Patient characteristics associated with frequent telehealth utilization in 2022: Evaluation of a national virtual integrated medical and behavioral health practice within the United States

Abstract

Objectives

As permanent telehealth policies are considered in the United States (U.S.), it is important to understand who uses telehealth most often following the pandemic. We described patients who used a national virtual care practice frequently, identified how they differed from patients who used it less often, and characterized the types of care frequent telehealth patients utilized.

Methods

We used video visit data for commercially-insured patients, aged 18+, from a national virtual integrated medical and behavioral health practice in 2022 in the U.S. Patients were categorized into three groups: one visit (’minimal use’), two to four visits (’some use’), and five or more visits (’frequent use’). We compared patient and geographic characteristics between the three groups and estimated an ordinary least squares linear regression to identify predictors of ‘frequent’ use relative to ‘minimal’ or ‘some’ use.

Results

The probability of being a frequent user declined with age (−0.4 percentage points (p.p.) per year; 95 % CI, −0.4 – −0.3), was higher for females (5.4 p.p.; 95 % CI, 4.1 – 6.7) and patients with greater clinical complexity (7.9 p.p. for highest relative to lowest quartile risk score; 95 % CI, 5.9 – 10.0), and lower for patients in the Northeast (−9.2 p.p.; 95 % CI, −15.5 – −2.9) or West (−3.2 p.p.; 95 % CI, −5.7 – −0.7) regions relative to the Southern region of the U.S. The five most common diagnoses were mental health conditions.

Conclusions

Our results highlight the need for comprehensive telehealth policy that enables access, particularly for patients who rely on it as their primary source of care.

1. Introduction

The COVID-19 pandemic quickly accelerated telehealth utilization, fueling virtual care innovation (Grandview Research, 2023). While telehealth utilization has decreased since its height (Healthcaredive, 2020), it remains a vital component of health care delivery (mHealth Intelligence, 2023). Past research has primarily compared telehealth users to non-users, with mixed results on the characteristics associated with utilization (Mehrotra and Uscher-Pines, 2022). However, few studies have examined the attributes of frequent users of telehealth compared to infrequent users, and little is known about telehealth use after 2021 when in-person care was once again considered a safe option. This study examined differences between adult patients who frequently utilized virtual care to those who used it less often within the United States.

2. Methods

We examined characteristics associated with frequent telehealth use among commercially-insured patients of a virtual integrated medical and behavioral health practice in 2022 in the United States (U.S.). The study was limited to 17,941 patients aged 18 or older with medical claims and access to the virtual practice through employer benefits throughout 2022, excluding 230 individuals with missing sex, region, or zip code. All patients in the study could access video-based virtual visits with minimal to no cost sharing. Patients were grouped by virtual visit frequency: one visit ('minimal use'), two to four visits ('some use'), and five or more visits ('frequent use').

Characteristics of interest included age, sex assigned at birth, U.S. census region, urban or rural location, residing in primary care or mental Health Professional Shortage Area (HPSA), 2021 Hierarchical Condition Categories (HCC) risk score (Kautter et al., 2014) as a proxy for clinical complexity, and number of 2022 primary care and behavioral health visits outside of the virtual practice. We also defined characteristics at the members’ geographic location, based on zip code tabulation area from the 2015–2019 American Community Survey (Social Determinants of Health Database, 2023), including the percent of residents by race and ethnicity, the percent of households with broadband, the percent with income below federal poverty level, the percent of household heads with less than 12 years education, and the percent with no vehicle.

We used the Agency for Healthcare Research and Quality chronic condition indicator crosswalk to categorize visits as chronic, acute, or both based on primary and secondary diagnoses (Chronic Condition Indicator Refined for ICD-10-CM, 2024). The remaining visits were categorized as ‘other’, which includes preventive visits (see supplement for International Classification of Diseases, Tenth Revision (ICD-10) codes) and visits with a missing diagnosis code. We also identified the top five most common diagnoses using the first three characters of the primary diagnosis ICD-10 code (e.g., F32 Depressive episode or major depressive disorder).

