Telerehabilitation During the COVID-19 Pandemic in Outpatient Rehabilitation Settings: A Descriptive Study

Abstract

Objective

COVID-19 has widely affected delivery of health care. In response, telerehabilitation (TR) has emerged as alternative care model. Aims were: (1) to describe baseline patient characteristics and available unadjusted outcomes for episodes of care administered during COVID-19 using TR versus traditional in-person care, and (2) to describe TR frequency levels by condition and telecommunication modes.

Methods

A descriptive retrospective observational design was used to report patient variables and outcomes including physical function, number of visits, and patient satisfaction, by TR frequency (few, most, or all visits) and telecommunication modes. Standardized differences were used to compare baseline characteristics between episodes with and without TR.

Results

Sample consisted of 222,680 patients (59% female; mean [SD] age = 55 [18] years). Overall TR rate was 6% decreasing from 10% to 5% between second and third quarters of 2020. Outcome measures were available for 90% to 100% of episodes. Thirty-seven percent of clinicians administered care via TR. Patients treated using TR compared with in-person care were more likely to be younger and live in large metropolitan areas. From those with TR, 55%, 20%, and 25% had TR during few, most, or all visits, respectively. TR care was administered equally across orthopedic body parts, with lower use for nonorthopedic conditions such as stroke, edema, and vestibular dysfunction. TR was primarily administered using synchronous (video or audio) modes. The rate of patients reported being very satisfied with their treatment results was 3% higher for no TR compared with TR.

Conclusions

These results provide new knowledge about to whom and how TR is being administered during the pandemic in outpatient rehabilitation practices throughout the United States. The database assessed was found to be suitable for conducting studies on associations between TR and diverse outcome measures, controlling for a comprehensive set of patient characteristics, to advance best TR care models, and promote high-quality care.

Impact

This study provided detailed and robust descriptive information using an existing national patient database containing patient health and demographic characteristics, outcome measures, and telerehabilitation (TR) administration data. Findings support the feasibility to conduct future studies on associations between TR care and patient outcomes, adjusting for a wide range of patient characteristics and clinical setting factors that may be associated with the probability of receiving TR. The finding of limited and decreasing use of TR over the study period calls for studies aimed to better understand facilitators and inhibitors of TR use by rehabilitation therapists during everyday practice to promote its use when clinically appropriate.

Administration, Descriptive Data, Occupational Therapy, Physical Therapy, Telerehabilitation

Introduction

The coronavirus disease 2019 (COVID-19) pandemic has widely affected all aspects of society and impacted delivery of physical therapy and other health care services internationally as well as in the United States.^1–4 In response, telerehabilitation (TR) has emerged as a promising alternative model to traditional in-person clinical visits. TR refers to clinical services administered at a distance using telecommunication.⁵ TR can be delivered using a variety of telecommunication media technologies with either real-time (synchronous) 2-way interactive mediums such as video and/or audio calls or asynchronous E-visits not in real time, for example, virtual check-ins, remote evaluations of recorded videos, or applications/links to exercises and educational materials.^4,6

Evidence exists on benefits and patient acceptance of TR care for patients with a wide variety of conditions including orthopedic (eg, low back, total joint arthroplasty),^5–8 neurological (eg, stroke, multiple sclerosis),^9–11 and widespread chronic pain syndromes (eg, fibromyalgia, rheumatoid arthritis).^12–14 Preliminary findings indicate that care delivered via TR in addition to or as replacement of in-person clinic visits was generally either equivalent to or yielded slightly better outcomes compared with usual in-person physical therapy care alone.^5,15 However, many authors recommend caution to avoid generalization and overestimation of these findings given methodological weaknesses in available studies, the heterogeneous nature of patient characteristics, variability in clinical conditions, and small sample sizes.^6,12,16,17

Evidence supporting the benefits and effectiveness of TR was mainly published prior to COVID-19. Since the onset of COVID-19, many state and federal regulatory and reimbursement policies were implemented to enhance the administration of TR care by rehabilitation therapy specialists.^18,19 No studies of TR using established large national patient databases that examine the actual implementation and administration of TR care in typical outpatient hospital and private practice physical therapy clinics across the United States were identified. Of interest, Miller et al² evaluated telehealth physical therapy implementation at the beginning of the pandemic (March 16 to May 16, 2020) and found that implementation of telehealth physical therapy during COVID-19 was feasible and acceptable by patients and physical therapists. However, the study was conducted within 1 large urban academic medical center and results may not be generalizable.

