key: cord-0764730-jxuk9qgx
authors: Gabel, Colleen K.; Nguyen, Emily; Karmouta, Ryan; Liu, Kristina Jing; Zhou, Guohai; Alloo, Allireza; Arakaki, Ryan; Balagula, Yevgeniy; Bridges, Alina G.; Cowen, Edward W.; Davis, Mark Denis P.; Femia, Alisa; Harp, Joanna; Kaffenberger, Benjamin; Keller, Jesse J.; Kwong, Bernice Y.; Markova, Alina; Mauskar, Melissa; Micheletti, Robert; Mostaghimi, Arash; Pierson, Joseph; Rosenbach, Misha; Schwager, Zachary; Seminario-Vidal, Lucia; Sharon, Victoria R.; Song, Philip I.; Strowd, Lindsay C.; Walls, Andrew C.; Wanat, Karolyn A.; Wetter, David A.; Worswick, Scott; Ziemer, Carolyn; Kvedar, Joseph; Mikailov, Anar; Kroshinsky, Daniela
title: Use of teledermatology by dermatology hospitalists is effective in the diagnosis and management of inpatient disease.
date: 2020-05-07
journal: J Am Acad Dermatol
DOI: 10.1016/j.jaad.2020.04.171
sha: f1b447dbca8f5cfd644e3965e7942f55df3ef40c
doc_id: 764730
cord_uid: jxuk9qgx

Abstract Background Patient outcomes are improved when dermatologists provide inpatient consults. Inpatient access to dermatologists is limited, illustrating an opportunity to utilize teledermatology. Little is known about the ability of dermatologists to accurately diagnose and manage inpatients using teledermatology, particularly utilizing non-dermatologist generated clinical data. Methods This prospective study assessed the ability of teledermatology to diagnose and manage 41 dermatology consults from a large urban tertiary care center utilizing internal medicine referral documentation and photos. Twenty-seven dermatology hospitalists were surveyed. Interrater agreement was assessed by the kappa statistic. Results There was substantial agreement between in-person and teledermatology assessment of the diagnosis with differential diagnosis (median kappa = 0.83), substantial agreement in laboratory work-up decisions (median kappa = 0.67), almost perfect agreement in imaging decisions (median kappa = 1.0), and moderate agreement in biopsy decisions (median kappa = 0.43). There was almost perfect agreement in treatment (median kappa = 1.0), but no agreement in follow-up planning (median kappa = 0.0). There was no association between raw photo quality and the primary plus differential diagnosis or primary diagnosis alone. Limitations Selection bias and single-center nature. Conclusions Teledermatology may be effective in the inpatient setting, with concordant diagnosis, evaluation, and management decisions.

-Inpatient access to dermatologists is limited, highlighting an opportunity to utilize 81 teledermatology within the inpatient setting. 82 -Teledermatology in the inpatient setting may be a clinically acceptable option for 83 diagnosis, evaluation, and management. This may represent a novel and effective option 84 for hospitals .  85   86  87  88  89  90  91  92  93  94  95  96  97  98  99  100  101  102  103  104  105  106  107  108  109  110  111  112  113  114  115  116  117  118  119 ABSTRACT: 120 121

Background: Patient outcomes are improved when dermatologists provide inpatient consults. 122

Inpatient access to dermatologists is limited, illustrating an opportunity to utilize 123 teledermatology. Little is known about the ability of dermatologists to accurately diagnose and 124 manage inpatients using teledermatology, particularly utilizing non-dermatologist generated 125 clinical data. 126

Methods: This prospective study assessed the ability of teledermatology to diagnose and manage 127 41 dermatology consults from a large urban tertiary care center utilizing internal medicine 128 referral documentation and photos. Twenty-seven dermatology hospitalists were surveyed. 129

