Telemedicine Use across Medical Specialties and Diagnoses

Abstract

Background The COVID-19 (coronavirus disease 2019) pandemic rapidly expanded telemedicine scale and scope. As telemedicine becomes routine, understanding how specialty and diagnosis combine with demographics to impact telemedicine use will aid in addressing its current limitations.

Objectives To analyze the relationship between medical specialty, diagnosis, and telemedicine use, and their interplay with patient demographics in determining telemedicine usage patterns.

Methods We extracted encounter and patient data of all adults who scheduled outpatient visits from June 1, 2020 to June 30, 2021 from the electronic health record of an integrated academic health system encompassing a broad range of subspecialties. Extracted variables included medical specialty, primary visit diagnosis, visit modality (video, audio, or in-person), and patient age, sex, self-reported race/ethnicity and 2013 rural–urban continuum code. Six specialties (General Surgery, Family Medicine, Gastroenterology, Oncology, General Internal Medicine, and Psychiatry) ranging from the lowest to the highest quartile of telemedicine use (video and audio) were chosen for analysis. Relative proportions of video, audio, and in-person modalities were compared. We examined diagnoses associated with the most and least frequent telemedicine use within each specialty. Finally, we analyzed associations between patient characteristics and telemedicine modality (video vs. audio/in-person, and video/audio vs. in-person) using a mixed-effects logistic regression model.

Results A total of 2,494,296 encounters occurred during the study period, representing 420,876 unique patients (mean age: 44 years, standard deviation: 24 years, 54% female). Medical diagnoses requiring physical examination or minor procedures were more likely to be conducted in-person. Rural patients were more likely than urban patients to use video telemedicine in General Surgery and Gastroenterology and less likely to use video for all other specialties. Within most specialties, male patients and patients of nonwhite race were overall less likely to use video modality and video/audio telemedicine. In Psychiatry, members of several demographic groups used video telemedicine more commonly than expected, while in other specialties, members of these groups tended to use less telemedicine overall.

Conclusion Medical diagnoses requiring physical examination or minor procedures are more likely to be conducted in-person. Patient characteristics (age, sex, rural vs. urban, race/ethnicity) affect video and video/audio telemedicine use differently depending on medical specialty. These factors contribute to a unique clinical scenario which impacts perceived usefulness and accessibility of telemedicine to providers and patients, and are likely to impact rates of telemedicine adoption.

Keywords

telemedicine - health - disparities

Background and Significance

The coronavirus disease 2019 (COVID-19) pandemic rapidly expanded telemedicine's scale and scope.[1] Our group previously performed a cross-sectional analysis of disparities within outpatient telemedicine use within our large academic health system in the Upper Midwest during a time period after the first COVID-19 surge. This prior work demonstrated that older age, rural status, lower socioeconomic status, and nonwhite race are associated with lower video-based telemedicine use.[2] We now aim to expand on this research further examining patterns of telemedicine use by medical specialty and primary visit diagnosis. It is necessary to consider both of these variables when examining use of telemedicine as certain clinical scenarios may lend themselves more easily to telemedicine than others. For example, even prior to the pandemic, behavioral health comprised most telemedicine visits in the United States given that these encounters are largely discussion-based.[3] Meanwhile, inability to conduct parts of the physical exam (e.g., auscultation or observation using special equipment; palpation) and minor procedures remain important limitations to telemedicine visits and remain a point of concern for clinicians.

Medical procedures require in-person visits; however, previous work has shown that highly procedural specialties, such as surgical subspecialties, still benefit from integration of telemedicine. One systematic review found telemedicine to be effective in preoperative assessment and diagnosis across a wide range of subspecialties.[4] Data from randomized controlled trials in orthopaedic surgery have found remote consultations, under proper conditions, are comparable with standard consultations in rates of referral to surgery and patient and provider satisfaction.[5] Some authors have investigated the eligibility for telemedicine of different clinical scenarios within particular subspecialties.[6] Few have performed a comprehensive, comparative analysis of relative frequency of telemedicine use across different medical specialties and diagnoses, particularly after the onset of the COVID-19 pandemic. Our previous work explored barriers to telemedicine by patient characteristics. In this work, we aim to examine specific clinical scenarios by patient demographic, diagnosis, and specialty, and their associations with the usage patterns for both provider and patient.

Under the Unified Theory of Acceptance and Use of Technology (UTAUT), the framework most commonly applied to telemedicine,[7] use of technology by an individual is a function of four key constructs: performance expectancy (perceived usefulness of the system for accomplishing the task at hand), effort expectancy (perceived ease of use of the system), social influence (perceived degree to which others believe one should use the system), and facilitating conditions (perceived degree of existing infrastructure to support use of the system).[8] [9] In telemedicine, the main predictors of acceptance are perceived usefulness, social influence, and overall attitude toward use.[7] As stated, the special circumstance of COVID-19 and associated shelter-in-place ordinances have already led to a significant decrease in in-person health care utilization while significantly increasing telemedicine encounters that is likely to persist.[10] [11] [12] In any given clinical scenario, characteristics of the medical specialty and diagnosis in question impact both the perceived usefulness and perceived ease of use of telemedicine, while patient characteristics may determine the social influence and attitudes toward different modalities of telemedicine. Whereas with some technologies there is a single end-user to consider, in the use of telemedicine, the perceptions of both patient and provider must be taken into account.

We aim to analyze the relationships between medical specialties, diagnoses, and telemedicine usage patterns, and their interplay among various demographic groups, within the framework of the UTAUT. As telemedicine becomes routine, understanding how specialty and diagnosis combine with demographics to impact telemedicine use may aid in addressing its current limitations and narrowing barriers to adoption in situations that benefit both clinicians and patients.