We used counts, proportions, averages and standard deviations to report member and visit characteristics. We performed unpaired, two-tailed t-tests to assess unadjusted differences between the patient groups. We also estimated an ordinary least squares linear regression to identify predictors of being a ‘frequent user’. The outcome was binary: ‘frequent user’ (five or more visits) vs. not a ‘frequent user’ (one to four visits). We chose a linear regression for ease of interpretation; the coefficients represent the expected change in the probability of being a frequent user associated with each covariate (Sommers et al., 2016). The regression model included all patient and geographic characteristics described above as covariates. We considered statistically significant findings to have a p-value of less than 0.05. The WIRB-Copernicus Group Institutional Review Board considered this study exempt from review under 45 CFR §46.104(d)(4) on May 4, 2023 because patients were not contacted and the data was de-identified.

3. Results

There were 17,941 patients with at least one virtual visit in 2022 of which 8,284 (46.2 %) had one visit (‘minimal use’), 6,358 (35.4 %) had two to four visits (‘some use’), and 3,299 (18.4 %) had five or more visits (‘frequent use’) (Table 1). Patients were 39.6 years old on average (SD±0.09). The majority were female (64.9 %), resided in the Southern region of the U.S. (61.0 %), and in urban areas (68.2 %). Few lived in primary care (8.0 %) and mental health HPSAs (19.6 %). Patients lived in zip codes where, on average, most residents had broadband access (82.8 %), were White (77.6 %), and were non-Hispanic (91.1 %), and in zip codes where 14.1 % of residents lived below the federal poverty level, 12.1 % of household heads had less than 12 years education, and 6.2 % of households did not own a vehicle.

Table 1.

Differences in characteristics by volume of virtual care visits, adult virtual care patients, United States, 2022.

	Any Virtual Care (N=17,941)	1 Visit 'Minimal Use' (N=8,284)	Minimal Use vs. Frequent Use1	2–4 Visits 'Some Use' (N=6,358)	Some Use vs. Frequent Use1	5 + Visits 'Frequent Use' (N=3,299)
Age in years (mean ± SD)	39.6 ± 0.09	40.4 ± 0.13	***	39.7 ± 0.15	***	37.2 ± 0.19
Sex assigned at birth (%):
Male	35.1 %	38.7 %	***	34.2 %	***	27.7 %
Female	64.9 %	61.3 %	***	65.8 %	***	72.3 %
Region of United States (%):
Northeast	1.0 %	1.1 %	***	1.1 %	***	0.5 %
South	61.0 %	58.6 %	***	62.5 %		64.3 %
West	11.0 %	11.7 %	***	11.2 %	***	9.0 %
Midwest	26.9 %	28.6 %	***	25.1 %		26.2 %
Urban/Rural (%):
Urban	68.2 %	67.3 %	***	68.4 %		70.0 %
Rural	31.8 %	32.7 %	***	31.6 %		30.0 %
Reside in professional shortage area (%)
Primary Care Professional Shortage Area	8.0 %	8.3 %		7.8 %		7.5 %
Mental Health Professional Shortage Area	19.6 %	20.5 %	**	19.0 %		18.5 %
Number of visits outside the virtual practice, in person or telehealth (mean ± SD)	1.75 ± 0.04	1.70 ± 0.04		1.75 + 0.09		1.87 ± 0.061
HCC score (%)2
Quartile 1 (lowest score)	20.9 %	22.1 %	***	20.2 %		19.6 %
Quartile 2	19.9 %	19.5 %	**	21.6 %	***	17.6 %
Quartile 3	15.9 %	17.0 %	***	16.4 %	***	12.4 %
Quartile 4 (highest score)	18.8 %	17.0 %	***	18.9 %	***	23.4 %
Missing HCC score	24.3 %	24.4 %	***	22.9 %	***	26.9 %
Proportion of households in zip code (avg % of residents):
Households with broadband of any type	82.80	82.89		82.69		82.79
Households below federal poverty level	14.12	13.97	***	14.08	***	14.56
Household head < high school education	12.11	11.99	**	12.20		12.26
Households with no vehicle	6.22	6.16		6.26		6.29
Proportion of residents in zip code tabulation by race (avg)
American Indian/Alaskan Native	0.53	0.52		0.58	***	0.47
Asian	2.98	3.16	***	2.91	**	2.64
Black	12.63	11.75	***	12.76	***	14.59
Multiple Races	3.72	3.74		3.71		3.72
Native Hawaiian/Pacific Islander	0.09	0.09		0.10		0.09
Other race	2.44	2.42	***	2.38	***	2.64
White	77.59	78.29	***	77.57	***	75.85
Proportion of residents in zip code who identify as Hispanic (Avg % of residents)	8.92	8.92		8.87		9.02

¹5 + visits Frequent Use group is the reference category; *** p < 0.01 and ** p < 0.05 based on unpaired, two-tailed t-tests.

