Outcomes of In-Person and Telehealth Ambulatory Encounters During COVID-19 Within a Large Commercially Insured Cohort

Key Points

Question  What is the association of telehealth vs in-person encounters with outcomes of care during the COVID-19 pandemic in the US?

Findings  In this cohort study of 40.7 million US commercially insured adults with acute clinical conditions, those with an initial telehealth encounter, compared with an in-person encounter, had higher odds for any follow-up encounter, an emergency department encounter, and in-patient admissions. For people with chronic conditions, the odds were lower for those with an initial telehealth encounter.

Meaning  The contrasting patterns of follow-up care among members receiving telehealth for acute and chronic conditions have implications for health services during and after the COVID-19 pandemic.

Importance  Since the start of the COVID-19 pandemic, few studies have assessed the association of telehealth with outcomes of care, including patterns of health care use after the initial encounter.

Objective  To assess the association of telehealth and in-person visits with outcomes of care during the COVID-19 pandemic.

Design, Setting, and Participants  This cohort study assessed continuously enrolled members in private health plans of the Blue Cross and Blue Shield Association from July 1, 2019, to December 31, 2020.

Main Outcomes and Measures  Main outcomes were ambulatory encounters per enrollee stratified by characteristics derived from enrollment files, practitioner claims, and community characteristics linked to the enrollee’s zip code. Outcomes of care were assessed 14 days after the initial encounters and included follow-up encounters of any kind, emergency department encounters, and hospitalizations after initial telehealth or in-person encounters.

Results  In this cohort study of 40 739 915 individuals (mean [SD] age, 35.37 [18.77] years; 20 480 768 [50.3%] female), ambulatory encounters decreased by 1.0% and the number of in-person encounters per enrollee decreased by 17.0% from 2019 to 2020; however, as a proportion of all ambulatory encounters, telehealth encounters increased substantially from 0.6% (n = 236 220) to 14.1% (n = 5 743 718). For members with an initial telehealth encounter for a new acute condition, the adjusted odds ratio was 1.44 (95% CI, 1.42-1.46) for all follow-ups combined and 1.11 (95% CI, 1.06-1.16) for an emergency department encounter. For members with an initial telehealth encounter for a new chronic condition, the adjusted odds ratios were 0.94 (95% CI, 0.92-0.95) for all follow-ups combined and 0.94 (95% CI, 0.90-0.99) for in-patient admissions.

Conclusions and Relevance  In this cohort study of 40.7 million commercially insured adults, telehealth accounted for a large share of ambulatory encounters at the peak of the pandemic and remained prevalent after infection rates subsided. Telehealth encounters for chronic conditions had similar rates of follow-up to in-person encounters for these conditions, whereas telehealth encounters for acute conditions seemed to be more likely than in-person encounters to require follow-up. These findings suggest a direction for future work and are relevant to policy makers, payers, and practitioners as they manage the use of telehealth during the COVID-19 pandemic and afterward.

Starting in 2020, the COVID-19 pandemic has resulted in notable changes in US health care. Much routine care was put on hold during the early months of the pandemic, whereas many hospitals were overwhelmed by patients seriously ill with COVID-19. At the same time, health insurance coverage of telehealth services was greatly expanded on an emergency basis to provide an alternative to in-person care without the risk of COVID-19 exposure. Thus, COVID-19 resulted in an unprecedented increase in the use of telehealth services during the early months of the pandemic.^1,2

Since the surge of COVID-19, several studies^2-14 have documented the expansion of telehealth services in the US, and few^15,16 have assessed the association of telehealth with outcomes of care, including patterns of health care use after the initial encounter. This study expands on previous work.² Here we assess the shifts in telehealth and ambulatory service use in the period after the initial implementation phase of telehealth. We document overall ambulatory encounters before and after the pandemic’s start within a large cohort of insured US patients. We also document key outcomes of care associated with an initial telehealth encounter compared with an in-person encounter. Finally, we assess the associations of the key patient, community, and health system characteristics with the likelihood of telehealth use.