Interrater agreement was assessed by the kappa statistic. 130

Results: There was substantial agreement between in-person and teledermatology assessment of 131 the diagnosis with differential diagnosis (median kappa = 0.83), substantial agreement in 132 laboratory work-up decisions (median kappa = 0.67), almost perfect agreement in imaging 133 decisions (median kappa = 1.0), and moderate agreement in biopsy decisions (median kappa = 134 0.43). There was almost perfect agreement in treatment (median kappa = 1.0), but no agreement 135 in follow-up planning (median kappa = 0.0). There was no association between raw photo quality 136 and the primary plus differential diagnosis or primary diagnosis alone. Teledermatology is the remote dermatologic assessment of patients, in real-time ("live 145 interactive"), by accessing stored data ("store-and-forward"), or a combination of the two 146 ("hybrid"), with worldwide applications.(1) Teledermatology has been studied in general triage, 147 consultation in remote locations, and monitoring of chronic skin conditions.(1) In addition to 148 increased access to dermatologists, potential benefits of store-and-forward teledermatology 149 include cost reduction due to fewer face-to-face (FTF) consultations,(2) reduced travel time and 150 opportunity cost due to missed work,(3-5) and reduced contagion spread amid infectious disease 151

Significant clinical evidence supports the outpatient use of store-and-forward 153 teledermatology.(2-10) In contrast, teledermatology has been studied in the inpatient setting to a 154 limited degree. A significant practice gap exists between the demand for inpatient dermatology 155 services and access to dermatologists,(11, 12) often a source of frustration for inpatient providers 156 and patients. Dermatology hospitalists represent a clinical group with expertise in complex 157 medical dermatology and the diagnosis and management of skin diseases affecting hospitalized 158 patients. Involvement of dermatology hospitalists in the care of hospitalized patients has been 159 found to improve patient outcomes.(13) In a subset of cases, inpatient teledermatology reduces 160 time for the primary medical team to receive a response for a dermatology consultation. (14) 161 Dermatologist interest in inpatient teledermatology is high. A survey of attending dermatologists 162 demonstrated that 61.5% agreed or strongly agreed that teledermatology helps inpatient care.(15) 163

Another study found that 95% of hospital and emergency department practitioners would utilize 164 a teledermatology consult service if available, however only 5% believed that teledermatology 165 would be equivalent to a face-to-face (FTF) consult.(16) This finding supports the need for additional studies evaluating inpatient teledermatology, which may shift perception and 167 encourage adoption of inpatient teledermatology. 168

This study investigates the diagnostic and management agreement between inpatient FTF and 169 store-and-forward teledermatology evaluations utilizing remote digital evaluations for hospital-170 based dermatology consultations. 171

Eligible patients for this study were admitted to Massachusetts General Hospital between July 174 and August 2013 and had a dermatology consultation staffed by a dermatology hospitalist with 175 more than six years of inpatient experience, defined as the Primary Dermatologist (PD). This 176 yielded a sample of 108 patients. Only those consultations with digital images and non-177 dermatology evaluations involving the dermatologic complaint were included. Cases were 178 selected if the accuracy of the PD's diagnosis was able to be confirmed based on testing, 179 response to therapy and final diagnosis at discharge. Based on these inclusion criteria, a total of 180 42 patients were initially included ( Figure 1 ). One case was excluded from analysis to preserve 181 the generalizability of study results,(17) as this patient presented with multiple concomitant 182 dermatologic complaints and the documentation did not specify the specific focus of the 183 dermatology consultation. 184

For teledermatology review, data abstractors not involved in the care of the included cases 185 packaged patient data into surveys by unique numerical patient identifiers. Each survey set 186 contained seven individual cases, randomly assigned to each survey set from the total case pool. 187

Each individual case contained the relevant history and physical exam notes generated by a non-188 dermatologic internal medicine or emergency medicine provider. In addition, all data such as 189 laboratory studies, imaging, microbiology, pathology, and digital images up to the day of the consult that would have been available to the PD were included. Finally, a 191 diagnosis/management questionnaire was included. The order of case examination within each 192 survey set was fixed across all TDs. Patient identifiers were uniquely created and stored safely. The surveys included the option to list a primary diagnosis as well as a maximum of three 205 differential diagnoses. The workup and management plans offered were as follows: (1) biopsy, 206

