We employed a convergent mixed methods design for this quasi-experimental pilot study. We also included a sample of patient encounters that were randomly assigned to a scribe (scribe present) or nonscribe (no scribe) group. To determine the effect of scribes, we collected patient surveys on perceptions of patient-provider communication and satisfaction with the encounter, video recordings of clinical encounters, and physician interviews focused on scribes. All methods were approved by the University of Michigan IRBMED Institutional Review Board (HUM00123396).

We recruited three physician-scribe pairs and 34 patients (19 and 15 randomized to scribed and nonscribed encounters, respectively). Only 31 recordings were obtained due to technical and user error, resulting in 17 scribed and 14 nonscribed recordings. Participants were predominantly of white race (79.4%) and male gender (67.7%), with income >US $50,000 (73.5%). Participants were, on an average, 51.1 years (SD 19.1) of age and equally divided between having a bachelor’s degree or higher education and some college or less education (Table 1).

Patients reported very high satisfaction with physician communication. Communication scores from the CAT were positively skewed, with no respondents rating anything less than “Good,” and the majority of items rated “Very Good” or ”Excellent.” Overall, 100% of respondents reported that their physicians’ communication was Excellent/Very Good in terms of greeting patients in a way that made them feel comfortable, treating patients with respect, showing interest in patients’ ideas about their health, understanding main health concerns, paying attention to the patient, giving as much information as the patient wanted, talking in terms that the patient could understand, checking to be sure the patient understood everything, discussing next steps, and showing care and concern. Neither the mean score of the CAT nor proportion of excellent responses had much variability; 21 of the 34 patients responded “Excellent” to all 14 questions (Table 2). No significant differences were found in communication scores between the scribed and nonscribed encounters.

All patients had positive assessments of their interaction with their physicians, their physician’s EHR use, and satisfaction with care. No significant differences were found between groups in terms of these aspects; however, response variability was low among both groups, with most items skewed positively (Table 3). The exception was only 52.6% of patients in the scribed encounters compared to 93% in the nonscribed encounters who indicated that physicians used the computer (P=.02).

The above mentioned findings were supported by video analysis, where both groups scored high on 12 of the 13 IAT domains, with little to no variability. The Introduction domain was discarded, as most recordings (n=23) started after physicians entered the room and presumably made their introductions. For the remaining 12 domains, three domains (Questions, Wait Time, and Concern) had no variability between groups, with all recordings coded as having the highest performance possible. An additional eight domains (Eye Contact with Patient, Nonverbal Communication, Listening, Organization, Information Gathering, Focus, Empathy, and Awareness of Unspoken Issues) had two or fewer videos in either group coded as 3. No encounter was coded as having the poorest performance (score of 1 or 2) on any domain. This suggests that physicians consistently demonstrated high performance across all IAT domains. The Closure domain had the largest difference between groups, with 50% of nonscribed encounters showing a score of 3 and 50% showing a score of 4; in addition, 18% of scribed encounters showed a score of 3 and 82% showed a score of 4. The presence of scribes was not associated with performing less (≤3 on IAT) than the highest category of performance (4 on IAT).

Video recordings revealed that scribes had little effect on encounters. Although the scribed encounters were slightly shorter, this difference was not significant (mean 15.6 [SD 5.4] min vs mean 16.5 (SD 6.7) min; P=.70). When scribes were present, physicians spent slightly less time looking at the computer (mean 16.1% [SD 15.5%]) than when scribes were absent (mean 29.8% [SD 23.7%]; P=.06). Neither the number of problems addressed nor the number of orders placed differed significantly between groups. Across all visits, the mean number of problems addressed was 3.4 (SD 1.4; scribe: mean 3.3 [SD 1.4] vs nonscribe: 3.6 [SD 1.6]; P=.60) and the mean number of orders placed was 0.9 (SD 1.0; scribe: mean 0.8 [SD 0.8] vs nonscribe: mean 1.1 (SD 1.2); P=.41). Linear regression results revealed that there were no significant differences between the scribed and nonscribed encounters with regard to the time to close charts.

Although it was standard protocol at this institution for medication orders to be entered by physicians or pended (but not signed) by medical assistants, scribes were restricted by institutional policy from pending medication orders. However, during the time of our data collection, the family medicine clinics where data collection occurred were participating in an institutional pilot that granted permission for scribes to pend (but not sign) medication orders during an encounter. There were 27 medication orders (10 renewals) across 18 encounters. Most medication orders were entered by the physician, and only nine were entered by the scribe, all of which were new orders. No medication errors were identified when comparing video recordings to EHR data.

Qualitative interviews yielded five major themes regarding scribes (also see Multimedia Appendix 2).

