Comparison of Telehealth and In-Person Behavioral Health Services and Payment in a Large Rural Multisite Usual Care Study

Abstract

Background:

Telehealth and in-person behavioral health services have previously shown equal effectiveness, but cost studies have largely been limited to travel savings for telehealth cohorts. The purpose of this analysis was to compare telehealth and in-person cohorts, who received behavioral health services in a large multisite study of usual care treatment approaches to examine relative value units (RVUs) and payment.

Methods:

We used current procedural terminology codes for each encounter to identify RVUs and Medicare payment rates. Mixed linear regression models compared telehealth and in-person cohorts on RVUs, per-encounter payment rates, and total-episode payment rates.

Results:

We found the behavioral health services provided by telehealth to have modest, but statistically significantly lower RVUs (i.e., less provider work in time spent and case complexity), per-encounter payments, and total episode payments than the in-person cohort. Despite Medicare rates discounting payments for nonphysician providers and the in-person cohort using clinical social workers more frequently, the services provided by the telehealth cohort still had lower payments. Thus, the differences observed are due to the in-person cohort receiving higher payment RVU services than the telehealth cohort, which was more likely to receive briefer therapy sessions and other less expensive services.

Conclusions:

Behavioral health services provided by telehealth used services with lower RVUs than behavioral health services provided in-person, on average, even after adjusting for patient demographics and diagnosis. Observed differences in Medicare payments resulted from the provider type and services used by the two cohorts; thus, costs and insurance reimbursements may vary for others.

Introduction

Behavioral health care providers are in short supply in many parts of the United States (U.S.), particularly in rural areas.^1
–3 Connecting rural residents in need of behavioral health services with mental health professionals at a distance through telehealth technology has evolved as an effective approach to addressing this need, as indicated by multiple systematic reviews.^4

–7

In addition to being equally clinically effective to in-person services for many conditions,^8,9 the telehealth modality is often touted as cost-saving. However, most evidence on telehealth cost-savings originates from avoided patient travel.^10,11 There are very few published studies of other cost components comparing telehealth with in-person behavioral health treatment.^4,11

These analyses use a novel approach to compare telehealth and in-person behavioral health care by examining the services that were delivered in a large multisite study of usual care treatment. There is a convention of assigning relative value units (RVUs)¹² to health care services, which form the basis for both government (e.g., Centers for Medicare & Medicaid Services [CMS]) and private insurance payments. By combining data across numerous diverse clinic sites, behavioral health services provided to patients could be directly compared between those receiving telehealth care versus those receiving similar types of in-person care to examine payment for provider work associated with time spent and case complexity using RVUs and Medicare payment rates.

Methods

PARTICIPATING SITES

The Office for the Advancement of Telehealth (OAT) within the Health Resources and Services Administration (HRSA) of the U.S. Department of Health and Human Services funded two grant programs: the Substance Abuse Treatment Telehealth Network Grant Program from September 2017 to August 2020 and the Evidence-Based Tele-Behavioral Health Network Program from September 2018 to August 2021. Through these 2 programs, 17 grantees provided behavioral health services to 95 clinic sites serving rural areas in 13 states. The grantees implemented behavioral health services using both in-person and synchronous video telehealth, which were tailored to the needs, resources, and capacity within their rural areas.

INSTRUMENTS

OAT funded the Rural Telehealth Research Center (RTRC) to serve as a data coordinating center for the grant-funded clinical programs. RTRC identified data elements, developed an Excel-based tool for standardized, anonymized data collection,¹³ and established data transfer and use agreements with each grantee. All involved entities secured Institutional Review Board Human Subject Review approval, as indicated. Office of Management and Budget clearance for the data collection activity was received in October 2019, and grantees provided data from then until July, 2021.

RESEARCH DESIGN

A nonrandomized prospective design involved treatment cohorts identified by clinical program sites. The telebehavioral health cohort included all patients who began telehealth treatment as part of either grant program during the data collection period. Grantees identified patients for the in-person cohort who had similar primary complaint or diagnosis and demographics, and who received comparable treatment (e.g., therapeutic approach and clinician type). Data collection extended for 3 months after treatment initiation for each patient and included data on all encounters during that period.

SAMPLE CRITERIA

The data set included information on 14,360 patients. Exclusion criteria removed patient data due to lack of treatment group assignment (n = 2), lack of encounter records (n = 811), or no completed encounter within the 3-month follow-up period (n = 2,310). This resulted in a study population with 11,237 patients and 37,059 encounters. Additional exclusion criteria further removed patients (n = 1,728) and encounters (n = 8,052) with no information on services provided (current procedural terminology [CPT] codes missing or not listed on the Medicare Physician Fee Schedule from the CMS website),¹⁴ and patients (n = 1,154) and their encounters (n = 4,321) with missing information on clinician type. The final analytical sample comprised 3,115 patients and 9,072 encounters in the telehealth cohort and 5,240 patients and 15,614 encounters in the in-person cohort.