TR services are likely to remain a standard mode for administration of care and represent a new normal for rehabilitation therapy practice during and after COVID-19.⁴ Therefore, continued TR research is required to explore how delivery of care has evolved in rehabilitation therapy practices in the United States as a result of COVID-19 and to inform future TR rehabilitation practices. Descriptive studies are needed to identify to whom and how TR care is being administered. To address these needs, Prvu Bettger and Resnik⁴ recently recommended rapid-cycle research, using large and existing patient database systems, to provide timely clinical insights about how TR care has affected rehabilitation therapy practice. Examining TR data documented by clinicians working in everyday clinical practice is recommended to best understand how COVID-19 has impacted rehabilitation therapy care models and to translate findings to identify best clinical practices using TR.^4,20

Our aims were to (1) describe baseline patient demographic and health characteristics and available unadjusted outcomes, for episodes of care administered using TR versus traditional in-person care documented during COVID-19, and (2) describe the TR frequency levels by conditions and by TR telecommunication modes.

Methods

Design and Data Collection

This was a descriptive study using retrospective, observational data from a large national patient database system collected routinely in outpatient rehabilitation therapy clinics in the United States. Data included diverse patient characteristics and standardized documentation of TR use in outpatient clinics throughout all 50 states. TR use documentation was standardized. The study was approved by Solutions IRB, a private institutional review board located in Yarnell, Arizona. Participating clinics routinely collect patient demographics, health characteristics, and outcomes using the Patient Inquiry software developed by Focus on Therapeutic Outcomes (FOTO),²¹ a Net Health company that provides outcomes management software solutions for rehabilitation therapists. Patients aged 14 to 89 years were included if their episode of care started no earlier than the fourth quarter 2019 and they were discharged from rehabilitation therapy care during the second quarter 2020 (May 1 to June 30) or third quarter 2020 (July 1 to September 30).

Telerehabilitation

TR data were collected using the following survey question: “How many of your current therapy visits have taken place over the internet or by phone (telehealth) instead of in the clinic?” Patient response categories were: none, few, most, or all. Response categories were defined as: none—when no visits used TR during the episode of care; few—when less than half of the total episode visits used TR; most—when half or more of total episode visits, but not all visits, used TR; and all—when all visits during the episode of care were administered using TR. For patients to accurately document the TR frequency used during his or her episode of care, the TR frequency data used for analyses were obtained from the patient’s discharge survey. This question was administered starting April 30, 2020. Subsequently on August 4, 2020 if the patient responded that TR was administered during the episode of care, a second patient-facing TR question was added to each follow-up FOTO survey: “Which of these was used in your telehealth care? (select all that apply).” Patient responses were: video call, audio call (without video), text or email messaging, links to video materials (like YouTube clips), and other. Patients were subsequently classified into 3 communication modes: (1) synchronous for patient responses video and/or audio call, (2) asynchronous for patient responses text or email messaging, links to video materials (like YouTube clips), and other, but no use of video and/or audio modes, and (3) mixed if the episode of care included both synchronous and asynchronous telecommunication modes.

Outcomes

The outcomes described were physical function (PF) change, number of treatment visits during the episode of care from intake to discharge, and patient satisfaction with treatment results at discharge. PF was assessed at intake and discharge using a set of patient-reported outcome measures developed using item response theory.^22–25 Measure administration mode was through computerized adaptive tests described previously in detail.^26–31 The item response theory model calibrated the PF scores into a linear metric from 0 (low) to 100 (high) functioning. Number of visits was used as a proxy to describe direct costs and health care usage incurred by TR use as recommended in a systematic review by van der Meij.³² Data on patient satisfaction with treatment results were collected using a question that was administered on every follow-up patient survey: “How satisfied were you with overall results of your treatment at this facility?” Patient response categories were: very satisfied, somewhat satisfied, neither satisfied nor dissatisfied, somewhat dissatisfied, or very dissatisfied.