Methods

Data Source and Variables

We extracted encounter and patient data of all adults who scheduled outpatient visits from June 1, 2020, to June 30, 2021, from the electronic health record (EHR) of an integrated academic health system encompassing a broad range of subspecialties and a large geographic region in the Upper Midwest. For patients meeting criteria, data were extracted from the University of Wisconsin Health instance of Epic's Clarity and Caboodle databases. Extracted variables included medical specialty, encounter-associated ICD-10 (International Classification of Disease, 10th Edition) code, modality (video telemedicine, audio telemedicine, or in-person), and patient demographic information including patient age, sex, self-reported race and ethnicity, 2013 rural–urban continuum code (RUCC), and Charlson comorbidity index (CCI). As in our previous paper, rurality was determined using each patient's documented county and state to derive a 2013 RUCC designation, with RUCC codes of 4 or greater designating nonmetropolitan, more rural counties.[13] “Race” and “ethnicity” henceforth refer to patient self-reported race and ethnicity. Patient portal activation status was also obtained.

Specialties were ranked by relative telemedicine (audio and video) visit volume. Six specialties (General Surgery, Family Medicine, Gastroenterology, Oncology, General Internal Medicine, and Psychiatry), evenly distributed throughout this ranking, were chosen for analysis by a team of physician informaticists of varied clinical backgrounds, based on representativeness and diversity of clinical conditions, procedural emphasis, and patient population ([Fig. 2]). The selected specialties thus include a surgical specialty, primary care specialty, Medicine subspecialty with an immunocompromised patient population, general medical specialty, and a largely discussion-based behavioral health specialty.

Statistical Analysis

Characteristics of patients who had at least one encounter via video, audio, and in-person modalities were compared using one-way ANOVA (analysis of variance) and Chi-squared (corrected p-value <0.05 considered statistically significant). p-Values were corrected for false discovery rate due to multiple hypothesis testing.

We next examined the primary visit diagnoses associated with the highest and lowest frequencies of telemedicine use (henceforth, “telemedicine” refers to a combination of video and audio telemedicine modalities). At our institution, the provider or practice decides which visit types are appropriate for video telemedicine, and which should be conducted in-person or by phone; patients always have the right to request an in-person visit. ICD-10 category consists of the first three characters of each ICD-10 code. From the 50 most common ICD-10 categories in each specialty, those associated with the most and least frequent use of in-person visits were identified and reported.

Finally, we analyzed the association between demographics including age, sex, rural versus urban, self-reported race and ethnicity, and patient portal activation status and the telemedicine modality of an encounter (video vs. audio/in-person and video/audio vs. in-person) using mixed-effects logistic regression models for each specialty. Each patient characteristic was added as a fixed effect, while the identity of the patient (represented as an anonymous, random study ID) was added as a random intercept. The patient characteristic variables were evaluated individually in a regression which included the random-effect patient identity variable, and those with a statistically significant association with the outcome were included in the multivariable regression.

Previously, we had reported aggregating encounter data at the patient level to perform a similar regression analysis. This approach requires summarizing the modality of the encounters in some way—for example, by selecting the highest level of technology used for any encounter (where video is higher than audio, which is higher than in-person), which was the method employed used in our previous work, as well as the aggregation method used in presenting the comparison in [Table 1]. However, all of these methods share the common pitfall of discarding potentially useful information about the range of modalities used across all encounters of a single patient. By employing a mixed-effects regression model on encounter-level data with patient identity modeled as a random effect, we were able to represent each unique encounter while accounting for the patient identity variable. Thus, the mixed-effects model was able to integrate the phenomenon of multiple encounter types for a single patient in a way that was more flexible than the fixed-effects model, and allowed us to use all encounter data rather than aggregating at the level of the patient.