Our results showed that scribes did not have any significant impact on measures of patient satisfaction or the encounter itself, suggesting that from a patient’s perspective, scribes are acceptable to patients.

Overall, patients were pleased with the medical care they received and were satisfied, regardless of scribe presence, which is consistent with the literature [13,14]. Although, Pozdnyakova et al also found no differences in patient satisfaction between patients in scribed and nonscribed encounters, they found that compared to patients aged ≥65 years, younger patients were more likely to find physicians attentive and provide more education when scribes were present [15]. Although our sample was small and had too little variability in patient satisfaction measures to find differences between older and younger patients, our findings merit further investigation.

Our study also showed that scribes had little impact on the encounter itself. When scribes were present, physicians spent more time in the encounter looking at the patient as opposed to the computer, a finding that was marginally significant. This is consistent with other similar findings in the literature [16]. Patients noticed that that their physician was the professional using the computer more frequently in nonscribed encounters as compared to scribed encounters (93% vs 52.6%), but this factor too had little impact on patients’ perceptions of the encounter. Although scribed encounters were, on an average, about 1 min shorter, this marginal efficiency was not statistically significant, which is contrary to other studies that found efficiencies in terms of physicians’ ability to see more patients per hour [16]. It is possible that our lack of a significant difference is due to our small sample size. Regardless, our findings support the idea that scribes do not lengthen clinical encounters and may even save time.

Our findings that scribes do not affect patient satisfaction and that they have little effect on the encounter itself are important. Physician participants reported positive experiences and satisfaction with their scribes, which has been reported elsewhere [8,14,17,18]. Administrative duties require substantial physician time and have affected physicians’ perceptions of ability to deliver high-quality care, career satisfaction, and burnout [19]. In fact, the demands of documentation and EHR use are a chief contributor to physician burnout [20,21]. In a large national study by Shanafelt et al, physicians’ satisfaction with EHRs and computerized physician order entry was generally low, and physicians who used these systems were at higher risk for professional burnout [22]. The topic of physician burnout has been gaining national attention, as it has been strongly correlated with health issues such as depression, drinking problems, and cardiovascular and digestive disorders as well as use of sedatives and overeating [23]. Previous research has shown that providers find scribes valuable and that they reduce documentation time [17]. Future work should seek to more clearly elucidate the relationship between scribes and physician burnout.

Although our quantitative results suggest support for scribes in clinical encounters, integrating results from our physician interviews highlight additional important context. The negative effect of scribe turnover was highlighted by nearly every physician. Scribe vendor services often employ young professionals in their gap year before medical school [24]. This creates a system where a cadre of scribes enters the workforce for 9-12 months before leaving for medical school, which creates a lack of consistency within the clinic and has significant transaction costs.

This discussion coincides with one of our providers theorizing about expanding the role of a scribe, allowing them to perform duties of an MA, or vice versa, where an MA can also assist in documentation. It is important to note that although we focused our research in a health system that relies on vendor scribes, other models have been reported in the literature, such as using nurses or MAs as scribes in addition to their regular duties [15,25]. The medical background and training for this cross-over type of role would need further investigation. Finally, although scribes cost money, they are often cost neutral [8]. McCormick et al. found that the return-to-investment ratio was greater than 6:1 when using scribes but had no effect on patient satisfaction [18].

An important limitation of our study was the small sample of physician-scribe pairs. With only three physicians enrolled, we were limited to their patients and only their insight, which limits generalizability. Moreover, we did not collect data from scribes. Future work should incorporate scribes as research participants to obtain their perspective on workflow and identify challenges in working with different physicians and specialties. Another limitation of our approach is the fact that physicians and scribes were completely aware of which encounters were being recorded (in most cases, it was the physician who started the recording). Because of this knowledge, it is possible that physicians and scribes may have altered their behavior within encounters. The only way to mitigate this possibility would be through discreet recording, which is not ethical or appropriate. It is possible that with more recorded encounters, any potential effects of recording may dissipate; however, this is not guaranteed. We must also note that despite the fact that survey responses were not shared with physicians, patient participant survey responses were not anonymized to study staff. Because responses were not completely anonymous to study staff, it is possible that patient participants may not have answered survey items truthfully, which may have contributed to the lack of variability and positive skew in survey results. Although scribes are often employed to help reduce documentation burdens among physicians, we did not design this study to verify whether that was the case. Despite finding a nonsignificant difference in encounter duration, with scribed encounters lasting for a slightly shorter duration than nonscribed encounters, we do not know if the presence of a scribe had an impact on documentation burden. Future work should investigate whether these theoretical efficiencies are actually established both inside and outside clinical encounters. Finally, we did not assess physician EHR literacy or comfort, which may have a possible effect on individuals’ use of scribes. Future work should seek to better understand the relationship between physician EHR literacy and scribe-related efficiencies.