Data Analyses

To address the first aim, standardized difference analytical methods were used to determine differences in baseline characteristics between those episodes with TR and those without TR. Standardized differences were calculated to compare means of continuous variables and prevalence of dichotomous variables as recommended by Austin.³³ Briefly for continuous variables the standardized difference was defined as:

$$\begin{align}& Standardized\ Difference=\frac{\mid{\overline{x}}_{group\ 1}-{\overline{x}}_{group\ 2}\mid }{SD_{pooled}} \end{align}$$

(1)

where |${\overline{x}}_{group\ 1}$| denotes the mean of the covariate in each group, and |${SD}_{pooled}$| denotes the full sample SD.

For dichotomous variables the standardized difference was defined as:

$$\begin{align}& Standardized\ Difference= \nonumber \\ & \frac{\mid{\hat{p}}_{group\ 1}-{\hat{p}}_{group\ 2}\mid }{\sqrt{\frac{{\hat{p}}_{group\ 1}\left(1-{\hat{p}}_{group\ 1}\right)+{\hat{p}}_{group\ 2}\left(1-{\hat{p}}_{group\ 2}\right)}{2}}} \end{align}$$

(2)

where |${\hat{p}}_{group}$| denotes the prevalence or mean of the dichotomous variable in each group.

Unlike P-values, standardized difference analyses are not influenced by sample size, and can be interpreted as an effect size, with values of 0.2, 0.5, and 0.8 proposed previously to represent thresholds of small, medium, and large effect sizes, respectively.³⁴ Standardized difference values less than 0.1 were suggested to represent clinically negligible differences.³³

For our second aim, we calculated the standardized difference between 2 orthopedic body parts that had the highest and lowest rates of TR use, allowing us to infer if TR was equally administered between all orthopedic body parts. Additionally, we calculated percentages for TR frequency levels, that is, few, most, or all, by telecommunication technology modes, that is, synchronous, asynchronous, and both (mixed) modes.

Results

Patient Sample

Our sample consisted of 222,680 episodes of care (59% female; mean age in years [SD], 25th, 75th percentile = 55 [18], 43, 69; age range = 14–89 years). Of those, 13,059 (6%) episodes incorporated some level of TR. Of interest, when percentages of episodes involving TR were compared between second quarter (May–June 2020) versus third quarter (July to September 2020), a higher percentage of episodes involving TR was observed during the second quarter (10%) versus the third quarter (5%). Data were contributed by 13,240 clinicians working in 3045 outpatient rehabilitation clinics located in all 50 US states. Of those, 37% of clinicians and 69% of clinics located in 49 states implemented and administered care using TR.

Patients With and Without TR

Baseline patient characteristics were compared between episodes with and without TR (Tab. 1). The standardized difference values suggest that many of the patient variable differences between the TR and no TR subgroups are not meaningful, that is, the standardized difference is less than 0.1. However, the standardized differences for 10 patient variables were considered important, with values greater than 0.1. For example, for those patients treated using TR compared with traditional in-person visit care: mean age was 4 years lower (standardized difference = 0.21), the rate of patients treated in private practice settings was 7% higher (standardized difference = 0.16), the rate of metropolitan core, a Rural-Urban Commuting Area classification^35,36 category, was 10% higher (standardized difference = 0.25), Medicare B age 65 or above was 8% lower (standardized difference = 0.19), exercise 3 times/wk was 5% higher (standardized difference = 0.10), arthritis was 7% lower (standardized difference = 0.14), high blood pressure was 7% lower (standardized difference = 0.14), and obesity was 6% lower (standardized difference = 0.12).

Table 1

Baseline Health and Demographic Patient Characteristics for the Full Sample and the Samples Not Using or Using Telerehabilitation (TR)^a

