Ensuring Primary Care Diagnostic Quality in the Era of Telemedicine

Telemedicine applications have been used for decades, most commonly in specific areas of medicine (eg, dermatology, pathology, radiology) and in specific contexts (eg, rural or other under-resourced areas).¹ The coronavirus disease 2019 (COVID-19) pandemic has markedly accelerated the expansion of the telemedicine platform, especially video visits, for primary care clinicians and practices relatively unfamiliar with its use. Although telemedicine usage may be receding from its peak pandemic rollout, clinicians will likely continue to employ these video visits for patient care. Within this context of new and widespread ongoing use, critical evaluation of telemedicine's benefits and harms related to diagnosis is necessary and linked to efforts to minimize potential diagnostic errors and to establish best clinical practices.

Potential for Diagnostic Error

Existing research regarding diagnostic accuracy in telemedicine has focused primarily on tele-dermatology or specific clinical contexts (eg, stroke identification).2, 3, 4 There has been far less examination of diagnostic accuracy in the broader range types and severity of diagnoses that typically present to primary care clinicians. Despite this, primary care clinicians have been thrust almost overnight, onto a fundamentally different platform for diagnosis and management of clinical questions: Does the patient appear well, mildly or moderately ill, or in need of emergency medical services? What tests, referrals, or procedures are needed or appropriate lacking a traditional physical examination?

There is an urgent need to evaluate the extent to which our current technological resources fall short of providing the truest clinical picture to inform these diagnostic decisions. This is not to say that face-to-face health care encounters unfailingly result in accurate diagnoses; indeed, there is robust literature to the contrary.⁵ However, in concordance with broader ongoing efforts to improve on diagnostic accuracy in medicine, we must recognize the specific challenges posed by telemedicine.^{6 , 7}

Diagnostic Advantages and Disadvantages

A 5-component structural model of primary care telemedicine encounters provides a framework to examine the elements impacting telemedicine diagnostic quality and error (Table ).

Table.

Factors Influencing Diagnostic Accuracy in Primary Care Telemedicine

Domain	Specifics
Patient	Technology availability, physical environment, social environment (privacy, distractions, embarrassment), health literacy, technical proficiency (experience with communication technologies), additional support from family members/others
Physician	Office versus home environment, staff support, clinical encounter workflow, EMR skills, health information access, clinical diagnostic experience, telemedicine training and proficiency, engagement versus burnout
Telemedicine/EMR Platform	Audio-video quality, bandwidth, connectivity time, embedded smart technologies, design for user ease and telemedicine integration and workflow, multiple other general features that support diagnosis8
Clinical Contexts	Visit agenda, patient-physician rapport, diagnostic problem urgency, complexity, and clinician familiarity, pre- and postvisit time, logistics, and support
Health System	Regulations, reimbursement, health system supports, scheduling, and productivity requirements (patients per hour), patient continuity, continuous learning, and improvement

EMR = electronic medical record.

The first component is the patient, whose video visit occurs in a specific context and environment, influenced by family and other social networks. In the secure environment of the home, patients may recount their illness and symptoms with greater detail, fostering more accurate diagnosis. Additionally, family member contributions to a patient's history may improve diagnosis, occurring more easily in the virtual platform (eg, “Honey, did you tell them about the sore on your foot?”). Likewise, examination of a patient's environment may provide diagnostic clues not visible in routine clinic visits (eg, spotting fall risks such as throw rugs in the home of an elderly patient). More challenging are situations in which the lack of privacy in the home may hinder discussion about sensitive topics such as anxiety, depression, or symptoms suggestive of sexually transmitted infections. Riskier still would be discussion of imminent threats to personal safety like domestic abuse.

The second component of telemedicine is the physician, who may be working less closely with the office-based health care team. In practice-based telemedicine, the introduction of video visits into outpatient clinics has changed workflow and team dynamics, creating environments where physicians may be working solo or with reduced support, altering routine diagnostic cues and data. Physicians in an outpatient practice may still have access to data such as prior patient records, but they are disadvantaged by loss of other data such as vital signs and clinical observations gathered by other health care team members. The current lack of standardization in telemedicine workflows and in comprehensive health care team training in the virtual setting may create wide variability in information-gathering and even in deciding which clinical complaints are appropriate for telemedicine encounters.