²HCC (Hierarchical Condition Categories) score required members to have continuous eligibility and medical claims data throughout 2021 and 2022. HCC scores were not calculated for members who did not meet that criterion.

Patients with frequent virtual care use differed from patients with minimal or some virtual care use. Relative to patients with minimal use, frequent users were younger (37.2 ± 0.19 vs. 40.4 ± 0.13; p < 0.01); more likely to be female (72.3 % vs. 61.3 %; p < 0.01), more clinically complex (23.4 % vs. 17.0 % in the highest HCC quartile; p < 0.01), more likely to live in the Southern region of the U.S. (64.3 % vs. 58.6 %; p < 0.01), more likely to live in urban areas (70.0 % vs. 67.3 %; p < 0.01), and less likely to live in mental HPSAs (18.5 % vs. 20.5 %, p < 0.05). Frequent users were also more likely to live in zip codes with a higher proportion of Black residents (14.6 % vs. 11.8 %; p < 0.01) and a lower proportion of White residents (75.9 % vs. 78.2 %; p < 0.01). Additional small but statistically significant differences include a higher proportion of residents living below the federal poverty level (14.6 % vs. 14.0 %; p < 0.01) with less than 12 years of education (12.3 % vs. 12.0 %; p < 0.05) and fewer identifying as Asian (2.6 % vs. 3.2 %; p < 0.01).

Similar patterns were found between patients with frequent and some use, although differences were smaller in magnitude and some were no longer statistically significant, including residing in a mental HPSA, residing in the Southern region of the U.S., and the proportion of residents within the zip code with less than a high school education.

After adjusting for all covariates, the probability of being a frequent user declined with age (−0.4 percentage points (p.p.) per year of age; 95 % CI, −0.4 – −0.3). The probability was higher for patients who were female (5.4 p.p.; 95 % CI, 4.1 – 6.7), had an HCC risk score in the highest quartile vs. the lowest quartile (7.9 p.p.; 95 % CI, 5.9 – 10.0), and lived in zip codes with a higher proportion of residents living below the federal poverty level (0.1 p.p.; 95 % CI, 0.0 – 0.3) and a higher proportion of residents identifying as ‘Other’ race (0.7 p.p.; 95 % CI, 0.0 – 1.3). The probability of being a frequent user was lower for patients who lived in the Northeast region (−9.2 p.p.; 95 % CI, −15.5 – −2.9) or West region (−3.2 p.p.; 95 % CI, −5.7– −0.7) relative to the Southern region of the U.S. (data not shown).

Patients with frequent use primarily accessed virtual care for mental health chronic conditions. Approximately 73.5 % of the frequent user visits were for chronic conditions compared to 17.1 % and 31.3 % of visits among those with minimal use and some use, respectively (Table 2). The five most common visit diagnoses among frequent users, which accounted for 54.3 % of all their visits, were mental health conditions: major depressive disorder, reaction to severe stress and adjustment disorders, other anxiety disorders, depressive disorder, and bipolar disorder. On the other hand, patients with minimal and some use of virtual care had greater variation among diagnoses codes; the top 5 diagnosis codes accounted for only 36 % and 31 % of visits, respectively. Both groups’ most common diagnosis was acute sinusitis followed by COVID-19. Frequent users’ visits were distributed similarly across the behavioral health and medical practices (46.8 % vs. 53.2 %), indicating they accessed both practices for their mental health needs (data not shown).

Table 2.

Proportion of visits by diagnosis category and ICD-10 codes, by type of virtual care user, United States, 2022.