ASSIGNMENT OF RVUs TO CPT CODES AND PHYSICIAN FEE SCHEDULE RATES TO RVUs

RVUs are a widely used measure of value assigned to each CPT code based on the extent of physician work, clinical and nonclinical resources, and expertise required to deliver a health care service to patients.¹² CMS uses RVUs to determine physician payment for each service and procedure as shown in the Medicare Physician Fee Schedule, which forms the basis of Medicare reimbursements and most commercial fee schedules.¹² CMS pays clinicians the same rate for telehealth and in-person services. For these analyses, we used the Medicare Physician Fee Schedule 2020 national RVUs.¹⁴ We chose 2020 because it was the midpoint of the data collection period.

We used the national RVUs, as an average across all geographic regions represented in the data. We used nonfacility fee rates because most grantees provided behavioral health services through outpatient clinics rather than hospitals. A CMS website provides a database that lists the Medicare Physician Fee Schedule for each CPT code.¹⁴ The Medicare Physician Fee Schedule multiplies the RVU for each CPT code by a conversion factor established by CMS each year. In 2020, the conversion factor was 34.8931.

MEDICARE PHYSICIAN FEE SCHEDULE ADJUSTMENT FOR VARIOUS CLINICIAN TYPES

Medicare adjusts the physician fee schedule for nonphysician clinicians and our analyses used these adjusted rates. For mental health services, Medicare reimburses physicians and clinical psychologists 100% of the published Medicare Physician Fee Schedule amount, reimburses physician assistants, nurse practitioners, and clinical nurse specialists 85% of the published amount, and reimburses clinical social workers 75% of the published amount. Medicare does not reimburse for services provided by licensed professional counselors, provisionally licensed professionals, licensed mental health counselors, or licensed marriage and family therapists.¹⁵

DATA ANALYSIS

Baseline characteristics of patients in the telehealth treatment group and the in-person treatment group were compared using chi-squared tests. RVUs and Medicare payment rates were compared between cohorts using t-tests. We then employed mixed linear regression models adjusted for covariates (i.e., age, sex, race, ethnicity, insurance type, and diagnosis), with clustering of standard errors at the level of the patient. In the primary outcome model, the dependent variable was per-encounter RVUs. In the secondary outcome model, the dependent variable was the per-encounter Medicare payment rate that includes Medicare adjustments by clinician type.

Initial encounters were more likely to involve diagnostic services that differ in the RVUs compared to subsequent follow-up encounters, so both models included a variable indicating whether the encounter was the first one or subsequent. In tertiary outcome analyses, we combined the payments across all encounters for each patient and the total payment amount per patient was used as the dependent variable. This model adjusted for the previous covariates and included an interaction term between the indicator variable for telehealth and the number of encounters for each patient.

SENSITIVITY ANALYSIS

Several sensitivity analyses were conducted to further explore approaches to applying payment rates to RVUs. To examine how the results would differ if Medicaid or private insurance payment rates were applied to all patients instead of Medicare payment rates, we used average adjustments extracted from the literature.^16,17 In doing so, we multiplied Medicare payment rates by 0.67 to approximate Medicaid payment rates and multiplied Medicare payment rates by 1.43 to approximate private insurance payment rates to bookend the Medicare payment rates with a high (private insurance) and low (Medicaid) estimate.

A second sensitivity analysis used these same adjustments, but applied them according to the insurance type for individual patients in the original models. Patients with dual Medicare/Medicaid, dual Medicare/private, and Tricare/Veterans Affairs insurance were assigned Medicare payment rates; patients in the self-pay/uninsured categories were assigned estimated private insurance payment rates; and patients with Medicaid were assigned estimated Medicaid payment rates.

Because data collection spanned from October 2019 to July 2021, a third sensitivity analysis was conducted after restricting the analytical sample to those patients who enrolled in the program after the declaration of the COVID-19 public health emergency in April 2020 and their encounters. A final sensitivity analysis excluded the 7.7% of patients who had any encounter in a modality (telehealth or in person) that differed from the cohort to which they were assigned.

Results

Descriptive statistics on the 3,115 patients in the telehealth cohort and the 5,240 patients in the in-person cohort are shown in Table 1. Chi-squared analysis indicated that the cohorts differed significantly on age category, race, ethnicity, insurance type (all p < 0.0001), and diagnosis (p < 0.001).

Table 1.