Table 1 Patient characteristics by medical specialty

Specialty	General Surgery				Family Medicine				Gastroenterology				Oncology				Internal Medicine				Psychiatry
	Video (N = 6,649)[a]	Audio (N = 2,787)[a]	In-person (N = 6,863)[a]	p [b]	Video (N = 57,183)[a]	Audio (N = 28,553)[a]	In-person (N = 86,123)[a]	p [b]	Video (N = 13,138)[a]	Audio (N = 5,351)[a]	In-person (N = 7,554)[ a ]	p [b]	Video (N = 5,608)[a]	Audio (N = 3,938)[a]	In-person (N = 3,131)[a]	p [b]	Video (N = 30,632)[a]	Audio (N = 12,983)[a]	In-person (N = 24,119)[a]	p [b]	Video (N = 13,309)[a]	Audio (N = 1,067)[a]	In-person (N = 86)[a]	p [b]
Age	53 (16)	62 (15)	53 (16)	<0.001	45 (21)	58 (20)	41 (24)	<0.001	51 (17)	62 (15)	58 (13)	<0.001	60 (15)	68 (13)	64 (13)	<0.001	52 (18)	64 (17)	52 (18)	<0.001	35 (20)	57 (20)	33 (28)	<0.001
Female sex	4,701 (71%)	1,723 (62%)	3,963 (58%)	<0.001	34,204 (60%)	15,939 (56%)	43,374 (50%)	<0.001	8,119 (62%)	2,984 (56%)	3,779 (50%)	<0.001	3,615 (64%)	2,287 (58%)	1,938 (62%)	<0.001	18,615 (61%)	7,559 (58%)	12,423 (52%)	<0.001	8,257 (62%)	661 (62%)	43 (50%)	0.084
Unavailable					0	0	7
Race				<0.001				<0.001				<0.001				0.2				<0.001				0.14
White	6,113 (93%)	2,549 (93%)	6,131 (92%)		51,941 (92%)	26,151 (93%)	74,848 (90%)		12,010 (93%)	4,886 (93%)	6,832 (93%)		5,257 (95%)	3,687 (95%)	2,927 (95%)		27,501 (91%)	11,812 (92%)	21,222 (89%)		12,093 (92%)	985 (93%)	82 (95%)
Black or African American	259 (3.9%)	140 (5.1%)	322 (4.8%)		2,366 (4.2%)	1,252 (4.4%)	3,642 (4.4%)		473 (3.7%)	256 (4.9%)	269 (3.7%)		161 (2.9%)	128 (3.3%)	84 (2.7%)		1,242 (4.1%)	608 (4.7%)	964 (4.1%)		572 (4.3%)	49 (4.6%)	2 (2.3%)
Asian	142 (2.2%)	37 (1.3%)	173 (2.6%)		1,635 (2.9%)	575 (2.0%)	3,558 (4.3%)		293 (2.3%)	83 (1.6%)	185 (2.5%)		93 (1.7%)	48 (1.2%)	39 (1.3%)		1,274 (4.2%)	344 (2.7%)	1,360 (5.7%)		376 (2.9%)	15 (1.4%)	1 (1.2%)
American Indian or Alaska Native	36 (0.5%)	23 (0.8%)	57 (0.9%)		371 (0.7%)	187 (0.7%)	705 (0.9%)		99 (0.8%)	34 (0.6%)	53 (0.7%)		28 (0.5%)	25 (0.6%)	16 (0.5%)		198 (0.7%)	80 (0.6%)	155 (0.7%)		92 (0.7%)	6 (0.6%)	1 (1.2%)
Native Hawaiian or Other Pacific Islander	10 (0.2%)	6 (0.2%)	13 (0.2%)		80 (0.1%)	38 (0.1%)	142 (0.2%)		15 (0.1%)	8 (0.2%)	12 (0.2%)		3 (<0.1%)	3 (<0.1%)	6 (0.2%)		53 (0.2%)	14 (0.1%)	53 (0.2%)		18 (0.1%)	0 (0%)	0 (0%)
Unavailable	89	32	167		790	350	3,228		248	84	203		66	47	59		364	125	365		158	12	0
Hispanic/Latino	198 (3.0%)	109 (3.9%)	351 (5.2%)	<0.001	2,103 (3.7%)	951 (3.4%)	5,575 (6.6%)	<0.001	382 (2.9%)	143 (2.7%)	289 (3.9%)	<0.001	114 (2.0%)	68 (1.7%)	83 (2.7%)	0.026	888 (2.9%)	306 (2.4%)	870 (3.6%)	<0.001	535 (4.0%)	29 (2.7%)	8 (9.3%)	0.008
Unavailable	53	15	95		435	188	2,129		178	58	140		47	38	42		206	64	222		73	5	0
Rural	1,029 (15%)	537 (19%)	1,291 (19%)	<0.001	5,645 (9.9%)	4,745 (17%)	11,545 (13%)	<0.001	1,596 (12%)	836 (16%)	620 (8.2%)	<0.001	1,229 (22%)	1,542 (39%)	1,153 (37%)	<0.001	1,779 (5.8%)	1,641 (13%)	2,344 (9.7%)	<0.001	823 (6.2%)	83 (7.8%)	11 (13%)	0.008
Patient portal activation	6,203 (93%)	1,961 (70%)	4,996 (73%)	<0.001	52,744 (92%)	20,808 (73%)	64,735 (75%)	<0.001	12,434 (95%)	3,735 (70%)	6,109 (81%)	<0.001	4,855 (87%)	2,263 (57%)	1,965 (63%)	<0.001	29,168 (95%)	9,751 (75%)	20,434 (85%)	<0.001	12,389 (93%)	684 (64%)	61 (71%)	<0.001
Unavailable	0	1	3		6	5	302		1	4	2		1	1	2		0	0	1		1	0	0
CCI	62 (40)	47 (41)	74 (34)	<0.001	78 (33)	61 (40)	86 (25)	<0.001	74 (34)	59 (38)	80 (26)	<0.001	29 (37)	21 (32)	33 (36)	<0.001	75 (34)	54 (39)	81 (28)	<0.001	87 (25)	61 (39)	83 (30)	<0.001
Unavailable	1	0	0		0	0	3						0	0	1

Abbreviation: CCI, Charlson comorbidity index.

^a Mean (SD) or n (%).

^b One-way analysis of variance; Pearson's chi-squared test; Fisher's exact test for count data with simulated p-value (based on 2,000 replicates) with false discovery rate correction for multiple testing.

This study was exempted from the University of Wisconsin Institutional Review Board review.

Results

Baseline Characteristics

A total of 2,494,296 encounters occurred during the study period, representing 420,876 unique patients (mean age = 44 years, SD = 24 years, 54% female). At the patient level, the occurrence of at least one video/audio telemedicine visit ranged from 49.9% of all visits for General Surgery to 99.4% for Psychiatry ([Fig. 1]). In all specialties, age was significantly associated with visit modality, with the audio modality generally older than in-person or video. Sex was significantly associated with visit modality in all specialties except Psychiatry; in all specialties, the video modality had the highest proportion of female patients; in five specialties (excepting oncology), the in-person modality had the highest proportion of male patients. Self-reported race differed among modalities within the specialties of general surgery, family medicine, gastroenterology and internal medicine; a slightly higher rate of white race seen was in the video and audio modalities. In every specialty, the in-person group had the highest average CCI (indicating the highest levels of morbidity), followed by the video group ([Table 1]).