(2) topical therapy, (3) systemic/oral therapy, (4) microbiology, (5) labs, (6) transfer to the burn 207 unit, if not already there, (6) recommend continued patient monitoring as an inpatient, and (7) 208 recommend follow-up as outpatient for dermatologic condition. Once the TD selected a 209 treatment plan, s/he was prompted for free-text details. Both the correct mode and type of 210 therapy was assessed. If the selected treatment differed between the PD and the TD but both 211 options were within the accepted standard of care for that disease, these treatments were 212 considered concordant. This was to minimize the effect of stylistic practice differences in 213 grading appropriateness. 214

The follow-up plan options were: (1) sign-off and no need for future follow-up either inpatient or 215 outpatient, (2) outpatient follow-up, no need for additional inpatient dermatology evaluations 216 ("sign off"), (3) no need to see the patient tomorrow, but evaluate if the primary team requests 217 and ensure outpatient follow-up planned, and (4) see the patient tomorrow and follow closely. 218

TDs rated their degree of comfort in managing the case as a dermatologist, as well as the quality 219 of each image. 220

Outcomes measured were concordance between the PD and the TDs for the following: primary 221 diagnosis, primary diagnosis plus differential diagnosis, decision to biopsy, laboratory work-up, 222

imaging, treatment, and follow-up plan. Primary outcomes were defined as primary plus 223 differential diagnostic concordance as well as management plan concordance, the rational of 224 which was to assess whether teledermatology could result in an appropriate work-up and 225 management leading to an effective outcome for the patient. Secondary outcomes were primary 226 diagnostic concordance alone, as well as concordance in work-up. 227

Primary diagnostic concordance was defined as agreement between the primary diagnosis 228 provided by the PD and the TD. Primary diagnostic plus differential diagnostic concordance was 229 defined as the PD's diagnosis being among the differential diagnosis of the TDs in cases when 230 the primary diagnosis was discordant. The diagnoses themselves, and not diagnostic family, were 231 used in calculating diagnostic concordance. 232

We calculated the prevalence-adjusted bias-adjusted kappa (20) to quantify the concordance 234 between a) the TDs' and PD's primary diagnosis, b) TDs' primary diagnosis plus differential diagnosis and PD's primary diagnosis, and c) TDs' and PD's management plan (separately for 236 each of the five domains: biopsy, work-up, imaging, treatment, and follow-up). The following 237 criteria were used to assess significance: values ≤ 0 as indicating no agreement, 0.01-0.20 as 238 none to slight, 0.21-0.40 as fair, 0.41-0.60 as moderate, 0.61-0.80 as substantial, and 0.81-1.00 239 as almost perfect agreement. (21) We evaluated the associations of the calculated concordance a) 240 and b) with TDs' years of experience and the reported photo quality rating, and the associations 241 of the calculated concordance c) with photo quality using the Pearson correlation coefficient. We 242 also evaluated the associations of TDs' level of comfort managing patients (with photos and 243 story alone) with photo quality and TDs' years of experience using the Wilcoxon rank sum test. 244

All were conducted using R version 3.6.1 (https://www.r-project.org/). 245 inflammatory (17.1%), neoplastic (7.3%), iatrogenic (4.9%) and traumatic (4.9%). 252

The TDs were 40.7% female and practiced in diverse academic institutions from all geographic 253 regions of the United States. The mean number of years' experience of each of the TDs was 7.0 254 (SD 1.2) ( Table 2) . Out of all cases, 45.1% of TDs felt comfortable managing the case as a 255 teledermatologist. The mean number of differential diagnoses per TD per individual case was 2.6 256 (SD 0.4).

There was fair concordance between PD and TD primary diagnosis alone (median concordance 258 66.7%, interquartile range (IQR) 57.1% to 78.6%; median kappa=0.33, interquartile range (IQR) 259 0.14 to 0.57), with substantial agreement between PD and TD primary plus differential diagnosis 260 (median concordance 91.7%, IQR 85.7% to 92.9%; median kappa=0.83, IQR 0.71 to 0.86). 261