Descriptive Statistics Comparing the Telehealth and In-Person Cohorts

VARIABLE	TELEHEALTH		IN PERSON		p
VARIABLE	n	%^a	n	%^a	p
Age group
<18 years	739	23.8	992	18.9	<0.0001
19–34 years	891	28.6	1,631	31.1
35–64 years	1,185	38.1	2,029	38.7
≥65 years	296	9.5	588	11.2
Unknown age	4	0.1	0	0.0
Sex
Male	1,080	34.7	1,864	35.6	0.183
Female	2,026	65.1	3,354	64.0
Unknown sex	9	0.2	22	0.4
Race^b
White	2,586	83.0	4,370	83.4	<0.0001
Black/African American	46	1.5	147	2.8
American Indian/Alaska Native	104	3.3	76	1.5
Asian/Native Hawaiian/Other Pacific Islander	10	0.4	57	1.1
Multiracial	29	0.9	45	0.9
Unknown race	340	10.9	545	10.4
Ethnicity
Hispanic/Latinx	177	5.7	146	2.8	<0.0001
Not Hispanic/Latinx	2,477	79.5	4,568	87.2
Unknown ethnicity	461	14.8	526	10.0
Primary insurer type
Medicaid	1,181	37.9	870	16.6	<0.0001
Medicare/Dual Medicare Medicaid	554	17.8	804	15.3
Government/Veterans Affairs	56	1.8	91	1.7
Private	999	32.1	2,658	50.7
Self-pay/uninsured	184	5.9	347	6.6
Other	20	0.6	32	0.6
Unknown insurer	121	3.9	438	8.4
Primary diagnosis at first encounter
Mental disorders due to known physiological conditions (F01–F09)	18	0.58	122	2.33	<0.001
Mental and behavioral disorders due to psychoactive substance use (F10–F19)	228	7.32	176	3.36
Schizophrenia, schizotypal, delusional, and other nonmood psychotic disorders (F20–F29)	74	2.38	77	1.47
Mood (affective) disorders (F30–F39)	1,274	40.90	1,776	33.89
Anxiety, dissociative, stress- related, somatoform and other nonpsychotic mental disorders (F40–F48)	963	30.91	2,491	47.54
Behavioral syndromes associated with physiological disturbances and physical factors (F50–F59)	11	0.35	33	0.63
Disorders of adult personality and behavior (F60–F69)	39	1.25	73	1.39
Pervasive and specific developmental disorders (F80–F89)	33	1.06	18	0.34
Behavioral and emotional disorders with onset usually occurring in childhood and adolescence (F90–F98)	203	6.52	236	4.50
Nonmental health primary diagnosis	170	5.46	141	2.69
Unknown or Intellectual disabilities (F70–F79)^c	102	3.27	97	1.85
n	3,115		5,240

Total may sum to >100% due to rounding.

Some race groups have been combined in this table due to small sample size, but are handled separately in the statistical analyses.

Intellectual disabilities (F70–F79) had a sample size <10 so those values have been combined with unknown in this table, but are handled separately in the statistical analyses.

Table 2 shows the Medicare Physician Fee Schedule amount for CPT codes in the dataset. The table also shows the percentage of such codes in the telehealth and in-person cohorts, some of which differed significantly, indicating that grantees provided different services to patients in the two cohorts, in turn producing different payment amounts for the two cohorts.

Table 2.

Medicare Physician Fee Schedule Amount for Current Procedural Terminology Codes and Percentage in the Two Cohorts

CPT CODE NUMBER AND LABEL	FEE SCHEDULE AMOUNT IN USD	TELEHEALTH %	IN PERSON %
90791 Psychiatric diagnostic evaluated w/o med services	$145.44	5.6%	11.8%
90792 Psychiatric diagnostic interview	$160.96	8.3%	4.8%
90832 Individual psychotherapy, 30 min	$71.10	13.9%	10.2%
90834 Individual psychotherapy, 45 min	$94.56	23.6%	21.1%
90837 Individual psychotherapy, 60 min	$141.47	13.0%	34.2%
90839 Psychotherapy for crisis, 60 min	$147.61	1.0%	0.1%
90846 Family psychotherapy w/o patient 50 min	$103.58	1.0%	0.1%
90847 Family psychotherapy w/patient 50 min	$107.19	2.2%	0.8%
90853 Group psychotherapy, 60 min	$28.15	6.7%	7.1%
96130 Psychological testing evaluated by physician/QHP 1st h	$121.98	0.0%	0.4%
96131 Psychological testing evaluated by physician/QHP each add hour	$93.83	0.0%	0.1%
96132 Neuropsychological testing evaluated by physician/QHP 1st h	$136.42	0.0%	0.8%
96158 Health behavior intervention, individual 1st 30 min	$68.21	0.7%	0.0%
98968 QHP telephone assessment & management 21–30 min	$41.14	0.7%	0.1%
99202 Outpatient E&M, new patient, 15–29 min (low complexity)	$77.23	0.3%	0.0%
99204 Outpatient E&M, new patient 45–59 min (moderate complexity)	$167.10	1.1%	0.1%
99205 Outpatient E&M, new patient 60–74 min (high complexity)	$211.12	0.9%	0.1%
99212 Evaluation & management, focused history	$46.20	0.2%	0.5%
99213 Evaluation & management, expanded history	$76.15	8.6%	5.6%
99214 Evaluation & management, detailed history	$110.43	11.6%	1.9%
99215 Evaluation & management, complex history (high complexity medical decision making)	$148.33	0.5%	0.3%
Other (rates used in analysis, but not shown here)		0.3%	0.1%

CPT, current procedural terminology; E&M, evaluation and management; QHP, qualified health professional.

