Structural Racism Shapes the Route Into Telehealth Video Visits
Telehealth did not distribute evenly across the primary care population examined in this study. Among 137,297 patients with completed telehealth visits over 26.5 months, race alone did not show a significant unadjusted effect on whether a visit happened by video or audio-only. Once individual and zip code level confounders entered the model, however, Black patients had a 6% lower video visit rate ratio, and the gap widened to 7% to 9% when interaction terms were added.
The pattern fits a staged view of telehealth access rather than a single binary choice between video and phone. Video visits require more than a clinic appointment slot. They depend on device access, broadband quality, prior patient portal use, technical confidence, and enough stability in the digital pathway to keep the encounter from collapsing into audio-only care. In that sense, the study treats telehealth as a chain of exposures, each one capable of widening inequity before the clinician ever appears on screen.
Broadband Conditions Did Not Explain the Video Gap
One of the most direct tests in the analysis focused on digital redlining, using broadband market data linked to the electronic health record. Broadband conditions were worse in areas with high neighborhood disadvantage and high Black residential composition, a finding consistent with a familiar pattern: digital infrastructure often tracks historical segregation and economic disinvestment.
Yet broadband conditions were not associated with video visit rates. The result narrows the explanatory field. It suggests that poor local broadband availability, while clearly unequal, did not by itself account for the lower likelihood of video use in this cohort. A clinic can sit inside a neighborhood with fragile infrastructure without the broadband metric fully predicting whether a specific telehealth encounter becomes a video visit. In practice, digital inequity appears to operate through more than raw connectivity.
That disconnect matters clinically. A patient may live in an area where broadband is objectively weaker, but the pathway from neighborhood network conditions to a completed video visit can be interrupted by portal status, prior experience with digital tools, device troubleshooting, or the way follow-up is organized inside the health system.
Portal Enrollment Followed a Different Racial Pattern
Path dependency offered a more layered explanation. Portal nonenrollment was more common among Black patients, which signals earlier disengagement from the digital workflow needed for seamless telehealth. The striking detail is where that barrier exerted its strongest effect: nonenrollment had a stronger negative association with video visit rates among White patients.
This result points to conditional behavior rather than a single uniform rule. A patient portal is not merely a login screen; it is an entry point for scheduling, messaging, previsit instructions, and often the small operational steps that make video feasible. The downstream effect of not being enrolled may therefore differ depending on how other barriers cluster in a given patient population. In some groups, not having portal access may be one obstacle among several. In others, it may become the decisive break in the sequence.
The study’s framing of path dependency is important because it shifts attention from one-time access metrics toward accumulated digital readiness. Long before the visit begins, the system has already sorted patients into different levels of preparedness for video care.
Technical Difficulties Pushed Encounters Toward Audio-Only Care
Technical difficulties showed the clearest bedside consequence. Black patients experienced more technical problems, and those problems were associated with greater odds of a subsequent audio-only visit, especially among Black patients. The progression is clinically intuitive: a failed video setup does not simply delay care; it can change the care modality entirely.
This is where sequential inequity becomes visible in real time. Video failure is not an abstract usability issue. It can shorten the clinical encounter, limit visual assessment, and reduce the clinician’s ability to observe respiratory effort, gait, rash, edema, or general appearance. Audio-only care can still be appropriate, but when technical trouble steers patients there unevenly, the system begins to reproduce the very disparities telehealth was supposed to reduce.
The study’s data suggest that the strongest racial differences may emerge after the patient has already entered the telehealth process. That is a harder problem than simple appointment availability. It means health systems need to watch for failure points during the visit itself, not only at scheduling or broadband access.
Telehealth Inequity Operates Across Stages, Not in One Place
The most clinically useful lesson from the analysis is its staging of structural racism. Digital redlining, portal nonenrollment, and technical difficulties did not behave as interchangeable explanations. They appeared at different points in the telehealth sequence and exerted different effects. Some barriers were associated with place. Some were associated with patient-system history. Some emerged during live use.
This sequencing has direct implications for primary care operations. If a health system measures only how many patients have broadband, it may miss the operational choke points that actually determine whether a video visit happens. If it measures only whether a portal exists, it may miss the failures that occur when patients try to use it. If it tracks only completed visits, it may miss the repeated effort and frustration required to get there. Structural racism in telehealth can therefore remain hidden inside metrics that look acceptable on the surface.
The study also reinforces an uncomfortable but necessary point: equal availability of telehealth technology does not guarantee equal usability. Two patients can be offered the same virtual visit, yet their chance of completing it by video can diverge based on neighborhood composition, prior digital engagement, and the likelihood that a technical issue will force a fallback to audio-only care.
Primary Care Systems Need Stage-Specific Monitoring
Health systems should monitor telehealth inequities across care stages rather than treating video use as a single outcome. The analysis supports attention to enrollment workflows, technical support, and the point at which video visits convert to audio-only care. Those are not administrative details. They are the mechanisms through which unequal access becomes routine practice.
Because the study used structured and unstructured electronic health record data linked with broadband market information, it also offers a useful template for surveillance. Systems already collect much of what is needed to detect patterns like these, but the data must be interpreted through an equity lens rather than a neutral operations lens. Otherwise, the digital front door can appear open even when some patients are repeatedly steered toward a narrower mode of care.
Funding came from an unrestricted grant from Google Health, with the funder absent from study design, data collection, analysis, interpretation, writing, and the decision to submit. The authors reported no competing interests. The remaining question for health systems is not whether telehealth inequity exists, but how quickly they can identify which stage of the process is producing it in their own patients.