to allow testing prior to the visit. Additionally, real-time technical support for troubleshooting video or, if troubleshooting was unsuccessful, rescheduling the visit as an audio-only telephone visit was made available to patients. Along with telehealth telemetry, demographic information such as age, sex, ethnicity, primary language spoken, and race was captured for each visit. Our primary hypothesis was that canceled/no-show visits, failed visits, and conversion to audio visits would be more common in a distinct cohort of patients. Beyond this, we hoped to identify in the telemetry data specific failed video visit scenarios that would inform future education, technical optimizations, and research.

Study Design

This was a retrospective review of video visit telemetry data and EHR appointment information for STVVs across UC Davis Health between September 1, 2020, and November 30, 2020. Demographic information included age, sex, ethnicity, primary language spoken, and race for each visit.

ExtendedCare, a third-party video visit platform that is fully integrated with the Epic Systems Inc EHR, was utilized for video visits. This software allows patients to test their computer, device, webcam, microphone, internet connection, and speaker in preparation for the STVV and take appropriate action if issues are identified. The stored test data and visit data from each EHR encounter were available for analysis using reporting tools and dashboards. Detailed network and device telemetry was available for each STVV, and root cause analysis was performed for questionable visits. Combining traditional EHR data with video connection telemetry for each visit gave us the ability to determine the outcome of STVVs more accurately than relying solely on reported outcomes or billing codes and, more importantly, gave us insight into the reasons for various outcomes.

METHODS

Detailed device and network telemetry is available to the support team for review after any visit that has been flagged for investigation. The support team investigation was categorized by the authors into 3 main visit outcome groups (video, failed, and excluded), as shown in the Figure. Each outcome can be uniquely identified by data stored in the EHR, which include video connection telemetry. A report was created and run daily with visit counts for each outcome. The reporting criteria used to identify each outcome were created so that a given visit could have only 1 outcome. Data were collected by running the outcome report with demographic variables over all STVVs in the study time range.

We studied all STVVs for our patient population. Visits occurred for patients ranging in age from 0 to 105 years. Patients’ demographics were verified as part of the standard registration process. Age was dichotomized as younger than 65 years or 65 years or older. Sex was defined as either male or female, as those in the groups defined as nonbinary or unknown did not have sufficient frequency data for analysis. Ethnicity was categorized as not Latino/Hispanic, Latino/Hispanic, or declined to state. Primary language was categorized as English or other. Race was categorized as White, African American or Black, Asian, other, or declined to state.

The primary outcome was video call outcome, which could be either successful or failed. A binomial generalized logistic regression estimated through a generalized estimating equation (GEE) approach was used to evaluate differences in outcomes among demographic characteristics, specifically age, sex, ethnicity, primary language spoken, and race. A GEE approach was used to account for potential correlations of outcomes within patient and within provider. An independent correlation structure was assumed for the working correlation matrix. Because this analysis was exploratory, all demographic variables, excluding primary language, were used in the multivariable binomial generalized logistic regression through a similar GEE approach. Primary language was not included because 40% of the video calls did not have one reported. All analyses were performed using PROC GEE in SAS 9.4 (SAS Institute Inc). All tests were 2-sided, and P values less than .05 were considered statistically significant. This study was approved by the UC Davis Health Institutional Review Board.

A secondary outcome was using visit telemetry to further stratify failed visits into detailed failed scenarios. Because the volume of each individual detailed failure scenario was so small, significantly more data would be required to analyze these and determine significant relationships than are available in this study.

RESULTS

UC Davis Health providers scheduled 47,065 unique telehealth video visits between September 1, 2020, and November 30, 2020. Telemetry data were available for 30,996 visits; of those, 27,273 (87.99%) were completed, with EHR data and video telemetry indicating a simultaneous video connection established between provider and patient. A total of 3723 visits (12.01%) were listed as failed STVVs.

In an effort to understand the technical hurdles that may lead to audio-only telephone encounters having a lower rate of no-show or canceled events, distinct video visit outcome events were analyzed. These are outlined in the Figure and Table 1.