Table 3 shows the CMS payment adjustment to the Medicare Physician Fee Schedule by clinician type in the first column. As shown in the table, the frequency of clinician type varied significantly across the two cohorts. This table shows that grantees used different types of clinicians to provide services to patients in the two cohorts, which affected Medicare payment amounts.

Table 3.

Medicare Physician Fee Schedule Amount by Frequency of Clinician Types in the Sample (n and %)

PROVIDER TYPE	PAYMENT RATE	TELEHEALTH n	TELEHEALTH %	IN PERSON n	IN PERSON %
Psychiatrist/other physician	100%	1,697	18.7%	1,587	10.2%
Clinical psychologist	100%	1,644	18.1%	487	3.1%
Psychiatric/mental health APP^*	85%	1,577	17.4%	852	5.5%
Nurse psychotherapist	85%	210	2.3%	3	0.0%
Clinical social worker	75%	2,403	26.5%	11,736	75.2%
Licensed professional counselor	0%	1,320	14.6%	949	6.1%
Provisionally licensed counselor	0%	221	2.4%	0	0.0%

APP, advanced practice provider (including Physician's Assistant and/or Advanced Practice Registered Nurse).

The two cohorts were similar in the mean number of encounters per patient during the first 3 months of treatment. The telehealth cohort averaged 2.91 encounters and the in-person cohort averaged 2.98 encounters (p-value = 0.32). As shown in Table 4, in unadjusted analyses, the telehealth cohort was provided behavioral health services with modest, but statistically significantly lower average RVUs, payment per encounter, and total payment amount than the in-person cohort (all p < 0.001).

Table 4.

Difference in Relative Value Units, Medicare Payment Amount per Encounter, and Total Payment per Patient Between the Telehealth and In-Person Cohorts

GROUP	ENCOUNTER-LEVEL ANALYSES			PATIENT-LEVEL ANALYSES
GROUP	ENCOUNTER n	RVUs PER ENCOUNTER MEAN (SD)^***	PAYMENT PER ENCOUNTER MEAN (SD)^***	PATIENT n	TOTAL PAYMENT PER PATIENT MEAN (SD)^***
Telehealth	9,072	2.92 (1.05)	$78.43 (49.64)	3,115	$228.40 (218.13)
In person	15,614	3.15 (1.04)	$84.37 (37.28)	5,240	$251.40 (227.64)

p < 0.05, ^** p < 0.01, ^*** p < 0.001. p-Value based on t-test.

RVUs, relative value units; SD, standard deviation.

The results of our primary and secondary outcome analyses are shown in Table 5. We found that after adjusting for covariates, the telehealth cohort was provided services with significantly lower RVUs (column 2) and Medicare payment amounts (column 3) than the in-person cohort (both p < 0.001). The third column indicates that, on average after adjusting for all covariates, the base Medicare payment amount for each encounter would have been $54.93, to which $24.79 would be added for the first encounter, and that telehealth encounters would have been paid $4.09 less (due to billing lower-RVU services) than in-person encounters.

Table 5.

Primary and Secondary Outcome Mixed Model Results with Random Effects at Patient Level for Relative Value Units per Encounter and for Medicare Payment Amounts per Encounter (n = 24,544 Encounters)

VARIABLE	RVU PER ENCOUNTER	PAYMENT PER ENCOUNTER
Telehealth	−0.17^*** [−0.21 to −0.15]	−4.09^*** [−5.74 to −2.43]
First encounter	0.62^*** [0.59 to 0.64]	24.79^*** [24.06 to 25.52]
Age group (65+ reference)
Age group ≥18	0.01 [−0.09 to 0.07]	−7.61^** [−11.14 to −4.08]
Age group 19–34	0.09^* [0.01 to 0.17]	2.33 [−1.01 to 5.67]
Age group 35–64	0.03 [−0.04 to 0.10]	1.37 [−1.71 to 4.45]
Sex (male as reference)
Female	−0.01 [−0.05 to 0.03]	0.91 [−0.71 to 2.53]
Race (White as reference)
Non-Whites	−0.09^* [−0.16 to −0.01]	2.12 [−1.15 to 5.40]
Unknown race	0.07^* [0.02 to 0.15]	2.33 [−0.54 to 5.21]
Ethnicity (Hispanic as reference)
Non-Hispanic	0.18^*** [0.08 to 0.28]	4.48^* [0.14 to 8.81]
Unknown ethnicity	0.28^*** [0.17 to 0.39]	7.55^** [2.97 to 12.13]
Insurance status (Medicare/dual as reference)
Medicaid	0.11^** [0.05 to 0.18]	8.35^*** [5.46 to 11.24]
Government/Veterans Affairs	0.02 [−0.12 to 0.16]	6.71^* [0.72 to 12.70]
Private	−0.03 [−0.09 to 0.03]	0.99 [−1.68 to 3.65]
Self pay/uninsured	−0.13^** [−0.22 to −0.04]	2.82 [−1.04 to 6.69]
Other	0.24^* [0.01 to 0.46]	7.48 [−2.36 to 17.33]
Unknown insurer	0.57^*** [0.48 to 0.67]	36.25^*** [32.28 to 40.22]
Primary diagnosis at first encounter
Mental disorders due to known physiological conditions (F01–F09)	0.55^*** [0.38 to 0.72]	16.20^*** [9.03 to 23.38]
Mental and behavioral disorders due to psychoactive substance use (F10–F19)	−0.15^* [−0.27 to −0.03]	−17.84^*** [−23.07 to −12.601]
Schizophrenia, schizotypal, delusional, and other nonmood psychotic disorders (F20–F29)	0.27^** [0.11 to 0.44]	19.78^*** [12.86 to 26.69]
Mood (affective) disorders (F30–F39)	0.36^*** [0.26 to 0.46]	18.97^*** [14.75 to 23.18]
Anxiety, dissociative, stress- related, somatoform and other nonpsychotic mental disorders (F40–F48)	0.30^*** [0.20 to 0.40]	12.50^*** [8.29 to 16.72]
Behavioral syndromes associated with physiological disturbances and physical factors (F50–F59)	0.31^* [0.04 to 0.57]	15.40^** [4.26 to 26.55]
Disorders of adult personality and behavior (F60–F69)	0.06 [−0.11 to 0.24]	14.51^*** [6.94 to 22.08]
Intellectual disabilities (F70–F79)	0.07 [−0.66 to 0.80]	−32.68 [−65.56 to 0.21]
Pervasive and specific developmental disorders (F80–F89)	0.61^*** [0.36 to 0.85]	34.09^*** [23.62 to 44.56]
Behavioral and emotional disorders with onset usually occurring in childhood and adolescence (F90–F98)	0.36^*** [0.24 to 0.49]	23.14^*** [17.82 to 28.46]
Constant	2.42^*** [2.27 to 2.56]	54.93^*** [48.60 to 61.25]

