br Primary analyses used an
Primary analyses used an intention-to-treat approach, which included all patients whose initial encounter was teledermatology, even if they received advice only and never had a face-to-face visit with a dermatologist or failed to complete their recommended dermatology appointment. This approach mimics the methods used for a randomized controlled trial. By including all patients, the analysis assessed whether dermatologists correctly classified patients who did and did not need a dermatology appointment based on their risk of Fasudil cancer, patient no-show rate, and risk of biopsy or skin cancer among those who had an appointment. To assess the number of face-to-face dermatologist visits needed to detect one skin cancer, we restricted to patients with a face-to-face visit14 and compared each of the teledermatology workflows to direct referral. All analyses were performed using SAS® version 9.4 (Cary , North Carolina).
This project was exempt from Institutional Review Board review.
We identified 111,117 potentially eligible primary care patients. We excluded 978 (0.8%) who had a dermatology visit during the 6 months before their primary care visit and then restricted to 59,279 patients (54%) who had a lesion. Of
Among patients with direct referral, 88% completed their dermatology appointment with average wait time of 7.3 days, 8% did not complete their appointment, 1% received advice without an appointment, and 3% had other dispositions. Among roving patients, 71% of consults were completed after the roving encounter, 27% had a follow-up visit in the dermatology clinic with average wait time of 5.8 days, and 2% had other dispositions. Among teledermatology patients, 49% received advice without an appointment (average, 5 hours), 39% received a dermatology appointment that was completed after an average wait time of 8.3 days, 6% were booked for a dermatology appointment that was not completed, and 6% had other dispositions. Considering the entire teledermatology population, 88% of dispositions were completed, and the average wait time to diagnosis was 3.3 days.
The intention-to-treat analysis included every patient who presented to primary care with a lesion. The frequency of biopsy was 35% among patients directly referred to dermatology (Table III). In multivariable analysis, the likelihood of a biopsy was highest for direct referral (all RR <1.0). The frequency of cancer detection was 12% for direct referral (Table III). In multivariable analysis, the RR for the association of diagnostic workflow with skin cancer detection was 1.02
(95% confidence interval [CI] 0.93-1.12) for roving, 1.09 (95% CI 1.02-1.16) for Stentor® with dermosc opy, and less than 1.0 for other teledermatology workflows. We also made comparisons after restricting to patients who received teledermatology, in which Stentor® with dermoscopy was associated with the highest likelihood of biopsy and cancer detection.
To assess the number of dermatology visits needed to detect one skin cancer, we restricted the analysis to those with a face-to-face visit to dermatology. All teledermatology workflows were associated with fewer dermatology visits to detect one skin cancer (Table IV). For example, Stentor® with dermoscopy required 39% fewer visits tha n direct referral (RR 0.61, CI 0.57-0.65).
We conducted a retrospective study to investigate the effectiveness and efficiency of four teledermatology and two face-to-face workflows in relation to risk of skin cancer detection, biopsy, and a visit to the dermatology clinic that did not result in a skin cancer diagnosis. Among direct referrals, 35% received a biopsy and 12% were diagnosed with skin cancer. We found that Stentor® with dermoscop y had 9% greater probability of cancer detection compared with direct referral (CI, 2% to 16%), was associated with 4% lower probability of biopsy (RR 0.96, CI 0.93-0.99), and required 39% fewer face-to-face visitsThese findings are consistent with anecdotal reports by dermatologists in our setting that they have a better view of the lesion with dermoscopy than with plain eyesight because of good image resolution with use of a large viewing screen. Alternatively, the consistent use of Stentor® with dermoscopy was the most common workflow used in our setting, and its consistent use may have led to a training effect for both the medical assistant who took the