There was substantial agreement in pursuing additional laboratory work-up (median concordance 262 85.7%, IQR 85.7% to 92.9%; median kappa=0.67, IQR 0.43 to 0.79), and almost perfect 263 agreement in imaging decisions (median concordance 100%, IQR 50.0% to 100.0%; kappa=1.0, 264 IQR, 0.0-1.0). There was moderate agreement in the decision to biopsy (median concordance 265 71.4%, IQR 53.6% to 85.7%; median kappa=0.43, IQR 0.07 to 0.71). There was almost perfect 266 agreement in treatment plans (median concordance 100%, IQR 85.7% to 100.0%; median 267 kappa=1.0, IQR 0.67 to 1.0). There was no agreement in the follow-up plan (median 268 concordance 50.0%, IQR 42.9% to 66.7%; median kappa=0.0, IQR -0.14 to 0.14). Figure 2 is a 269 pair of histograms depicting the distribution of kappa values for agreement between the TDs' and 270 the PD's primary diagnosis (Figure 2A) , and primary plus differential diagnosis ( Figure 2B) . 271

There was no association between experience of the TD and primary plus differential diagnostic There was no association between either raw photo quality and the primary plus differential 280 diagnosis (correlation=0.008; 95% CI, -0.18-0.19), or primary diagnostic concordance alone 281 (correlation=-0.07; 95% CI, -0.12-0.25). The Wilcoxon rank sum test of the TDs' comfort with 282 managing the case and years of experience indicated that TDs with fewer years of experience 283 were more likely to feel comfortable managing the patients as a teledermatologist (p=0.04). 284

This study illustrates that store-and-forward teledermatology may be reliable in the academic 286 inpatient setting, with strong agreement between PD and TD for diagnosis, work-up, and 287

The high concordance of primary plus differential diagnosis is in-line with prior outpatient 289 literature, (8, 22) with studies demonstrating diagnostic concordance ranging from 41% to 100% 290 for store-and-forward cases.

(2) This finding builds upon limited studies evaluating the use of 291 teledermatology in the inpatient setting.(12, 23, 24) As with prior study,(2) diagnostic 292 concordance improved when the differential diagnosis was taken into account. 293

The decision by TDs to pursue work-up in this study was highly concordant, with substantial 294 agreement in the laboratory work-up desired. However, there was only moderate agreement in 295 the decision to biopsy, which is in contrast with a prior inpatient teledermatology study finding a 296 >95% concordance in assessing need for biopsy.(12) This may be due to stylistic practice 297 differences or individual comfort level. 298

The treatment plans offered by the TDs were highly concordant with those of the PD, suggesting 299 that the outcomes of each patient may have been the same if managed by teledermatology, even 300 in cases where the primary diagnosis differed. This may be due to the high concordance of 301 primary plus differential diagnosis, leading to treatment plans applicable to multiple diagnoses.

The baseline inter-dermatologist variability that occurs even with face-to-face consultations must 303 also be taken into consideration, as a previous study of face-to-face, clinic-based dermatologists 304 has found diagnostic testing to be 85% concordant, medical-based therapy to be 85% concordant, 305 and clinic-based therapy 77% concordant, respectively.(22) Thus, some degree of discordance 306 may be expected. 307

The lack of concordance between TDs and the PD for follow-up plans suggests that in-person 308 evaluation may be needed prior to disposition planning. Stylistic differences also likely played a 309 role. Patient-specific factors may go into disposition planning, such as access to resources and 310 health literacy, which may contribute to the discordance between the PD and the TDs. Further 311 study of follow-up planning is needed to elucidate whether teledermatology may be reliable for 312 this use. 313

Photo quality was not associated with primary diagnostic concordance or primary plus 314 differential diagnostic concordance. This suggests that even in cases in which image quality is 315 suboptimal, the reliability of teledermatology may not be impacted. However, while the authors 316 utilized images from heterogeneous sources, many photos utilized in the study surveys met the 317 minimum standards recommended for teledermatology.(25) Additionally, assessment of image 318 quality was not broken down into detailed components, such as lighting, focus, or capture of 319 clinically-relevant information. Photo quality and training in obtaining photos may be needed to 320 ensure good capture of the relevant areas when implementing teledermatology, as the study 321 photos were captured by dermatology resident physicians. 322

There was no association between experience of the teledermatologist and diagnostic 323 concordance, illustrating the generalizability of teledermatology across all ages of practicing 324