95% CIs in brackets ^* p < 0.05, ^** p < 0.01, ^*** p < 0.001.

CI, confidence interval.

Sensitivity analyses adjusting for Medicaid- and private insurance-estimated payment rates yielded results that were consistent with this model, as did sensitivity analyses limiting the data to the period during the COVID-19 public health emergency. Overwhelmingly, patients only received treatment using the modality to which they were assigned; however, 7.7% of patients did switch modality for at least one encounter. Sensitivity analyses excluding these patients did not change the results.

The results of our tertiary outcome analyses are shown in Table 6. In contrast to the encounter-level models shown above, the tertiary outcome analysis was at the patient level and used total payment amounts as the dependent measure factoring in the patient's number of encounters. After adjusting for covariates, the telehealth cohort had significantly lower (p < 0.001) total Medicare payment rates than the in-person cohort. These analyses suggest that an average rate for each encounter was $63.96, and that adjusting for all covariates, the telehealth cohort would receive Medicare payments $8.20 less for an episode of care, on average, than the in-person cohort. Sensitivity analyses with Medicaid and private insurance estimated rates yielded results that were consistent with this model, as did sensitivity analyses limiting the data to the period during the COVID-19 public health emergency and sensitivity analyses excluding patients with treatment outside their assigned cohort.

Table 6.

Tertiary Outcome Model with Covariates (n = 8,130 Patients) on Total Medicare Payment Amounts Combining Across Encounters

VARIABLE	TOTAL PAYMENT
Per telehealth encounter^a	−8.20^*** [−9.69 to −6.72]
Per encounter	63.96^*** [62.79 to 65.13]
Age group (65+ as reference)
Age group ≤18	−22.10^** [−35.91 to −8.29]
Age group 19–34	3.80 [−9.19 to 16.79]
Age group 35–64	6.46 [−5.50 to 18.41]
Sex (male as reference)
Female	5.32 [−1.03 to 11.67]
Race (White as reference)
Non-Whites	3.50 [−9.20 to 16.19]
Unknown race	−4.51 [−15.76 to 6.74]
Ethnicity (Hispanic as reference)
Non-Hispanic	2.06 [−14.84 to 18.96]
Unknown ethnicity	7.72 [−10.08 to 25.51]
Insurance status (Medicare/dual as reference)
Medicaid	7.12 [−4.22 to 18.46]
Government/VA	9.10 [−14.42 to 32.62]
Private	1.49 [−8.94 to 11.92]
Self-pay/uninsured	20.22^** [5.17 to 35.27]
Other	7.28 [−31.42 to 45.97]
Unknown insurer	−3.50 [−18.42 to 11.41]
Primary diagnosis at first encounter
Mental disorders due to known physiological conditions (F01–F09)	−13.34 [−40.72 to 14.03]
Mental and behavioral disorders due to psychoactive substance use (F10–F19)	−154.16^*** [−174.94 to −133.38]
Schizophrenia, schizotypal, delusional, and other nonmood psychotic disorders (F20–F29)	18.53 [−8.28 to 45.34]
Mood [affective] disorders (F30–F39)	22.62^** [6.38 to 38.85]
Anxiety, dissociative, stress- related, somatoform and other nonpsychotic mental disorders (F40–F48)	2.35 [−13.82 to 18.52]
Behavioral syndromes associated with physiological disturbances and physical factors (F50–F59)	3.96 [−39.21 to 47.12]
Disorders of adult personality and behavior (F60–F69)	−8.78 [−38.59 to 21.03]
Intellectual disabilities (F70–F79)	−63.71 [−184.29 to 56.87]
Pervasive and specific developmental disorders (F80–F89)	59.72^** [18.84 to 100.59]
Behavioral and emotional disorders with onset usually occurring in childhood and adolescence (F90–F98)	29.40^** [8.73 to 50.06]
Constant	51.42^*** [27.35 to 75.49]