The data set for this cohort study came from the repository of Blue Health Intelligence, a licensee of the Blue Cross and Blue Shield Association. The data set included claims files for commercial health plan members who used the commercial plan as their primary coverage. The study period was July 1, 2020, to December 31, 2020. eFigure 1 in the Supplement presents the sample selection and subgroup identification. We compared July to December 2020 with the same period in 2019 to account for seasonality. The study population was limited to members who were continuously enrolled from July 1, 2019, to December 31, 2020, and were covered through employer-based, Patient Protection and Affordable Care Act, and other private health insurance plans. The institutional review board of the Johns Hopkins Bloomberg School of Public Health approved this study as being exempt because it was a secondary analysis of de-identified data; therefore, patient informed consent was not required. This study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.¹⁷

Our unit of analysis was an ambulatory encounter with a telehealth-eligible service, excluding in-patient and emergency department (ED) services. We defined an encounter as a patient seeing a specific practitioner on a specific date and at a specific place of service. We identified telehealth-eligible services using the Current Procedural Terminology or Healthcare Common Procedure Coding System codes that, based on payer policy, were eligible for telehealth coverage,¹⁸ subdividing telehealth-eligible services between those provided in-person or via telehealth. eTables 1 and 2 in the Supplement present the list of telehealth-eligible services with the associated codes and designated codes for the place of service. We classified a telehealth-eligible service as provided via telehealth only when an appropriate modifier was present. We considered telehealth to include any synchronous service provided on a remote basis, whether via video or telephone. eTable 3 in the Supplement presents the approach to designating telehealth services and modality type. In the case of multiple claim lines per encounter (mean of 1.13), we selected the claim line with the highest allowed charge for each unique encounter. eTable 4 in the Supplement presents the study-designated specialty using practitioner specialty codes.

We captured information on the members’ demographic characteristics from enrollment files and mapped study members to the census region, state, county, and urban vs rural categorizations.¹⁹ Data on race and ethnicity were not available from the data repository. We documented the Elixhauser Comorbidity Index²⁰ for each member based on all ambulatory care primary diagnoses noted during the 12 months of 2019.²¹

We calculated a rolling 7-day mean of new COVID-19 cases to compare different counties’ COVID-19 rates, using the data repository for the Johns Hopkins University COVID-19 Visual Dashboard,²² defining a COVID-19 hotspot of residence as any county where the rolling 7-day mean of COVID-19 cases was in the top decile throughout the country at any point during the study period. We assigned a member’s residence to 1 of 4 levels of social deprivation based on a national ranking of the Area Deprivation Index.²³ We assessed internet connectivity using data from the 2019 US Census American Community Survey with a roll-up of census tract rates to zip code rates.²⁴

We categorized the insurance type as standard preferred provider organization, high deductible (ie, >$1000), or health maintenance organization. Finally, we classified each encounter as involving a new patient if the enrollee had not visited the billing provider organization within the past 3 years and as a new condition if there had not been an encounter for that primary diagnosis within the past 3 years.²⁵ All members of the cohort were continuously enrolled at least from July 2019, but to identify a new patient or a new condition we reviewed data from 2016 onward for those members who were enrolled longer.

We assessed patterns of health care use 14 days after initial encounters, defining the use as follow-up encounters of any kind, ED encounters, or hospitalizations. We hypothesized that if a telehealth encounter was not as effective as an in-person encounter in addressing outstanding clinical issues, a higher rate of postindex follow-up encounters would be observed. We used the primary diagnosis reported for the encounter to characterize the clinical issues involved and used the list of ambulatory care sensitive (ACS)²⁶ conditions,²¹ conditions that might result in an avoidable hospitalization if ambulatory care is inadequate in some respect.

We assessed the ambulatory encounters for each set of the 3-digit International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes for ACS conditions. From the cohort of members who had 1 or more encounters during the study period, we counted the number of members who had an encounter for an ACS condition. To eliminate rare conditions, we limited the list to conditions that were diagnosed in at least 5000 members. To ensure adequate use of telehealth among the conditions, we required at least 10% of the members diagnosed with an ACS condition to have at least 1 telehealth claim associated with the condition. eTable 5 in the Supplement presents the approach to selecting acute and chronic ACS conditions for the assessment of use patterns.

We required members to have no condition-related preexisting care, defined as no encounters for the condition 90 days before the initial encounter. We categorized the ACS conditions to acute and chronic to account for the potential difference in telehealth and in-person encounters because of the nature of the conditions.