Completed Visits

Table 2 captures the baseline characteristics of each patient per included individual STVV during the study period. A total of 19,141 (61.8%) encounters were with female patients, and 11,833 (38.2%) were with male patients. Regarding self-identified ethnicity, 3722 (12.4%) encounters were with Hispanic/Latino patients, 1149 (3.8%) were with patients who declined to state their ethnicity, and 25,162 (83.8%) were with patients who were not Hispanic/Latino. The majority of encounters (20,379 [68.1%]) were with patients who self-identified as White; the remaining encounters were 2005 (6.7%) with African American/Black patients, 2378 (7.9%) with Asian patients, 1121 (3.7%) with patients who declined to state their race, and 4060 (13.6%) with patients of other race. Nearly every encounter (19,418 [97.4%]) had the individual’s primary language listed as English. Approximately one-fourth of the encounters (8868 [28.6%]) were listed as with a patient 65 years or older.

Failed: Video Visits That Did Not Connect as Planned

Univariate analyses were performed to determine which demographic variables predicted failure of a video visit. Table 3 shows the results of univariate analyses for demographic characteristics in relation to failure of video visit. Older patients had higher odds of a video failure, as did those who self-reported their race as African American/Black compared with those who self-reported their race as White. Those who self-reported their race as Asian had lower odds of video failure than those who self-reported their race as White. No significant differences could be seen in the outcome of the video visit based on self-reported ethnicity, sex, or primary language spoken.

Overall, 12.01% of visits were considered failed. These included visits explicitly flagged by providers as failed, visits for which telemetry exists but there was no indication of a simultaneous connection, and visits changed to audio-only telephone encounters after a video connection attempt. Excluded from failure analysis were visits canceled, visits no-showed with no technical evidence of a connection attempt, and visits completed with an alternate modality for which telemetry was not available. Results seen in Table 4 show that after adjusting for other baseline demographic variables, patients older than 65 years were 102% more likely to have an outcome of failed rather than a completed video visit than those younger than 65 years (95% CI, 1.87-2.17). Those who were African American/Black were 56% more likely to have an outcome of a failed video visit than those who were White (95% CI, 1.37-1.78). Those who were Asian were 19% less likely to have an outcome of failed video visit than those who were not Asian (95% CI, 0.70-0.94). Those who were Hispanic/Latino were 16% more likely to have an outcome of failed video visit than those who were not Hispanic/Latino (95% CI, 1.02-1.31).

DISCUSSION

The COVID-19 pandemic provided a catalyst for the rapid expansion of the use of telehealth. Health systems all over the world created workflows to accommodate a telehealth platform for continued access to health care. These telehealth encounters were primarily delivered as scheduled visits conducted either via audio only or with both audio and video. As the pandemic persisted, appropriate concern was raised about equal opportunity and access for patients to receive health care when utilizing a telehealth platform and that these 2 methods of connecting may not be equal.

Telehealth has proven valuable in increasing access to medical care via reduction of no-show rates,7 but unfortunately, studies have highlighted the disparities that appear to be associated with access utilizing a telehealth modality. One early study from a tertiary care center evaluated the patterns of successful clinic appointments for 148,402 unique patients. The researchers identified that age, race, primary language, and household income were all independently associated with likelihood of success with a telehealth encounter. Additionally, they identified that age older than 75 years, female sex, African American/Black race, Latinx ethnicity, and household income below $50,000 were independently associated with choosing audio-only telephone over video-based telehealth.6 A similar pattern was reported by researchers from Oregon, who identified that older patients, non-English speakers, and African American/Black patients preferred audio-only telephone visits to video-based telehealth.8 Moreover, some study findings have suggested that health care staff may develop a bias toward certain populations who successfully complete telehealth visits.9

Building on this reported research, our team elected to evaluate whether disparities exist at the level of technical success STVVs. Although other studies, as pointed out earlier, have established the potential for varied utilization of telehealth among distinct cohorts of patients, no study has been conducted to evaluate the success rate of video visits from a technical standpoint. Utilizing the unique advantage of the UC Davis Health video visit platform, our team was able to identify specific technical events during the time of a video visit and analyze the distribution of these technical events among our diverse patient population. Our investigation identified with univariate analysis that age followed by African American/Black race had the highest odds of a technical failure, whereas female patients and Asian patients were more likely than others to have successful video visits. In multivariate analysis, Latino ethnicity was also additionally identified as a risk factor for video visit technical failure.

Limitations

This study is focused on the population of patients who chose a video visit and then subsequently completed it or had technical difficulties. As such, there is selection bias in this study, which may help to explain why language did not bear out to be a significant factor. Simply, a non–English-speaking patient may have chosen to not have a video visit. Additional research on the patterns of use of video visits may be helpful to understand some of the specific factors that may lead to a technical failure, including connectivity, digital technology aptitude and familiarity, and preparation for the video visit encounter.

