Choice, Transparency, Coordination, and Quality Among Direct-to-Consumer Telemedicine Websites and Apps Treating Skin Disease

Importance  Evidence supports use of teleconsultation for improving patient access to dermatology. However, little is known about the quality of rapidly expanding direct-to-consumer (DTC) telemedicine websites and smartphone apps diagnosing and treating skin disease.

Objective  To assess the performance of DTC teledermatology services.

Design and Participants  Simulated patients submitted a series of structured dermatologic cases with photographs, including neoplastic, inflammatory, and infectious conditions, using regional and national DTC telemedicine websites and smartphone apps offering services to California residents.

Main Outcomes and Measures  Choice of clinician, transparency of credentials, clinician location, demographic and medical data requested, diagnoses given, treatments recommended or prescribed, adverse effects discussed, care coordination.

Results  We received responses for 62 clinical encounters from 16 DTC telemedicine websites from February 4 to March 11, 2016. None asked for identification or raised concerns about pseudonym use or falsified photographs. During most encounters (42 [68%]), patients were assigned a clinician without any choice. Only 16 (26%) disclosed information about clinician licensure, and some used internationally based physicians without California licenses. Few collected the name of an existing primary care physician (14 [23%]) or offered to send records (6 [10%]). A diagnosis or likely diagnosis was proffered in 48 encounters (77%). Prescription medications were ordered in 31 of 48 diagnosed cases (65%), and relevant adverse effects or pregnancy risks were disclosed in a minority (10 of 31 [32%] and 6 of 14 [43%], respectively). Websites made several correct diagnoses in clinical scenarios where photographs alone were adequate, but when basic additional history elements (eg, fever, hypertrichosis, oligomenorrhea) were important, they regularly failed to ask simple relevant questions and diagnostic performance was poor. Major diagnoses were repeatedly missed, including secondary syphilis, eczema herpeticum, gram-negative folliculitis, and polycystic ovarian syndrome. Regardless of the diagnoses given, treatments prescribed were sometimes at odds with existing guidelines.

Conclusions and Relevance  Telemedicine has potential to expand access to high-value health care. Our findings, however, raise concerns about the quality of skin disease diagnosis and treatment provided by many DTC telemedicine websites. Ongoing expansion of health plan coverage of these services may be premature. Until improvements are made, patients risk using health care services that lack transparency, choice, thoroughness, diagnostic and therapeutic quality, and care coordination. We offer several suggestions to improve the quality of DTC telemedicine websites and apps and avoid further growth of fragmented, low-quality care.

Dermatology was an early adopter of telemedicine technologies, and a substantial evidence base has demonstrated diagnostic concordance,^1-3 improved cost,⁴ and similar outcomes to in-person care^5,6 for certain clinical scenarios. These positive findings are primarily from store-and-forward (asynchronous) teleconsultation models, where a referring clinician sends a patient’s history and images to a dermatologist for recommendations.

Direct-to-consumer (DTC) teledermatology, while less studied, is expanding rapidly. Direct-to-consumer telemedicine companies have raised substantial investor capital⁷ and marketed themselves heavily to consumers, employers, and health plans. Many additional DTC telemedicine websites with a narrower focus on dermatology have also emerged. While most initially relied on fees paid by individual patients who sought convenience or inexpensive care, large DTC services are increasingly contracting with major health plans to provide telecare to millions of enrollees.⁸ Meanwhile, many of these payers still exclude asynchronous telemedicine services provided by a patient’s existing health care team.

While the number of DTC telemedicine website visits in 2015 reached 1.25 million,⁹ relatively little is known about the quality of these services. A recent study assessing DTC telemedicine consultations for primary care conditions revealed a lack of appropriate diagnostic testing and poor antibiotic stewardship.¹⁰ Reviews of teledermatology services showed failures to collect key medical history¹¹ and raised concerns about use of international physicians and advertising prominently featuring easily obtained prescriptions.¹²

National physician groups, many of which support expanded telemedicine, have questioned whether these companies are giving patients choice, disclosing clinician credentials, collecting adequate medical history, following evidence-based guidelines, coordinating care with existing treating physicians, and establishing protocols for local referrals.^13-16 We assessed the performance of DTC telemedicine websites and apps by submitting a series of structured dermatologic consultations.

Box Section Ref ID

We performed a series of internet searches seeking DTC telemedicine websites and smartphone apps accepting patient-originated visits for skin conditions. For inclusion, websites had to offer services to California residents and mention skin disease among treated conditions. Websites were excluded if they required live-interactive video, a specific enrolled employer, or communication from a referring clinician. We also excluded websites with just 1 clinician, only serving a clinic’s existing patients, or primarily offering custom medication compounding.