1 Visit 'Minimal Use' (N=8,284)		2 to 4 Visits 'Some Use' (N=6,358)		5 or more Visits 'Frequent Use' (N=3,299)
Diagnosis code or category	Visit %	Diagnosis code or category	Visit %	Diagnosis code or category	Visit %
Proportion of visits by diagnosis category
Acute only	81.7 %	Acute only	66.7 %	Chronic only	58.9 %
Chronic only	10.3 %	Chronic only	20.4 %	Acute only	25.2 %
Both acute and chronic	6.9 %	Both acute and chronic	10.9 %	Both acute and chronic	14.6 %
Other	1.2 %	Other	2.0 %	Other	1.3 %
Proportion of visits by top 5 most common ICD-10 codes
J01 Acute sinusitis	11.8 %	J01 Acute sinusitis	10.2 %	F33 Major depressive disorder	24.9 %
U07 COVID-19	7.0 %	U07 COVID-19	5.8 %	F43 Reaction to severe stress and adjustment disorders	10.4 %
J06 Acute upper respiratory infections	6.6 %	F33 Major depressive disorder	5.8 %	F41 Other anxiety disorders	9.6 %
N39 Other disorders of urinary system	5.5 %	N39 Other disorders of urinary system	5.1 %	F32 Depressive disorder	4.8 %
J02 Strep	4.8 %	J06 Acute upper respiratory infections	4.5 %	F31 Bipolar disorder	4.6 %

Diagnosis category (acute, chronic, other) considered both primary and secondary diagnosis codes. Proportion of visits by ICD-10 codes is primary diagnosis only. Other includes missing ICD-10 or ICD-10 categorized as preventive visit (see Supplement). ICD-10 = International Classification of Diseases, Tenth Revision (ICD-10).

4. Discussion

In 2022, following the peak of the COVID-19 pandemic, over 18 % of patients with at least one telehealth visit had five or more visits. Virtual care became their primary source of care, predominantly for mental health conditions, distributed nearly equally between the behavioral health and primary care practices available to them. Frequent users of virtual care were younger, had higher HCC scores, were more likely to be female, and were more likely to live in the Southern region of the U.S. compared to less frequent users. Other studies have found young females were more likely than their older or male counterparts to use telehealth at all, and that most telehealth was for mental health care (Eberly et al., 2020, Harju and Neufeld, 2022). However, our findings that frequent telehealth users were more likely than less-frequent telehealth users to live in the Southern region of the U.S. and were slightly more likely to live in low income areas differs from studies characterizing users vs. non-users (Patel et al., 2021, Lee et al., 2023).

Despite representation in all but three U.S. states, results are not generalizable outside of commercially-insured adults and were based on a sample of patients from one virtual practice. Regardless, this study demonstrates the potential for a virtual primary care and behavioral health care integrated practice to be an effective tool for providing ongoing care, particularly mental health care, to a diverse group of patients who reside in areas with greater barriers to care. The inclusion of virtual practices for increased access to mental chronic health care must be considered as Congress debates extending many of the current Medicare telehealth flexibilities through the end of 2026 (Beavins, 2024).

Funding sources and conflicts of interest statement

Fredric Blavin reports financial support was provided by Robert Wood Johnson Foundation. Laura Barrie Smith reports financial support was provided by Robert Wood Johnson Foundation. Claire O’Brien reports financial support was provided by Robert Wood Johnson Foundation. Jaclyn Marshall and Ami Parekh report a relationship with Included Health that includes: employment and equity or stocks. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

CRediT authorship contribution statement

Jaclyn Marshall: Writing – original draft, Project administration, Methodology, Investigation, Data curation, Conceptualization. Fredric Blavin: Writing – review & editing, Supervision, Resources, Methodology, Funding acquisition. Claire O’Brien: Writing – review & editing, Visualization, Software, Formal analysis, Data curation. Ami Parekh: Writing – review & editing, Resources. Laura Barrie Smith: Writing – review & editing, Supervision, Resources, Methodology, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Fredric Blavin reports financial support was provided by Robert Wood Johnson Foundation. Laura Barrie Smith reports financial support was provided by Robert Wood Johnson Foundation. Claire O’Brien reports financial support was provided by Robert Wood Johnson Foundation. Jaclyn Marshall and Ami Parekh report a relationship with Included Health that includes: employment and equity or stocks. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.