95% CIs in brackets ^* p < 0.05, ^** p < 0.01, ^*** p < 0.001.

This is the interaction terms between telehealth and the number of encounters for each patient.

Discussion

The primary comparison in the analyses indicated that providers treating patients using telehealth did so at lower RVUs (i.e., lower provider work associated with time spent and case complexity), lower Medicare payment amounts, and lower total treatment episode payment amounts compared to the in-person treatment cohort. Other analyses help to interpret these differences. First, the two treatment cohorts both averaged just under three treatment episodes per patient, which did not differ significantly, indicating that the differences in RVUs and payment amounts were not due to differences in the number of treatment sessions.

Second, the pattern of RVUs indicates that the in-person cohort was more likely to receive 60-min individual therapy at a higher RVU and payment than the telehealth cohort. In contrast, the telehealth cohort was more likely to receive 30- to 45-min individual therapy or evaluation and management (E/M) services at a lower RVU and payment than the in-person cohort.

Third, a notable difference between the cohorts was the type of clinician providing the services. A much higher percentage of services were delivered by clinical social workers in the in-person cohort than in the telehealth cohort. In contrast, the telehealth cohort received services from a more diverse set of clinicians, including psychiatrists/other physicians, clinical psychologists, psychiatric/mental health advanced practice providers (APPs), and licensed professional counselors.

All, but one of these clinician categories has a Medicare payment rate that is higher than the payment rate for clinical social workers. There are also connections between clinician type and CPT code dictated by licensure requirements. For example, E/M codes are only within the scope of practice for physicians and APPs, leading to more E/M CPT codes for the telehealth cohort that had more encounters with psychiatrists and psychiatric/mental health APPs. Interestingly, the in-person cohort used clinicians who received lower Medicare payment amounts (which are adjusted by clinician type) than the telehealth cohort. Thus, the findings that payments per encounter and total episode payments were lower in the telehealth cohort cannot be explained by Medicare payment adjustments by clinician type, and instead are due to the use of services with lower RVUs.

Shorter encounter duration with telehealth visits has been observed in several applications, including pre-operative and post-operative visits.^18

–21 The application most similar to ours was cognitive behavioral therapy for insomnia²² where treatment fidelity, outcomes, and therapeutic alliance were similar, but telehealth sessions averaged 10 min shorter than in-person encounters. Greater efficiency in telehealth visits has been attributed to less small talk, getting down to work almost immediately, less clinician distraction, more directed and focused delivery, and eliminating time spent on rooming the patient and setup.^20,22,23 In these analyses, differences in the frequency of specific services and their associated billing times that the two cohorts received explain the higher RVUs in the in-person cohort than the telehealth cohort, which in turn help explain the differences in Medicare payment amounts, both per encounter and for the entire course of treatment episode.

The most important factor to consider when interpreting these findings is the context of this study. The data were pooled, coming from 17 OAT-funded grantees who were charged with delivering needed behavioral health services to patients living in rural and underserved areas. For data analysis purposes, the grantees were asked to collect data on a similar set of patients who received comparable behavioral health services in person. These are both convenience samples, and the research design relied on a pragmatic, usual care approach. Thus, these data reflect how telehealth and in-person behavioral health services are often delivered in a variety of care settings across a wide swath of rural regions of the U.S.

No attempt was made to systematize the services or specific therapeutic approaches, leaving treatment choices up to the grantees who were directed to serve their populations. Although the grantees used various therapeutic approaches, previous analyses of subsets of patients in this study indicated that those with depression, anxiety, or substance use showed improvement in clinical effectiveness as measured by the Patient Health Questionnaire (PHQ-9), Generalized Anxiety Disorder (GAD-7) scale, and Drug Use Disorders Identification Test (DUDIT-C), which did not differ significantly between the telehealth and in-person cohorts.^24,25 Thus the differences in services observed in the analyses described here did not result in differences in effectiveness, as measured by relevant clinical instruments. And those findings are consistent with multiple previous studies that reported similar effectiveness for behavioral health services delivered by telehealth or in person.^4

–7

Another factor to consider when interpreting the findings is the approach to assign value and payment amounts to the services provided. We report RVUs as a method to value the type of services provided because this is a standardized system that has been long used by CMS and private payers for establishing payment amounts. We also rely on the Medicare Physician Fee Schedule approach for assigning a payment amount to the RVUs. Although Medicare was the primary insurer for only a portion of the patients, this approach is transparent, which makes it popular in this type of analysis.