Six simulated structured dermatologic cases were constructed, each including patient demographics, history of present illness (HPI), review of systems (ROS), and 3 photographs of skin eruptions or lesions. The photographs were mostly obtained from publicly available online image search engines, which any patient wanting to obtain prescriptions from a DTC telemedicine website could access, copy, and submit. To simulate the information that real patients present during both in-person and virtual interactions with clinicians, each universally submitted HPI (what a patient might think is relevant) was supplemented with additional HPI and ROS data that each patient would offer if asked or if there was any place in the submission process to include it. The highlights of each case are summarized in Table 1.

Study personnel accessed each website and app between February 4 and March 11, 2016, posing as patients (all cases were separately submitted to each website, except that acne-only websites received just 2 cases). Study personnel claimed to be uninsured and paid fees using Visa gift debit cards in the names of the simulated patients. No study personnel provided any false government-issued identification cards or numbers. When given a choice of clinician, we selected with the following priority: dermatologist > primary care physician > other physician > nurse practitioner > physician assistant. The initial HPI was submitted in its entirety if there was any opportunity to enter it, and was supplemented with the additional history if opportunities were given to enter an ROS or if relevant additional questions were asked. At least 1 of the photographs was submitted if there was any way to do so, and additional photographs were also uploaded if the website specified that 2 or 3 photographs were preferable.

We recorded data from each encounter related to fees paid, response times, choice of clinician, transparency of credentials, clinician location, demographic and medical data requested, diagnoses given, treatments prescribed, adverse effects discussed, referrals made, and medical records made available. When provided with a clinician name and location but not credentials, we used searches of state licensure databases, national board certification databases, and other websites to attempt to establish licensure and board certification status. No prescriptions were ever procured by study personnel.

The institutional review board of the University of California, San Francisco, reviewed and approved the study protocol.

Sixteen DTC telemedicine websites and apps met inclusion criteria and responded to at least 1 case (Box 1). A few others met inclusion criteria but failed to respond to requests or had portal error messages that prevented successful case submission (Figure).

Box Section Ref ID

There were 62 clinical cases with a successful submission and response. Nonresponders were the result of scope limitations, payment problems (a few websites accepted gift debit cards on some submissions but not others), and inconsistent replies (some websites only intermittently replied). No website asked for photographic identification or raised any concerns that the patient had used a pseudonym or submitted false photographs.

The price per consultation varied—most charged a fixed cost per visit ($35-$95), while others charged monthly memberships ($9 per 2 weeks to $49 monthly). Response times were rapid (0-8 days), with 87% of encounters receiving replies on the same or following day.

Websites allowed patients to select a clinician on 20 of 62 encounters (32%), while the remaining 42 (68%) assigned a clinician without any patient choice. We were able to determine the practice type of the clinician in all but 5 encounters (57 of 62 encounters). Physician backgrounds included dermatology (27), internal medicine (5), emergency medicine (3), family medicine (3), obstetrics/gynecology (1), cardiology (1), pain management (1), and physical medicine/rehabilitation (1). The nonphysician backgrounds included family nurse practitioners (3), physician assistants from dermatology settings (3), and a nurse practitioner in emergency medicine. The rest were primary care physicians in India (6) and dermatologists in Sweden (2).

For the 49 identified US-based clinicians, we were able to confirm that all were licensed in California. Only 16 of the encounters (26%) disclosed information about licensure. Board certification was disclosed for 20 of the 42 known domestic MD encounters, and we were able to subsequently verify American Board of Medical Specialties (ABMS) board certification in the declared specialty for 39 of these.

The clinician’s geographic practice location was disclosed in 38 encounters (61%). We were able to determine the location for 57 encounters, 35 of which were in California (11 within 100 miles of the patient). The remaining locations included India (6), Minnesota (4), Colorado (3), Illinois (2), Washington, DC (2), Vermont (2), Sweden (2), and New York (1).

The 8 internationally based physicians who were not California-licensed and provided responses worked for First Opinion or First Derm, both of which are headquartered in California. First Derm claims their physicians are “board-certified dermatologists,” but we were unable to find evidence that they were either licensed in California or were ABMS certified. These websites post a variety of disclaimers that they do “not provide health care services,” and do “not create a physician-patient relationship,” but we found that they did make diagnoses and offer specific treatment recommendations (without sending prescriptions to pharmacies).