Fig. 1 Percentage of patients using visit modality, by specialty. Stacked bar chart shows relative percentages of patients who had visits of each modality. Patients who had at least one visit were counted in the “video” bar; at least one audio visit (and no video visits) in the “audio” bar, and the remainder of patients in the “in-person” bar.

Modality by Diagnosis Category

ICD-10 categories associated with most and least frequent telemedicine use (video or audio) are shown in [Table 2] for General Surgery and Internal Medicine. In General Surgery, the diagnoses least frequently evaluated via telemedicine often necessitated a physical examination or minor procedure (hemorrhoids, pilonidal cysts, benign lipomas, abnormal breast imaging); diagnoses most frequently evaluated via telemedicine included those for which treatment is nonoperative (e.g., rib fractures) or workup does not rely heavily on physical examination (diaphragmatic hernia and foregut pathology). In Internal Medicine, diagnoses least frequently evaluated via telemedicine similarly required physical contact with the patient (e.g., external ear disorders) while diagnoses most frequently evaluated via telemedicine were all behavioral health-related. Of note, some artifact occurred in the ranking of Internal Medicine diagnoses least frequently evaluated by telemedicine, as many primary encounter diagnoses were populated with “encounter for immunization” or “deficiency of other B group vitamins” because routine immunizations and vitamin supplementation often took place before patients saw their providers ([Table 2]).

Table 2 Most versus least common ICD-10 categories of primary diagnoses associated with in-person visit, out of all encounters for the category: (A) General Surgery; (B) Internal Medicine

Least frequent telemedicine use			Most frequent telemedicine use
ICD-10 category	n (%)	Total	ICD-10 category	n (%)	Total
A. General surgery
L98: Other disorders of skin and subcutaneous tissue, not elsewhere classified	224 (100.0)	224	K44: Diaphragmatic hernia	142 (41.3)	344
K64: Hemorrhoids and perianal venous thrombosis	1,598 (99.8)	1,601	S22: Fracture of rib(s), sternum, and thoracic spine	171 (43.1)	397
L05: Pilonidal cyst and sinus	532 (99.8)	533	K22: Other diseases of esophagus	109 (53.4)	204
D17: Benign lipomatous neoplasm	800 (99.8)	802	K21: Gastro-esophageal reflux disease	260 (59.1)	440
R92: Abnormal and inconclusive findings on diagnostic imaging of breast	1,753 (99.7)	1,758	K31: Other diseases of stomach and duodenum	240 (65.0)	369
B. Internal medicine
Z23: Encounter for immunization	15,844 (99.9)	15,853	F43: Reaction to severe stress, and adjustment disorders	352 (9.8)	3,602
H61: Other disorders of external ear	855 (99.3)	861	F33: Major depressive disorder, recurrent	335 (11.5)	2,913
Z01: Encounter for other special examination without complaint, suspected or reported diagnosis	2,953 (98.3)	3,003	F41: Other anxiety disorders	2,096 (15.9)	13,193
E53: Deficiency of other B group vitamins	1,506 (96.9)	1,554	F32: Depressive episode	841 (18.6)	4,524
Z11: Encounter for screening for infectious and parasitic diseases	1,813 (96.7)	1,874	F90: Attention-deficit hyperactivity disorders	206 (26.5)	776

Abbreviation: ICD-10, International Classification of Disease, 10th Edition.

Note: ICD-10 categories associated with most and least frequent telemedicine use (video or audio) are shown in [Table 2] for General Surgery and Internal Medicine. ICD-10 category consists of the first three characters of each ICD-10 code.

In gastroenterology, screening for malignant neoplasms and liver fibrosis/cirrhosis were among the most frequently done in-person; acute pancreatitis and irritable bowel syndrome were among the least frequent. In oncology, malignant ovarian and uterine neoplasms were among the most frequent, while malignant neoplasms of brain, connective and soft tissue, and melanoma were among the least frequent. In psychiatry, visits for psychotic disorders and schizophrenia most frequently occurred in-person, while most other mental health diagnoses had virtually no in-person visits.

Association between Demographics and Visit Modality—Mixed-Effects Regression Model

Video versus Nonvideo (Audio/In-Person)

[Table 3] displays odds ratios (ORs), confidence intervals (CIs), and p-values from the mixed-effects regression model for the video versus nonvideo (audio/in-person) outcome. In General Surgery and Gastroenterology, rural RUCC independently increased the odds of an encounter taking place via video modality. In all other specialties, rural RUCC decreased the odds of a video encounter.