There appeared to be a disconnect between concordance and the TDs' level of comfort in 326 managing each case as a teledermatologist. The TDs considered themselves comfortable less 327 than half of the time; however, their survey responses often aligned with the PD. This may be in 328 part due to the novelty of teledermatology. The TDs with fewer years of experience were more 329 likely to feel comfortable managing the case, aligning with prior literature,(26) reflecting an 330 opportunity to utilize teledermatology even in novice practice settings. Similarly, 331 teledermatology exposure in residency may correlate with comfort of use,(27) suggesting that 332 early incorporation of teledermatology in training may facilitate its implementation. 333

One of the greatest strengths of this study is the large sample size of TDs, mimicking the 334 heterogeneity of applying teledermatology to real-life practice settings. The distribution of 335 diagnoses included in this study reflects that of common dermatology consultations.(13) 336

Limitations of this study include its single-center nature and the fact that dermatology residents 337 captured the clinical photos. The dermatology residents may have had a more thorough 338 understanding of how to obtain a high-quality dermatology photo than non-dermatology staff, 339 who would be submitting the teledermatology consult in real-life. Training of non-dermatology 340 staff in obtaining high-quality images may be needed. On the other hand, camera technology has 341 likely improved today and may lead to heightened quality of photos in today's use of 342 teledermatology. Further study is needed to determine best practices for implementing an 343 inpatient teledermatology program. 344

In conclusion, teledermatology may be effective for managing dermatologic disease in the 345 inpatient setting and leads to highly concordant diagnostic, work-up, and management decisions 346 when performed by experienced inpatient dermatologists. This may represent a novel and 347 effective option for community hospitals and may be particularly applicable during times of concern for spread of infectious disease, such as during the 2019-2020 outbreak of the severe 349 acute respiratory syndrome coronavirus 2 (SARS-CoV-2). 350 351 

Telemedicine in dermatology: findings and experiences 368 worldwide -a systematic literature review

Two Decades of 371 Teledermatology: Current Status and Integration in National Healthcare Systems

Teledermatology for diagnosis and management of skin conditions: a systematic review

Store-and-forward 379 teledermatology results in similar clinical outcomes to conventional clinic-based care

Mobile 382 teledermatology for skin tumour screening: diagnostic accuracy of clinical and dermoscopic 383 image tele-evaluation using cellular phones

The substitution 385 of digital images for dermatologic physical examination

Comparison of skin 387 biopsy triage decisions in 49 patients with pigmented lesions and skin neoplasms: store-and-388 forward teledermatology vs face-to-face dermatology

Teledermatology: A Review and Update

Practice Gaps. Improving accessibility to inpatient dermatology through 392 teledermatology

The 394 reliability of teledermatology to triage inpatient dermatology consultations

Consultation With Accuracy of Cutaneous Disorder Diagnoses in Hospitalized Patients: A 398 Multicenter Analysis

Teledermatology as a means to improve access to 400 inpatient dermatology care

Inpatient teledermatology: current state and 402 practice gaps

Consultative teledermatology in the emergency 404 department and inpatient wards: A survey of potential referring providers

Generalizability: the trees, the forest, and the low-hanging fruit

Research electronic data 409 capture (REDCap)--a metadata-driven methodology and workflow process for providing 410 translational research informatics support

The REDCap 412 consortium: Building an international community of software platform partners

Bias, prevalence and kappa

Interrater reliability: the kappa statistic

Reliability and 419 accuracy of dermatologists' clinic-based and digital image consultations

Inpatient teledermatology: Diagnostic and therapeutic 422 concordance among hospitalist, dermatologist, and teledermatologist using store-and-forward 423 teledermatology

Incorporating teledermatology into 425 emergency medicine

Practice 427 Guidelines for Teledermatology

A 429 survey-based study of diagnostic and treatment concordance in standardized cases of cellulitis 430 and pseudocellulitis via teledermatology

Exposure to teledermatology and resident preparedness for 432 future practice: results of a national survey

Capsule summary:-Inpatient access to dermatologists is limited, highlighting an opportunity to utilize teledermatology within the inpatient setting.-Teledermatology in the inpatient setting may be a clinically acceptable option for diagnosis, evaluation, and management. This may represent a novel and effective option for hospitals.