No ethics approval was required for this study. The study participants (ie, pathologists and residents) agreed to—and coauthored—the study.

This study involved 17 participants who were divided into the following 4 groups or career levels, based on their pathology experience: (1) senior (pathologists with >20 years of experience, n=2), (2) expert (pathologists with 10-20 years of experience, n=5), (3) junior (pathologists with <10 years of experience, n=6), and (4) resident (1st year, n=1; 2nd year, n=3). Each of the 17 participants evaluated 25 digital cases, with a total of 425 digital images examined in the study. Overall, 1445 questions were examined (ie, 85 questions per participant) in the study.

In addition to their own diagnostic tasks, which were not considered in this study, the pathologists and residents received (1) a set of digital cases within the area of general surgical pathology, (2) specific questions to be addressed while reviewing the cases, and (3) a survey about their digital experience.

We set up 5 sets of digital cases representing 3 different specialties (breast: n=2; urology: n=1; and gastrointestinal: n=2) and assigned them to each study participant. Each test comprised 5 cases, represented by one or more slides of a single case that was previously diagnosed using conventional microscopy by the referral pathologist at our institution. The information reported about the original diagnosis was considered as the gold standard. To cover a spectrum of conditions overlapping the routine situation, we considered biopsy and surgical specimens (specimen type). Cases were digitalized using the Aperio AT2 scanner (Leica Biosystems) and visualized using the WebViewer APERIO ImageScope (version 12.1). The slides used for the tests were from 8 nontumoral and 17 tumoral cases. Of the tumoral cases, 7 tumors were benign and 10 were malignant; all malignant tumors were infiltrative and equally distributed between grade 2 and grade 3; 14 cases were biopsy and 11 were surgical.

Participants answered (all or some) of the following questions (ie, categories of diagnostic task), for each case: (1) Is it neoplastic or negative for neoplasia? (2) Is it a malignant (in situ or infiltrative) or a benign neoplasia? (3) What is the histopathological diagnosis? (4) What is the histotype of the lesion? (5) What is the grade of the lesion? Questions 1 and 3 were answered for all cases, question 2 was answered only for neoplastic lesions, and questions 4 and 5 were answered for malignant neoplasms.

To model data clusters, we used a varying effects, multilevel (hierarchical) model [14-16]. The rate of wrong answers (W_i) was modeled as a Bernoulli distribution:

For each pathologist (PID), their career level (LEVEL), the specific diagnostic question (CATEGORY), the specimen type (SPECIMEN), and the subspecialty of the case (SPECIALTY), we used the logit link function and modeled the varying intercepts as follows:

The prior distribution for the intercepts and SD values were as follows:

The hyperpriors for the hyperparameters average pathologist and σ_α were set as follows:

The SD value for was set at 1.5 since it produces a flat (weakly regularizing) prior after logit transformation [16,18]; moreover, we used an exponential distribution to model SD, because this assumes the least, for maximum entropy reasons [16,25-28], given the fact that σ is a nonnegative continuous parameter. To assess the validity of priors, we run prior predictive simulation of the model [16,29,30] (see Table S1 in Multimedia Appendix 1, and Multimedia Appendices 2 and 3). To limit divergent transitions, we reparametrized the models with a noncentered equivalent form [31,32]. Models were fit using Stan (a probabilistic programming language) and R [33,34]. Full anonymized data and custom code can be found in the public repository SmartCovid hosted on Github [35].

The survey was inspired by previous published works [36-38]. Briefly, the survey included 17 questions in a randomized order for all the pathologists, covering 3 fields: (1) attitude towards DP, (2) confidence in using DP solutions, and (3) satisfaction with DP. The survey was sent at the end of the digital experience. Pathologists were requested to answer the questions using a Likert scale, with scores ranging from 1 (strongly disagree) to 5 (strongly agree). The results were reported as the proportion of pathologists who assigned each single value of the Likert scale.

The pathologists answered 1345 of the total 1445 questions (100 missing answers in total), of which 1228 (91.30%) corresponded to the original diagnoses and were considered correct. Table 1 depicts the errors among each group of the 5 different categories recorded, and Figure 1 highlights the median (IQR) values of those categories. Considerable variation was observed among the performances of each pathologist, ranging from an error rate of 0.01 (1/67, Pathologist #4) to 0.32 (26/81, Pathologist #13), with a collective median error of 0.07 (IQR 0.04-0.11). This performance variation was tapered once the same data were considered after filtering among the different career levels, yielding the same median of 0.07, but a narrower IQR of 0.07-0.10. Moreover, some diagnostic tasks were more error prone than others; for instance, histotyping of the lesions had a very low rate of errors 0.01 (2/160), whereas grading was a more error-prone task with an error rate of 0.18 (27/147). The specimen type also resulted in different error rates, with surgical specimens easier to diagnose, with an error rate of 0.06 (40/716), than biopsy specimens, with a 2-fold error rate at 0.12 (77/629).

Differences in error rates for two important tasks—differentiation between neoplastic and nonneoplastic processes and that between benign and malignant neoplastic processes—were observed among pathologists at different career levels and for different specimen types. The same error profile was observed across career levels, although the former task had a higher error rate (Figure 2A). However, even though the differentiation of a neoplastic process from a nonneoplastic one might be more challenging on a biopsy specimen, the distinction between a benign and malignant neoplasm was done with the same error rate regardless of the specimen type (Figure 2B). Differences in the prevalence of errors among individual pathologists and those at different career levels, as well as across diagnostic tasks, specimen type, and case subspecialty, are further highlighted in Multimedia Appendices 4 and 5.