Table 3 Video versus audio/in-person, mixed-effects regression analysis

Characteristic	General Surgery			Gastroenterology			Oncology			Internal Medicine			Psychiatry			Family Medicine
	OR	(95% CI)	p-Value	OR	(95% CI)	p-Value	OR	(95% CI)	p-Value	OR	(95% CI)	p-Value	OR	(95% CI)	p-Value	OR	(95% CI)	p-Value
Rural	1.58	(1.03–2.42)	0.04	3.72	(2.86–4.83)	<0.001	0.55	(0.42–0.72)	<0.001	0.48	(0.43–0.53)	<0.001	0.82	(0.82–0.82)	<0.001
Age				0.90	(0.89–0.90)	<0.001	0.96	(0.95–0.97)	<0.001	0.97	(0.97–0.97)	<0.001	0.95	(0.94–0.95)	<0.001	0.99	(0.98–0.99)	<0.001
Male sex				0.54	(0.46–0.64)	<0.001	0.62	(0.49–0.79)	<0.001	0.82	(0.78–0.86)	<0.001	1.40	(1.40–1.40)	<0.001	0.80	(0.75–0.86)	<0.001
Hispanic/Latino				0.59	(0.35–0.99)	0.05	0.46	(0.18–1.14)	0.09	0.80	(0.68–0.95)	0.01	1.43	(1.43–1.43)	<0.001	0.71	(0.58–0.86)	<0.001
Patient portal activated				4.56	(3.43–6.07)	<0.001	2.37	(1.78–3.17)	<0.001	5.30	(4.82–5.82)	<0.001	82.70	(82.7–82.7)	<0.001	1.83	(1.64–2.04)	<0.001
Race
White (ref.)				–	–		–	–		–	–		–	–		–	–
American Indian Alaska Native				0.87	(0.32–2.35)	0.80	0.76	(0.15–3.94)	0.70	0.87	(0.63–1.21)	0.40	2.68	(2.67–2.68)	<0.001	1.01	(0.65–1.58)	>0.9
Asian				0.43	(0.25–0.75)	0.00	1.44	(0.57–3.64)	0.40	0.86	(0.76–0.98)	0.02	2.99	(2.99–2.99)	<0.001	0.77	(0.62–0.94)	0.01
Black/African American				0.63	(0.41–0.97)	0.04	0.31	(0.15–0.65)	0.00	0.81	(0.71–0.92)	0.00	0.59	(0.59–0.59)	<0.001	0.90	(0.75–1.07)	0.20
Native Hawaiian/Other Pacific Islander				0.33	(0.04–2.99)	0.30	0.33	(0.00–28.1)	0.60	1.02	(0.55–1.90)	>0.9	3.72	(0.13–104)	0.40	0.74	(0.28–1.99)	0.60

Abbreviations: CI, confidence interval; OR, odds ratio.

Note: Mixed-effects regression models by specialty; outcome was video modality versus audio/in-person; patient identity was modeled as a random effect. The patient characteristic variables were evaluated individually in a regression which included the fixed-effects patient identity variable, and those with a statistically significant association with the outcome were included in the multivariable regression. p-Values are corrected for false discovery rate.

Within General Surgery, only rural RUCC significantly affected the odds of video modality use; the remaining patient characteristics did not significantly impact the odds. Among the remaining specialties, each 1-year increase in patient age decreased the odds of video modality in all specialties. Male sex independently decreased the odds of video modality in all specialties except for Psychiatry, where the odds of video modality increased. Hispanic/Latino ethnicity independently decreased the odds of video in all specialties except for Psychiatry, where the odds of video modality increased, and Oncology, where the OR was not statistically significant. Patient portal activation independently increased the odds of video modality in all specialties.

Race significantly and independently affected odds of video modality use in every specialty except General Surgery. Among the remaining specialties, American Indian/Alaska Native race increased the odds of a video visit in Psychiatry only. Asian race significantly decreased the odds of video modality use in Gastroenterology, Internal Medicine, and Family Medicine, but increased the odds of video use in Psychiatry. Black/African American self-reported patient race significantly decreased the odds of video modality use in every specialty except General Surgery and Family Medicine.

Telemedicine (Video/Audio) versus In-Person

[Table 4] displays ORs, CIs, and p-values of the mixed-effects regression model for the telemedicine (video/audio) versus in-person outcome. In this model, a rural RUCC increased the odds of telemedicine use in comparison to urban for a given encounter, except in Psychiatry (where it decreased the odds) and Family Medicine (where the result was not statistically significant). Similar to the video versus nonvideo model, in General Surgery, RUCC was the only patient characteristic that impacted the odds of telemedicine use.

Table 4 Mixed-effects regression—video/audio versus in-person analysis

Characteristic	General Surgery			Gastroenterology			Oncology			Internal Medicine			Psychiatry			Family Medicine
	OR	(95% CI)	p-Value	OR	(95% CI)	p-Value	OR	(95% CI)	p-Value	OR	(95% CI)	p-Value	OR	(95% CI)	p-Value	OR	(95% CI)	p-Value
Rural	1.88	(1.33–2.64)	<0.001	1.41	(1.31–1.52)	<0.001	12.60	(10.0–15.9)	<0.001	1.29	(1.14–1.47)	<0.001	0.73	(0.68–0.78)	<0.001
Age				1.00	(1.00–1.00)	<0.001	0.92	(0.91–0.92)	<0.001				0.99	(0.99–1.00)	<0.001	0.99	(0.98–0.99)	<0.001
Male sex				0.71	(0.68–0.75)	<0.001	0.40	(0.36–0.46)	<0.001	0.65	(0.58–0.74)	<0.001	0.76	(0.73–0.79)	<0.001	0.71	(0.53–0.97)	0.03
Hispanic/Latino				0.60	(0.53–0.69)	<0.001	0.54	(0.37–0.80)	0.00	0.43	(0.28–0.68)	<0.001	0.82	(0.73–0.92)	<0.001
Race
White (ref.)				–	–		–	–		–	–		–	–
American Indian Alaska Native				0.88	(0.65–1.18)	0.40	0.85	(0.42–1.76)	0.70	0.87	(0.40–1.87)	0.70	0.92	(0.73–1.15)	0.50
Asian				0.54	(0.47–0.63)	<0.001	0.41	(0.27–0.61)	<0.001	1.59	(0.98–2.57)	0.06	0.74	(0.68–0.82)	<0.001
Black African American				0.97	(0.86–1.10)	0.70	1.00	(0.73–1.38)	>0.9	0.50	(0.35–0.70)	<0.001	0.93	(0.85–1.02)	0.13
Native Hawaiian/Other Pacific Islander				0.76	(0.40–1.46)	0.40	0.39	(0.08–1.97)	0.30	0.15	(0.02–1.36)	0.09	0.76	(0.48–1.20)	0.20
Patient portal activated							0.40	(0.34–0.49)	<0.001	0.86	(0.76–0.99)	0.03	1.32	(1.25–1.40)	<0.001	3.22	(2.08–4.98)	<0.001

Abbreviations: CI, confidence interval; OR, odds ratio.