Medicare applies certain rules that adjust payment rates by the clinician type, and this adjustment affects the per-encounter and total episode payment amounts (but not the RVU models, which were consistent with the other outcome models). Other public payers (e.g., Medicaid) and private insurers use different approaches to assigning payment to RVUs and if those were known and applied, the findings might differ.

Limitations include that the study relied on grantees to provide usual care that met the needs of their treatment populations. In doing so, the use of convenience samples may have introduced potential bias resulting from both patient and clinician choice about treatment, which were not accounted for in this design. In addition, we relied on Medicare payment approaches to estimate the cost of treatment because actual payments by Medicare and other sources of insurance were not available in the dataset.

These analyses found modest, but statistically significantly lower RVUs, per-encounter Medicare payment, and total episode Medicare payment in the telehealth cohort compared to the in-person cohort who received behavioral health services. The findings add to previous cost studies that focused on avoided travel or equipment acquisition. This methodology relied on using CPT codes to identify RVUs and then attributing Medicare payment rates to those RVUs. Although the Medicare payment approach discounts payment for nonphysician providers and the in-person cohort used clinical social workers more frequently, the telehealth cohort still had lower payment rates.

Thus, the differences observed are due to the in-person cohort receiving services assigned higher RVUs than the telehealth cohort. Delivering 60-min individual therapy is a standard approach for in-person treatment, but alternative, less costly therapeutic approaches are often employed by telehealth practitioners. The grantees were able to involve a diverse set of more highly reimbursed clinicians for the telehealth cohort, but those clinicians provided services that may be considered more efficient, as captured by billed CPT codes. The large, multisite study increases generalizability, but of course, individual clinics using different clinicians and services may see different results.

Conclusions

These analyses compared telehealth and in-person cohorts who received behavioral health services that were delivered in a large multisite study of usual care treatment approaches. We found that behavioral health services provided using telehealth used services that had lower RVUs than behavioral health services provided in person. Using Medicare approaches to assign payment rates to RVUs, we found that the telehealth cohort had lower per-encounter payments and total episode payments than the in-person cohort. The differences observed are due to the in-person cohort receiving higher payment RVU services such as 60-min individual therapy than the telehealth cohort, which was more likely to receive briefer therapy sessions and other less expensive services from a more diverse pool of clinician type.

Footnotes

Disclosure Statement

No competing financial interests exist.

Funding Information

This study was supported by the OAT, HRSA, U.S. Department of Health and Human Services (HHS) to the RTRC under cooperative agreements no. UICRH29074 and no. U3GRH40003. The Evidence Based Tele-Behavioral Health Network Program provided funding for delivery of telebehavioral health to these grantees providing services with the following grant numbers: Athol Hospital/Heywood Healthcare G01RH32149, Avera Health G01RH32150, Baptist Health Corbin Foundation G01RH32151, CentraCare Health System G01-RH-32152, Greater Oregon Behavioral Health, Inc., G01RH32153, Indiana Rural Health Association G01RH32154, Lester E. Cox Medical Center G01RH32155, Primary Health Network G01RH32156, Texas A&M University G01RH32158, University of California, Davis G01RH32159, University of Kansas Hospital G01RH32160, University of Maryland G01RH32161, University of Minnesota G01RH32157, and West Virginia University Research Corporation G01RH32162. The Substance Abuse Treatment Telehealth Network Grant Program provided funding for delivery of tele-substance abuse treatment to these grantees providing services with the following grant numbers: Avera Health H1W-RH-31446, Union Hospital, Inc., H1W-RH-31447, and Westbrook Health Services, Inc., H1W-RH-31448. The information, conclusions, and opinions expressed are those of the authors and no endorsement by OAT, HRSA, or HHS is intended or should be inferred.

References

Larson

, Patterson

, Garberson

, et al. Supply and Distribution of the Behavioral Health Workforce in Rural America. WWAMI Rural Health Center: Seattle WA; 2016. Available from: https://familymedicine.uw.edu/rhrc/publications/supply-and-distribution-of-the-behavioral-health-workforce-in-rural-america/ [Last accessed: March 30, 2023 ].

Rural Health Information Hub. Rural mental health; 2021 [March 18, 2022]. Available from: https://www.ruralhealthinfo.org/topics/mental-health [Last accessed: August 2, 2022 ].

Kaiser Family Foundation. Mental health care health professional shortage areas (HPSAs); 2017. Available from: https://www.kff.org/other/stateindicator/mental-health-care-health-professional-shortage-areashpsas/?currentTimeframe=0&sortModel=%7B%22colId%22:%22 Location%22,%22sort%22:%22asc%22%7D [Last accessed: August 2, 2022 ].

Barnett

, Goulding

, Casetta

, et al. Implementation of telemental health services before COVID-19: Rapid umbrella review of systematic reviews. J Med Internet Res, 2021; 23(7):e26492; doi: 10.2196/26492

Hilty

, Ferrer

, Parish

, et al. The effectiveness of telemental health: A 2013 review. Telemed e-Health, 2013; 19(6):444–454.