The demographic and clinical data collected by the websites are summarized in Table 2. Quiz Ref IDIn very few of the encounters did the clinician ask the name of an existing primary care physician (23%) or dermatologist (2%), and only 6 of 62 (10%) ultimately offered to send records to a current member of the patient’s clinical team. While all encounters allowed the patient to give some history, Quiz Ref IDonly 21 sought any ROS (34%). Forty-six (74%) collected medication or allergy lists, and only 11 (52%) asked female patients about pregnancy or lactation status. Most encounters allowed our patients to submit photographs, requesting 1 to 3 images. Clinicians in 4 encounters, however, provided diagnoses based solely on a submitted medical history without asking for or offering the chance to upload any images. These diagnoses without images were provided by a board-certified physical medicine/rehabilitation physician (1), board-certified internist (1), and physicians located in India (2).

A diagnosis or likely diagnosis was proffered in 48 (77%) of encounters, with the remainder of patients being referred to a local physician for diagnosis. Patients referred for diagnosis were given suggested name(s) of practicing local clinicians in only 2 of 14 instances. No patients were sent to a laboratory for studies. A differential diagnosis was listed in 8 of 48 cases with diagnoses (17%), and 2 of 14 cases without diagnoses (14%).

Quiz Ref IDAmong 48 encounters with diagnoses, 31 (65%) were given prescription medications, and 9 (19%) were told to seek a local physician for treatment (3 gave suggested names). The remainder were reassured, treated with over-the-counter medications, or had prescriptions recommended but not provided. When prescriptions were provided, the clinician disclosed relevant adverse effects or risks in 10 of 31 encounters (32%). Among 14 encounters where a female patient was prescribed a pregnancy class C or higher drug, pregnancy risks were disclosed in 6 (43%).

In the case of polycystic ovarian syndrome (PCOS) with inflammatory acne, the 28-year-old patient was not asked relevant questions about her hypertrichosis or irregular menses by a single website (1 website inquired about menses only to check pregnancy status before prescribing). Every clinician diagnosed acne, including 1 who did not ask for photographs, and none raised the issue of hirsutism, androgen excess, or PCOS. Two of 12 websites responding to this case recommended in-person follow-up for treatment, and 1 recommended use of only mild soaps, moisturizers, and cool showers. The patient received prescriptions for oral antibiotics (from 8 websites), topical retinoids (6), and topical antibiotics (5). None of the clinicians discussed hormonal treatment options.

In the case of secondary syphilis with unusual plaques, the patient was not asked about his recent fever (even when he proactively reported it on ROS), and no clinician seemed concerned that his diffuse eruption began so acutely only 3 weeks previously without a history of prior skin eruptions (uncommon in patients with large-plaque psoriasis). Seven of the 8 clinicians diagnosed psoriasis, including 1 who did not ask for photographs. One clinician (an emergency physician) made no diagnosis and referred the patient to see a local dermatologist, and another (a Sweden-based physician) who diagnosed psoriasis suggested that the patient see a local dermatologist for treatment. This patient received prescriptions for class I or II topical steroids in 5 encounters and was told to use moisturizers and take lukewarm baths in another.

In the case of gram-negative folliculitis, 1 clinician (a board-certified dermatologist) referred the patient to a local dermatologist for a culture, and another (an India-based physician) sent him to his primary care physician for diagnosis. The remaining 10 clinicians all diagnosed acne, seemingly unconcerned by the sudden appearance of pustules around the mouth after lengthy doxycycline use (a common presentation of gram-negative folliculitis caused by Proteus, Klebsiella, or Serratia). This patient was prescribed trimethoprim-sulfamethoxazole (by 2 clinicians), minocycline (by 3), continued doxycycline (by 1), topical antibiotics (by 3), and topical retinoids (by 2). One clinician suggested only topical retinol, and another recommended reducing his sugar intake and applying moisturizer.

In the patient with 1 nodular melanoma and 1 seborrheic keratosis, 11 of 14 clinicians referred the patient for an in-person examination of at least 1 of his lesions. The others (2 dermatologists and a physician assistant) reassured him that both lesions were benign.