Baseline Characteristics	Total Sample (N = 222,680)	Sample Not Using TR (n = 209,621)	Sample Using TR (n = 13,059)	Standardized Differenceb
Number of providers
Clinicians	13,240	12,878	4943	−
Clinics	3045	3012	2096
States	50	50	49
Physical function score at intake
Mean [SD]	48.5 [15.5]	8.6 [15.5]	48.0 [16.0]	0.04
Median (25th; 75th percentiles)	49.0 (39.0; 58.6)	49.0 (39.1; 58.6)	48.5 (38.0; 58.6)
Age, y
Mean [SD]	54.9 [18.2]	55.1 [18.2]	51.3 [18.1]	0.21
Median (25th; 75th percentiles)	58 (43; 69)	58 (43; 69)	53 (37; 66)
Age groups (column %)
14 to <18	3.2	3.2	4.0	0.04
18 to <45	23.9	23.5	30.8	0.16
45 to <65	36.6	36.5	37.2	0.01
65 to <75	23.4	23.7	19.4	0.11
75 to 89	12.8	13.1	8.7	0.14
Sex: female, %	59	58	62	0.07
Practice type (column %)
Hospital outpatient dept	28.3	28.8	21.5	0.17
Physical therapist private practice	70.6	70.2	77.3	0.16
Other	1.1	1.1	1.2	0.01
Rural–Urban Commuting Area (RUCA) (column %)
Metropolitan core (primary flow within an urbanized area; population density >50,000)	80.0	79.4	88.6	0.25
Metropolitan (primary flow to an urbanized area)	5.3	5.4	3.2	0.11
Micropolitan (primary flow within or to a large urban cluster: population density 10,000–50,000)	9.0	9.2	5.4	0.15
Small town (primary flow within or to a small urban cluster: population density 2500–9999)	4.2	4.3	2.0	0.13
Rural areas	1.6	1.7	0.7	0.09
Acuity (column %)
0–7 d	5.4	5.5	5.2	0.01
8–14 d	6.5	6.5	6.7	0.01
15–21 d	8.7	8.7	8.4	0.01
22–90 d	27.7	27.5	29.5	0.04
91 d to 6 mo	15.7	15.7	16.0	0.01
Over 6 mo	36.0	36.1	34.2	0.04
Payer (column %)
HMO, preferred provider	46.5	46.2	50.9	0.09
Medicare B age 65 or above	22.6	23.0	15.4	0.19
Workers’ compensation	6.9	6.8	8.3	0.06
Medicaid	4.9	4.9	4.6	0.01
Indemnity insurance	3.7	3.8	3.6	0.01
Medicare B under age 65	2.4	2.5	2.2	0.02
Medicare A	1.3	1.3	1.2	0.01
No fault, auto insurance	1.1	1.1	1.6	0.04
Patient	0.6	0.6	0.6	0.00
Other (litigation, Medicare C, school, no charge, early intervention, commercial insurance)	9.9	9.8	11.6	0.06
Surgical history (column %)
No related surgery	69.7	69.8	68.6	0.03
1 related surgery	23.0	23.0	24.1	0.03
2 related surgeries	4.6	4.6	4.8	0.01
3 or more related surgeries	2.6	2.6	2.5	0.01
Postsurgical procedures, %	15.3	15.3	15.8	0.01
Exercise history (column %)
At least 3×/wk	42.1	41.8	46.8	0.10
1–2×/wk	23.7	23.7	24.3	0.01
Seldom or never	34.1	34.5	28.9	0.12
Medication use at intake, %	58.8	58.7	60.4	0.04
Previous treatment, %	62.1	62.2	60.5	0.03
Number of comorbidities
Mean [SD]	4.4 [3.2]	4.4 [3.2]	4.1 [3.1]	0.11
Median (25th; 75th percentiles)	4 (2; 6)	4 (2; 6)	3 (2; 6)
Specific comorbidities, %
Allergy	26.2	26.1	27.6	0.03
Angina	0.9	0.9	0.8	0.02
Anxiety or panic disorders	16.3	16.2	18.0	0.05
Arthritis	42.6	43.0	36.1	0.14
Asthma	10.7	10.6	11.9	0.04
Back pain (neck pain, low back pain, degenerative disc disease)	52.9	53.0	51.4	0.03
Cancer	8.5	8.6	7.3	0.05
Chronic obstructive pulmonary disease (COPD)	3.3	3.4	2.5	0.05
Congestive heart failure	4.5	4.6	3.4	0.06
Depression	16.2	16.2	17.1	0.03
Diabetes type 1 or 2	13.4	13.5	11.4	0.06
Gastrointestinal	14.8	14.8	14.0	0.02
Headaches	19.8	19.7	21.4	0.04
Hearing	5.7	5.8	4.1	0.08
Hepatitis/HIV-AIDS	0.8	0.8	0.8	0.00
High blood pressure	36.4	36.8	30.2	0.14
Heart attack (myocardial infarction)	2.5	2.5	1.9	0.04
Incontinence	5.5	5.5	5.1	0.02
Kidney, bladder, prostate, or urination problems	9.2	9.3	7.9	0.05
Neurological disease	1.7	1.7	1.7	0.00
Obesity (BMI ≥30)	40.7	41.0	35.4	0.12
Osteoporosis	8.2	8.3	7.6	0.02
Other disorders	3.0	3.0	3.3	0.02
Peripheral vascular disease (or claudication)	1.5	1.5	1.1	0.03
Previous accidents (motor vehicle, work, or other accident)	11.6	11.6	12.6	0.03
Previous surgery	39.9	40.2	36.1	0.08
Prosthesis/implants	8.8	8.9	6.7	0.08
Sleep dysfunction	15.6	15.6	15.8	0.00
Stroke or transient ischemic attack	3.9	4.0	3.0	0.06
Visual impairment	8.8	8.9	6.8	0.08
Pacemaker	1.5	1.5	1.2	0.03
Seizures	1.4	1.4	1.4	0.00
Condition type (column %)
Low back	21.1	21.1	21.0	0.00
Shoulder	18.1	18.0	20.0	0.05
Knee	17.8	17.8	17.0	0.02
Foot and ankle	9.6	9.6	9.7	0.00
Hip	8.8	8.9	8.3	0.02
Neck	8.5	8.6	8.2	0.01
Elbow, wrist, hand	8.2	8.1	9.0	0.03
Thoracic	1.7	1.7	2.0	0.02
Vestibularc	1.7	1.8	1.1	0.05
Other	4.3	4.4	3.7	0.00