Note: Mixed-effects regression models by specialty; outcome was video/audio modality versus in-person; patient identity was modeled as a random effect. The patient characteristic variables were evaluated individually in a regression which included the fixed-effects patient identity variable, and those with a statistically significant association with the outcome were included in the multivariable regression. p-Values are corrected for false discovery rate.

Among the remaining specialties, each 1-year increase in patient age independently decreased the odds of telemedicine use, though the effect was smaller than observed in the video versus nonvideo model. Male sex independently decreased the odds of telemedicine use in every specialty, and Hispanic/Latino ethnicity decreased the odds of telemedicine use in every specialty except Family Medicine. Asian race here significantly decreased the odds of telemedicine use in Gastroenterology, Oncology, and Psychiatry, but increased the odds of telemedicine use in Internal Medicine. Black/African American race significantly decreased the odds of telemedicine use in Internal Medicine; it did not significantly affect telemedicine use in other specialties.

Discussion

Within the UTAUT, adoption and use of telemedicine depend on the behavioral intention and perception of the technology on the part of either the patient, provider, or both. Medical specialty, diagnosis, and patient characteristics likely affect the four key constructs predicting overall attitudes toward telemedicine: performance expectancy, effort expectancy, social influence, and facilitating conditions.[9] We now describe how our findings relate to each of these constructs.

Performance Expectancy

Performance expectancy is defined as “the degree to which an individual believes that using the system will help him or her to attain gains in job performance.”[9] On the part of the provider, medical specialty and diagnosis appear to influence performance expectancy, as telemedicine facilitates the management of some clinical scenarios more than others. Medical diagnoses requiring physical examination or minor procedures are more likely to be conducted in-person. In General Surgery, initial workup of hemorrhoids and pilonidal cysts relies on the physical exam, while that of many foregut disorders (e.g., gastroesophageal reflux disease) largely consists of specialized tests completed elsewhere prior to the initial consultation. Similarly, otoscopic examination is required for diagnosis of ear disorders in Family Medicine and Internal Medicine, while acute sinusitis, a diagnosis based mostly on patient-reported symptoms, rarely requires an in-person visit. Even simple injections such as B vitamin supplements and immunizations cannot currently take place remotely. Unsurprisingly, in Internal Medicine, Family Medicine, and Behavioral Health, nonacute behavioral health diagnoses can often be managed remotely. Within Oncology, provider practices may vary: Gynecologic Oncology functions as a surgical subspecialty and may be less likely to use telemedicine, while Medical Oncology patients undergoing chemotherapy are at high infection risk due to immunocompromise, and telemedicine may be preferred to avoid unnecessary exposure but sometimes a physical examination is required to assess performance status or palpate a mass. In Psychiatry, where telemedicine dominates the encounter landscape, evaluation of psychotic disorders more frequently occurs in-person.

We currently reside in a relatively untested period of telemedicine—little guidance surrounds which clinical scenarios or visit types are appropriate for telemedicine. Describing current practice patterns may enable future research on appropriateness, which will allow a more focused examination of performance expectancy in any given clinical scenario.

Effort Expectancy

Effort expectancy is “the degree of ease associated with consumers' use of technology.”[9] Our data support higher effort expectancy for video telemedicine use (and decreased ease of use) among rural patients. While rural RUCC increases the odds for video modality use in General Surgery and Gastroenterology, it decreases the odds in all other specialties. However, rural RUCC increases the odds of telemedicine use (audio and video) in every specialty, which suggests heavier use of audio-based telemedicine is responsible. A possible explanation is that for rural patients, video visits may take the place of a visit that might be conducted in-person for an urban patient, particularly in General Surgery and Gastroenterology, which are uniquely procedural specialties.

Socioeconomic inequality is a key challenge for telemedicine adoption, whereas clinical opportunity is a key opportunity.[12] A purported benefit of telemedicine is an expansion of options available to rural patients. Thus, the paradoxical trend of lower rates of video modality use in most specialties may point to a gap in infrastructure (including broadband Internet access) and relative digital literacy and comfort with video-based telemedicine in the rural population. In a rural nursing home setting, a lack of technology training, poor video and sound quality, and connectivity issues were perceived as pitfalls to telemedicine.[14] Others and we have shown that patients who use video telemedicine are younger, and more likely to be white and have private insurance.[15] Improving facilitating conditions for other demographic groups will thus likely improve telemedicine adoption in the rural population.

Quality improvement initiatives to identify gaps across domains that inhibit telemedicine use can increase preferences for video telemedicine use in these groups over time.[16] Our institution has actively implemented these initiatives, collecting patient satisfaction scores and correlating these data with qualitative studies into patients' perceived barriers to telehealth usage, which we hope will inform further outreach efforts within our center and beyond.

Social Influence

Social influence is “the extent to which consumers perceive that important others (e.g., family and friends) believe they should use a particular technology.”[9] At our institution, the decision regarding visit modality is dependent on both provider practice and patient. In some cases use of in-person versus telemedicine modality is protocolized, but patient preferences are taken into account, and patients are always given the right to request an in-person visit. Thus, the provider or practice offers the modality or modalities appropriate and available for the diagnosis and visit type (new, follow-up, postoperative, etc.), and the patient ultimately chooses between them. The scenarios for which patients are offered telemedicine are mainly driven by each specialty group. Thus, visit modality is an interaction between offered modality and patient preference.