Third, the telemedicine platform itself introduces potential for diagnostic error. On the one hand, video visits may enhance diagnostic timeliness by removing access barriers and shortening delays in initial diagnostic evaluation. Indeed, the convenience and expanded access to care through telemedicine has been associated with high patient satisfaction.⁹ However, the sudden and substantial demand for telemedicine during the pandemic has led to use of readily available consumer-grade videoconferencing platforms (Zoom, Facetime, and Skype), devoid of any “smart” technology that might enhance clinician observation or diagnostic accuracy. Such platforms were originally designed to connect individuals socially or in office meeting conditions—not for Health Insurance Portability and Accountability Act (HIPAA)-compliant, high-resolution clinical observation, on which a patient's medical encounters and diagnosis may depend.¹⁰ Additionally, the diagnostic utility of video visits depends on both clear audio and image quality. Either feature may be impaired by an unstable Internet connection, limited capacity of a patient's device, or disrupting features of a patient's environment (poor lighting, background noise), all of which may contribute to diagnostic error. Further, a clinician's inability to expedite connectivity/technology issues required for video teleconferencing may consume 5-10 minutes of a 15-minute appointment, obviously compromising valuable encounter and diagnostic assessment time.

Fourth, the context of the video visit itself can significantly influence the quality of care provided. Certain types of clinical encounters may be better suited for the telemedicine platform. For example, telemedicine visits regarding follow-up for patients’ chronic conditions such as diabetes or high blood pressure with their established primary care physician may be conducted remotely with less chance of clinical error compared with a visit involving an unknown patient-clinician dyad with acute symptoms (ie, shortness of breath or chest pain). Explicit evidence-based criteria regarding which types of encounters are appropriate for primary care telemedicine visits have not been established. Challenges in building rapport, especially for a new patient encounter, may also introduce situations where a patient's level of comfort may lead to less disclosure of information and hinder an accurate diagnostic assessment.

Fifth, the system through which telemedicine currently operates in the United States was built and disseminated quickly to meet emergent needs. Regulations hindering ease of telemedicine use were initially relaxed.¹¹ However, the ways by which legal and financial supports for telemedicine evolve moving forward will shape potential for diagnostic error reduction in the future. For example, a downgraded payment structure for video visits may directly affect a clinician's time and resources to accurately diagnose a patient's condition. Similarly, health care system support and training of end users to wield this new medium have the potential to lessen—or exacerbate—possible threats to diagnostic accuracy and patient safety.

Setting a Research Agenda

The preceding model provides a framework to examine factors that may influence diagnostic error in telemedicine. Examining these components could guide explorations of the multidimensional processes relevant to telemedicine practice, ranging from psychosocial  (eg, patient reluctant to disclose symptoms within earshot of others) to health professional education (eg, deficits in physician training and telemedicine diagnosis) to improving the telemedicine platform (eg, “smart information technology”) to contextual factors (eg, establishing appropriate triage algorithms) to systemic issues (eg, adequate time allotted per encounter and resources devoted to infrastructure development and training by end-users).

Traditional strategies for decreasing diagnostic error must be reconceptualized, redesigned, and tailored to the specific capacities and limitations of video visits. Creating, implementing, and evaluating these strategies will require cross-disciplinary insights and collaboration. Applying concepts and methods from social psychology and communication science will be needed to provide insights for optimizing communication through virtual media. Even something as seemingly basic as teaching patients to adjust the camera on their device may be decisive for enhanced clues needed for accurate diagnosis (ie, showing a dermal lesion or jugular venous distention clearly).

Improvements in video platform technology may help avert diagnostic errors due to poor visual or auditory clarity but are likely marginal compared to broader nontechnical issues. Borrowing from industrial and systems engineering principles, innovative larger system workflow redesign that redirect specific clinical tasks or scenarios to or away from exclusive reliance on telemedicine platforms (eg, leveraging home visiting nurses or physical therapists or shorten follow-up intervals to better operationalize “the test of time”). There is limited research to date measuring similarities and differences in diagnostic accuracy for specific, common conditions via virtual platforms compared with face-to-face encounters.¹² However, at the very least, the inability to form a corporeal connection and perform certain physical examination assessments over video platforms will likely continue to pose a major limitation for reliable diagnosis and introduce opportunities for medical error and mismanagement of both acute and chronic conditions.

Embracing the Future

Although the conveniences of telemedicine use for both primary care patients and clinicians are many, benefits must be carefully weighted with limitations, particularly for medical diagnosis. Telemedicine encounters do not happen in a vacuum. Pre-encounter information solicitation as well as more frequent and systematic postvideo encounter follow-up can surely enhance diagnostic thoroughness and safety. The framework of management reasoning—defined as the process of making decisions about patient management, bundling choices about treatment, follow-up visits, future testing, and allocation of limited resources—may aid reframing telemedicine diagnosis and the clinical reasoning behind it.¹³ Ultimately a directed focus on understanding the extent and sources of potential diagnostic error in telemedicine coupled with mitigation of its modifiable factors will allow an expanded capacity to provide safe, patient-centered primary care services.