^aValues are percentages unless noted otherwise. Column % sum may not be exactly 100% due to rounding. BMI = body mass index; HMO = health maintenance organization; TR = telerehabilitation.

^bStandardized differences between those with or without TR; values highlighted in bold represent an important difference. Variables were included for those variables that had a frequency or TR use of at least 1%.

^cVestibular is the only nonorthopedic condition included due to ≥1% threshold; “Other” includes neurological and lymphedema conditions with <1% frequency, eg, stroke, edema.

Outcomes

Unadjusted patient outcomes at discharge for the full sample and the samples using or not using TR for PF change, number of visits, and patient satisfaction are presented in Table 2. Briefly, PF change varied depending on condition and TR use, ranging between 2 and 23 points. Of interest, patients experiencing thoracic, vertigo, stroke lower extremity, and lower and upper extremity edema conditions who received TR reported 2.2, 2.9, 3.9, 6.5, and 9.1 unadjusted PF points less at discharge than those not receiving TR, respectively. Patients with stroke upper extremity impairments who received TR reported 7.1 unadjusted PF points more at discharge than those not receiving TR. Number of visits at discharge had a mean of 13 (SD = 9, median = 11). Patient satisfaction ratings were available for 90% of the sample. The rate of patients reported being very satisfied with their treatment results was 3% higher for no TR than TR.

Table 2

Unadjusted Patient Outcomes at Discharge for the Full Sample and the Samples Not Using or Using Telerehabilitation (TR)^a

Impairment Type	Total Sample (N = 222,680)	Sample Not Using TR (n = 209,621)	Sample Using TR (n = 13,059)
Unadjusted physical function change, mean [SD]
Low back	16.1 [15.9]	16.2 [15.8]	15.8 [16.6]
Shoulder	19.2 [17.5]	19.1 [17.5]	20.0 [18.2]
Knee	22.1 [17.8]	22.1 [17.7]	23.1 [18.4]
Foot and ankle	19.5 [16.6]	19.5 [16.6]	20.7 [17.3]
Hip	17.8 [16.8]	17.8 [16.7]	17.9 [17.4]
Neck	14.0 [14.8]	14.0 [14.8]	14.1 [14.8]
EWH	19.4 [16.4]	19.4 [16.3]	19.7 [17.1]
Thoracic	15.9 [16.7]	16.0 [16.7]	13.8 [16.0]
Vertigo	19.6 [21.1]	19.7 [21.1]	16.8 [21.1]
Stroke lower extremity	12.9 [14.7]	13.0 [14.8]	9.1 [10.6]
Stroke upper extremity	12.3 [14.6]	12.0 [14.6]	19.1 [13.3]
Lower quadrant edema	8.7 [11.9]	8.8 [11.8]	2.3 [14.4]
Upper quadrant edema	13.8 [15.1]	13.9 [15.1]	4.9 [9.7]
Other	11.4 [15.1]	11.5 [15.2]	9.3 [13.6]
Number of visits
Mean [SD]	13.1 [8.5]	13.0 [8.4]	14.2 [10.1]
Median (25th; 75th percentiles)	11 (7; 17)	11 (7; 17)	12 (6; 19)
Satisfaction data with overall results (column %)b
Very satisfied	80.2	80.4	77.0
Somewhat satisfied	7.6	7.5	9.2
Neither satisfied nor dissatisfied	1.6	1.6	2.1
Somewhat dissatisfied	0.3	0.3	0.5
Very dissatisfied	0.1	0.1	0.1
Missing satisfaction data	10.1	10.1	11.1

^aEWH = elbow wrist hand.