For selected ACS conditions, we used multivariate logistic regression to calculate odds ratios (ORs) and 95% CIs to assess the independent effect of the type of initial encounter on the likelihood of any 14-day follow-up encounters, adjusted for various member characteristics, including age, sex, Elixhauser Comorbidity Index based on 2019 ambulatory encounters, condition at the initial encounter, urban vs rural location of residence, member’s initial encounter coinciding in a COVID-19 hotspot, Area Deprivation Index, and internet connectivity levels. For chronic conditions, we included an encounter-level severity adjustment based on the primary diagnosis of the initial encounter. eTable 6 in the Supplement presents the assignment of a severity level to each complete ICD-10 primary diagnosis within a 3-digit ICD-10 category. This level was assigned based on clinical judgment as to the relative severity of the complete ICD-10 code within the 3-digit category.

Only members with at least 1 telehealth-eligible service and at least 1 encounter in 2020 were included in the regression analysis (n = 1 086 720). The characteristics of this population are presented in eTable 7 in the Supplement. We performed a second logistic regression including an interaction term between the type of initial encounter and the specific ACS condition to address how telehealth effectiveness may vary by each clinical condition type. We used SAS, version 9.4 (SAS Institute Inc) to conduct the analyses.²⁷

In this cohort study of 40 739 915 individuals (mean [SD] age, 35.37 [18.77] years; 20 480 768 [50.3%] female and 20 259 147 [49.7%] male), ambulatory encounters decreased by 1.0% and the number of in-person encounters per enrollee decreased by 17.0% from 2019 to 2020 (Table 1). As a proportion of all ambulatory encounters, telehealth encounters increased substantially from 0.6% (n = 236 220) to 14.1% (n = 5 743 718). eTable 8 in the Supplement presents the breakdown of the summer and fall trends, and eFigures 2 and 3 in the Supplement present weekly trends for total ambulatory encounters and the percentage delivered via telehealth in 2020.

Table 2 presents the unadjusted member-level rates of total and telehealth ambulatory encounters according to member characteristics, and eTable 9 in the Supplement presents the breakdown of the summer and fall trends. In 2020, the highest telehealth uptake was seen in those aged 18 to 34 years (25.3% of per person encounters) and 35 to 49 years (19.0% of per person encounters), and more telehealth services were used in patients with a comorbidity index of 2 (19.6% of per person encounters) and 3 or higher (18.7% of per person encounters). Table 3 presents unadjusted encounter-level analyses stratified by encounter characteristics, and eTable 10 in the Supplement presents the breakdown of the summer and fall trends. The overall encounters decreased from a mean of 2.46 encounters per person in 2019 to 2.46 in 2020; however, for behavioral health encounters, the per person encounters increased from 0.32 in 2019 to 0.36 in 2020.

Table 4 presents the results of the logistic regression model. In the cohort with acute ACS conditions, the adjusted ORs for those with an initial telehealth encounter, compared with in-person, were 1.44 (95% CI, 1.42-1.46) for a follow-up encounter of any kind and 1.11 (95% CI, 1.06-1.16) for an ED encounter. eTable 11 in the Supplement presents a summary of ACS conditions and subsequent use patterns during 2020. Among the cohort with chronic ACS conditions, the adjusted ORs were 0.94 (95% CI, 0.92-0.95) for follow-up encounters of any kind and 0.94 (95% CI, 0.90-0.99) for hospitalization. Table 5 presents the patterns of subsequent use among those with acute and chronic ACS conditions by the clinical condition. For instance, among those having an acute upper respiratory tract infection episode, the ORs for members with an initial telehealth encounter, compared with in-person encounters, were 1.65 (95% CI, 1.61-1.68) for a follow-up encounter of any kind, 1.18 (95% CI, 1.10-1.25) for an ED follow-up encounter, and 1.11 (95% CI, 1.03-1.20) for hospitalization. Among patients with a chronic ACS condition, the ORs were 0.86 (95% CI, 0.84-0.88) for members with essential hypertension and 0.63 (95% CI, 0.50-0.80) for members with heart failure who had initial telehealth encounters, compared to in-person encounters, for a follow-up encounter of any kind.