The young woman with eczema herpeticum blisters (a potentially life-threatening infection caused by herpes simplex virus erupting in eczema) was diagnosed as having an ordinary eczema flare or contact allergies in 7 of 9 encounters, but 2 clinicians (1 dermatologist, 1 unidentified) raised concerns about infection and referred her to be seen in person (though they did not suggest any timeline or individuals she could see). She was prescribed oral prednisone (2), topical steroids (2), and antihistamines (1). Of the remaining 2 websites that diagnosed eczema, one recommended that topical steroids be prescribed in person, and the other recommended a vegan diet and suggested follow-up with an allergist. The 2 clinicians recommending prednisone, which can worsen eczema herpeticum when given without antiviral drugs, and which clinical guidelines recommend avoiding even for uninfected atopic dermatitis, were a doctor of osteopathic internal medicine and a family nurse practitioner.

The patient with stasis dermatitis was properly diagnosed by all 7 websites where his case was successfully submitted. He was prescribed topical clobetasol (3) or hydrocortisone (2). One website suggested clobetasol but did not provide a prescription, and 1 website suggested he see a local physician for a cardiac workup.

When implemented appropriately, DTC telemedicine websites can improve access to quality care for patients facing geographic, mobility, or financial constraints. Quiz Ref IDWhile a large body of evidence supports the use of teleconsultation for skin disease between a referring clinician and dermatologist, our study raises significant concerns about the rapidly expanding use of direct-to-consumer telemedicine.

The DTC telemedicine websites we consulted, many with a large and growing national footprint, provided care that often lacked patient choice of clinician, transparency of clinician credentials, or care coordination with existing health care teams. Many diagnoses were proffered without reasonable attempts to ask basic medical history follow-up questions, treatment recommendations sometimes contradicted evidence-based guidelines, and prescriptions frequently lacked basic disclosure of adverse effects, risks, or pregnancy warnings. When referrals to in-person care were made, they rarely included suggested clinician names. No DTC telemedicine websites attempted to confirm the identity of patients or raised doubts about the authenticity of photographs.

Patient choice of their treating physician is part of our medical code of ethics,¹⁷ and we were surprised that these websites with multiple clinicians on staff assigned a clinician without patient choice in most encounters. Transparency about licensing and credentials is inherent in patient choice and informed consent, and most services failed to provide licensure or board certification information during encounters. Despite claims that they were not providing health care services, we believe that 2 DTC telemedicine websites headquartered in California but using foreign clinicians were engaged in the practice of medicine without a state license, as they clearly provided diagnoses and treatment recommendations.

Our findings also suggest that treatment provided by these websites is likely to exacerbate care fragmentation, since only 23% asked for the names of existing primary care physicians, and only 10% offered to provide medical records to existing clinicians. In an era when traditional health care delivery is working to “minimize silos” (ie, individuals, groups, organizations, or systems that operate in isolation, lacking adequate communication with other groups) and improve inadequate care coordination, it makes little sense to design new care delivery channels without including an ability for patients’ longitudinal care teams to know about new diagnoses made and treatments initiated. When DTC telemedicine websites determined that in-person care was necessary, they only suggested where patients might go in a small minority of cases. None offered to actually facilitate a referral to a local physician (except in 1 instance where the clinician self-referred), even when some of the clinical scenarios involved urgent medical conditions in which ensuring follow-up was vital. This lack of care coordination may be related to the substantial geographic distance of many of the clinicians from our simulated patients and their possible lack of knowledge of the local health care system.

Many patients were prescribed oral or topical medications, and only one-third were told about relevant risks or adverse effects. Fewer than half of pregnancy category C or higher prescriptions given to women of childbearing age included any discussion of pregnancy risks or avoidance.

One dominant pattern emerged when examining the diagnostic accuracy of these websites with our clinical scenarios. For the 2 cases in which photographs alone were adequate to easily make a correct diagnosis (stasis dermatitis and melanoma), most of the DTC telemedicine websites performed well. However, when further engaging the patient for basic history elements was a necessary part of the interaction, diagnostic performance was poor.

None of the clinicians asked the 28-year-old woman with inflammatory acne about her irregular menses or hirsutism, so none raised concerns about PCOS or androgen excess or initiated appropriate treatment. The young man with highly infectious secondary syphilis was not asked about his recent fevers, attention was not paid to the unusually sudden onset he described, and all but 1 of the websites accepted the diagnosis of psoriasis the patient himself offered. This not only left him at substantial risk from untreated syphilis, but was a public health failure given the urgent need for contact tracing and risks of ongoing transmission. The patient with gram-negative folliculitis was not given a chance to say that the rampant new pustules around his mouth were unlike the acne he had before. The patient with eczema herpeticum faces the risk of disseminated infection and even death (especially when given systemic steroids without antiviral drugs) because most of the DTC telemedicine clinicians did not ask about the stinging vesicles shown in her photographs.