CONCLUSIONS

Using video visit telemetry to understand the patterns of failure can support more targeted interventions to improve the quality and success of telehealth visits. For instance, missed connections can be prevented by enhancing status updates and communication with patients as they wait for providers who may be running late. In addition, patients who have video visits converted to phone visits can be flagged for follow-up and testing to ensure they have successful video visits in the future. Having access to technical success data correlated to demographics, as in this study, can target these interventions to have the greatest impact.

Future research on telehealth should make use of telemetry where possible to identify not just the failure of telehealth visits but also the causes of that failure, including factors such as language, access to reliable broadband, and comfort with technology devices. A large set of data should be able to identify, for instance, whether missed connections or no provider connections are significantly more likely for demographic subgroups. Such a finding could shed light on biases in process while excluding from consideration the much more common explanation of “technical limitations” of certain populations. This would allow for targeted interventions as well as monitoring effectiveness of interventions stratified by demographic.

A second area of future research should be to develop a risk assessment tool for identifying a priori the potential for technical failure and intervening ahead of visits to help ensure success. Such a tool could be deployed not only to improve the rates at which patients successfully complete video visits but also to proactively identify patients for whom video is not an option. This could ensure that the optimal modality is utilized and the best possible outcome is achieved for each individual patient visit. 

Author Affiliations: UC Davis Health (DS, MLM, JF, JW, MA), Sacramento, CA.

Source of Funding: National Center for Advancing Translational Sciences, National Institutes of Health (NIH), through grant number UL1 TR001860. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Author Disclosures: The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (DS, JW, MA); acquisition of data (DS); analysis and interpretation of data (DS, MLM, JF, MA); drafting of the manuscript (DS, MLM, JF, JW); critical revision of the manuscript for important intellectual content (DS, MLM, MA); statistical analysis (DS, JF); obtaining funding (JW); administrative, technical, or logistic support (JW, MA); and supervision (JW).

Address Correspondence to: Daniel Stein, MS, UC Davis Health, 10850 White Rock Rd, Rancho Cordova, CA 95670. Email: dstein@ucdavis.edu.

REFERENCES

1. Jabbarpour Y, Jetty A, Westfall M, Westfall J. Not telehealth: which primary care visits need in-person care? J Am Board Fam Med. 2021;34(suppl):S162-S169. doi:10.3122/jabfm.2021.S1.200247

2. Tuckson RV, Edmunds M, Hodgkins ML. Telehealth. N Engl J Med. 2017;377(16):1585-1592. doi:10.1056/NEJMsr1503323

3. Totten AM, Hansen RN, Wagner J, et al. Telehealth for Acute and Chronic Care Consultations. Agency for Healthcare Research and Quality; 2019. Accessed March 12, 2021. https://www.ncbi.nlm.nih.gov/books/NBK547239/

4. Totten AM, McDonagh MS, Wagner JH. The evidence base for telehealth: reassurance in the face of rapid expansion during the COVID-19 pandemic. Agency for Healthcare Research and Quality. May 14, 2020. Accessed March 12, 2021. https://effectivehealthcare.ahrq.gov/products/telehealth-expansion/white-paper

5. Rodriguez JA, Saadi A, Schwamm LH, Bates DW, Samal L. Disparities in telehealth use among California patients with limited English proficiency. Health Aff (Millwood). 2021;40(3):487-495. doi:10.1377/hlthaff.2020.00823

6. Eberly LA, Kallan MJ, Julien HM, et al. Patient characteristics associated with telemedicine access for primary and specialty ambulatory care during the COVID-19 pandemic. JAMA Netw Open. 2020;3(12):e2031640. doi:10.1001/jamanetworkopen.2020.31640

7. Sumarsono A, Case M, Kassa S, Moran B. Telehealth as a tool to improve access and reduce no-show rates in a large safety-net population in the USA. J Urban Health. 2023;100(2):398-407. doi:10.1007/s11524-023-00721-2

8. Sachs JW, Graven P, Gold JA, Kassakian SZ. Disparities in telephone and video telehealth engagement during the COVID-19 pandemic. JAMIA Open. 2021;4(3):ooab056. doi:10.1093/jamiaopen/ooab056

9. Phimphasone-Brady P, Chiao J, Karamsetti L, et al. Clinician and staff perspectives on potential disparities introduced by the rapid implementation of telehealth services during COVID-19: a mixed-methods analysis. Transl Behav Med. 2021;11(7):1339-1347. doi:10.1093/tbm/ibab060