Diagnostics of the model’s fit are shown in Multimedia Appendices 6, 7, and 8. The analysis reported a good overall performance: the average pathologist showed a negative mean coefficient of -1.8 with most of the posterior probability mass below 0 (given the model structure, positive values reflect the probability of making errors; Table S2 in Multimedia Appendix 1). The pathologists’ individual performances and their career levels were the variables that showed less variance in predicting the error rate, whereas the specimen type, case subspecialty, and the particular type of task collectively showed more variance (Multimedia Appendix 9). Hence, we simulated the performance of an average pathologist at different career levels; this prediction shows better performance among pathologists at intermediate career levels of career (Figure 3).

Most pathologists reported a very good score (ie, 4 or 5 indicating they “moderately agree” and “strongly agree,” respectively) for their attitude toward DP (44/68, 64%), confidence in DP (75/119, 63%), and satisfaction with DP (56/102, 54.9%). A detailed analysis of these parameters showed that the residents reported the highest value for confidence, junior pathologists reported the highest values for attitude and satisfaction, whereas expert and senior pathologists reported relatively lower levels of confidence in and satisfaction with DP (Figure 4).

Our study showed an overall discordant rate of 9% among diagnoses performed using digital slides and those performed using the gold standard (ie, conventional microscopy). However, when we considered the different diagnostic tasks, this rate dropped to less than 5% in the category “benign versus malignant tumor”, which is probably the most clinically impacting category among the other diagnostic tasks. A systematic review of 38 pertinent studies published before 2015 reported a 7.6% overall discordance rate between digital and glass slide diagnoses. Among these studies, 17 studies reported a discordant rate higher than 5%, and 8 reported a discordant rate higher than 15% [39]. A later reanalysis of the same series fixed the overall discordance rate to 4% and major discrepancies to 1% [40]. A more recent review, covering studies published until 2018, reported a disagreement ranging from 1.7% to 13% [10]. Two multicentric, randomized, non-inferiority studies reported major discordant rates of 4.9% [41] and 3.6% [42] between diagnoses done by digital and glass slides. Furthermore, a study from a single, large academic center reported an overall diagnostic equivalency of 99.3% [43]. The same group was also the first to report about the use of DP during COVID-19 with an overall concordance of 98.8% [44]. Thus, despite our challenging approach to DP, the diagnostic performance we recorded was consistent with previous reports—a result that further supports the transition to DP.

In our study, a high proportion of errors was generated in small biopsy specimen type (12.2%) and diagnostic tasks involving tumor grading (23%). These results are consistent with those of the review by Williams et al [40]. The latter showed that 21% of all errors concerned grading or histotyping of malignant lesions, whereas 10% of the errors could be ascribed to the inability to find the target.

Moreover, recent studies have consistently reported high, intermediate, and low discordant rates for bladder, breast, and gastrointestinal tract specimens, respectively [41,42]—a finding suggesting intrinsic difficulties of specific areas. In contrast, we observed 4%, 8%, and 12% of discrepancies for urology, gastrointestinal tract, and breast specimens. This result could be attributed to a nonrandom selection of the cases and might represent a study limitation, biasing the value of the coefficients of specific parameters of the case subspecialty, similar to those of diagnostic tasks and the specimen type. However, these characteristics were excluded in the posterior predictive simulation, which was intended to represent how the different career levels might impact the pathologists’ performance, after adjusting for all other factors.

As compared by the study by Hanna et al [44], our readiness to undertake digital diagnostic tasks was far from being mature in March 2020, and this study was specifically designed to identify and illustrate the effects of such a sudden adoption of DP—something that had never been investigated before. Our results suggest that this abrupt transition might not influence the adoption of and performance with DP. However, different factors seem to be involved. In particular, data concerning major discrepancies between diagnoses using DP and gold standard methods disclosed an interesting feature. Both in the distinction of neoplastic versus non-neoplastic lesions and of benign versus malignant tumors, the worst results obtained were among residents and senior pathologists–2 contrasting categories in terms of pathologists’ working experience. Therefore, these survey results might suggest an explanation to this paradoxical result: senior pathologists felt ready to diagnose a pathological process using a digital approach (ie, positive attitude) but were less prepared to use digital devices (ie, low confidence). Residents, in turn, had a high predisposition to using a digital device (ie, high confidence) but also had some concerns about diagnosis of a pathological process (ie, poor attitude). The hypothesis that senior pathologists were limited by a digital gap was supported by another finding: once they decided a lesion was malignant, they demonstrated the best performance with regard to tumor grading. By contrast, residents made several errors, likely due to their limited working experience. Lastly, even if expert pathologists showed a good diagnostic performance, they had the lowest level of satisfaction in DP. This result suggests that DP can be adopted rapidly for practical purposes. However, it also highlights a critical point of the process that needs to be addressed, possibly with adequate training or user-friendly equipment, and warrants further investigations.

Our study describes how the abrupt transition to DP affected the quality of diagnoses and qualitatively gauged the psychological aspects of this abrupt transition. Moreover, our study model highlighted the potential causes for these challenges and might inform what could be expected in other laboratories. In conclusion, the exceptional conditions dictated by the COVID-19 pandemic highlighted that DP could be adopted safely for diagnostic purposes by any skilled pathologist, even abruptly.