Langarizadeh

, Tabatabaei

, Tavakot

, et al. Telemental health care, an effective alternative to conventional mental care: A systematic review. Acta Inform Med, 2017; 25(4):240–246.

Schaffer

, Nakrani

, Pirraglia

. Telemental health care: A review of efficacy and interventions. Telehealth Med Today, 2021; doi: 10.30953/tmt.v4.218

Ekeland

, Bowes

, Flottorp

. Effectiveness of telemedicine: A systematic review of reviews. Int J Med Inform, 2010; 79(11):736–771.

Snoswell

, Ghelberg

, De Guzmn

, et al. The clinical effectiveness of telehealth: A systematic review of meta-analyses from 2010 to 2019. J Telemed Telecare, 2021 [Online ahead of print]; doi: 10.1177/1357633X211022907

10.

Snoswell

, Taylor

, Comans

, et al. Determining if telehealth can reduce health system costs: Scoping review. J Med Internet Res, 2020; 22(10):e17298; doi: 10.2196/17298

11.

Naslund

, Mitchell

, Joshi

, et al. Economic evaluation and costs of telepsychiatry programmes: A systematic review. J Telemed Telecare, 2022; 28(5):311–330; doi: 10.1177/1357633x20938919

12.

AAPC. What are relative value units (RVUs)? Available from: https://www.aapc.com/practice-management/rvus.aspx

13.

Ward

, Fox

, Merchant

, et al. Measure and data element identification for the HRSA evidence-based tele-behavioral health network program and the HRSA substance abuse treatment telehealth network grant program. Rural Telehealth Research Center; April 2020. Available from: https://ruraltelehealth.org/research-and-policy-briefs.php [Last accessed: October 7, 2022 ].

14.

CMS.gov. Search the Physician Fee Schedule; 2022. Available from: https://www.cms.gov/medicare/physician-fee-schedule/search?Y=2&T=4&HT=0&CT=0&H1=96152&M=5 [Last accessed: August 2, 2022 ].

15.

CMS.gov. Medicare mental health. How to use the MPFS look-up tool; 2022. Available from: https://www.cms.gov/files/document/mln1986542-medicare-mental-health.pdf [Last accessed: August 2, 2022 ].

16.

Kaiser Family Foundation. Medicaid-to-Medicare fee index; 2022, June 22. Available from: https://www.kff.org/medicaid/state-indicator/medicaid-to-medicare-fee-index/?currentTimeframe=0&selectedDistributions=primary-care&sortModel=%7B%22colId%22:%22Location%22,%22sort%22:%22asc%22%7D [Last accessed: August 2, 2022 ].

17.

Kaiser Family Foundation. How much more than Medicare do private insurers pay? A review of the literature—Issue Brief; 2020, May 4. Available from: https://www.kff.org/report-section/how-much-more-than-medicare-do-private-insurers-pay-a-review-of-the-literature-issue-brief [Last accessed: August 2, 2022 ].

18.

Soegaard Ballester

, Scott

, Owei

, et al. Patient preference for time-saving telehealth postoperative visits after routine surgery in an urban setting. Surgery, 2018; 163:672–679.

19.

Mullen-Fortino

, Rising

, Duckworth

, et al. Presurgical assessment using telemedicine technology: Impact on efficiency, effectiveness, and patient experience of care. Telemed e-Health, 2019; 25(2):137–142.

20.

O'Donnell

, Haberli

, Muniz Martinez

, et al. Telehealth visits after shoulder surgery: Higher patient satisfaction and lower costs. JAAOS Glob Res Rev, 2022; 6:e22.00119; doi: 10.5435/JAAOSGlobal-D-22-00119

21.

Silva

, Delfossea

, Aceves-Martina

, et al. Telehealth: A novel approach for the treatment of nondisplaced pediatric elbow fractures. J Pediatr Orthop, 2019; 28:542–548.

22.

Arnedt

, Conroy

, Mooney

, et al. Telemedicine versus face-to-face delivery of cognitive behavioral therapy for insomnia: A randomized controlled noninferiority trial. Sleep J, 2021:1–11; doi: 10.1093/sleep/zsaa136

23.

Henry

, Block

, Ciesla

, et al. Clinician behaviors in telehealth care delivery: A systematic review. Adv in Health Sci Educ, 2017; 22:869–888.

24.

McCord

, Ullrich

, Merchant

KAS

, et al. Comparison of in-person vs. telebehavioral health outcomes from rural populations across America. BMC Psychiatry, 2022; 22(1):778; doi: 10.1186/s12888-022-04421-0

25.

Neufeld

, McCord

, Ullrich

, et al. Pragmatic comparative evaluation of telehealth vs. in-person services for substance use in rural America. In: Presented at the 3rd Edition of Global Conference on Addiction Medicine, Behavioral Health and Psychiatry. October 24–26, 2022. Orlando, FL.