When initial telehealth encounters were compared with initial in-person encounters within comparable 3-digit ICD-10 disease categories, patients tended to have higher diagnostic severity levels. For example, 10.1% of telehealth encounters had a severity level of 2.0% vs 7.4% of in-person encounters (eTable 12 in the Supplement). This finding suggests the possibility that telehealth may be used preferentially for sicker patients, perhaps as a quicker way to treat patients with greater need.

A previous study² examined the association of COVID-19 and changing coverage policies with the uptake of telehealth for a commercially insured population during the initial phase of the pandemic (March to June 2020). In this study, we extended the time frame to the next phase of the pandemic (July to December 2020). During this period of more established telehealth use, we undertook an analysis within our insured cohort that allowed us to assess follow-up encounters among comparable episodes of care delivered via telehealth vs in-person encounters.

Our findings describe a period of continuity during the second half of 2020, after the marked changes that occurred during the first half of the year. From March to June 2020, there was initially a precipitous decrease in in-person ambulatory encounter rates accompanied by a marked increase in telehealth encounters.^2,28 Clinicians and patients adapted quickly, and by the end of June, the persistent decrease in in-person office encounters was fully offset by a corresponding increase in telehealth encounters.² This picture was similar to those identified in other reports^{3-6,13,14,28-31} on telehealth expansion in the early months of the COVID-19 pandemic.

In this study, we followed up a well-defined, continuously insured cohort of patients to offer an assessment of factors associated with changing patterns of telehealth use beyond the initial months of telehealth implementation.¹² In contrast with the early months of the pandemic, patterns of use for July to December 2020 stabilized, with ambulatory encounters per health plan member being approximately the same for the summer of 2020 vs 2019. The mix of telehealth modalities (video, telephone, and other) also was relatively stable in March to December 2020.² Although ambulatory encounters per member (in-person encounters plus telehealth equivalents) and the percentage of encounters via telehealth were relatively stable during the summer and fall of 2020, some subpopulation of members had lower encounter rates in 2020 compared with 2019. In 2020, the numbers of encounters for members aged 0 to 17 years were lower than comparable rates for 2019.³² Decreased well-child visits associated with lower 2020 birth rates may explain part of this difference.³³ The telehealth rate for this age group also was lower than for the others, in part explaining the lower overall encounters. This result suggests that telehealth may not be viewed as equivalent to in-person encounters for younger patients or possibly for the type of problems that drive many pediatric encounters (eg, otitis media). These findings were similar to those detailed by Schweiberger et al,³² who identified a decreased number of pediatric care encounters because of fewer problem-focused encounters, with notably fewer infection-related encounters 7 months after the beginning of the pandemic. Ambulatory encounter rates were also decreased in the summer and fall of 2020 for older members (≥50 years of age) but to a lesser extent than for children and adolescents.

Encounter rates by the member and encounter characteristics illustrated that treatment patterns were modestly different in 2020 compared with 2019, despite the availability of telehealth as an alternative. Conversely, increased telehealth use in 2020 (summer and fall) compared with 2019 appeared to have prevented what would otherwise have been a precipitous decrease in ambulatory care triggered by COVID-19.

We provide a comparison of key outcomes of care 14 days after initial telehealth vs in-person encounters for a set of acute and chronic ACS conditions. Although other studies^15,16 have assessed such an association among a small sample of patients, we expanded this work and assessed the outcomes for acute and chronic ACS conditions separately and beyond in-person hospitalization. Our results showed that the use of telehealth services for the management of chronic ACS conditions appeared to be comparable to in-person encounters concerning the need for follow-up. However, patients with an initial telehealth encounter for 1 of the acute ACS conditions appeared to require additional follow-up compared with patients with an initial in-person ambulatory encounter.

From our analysis of episodes of care for specific presenting conditions, we identified that follow-up encounters after an index telehealth encounter were substantially more common for acute respiratory infections. Such increased follow-up was not more likely for other types of acute conditions (eg, acute pyelonephritis), suggesting that this difference could reflect concerns associated with the ongoing pandemic. Given that symptoms of these respiratory infections may be similar to those of COVID-19, one explanation of the higher number of follow-up encounters after an initial telehealth encounter of these types could reflect follow-up linked to suspected COVID-19 (eg, testing or ensuring adequate patient recovery). In contrast, follow-up care was generally less frequent after an initial telehealth encounter for a chronic condition than for an in-person encounter for the same condition.