Patient characteristics such as age, sex, and race consistently influence telemedicine use across most medical specialties ([Tables 2] and [3]). Older age, male sex, and nonwhite race were generally associated with lower rates of telemedicine use across most specialties. Social attitudes toward telemedicine, and technology in general, in certain communities, coupled with effort expectancy, may well explain the associations we observe between characteristics above and acceptance of telemedicine. For providers, agreements and guidelines within practices and groups, as well as those of other practices in the system, will largely comprise the social influences on each individual provider for offering and conducting visits by telemedicine.

Fig. 2 Ranking of medical specialties by proportion of visits conducted via telemedicine (video/audio). Medical specialties ranked by the proportion of visits within the study period that took place via audio or video telemedicine. The horizontal black lines divide the quartiles. Medical specialties roughly evenly distributed among the quartiles were chosen for further analysis.

Facilitating Conditions

Facilitating conditions refers to “consumers' perceptions of the resources and support available to performing a behavior.”[9] As noted above, Psychiatry, among the specialties examined, by far makes the most use of telemedicine, as was the case even prior to the start of the COVID-19 pandemic. Favorable facilitating conditions, namely, a more mature and accessible infrastructure for telemedicine (video and audio), may exist within Psychiatry.

A seemingly paradoxical pattern is the reversal of multiple trends within the Psychiatry specialty. Male sex decreased the odds of both video modality use and telemedicine use in most specialties, with the exception of Psychiatry, in which male patient sex increased the odds of video modality. Hispanic/Latino ethnicity followed a similar trend. Self-reported American Indian/Alaska Native race also increased the odds of video modality use compared with White race. Self-reported Asian race decreased the odds of telemedicine use in most specialties compared with White race in Psychiatry only; it also decreased the odds of video modality use in Internal Medicine, but increased the odds of video use in Psychiatry compared with the reference White patient group.

We are unable to fully explain why numerous demographic groups use the video modality in Psychiatry more commonly than expected, while using less telemedicine in general compared with other groups. The answer may lie in the heterogeneity of the patient population seeking psychiatric care. In 2019, 19.2% of adults in the United States received any mental health treatment, including 15.8% who took medication for mental health.[17] The rate of mental health-related physician office visits to psychiatrists was higher than the rate of visits to primary care physicians in all adult age groups except 65 years and over.[18] The wide range of mental illness nature and severity implies significant heterogeneity in the patient population. Thus, it may well be the case that in multiple demographic groups, the subset of patients who seek and are offered mental health care through telemedicine may represent a smaller segment of that population in general, but may disproportionately possess the facilitating conditions, including means, attitudes, and education, associated with preferential use of the video modality.[19]

Finally, within Psychiatry, where telemedicine dominates the encounter landscape, evaluation of psychotic disorders and schizophrenia more frequently occur in-person. This is concordant with previous work evaluating telemedicine and mental health, possibly because people with these mental illnesses experience discomfort or challenges related to telemedicine use, or in some situations require in-person care.[19] [20] Thus, even within mental health, it is important to promote telemedicine access with a focus on improving the telemedicine experience for patients with psychotic disorders and schizophrenia, while maintaining in-person care to those who need it.

Alignment of Patient and Provider Perceptions

Providers choose which conditions are suitable or unsuitable for telemedicine use and visit modality. Patients always have the right to request an in-person visit. The alignment of patient and provider perceptions leads to telemedicine use. Our findings inform how patient demographics, medical specialty, and diagnosis may affect each of these perceptions, which are in turn mutually dependent. [Fig. 3B] shows a potential schematic explaining our findings in the context of the UTAUT, with the original figure from Venkatesh et al given in [Fig. 3A] for reference. Further work will be required to elucidate the strength and significance of each of these potential associations.

Fig. 3 (A) Original UTAUT model. (B) Possible interactions between patient demographics, diagnosis, and medical specialty on patient and provider use behavior within UTAUT. UTAUT, Unified Theory of Acceptance and Use of Technology.

Limitations

This study was mainly limited by its retrospective nature and the data endpoints available in the EHR. For example, the ICD-10 category was used as a proxy for diagnosis, and allowed for broader generalizability and simplification of the analysis, but on occasion, the categories themselves are not intended to be informative. The six specialties examined were chosen through consensus, introducing an unavoidable level of bias in these selections; we aimed for an illustrative, rather than a comprehensive, analysis of medical specialty and telemedicine use. We plan to expand on these areas in future studies and perform a more detailed analysis of patient barriers to usage, influence of the visit type (new, follow-up, etc.) as well as the various scheduling practices. We believe that our work will aid in determining best practices, appropriateness of care, and guidelines for telemedicine.

Conclusion

In a large integrated academic health system, medical specialty and diagnosis impacted the frequency of telemedicine use versus in-person care in expected patterns, likely depending on the need for physical exams. These factors contribute to a unique clinical scenario which impacts both performance and effort expectancy. Our findings with regard to the interaction between specialty, diagnosis, and demographics combine to influence telemedicine use may aid in addressing current limitations and narrowing barriers to adoption. More work will be needed to determine how to address persistent disparities observed in sex and race across most specialties, including use of qualitative methods, which has begun at our institution to identify and address gaps in access.

Clinical Relevance Statement

Medical specialty and diagnosis likely interact with patient demographics to impact telemedicine adoption by both providers and patients, by modulating expectations surrounding the performance of telemedicine, and the effort required to utilize it in any given scenario. We believe that our work will aid in determining best practices, appropriateness of care, and guidelines for telemedicine.

Multiple-Choice Questions

1. What is true regarding primary diagnoses of visits more likely to be conducted in-person rather than by telemedicine?

   • Primary diagnosis for a visit does not affect the likelihood of a visit occurring through telemedicine versus in-person.

   • Nonacute behavioral health diagnoses are associated with the lowest rates of telemedicine use.