Our study has a significant limitation: we are unable to assess whether clinicians seeing these patients in traditional in-person encounters would have performed better on diagnostic accuracy. However, the additional medical history necessary to make the diagnoses (eg, fevers, irregular menses) are the types of information that, in our experience, typically emerge in the give-and-take of obtaining a history in the office setting. This give-and-take could certainly be replicated in a store-and-forward telemedicine environment, especially if the clinician takes time to engage the patient, but the DTC telemedicine websites we queried typically offered diagnoses and prescriptions immediately after 1 form-driven submission, without sending meaningful, appropriate follow-up questions to our simulated patients. In addition, diagnostic inaccuracy is only 1 of many concerning outcomes we reported.

Another limitation of our study is that the design forced us to exclude any DTC telemedicine websites or apps that required live interactive video. However, only 3 DTC telemedicine websites from our initial search had to be excluded for requiring video. We also excluded platforms serving only a clinic’s existing patients, which may be expected to offer more coordinated care.

Finally, we cannot fully guarantee the photographs acquired from the internet are of patients with biopsy or laboratory proof of the listed diseases. However, the fact that most DTC telemedicine websites failed to consider a diagnosis of eczema herpeticum in a young woman with eczema and new oozing, stinging blisters seems far more important than whether we have evidence of a positive Herpes culture from this woman with a classic presentation in her photographs.

Telemedicine has potential to expand access, and the medical literature is filled with examples of telehealth systems providing quality care. Our findings, however, raise doubts about the quality of skin disease diagnosis and treatment being provided by a variety of DTC telemedicine websites and apps. Quiz Ref IDThe substantial investment of health plans in providing enrollees with insurance-covered access to these DTC telemedicine websites, as well as rapidly expanding availability for patients willing to pay out of pocket over the internet, may be premature. The problem may be exacerbated by several health plans’ willingness to pay large national DTC telemedicine providers while excluding asynchronous telemedicine provided by a patient’s existing specialists.

We believe that DTC telemedicine can be used effectively, but it is best performed by physicians and team members who are part of practices or regional systems in which patients already receive care. This allows telemedicine clinicians to view a patient’s existing medical records, communicate with existing health care team members, provide in-person follow-up when needed, and be accountable for the care they deliver. A regional clinician network can provide a safety net for patients when errors in diagnosis and treatment occur.

Based on our findings, we believe that DTC telemedicine services should, at a minimum, adopt several standard practices (Box 2). We hope that studies such as ours will help DTC telemedicine companies and participating clinicians improve their services so that patients seeking these newer channels of care delivery can receive transparent, ethically provided, coordinated, high-quality care. In the meantime, we also hope that policymakers, regulators, and payers will find ways to identify and encourage telemedicine services that abide by these principles, while discouraging the growth of DTC telemedicine websites and apps that provide fragmented, low-quality care.

Box Section Ref ID

Corresponding Author: Jack S. Resneck Jr, MD, Department of Dermatology, University of California, San Francisco, Box 0316, San Francisco, CA 94143-0316 (resneckj@derm.ucsf.edu).

Accepted for Publication: April 23, 2016.

Published Online: May 15, 2016. doi:10.1001/jamadermatol.2016.1774.

Author Contributions: Dr Resneck had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Resneck, Abrouk, Lee, Kovarik, Edison.

Acquisition, analysis, or interpretation of data: Resneck, Abrouk, Steuer, Tam, Yen.

Drafting of the manuscript: Resneck, Abrouk.

Critical revision of the manuscript for important intellectual content: All authors.

Obtained funding: Resneck.

Administrative, technical, or material support: Abrouk, Lee.

Study supervision: Abrouk, Lee, Kovarik.

Conflict of Interest Disclosures: Dr Resneck serves on the Board of Trustees of the American Medical Association. Drs Resneck, Lee, Kovarik, and Edison serve on the Telemedicine Task Force of the American Academy of Dermatology Association. Dr Lee serves as a physician performing consultations for Direct Dermatology but did not participate in any of the telemedicine consultations encompassed in this study. No other disclosures are reported.

Funding/Support: The American Academy of Dermatology Association provided funding for the debit cards used in the study.

Role of the Funder/Sponsor: The American Academy of Dermatology Association had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The views expressed in this article are those of the authors and do not necessarily represent the views of the American Medical Association or the American Academy of Dermatology Association (Dr Resneck is a trustee of the American Medical Association).