This study has some limitations. Our results comparing the association of telehealth with follow-up care should be interpreted with caution. Although we applied multivariate regression modeling to account for a range of factors that could bias the results, some uncontrolled confounding bias might remain. One example would be the bias in the choice of telehealth and in-person encounter. For instance, a clinician may choose to provide a telehealth encounter to perform an initial assessment of a patient with mobility limitations before recommending an ED encounter or hospitalization. The finding of patients with complex clinical conditions (higher diagnostic severity levels) having more telehealth than in-person encounters support this concern. In addition, the end points used to compare telehealth and in-person encounters were limited in scope and temporality. Other clinically important end points for comparison would include the use of relevant laboratory tests and medication use and adjustments. Another alternative approach would be assessing additional models of care that integrate telehealth (eg, a telehealth and in-person hybrid model) and their effect on outcomes of care. Moreover, the time window for monitoring outcomes could be extended beyond 14 days, and a broader range of conditions could be studied. The nationally representative database we used did not include Medicare, Medicaid, or uninsured patients. Thus, experiences among these special needs patients could be different from those we documented in this study.

In this cohort study of 40.7 million US commercially insured adults telehealth accounted for a large share of ambulatory encounters at the peak of the COVID-19 pandemic and remained prevalent after infection rates subsided. We identified many patient, practitioner, and community factors associated with the higher telehealth use, and trends were similar to those observed during the early months of the pandemic. Moreover, we compared the clinically relevant outcomes for telehealth vs in-person encounters in a nationally representative population, which extends our knowledge in terms of assessing the association of telehealth with outcomes of care. This study found that the use of telehealth services for the management of chronic ACS conditions was comparable, or even more efficient, than in-person care when follow-up encounters were assessed. On the other hand, patients with an initial telehealth encounter for acute ACS conditions appeared to require additional follow-up. This trend was observed especially for acute respiratory-related conditions, which potentially could be confounded by concerns over COVID-19 rather than the less complicated acute non–COVID-19 diagnosis. These findings suggest a direction for future work and are relevant to policy makers, payers, and practitioners as they manage the use of telehealth during the pandemic and afterward.

Accepted for Publication: March 7, 2022.

Published: April 26, 2022. doi:10.1001/jamanetworkopen.2022.8954

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2022 Hatef E et al. JAMA Network Open.

Corresponding Author: Elham Hatef, MD, MPH, Division of General Internal Medicine, Department of Medicine, Johns Hopkins University, 624 N Broadway, Room 502, Baltimore, MD 21218 (ehatef1@jhu.edu).

Author Contributions: Mr Lans had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Hatef, Bandeian, Lasser, Goldsack, Weiner.

Acquisition, analysis, or interpretation of data: Hatef, Lans, Bandeian, Lasser, Weiner.

Drafting of the manuscript: Hatef, Lans, Lasser.

Critical revision of the manuscript for important intellectual content: Hatef, Bandeian, Goldsack, Weiner.

Statistical analysis: Lans, Bandeian.

Obtained funding: Weiner.

Administrative, technical, or material support: Lasser, Goldsack, Weiner.

Supervision: Hatef, Weiner.

Conflict of Interest Disclosures: Drs Hatef, Lasser, and Weiner had a contract with the American Telehealth Association during the conduct of the study. No other disclosures were reported.

Funding/Support: This study was funded by the Center for Population Health Information Technology at Johns Hopkins Bloomberg School of Public Health (Drs Hatef, Lasser, and Weiner) and Blue Health Intelligence and partially funded by the American Telemedicine Association (Drs Hatef, Lasser, and Weiner).

Role of the Funder/Sponsor: The Center for Population Health Information Technology at Johns Hopkins Bloomberg School of Public Health and Blue Health Intelligence had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. The American Telemedicine Association reviewed and commented on the conduct of the study, but did not play any other role in the study design; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The content of this article is the responsibility of the authors and does not necessarily represent the positions of the Johns Hopkins University or Blue Health Intelligence or the Blue Cross and Blue Shield Association or the American Telemedicine Association.

Additional Contributions: Lauren Tansky, BS, Center for Population Health Information Technology, provided technical support in the preparation of tables. She was compensated for her work. We are grateful to Blue Health Intelligence and the health plans who contributed to this database for making these data available.