   • Diagnoses where parts of the evaluation rely on the physical exam are least likely associated with in-person visits.

   • Diagnoses where parts of the evaluation rely on the physical exam, or require minor procedures, are more likely to be conducted in-person.

   Correct Answer: The correct answer is option d. In all specialties, we found that diagnoses that required physical examination or minor procedures as part of workup and management were more likely to be conducted in-person rather than by video or audio telemedicine.

2. How was rural RUCC associated with odds of video telemedicine use, by specialty?

   • Rural RUCC increased the odds for video telemedicine use in General Surgery and Gastroenterology and decreased the odds in all other specialties.

   • Rural RUCC decreased the odds for video telemedicine use in General Surgery and Gastroenterology and increased the odds in all other specialties.

   • Rural RUCC decreased the odds of video telemedicine use in all specialties.

   • Rural RUCC increased the odds of video telemedicine use in all specialties.

   Correct Answer: The correct answer is option a. Rural RUCC increased the odds for video modality use in General Surgery and Gastroenterology, while decreasing the odds in all other specialties. However, rural RUCC increased the odds of telemedicine use (audio and video) in every specialty, which suggests heavier use of audio-based telemedicine is responsible. A possible explanation is that for rural patients, video visits may take the place of a visit that might be conducted in-person for an urban patient, particularly in General Surgery and Gastroenterology, which are uniquely procedural specialties.

Conflict of Interest

A.J.T. is a family member of Epic Systems.

Protection of Human and Animal Subjects

This study was exempted from the University of Wisconsin Institutional Review Board (IRB) review.

• References

• 1 Demeke HB, Merali S, Marks S. et al. Trends in use of telehealth among health centers during the COVID-19 pandemic - United States, June 26-November 6, 2020. MMWR Morb Mortal Wkly Rep 2021; 70 (07) 240-244
• 2 Hsiao V, Chandereng T, Lankton RL. et al. Disparities in telemedicine access: a cross-sectional study of a newly established infrastructure during the COVID-19 pandemic. Appl Clin Inform 2021; 12 (03) 445-458
• 3 Barnett ML, Huskamp HA. Telemedicine for mental health in the United States: making progress, still a long way to go. Psychiatr Serv 2020; 71 (02) 197-198
• 4 Asiri A, AlBishi S, AlMadani W, ElMetwally A, Househ M. The use of telemedicine in surgical care: a systematic review. Acta Inform Med 2018; 26 (03) 201-206
• 5 Makhni MC, Riew GJ, Sumathipala MG. Telemedicine in orthopaedic surgery: challenges and opportunities. J Bone Joint Surg Am 2020; 102 (13) 1109-1115
• 6 McCool RR, Davies L. Where does telemedicine fit into otolaryngology? An assessment of telemedicine eligibility among otolaryngology diagnoses. Otolaryngol Head Neck Surg 2018; 158 (04) 641-644
• 7 Harst L, Lantzsch H, Scheibe M. Theories predicting end-user acceptance of telemedicine use: systematic review. J Med Internet Res 2019; 21 (05) e13117
• 8 Davis FD. Perceived usefulness, perceived ease of use, and user acceptance of information technology. Manag Inf Syst Q 1989; 13 (03) 319-340
• 9 Venkatesh V, Morris MG, Davis GB, Davis FD. User Acceptance of Information Technology: Toward a Unified View. Manag Inf Syst Q 2003; 27 (03) 425-478
• 10 Cantor J, Sood N, Bravata DM, Pera M, Whaley C. The impact of the COVID-19 pandemic and policy response on health care utilization: evidence from county-level medical claims and cellphone data. J Health Econ 2022; 82: 102581
• 11 deMayo R, Huang Y, Lin ED. et al. Associations of telehealth care delivery with pediatric health care provider well-being. Appl Clin Inform 2022; 13 (01) 230-241
• 12 Alipour J, Hayavi-Haghighi MH. Opportunities and challenges of telehealth in disease management during COVID-19 pandemic: a scoping review. Appl Clin Inform 2021; 12 (04) 864-876
• 13 United States Department of Agriculture. Rural-urban continuum codes. 2020. Accessed January 4, 2023 at: https://www.ers.usda.gov/data-products/rural-urban-continuum-codes/
• 14 Powell KR, Alexander GL. Consequences of rapid telehealth expansion in nursing homes: promise and pitfalls. Appl Clin Inform 2021; 12 (04) 933-943
• 15 Luo J, Tong L, Crotty BH. et al. Telemedicine adoption during the COVID-19 pandemic: gaps and inequalities. Appl Clin Inform 2021; 12 (04) 836-844
• 16 Jelinek R, Pandita D, Linzer M, Engoang JBBN, Rodin H. An evidence-based roadmap for the provision of more equitable telemedicine. Appl Clin Inform 2022; 13 (03) 612-620
• 17 Terlizzi EP, Zablotsky B. Mental health treatment among adults: United States, 2019. NCHS Data Brief 2020; (380) 1-8
• 18 Cherry D, Albert M, McCaig L. Mental health-related physician office visits by adults aged 18 and over: United States, 2012–2014. 2018; NCHS Data Brief (311) 1-8
• 19 Sizer MA, Bhatta D, Acharya B, Paudel KP. Determinants of telehealth service use among mental health patients: a case of rural Louisiana. Int J Environ Res Public Health 2022; 19 (11) 6930
• 20 Zhu JM, Myers R, McConnell KJ, Levander X, Lin SC. Trends in outpatient mental health services use before and during the COVID-19 pandemic. Health Aff (Millwood) 2022; 41 (04) 573-580

Publication History

Received: 09 July 2022

Accepted: 15 December 2022

Article published online:
01 March 2023