^bPatient satisfaction data were available for 89.9% (n = 200,092) of the sample.

TR Frequency and Mode

Percentages of TR use by frequency levels and body part or care type are presented in Table 3. From those patients that received TR, 55%, 20%, and 25% had TR during few, most, or all visits, respectively. To determine if TR was equally administered between all orthopedic body parts, we calculated the standardized difference comparing the rate of no TR between hip, which had the highest rate of no TR (94.5%), and thoracic, which had the lowest percentage of no TR (93.4%). The standardized difference was trivial (0.05), inferring that TR was equally administered between all orthopedic body parts. We also observed TR administration for vestibular, stroke, edema, and other conditions; however, this sample was small (n = 638) compared with the orthopedic sample that received TR (n = 12,421). TR was used less frequently for nonorthopedic conditions.

Table 3

Telerehabilitation Frequency Levels by Body Part or Care Type^a

Body Part or Care Type Conditions	None	Few	Most	All	Total n (100%)
Low back	44,264 (94.2)	1440 (3.1)	557 (1.2)	742 (1.6)	47,003
Shoulder	37,793 (93.5)	1501 (3.7)	520 (1.3)	589 (1.5)	40,403
Knee	37,404 (94.4)	1267 (3.2)	420 (1.1)	539 (1.4)	39,630
Foot and ankle	20,126 (94.1)	683 (3.2)	282 (1.3)	305 (1.4)	21,396
Hip	18,557 (94.5)	592 (3.0)	212 (1.1)	281 (1.4)	19,642
Neck	17,946 (94.4)	573 (3.0)	205 (1.1)	288 (1.5)	19,012
EWH	17,081 (93.6)	592 (3.2)	249 (1.4)	328 (1.8)	18,250
Thoracic	3615 (93.4)	145 (3.7)	48 (1.2)	63 (1.6)	3871
Vertigo	3693 (96.1)	85 (2.2)	30 (0.8)	35 (0.9)	3843
Stroke lower extremity	673 (96.4)	15 (2.1)	3 (0.4)	7 (1.0)	698
Stroke upper extremity	390 (95.6)	13 (3.2)	3 (0.7)	2 (0.5)	408
Lower quadrant edema	501 (97.9)	10 (2.0)	−	1 (0.2)	512
Upper quadrant edema	339 (98.5)	3 (0.9)	−	2 (0.6)	344
Other	7239 (94.4)	240 (3.1)	90 (1.2)	99 (1.3)	7668
Total	209,621 (94.1)	7159 (3.2)	2619 (1.2)	3281 (1.5)	222,680

^aValues are n (row %). EWH = elbow wrist hand.

There were 2634 episodes of care that included the second TR survey question regarding type of telecommunication technology modes. Most episodes with TR included a video call communication mode (71%), followed by 18%, 15%, 14%, and 12%, for text messaging, educational links, other modes, and audio call, respectively. The percentages per TR mode groups were 60%, 21%, and 19% for synchronous, asynchronous, and mixed, respectively (Tab. 4). The interaction between TR mode and TR intensity levels varied. For few TR frequency level, synchronous mode was used more frequently (69%) compared with most (53%) or all (46%) TR frequency levels, yet asynchronous mode was used more frequently for all and most (26%–27%) versus few (16%) TR frequency levels.

Table 4

Telerehabilitation (TR) Frequency Levels by Telecommunication Delivery Modes^a

TR Telecommunication Delivery Mode	Few	Most	All	Total
Synchronous mode (video or audio)b	966 (69)	247 (53)	359 (46)	1572 (60)
Asynchronous mode (no video or audio)c	229 (16)	122 (26)	211 (27)	562 (21)
Mixed modesd	199 (14)	95 (20)	206 (27)	500 (19)
Total n (100%)	1394	464	776	2634

^aValues are n (column %), and may sum to 99%–101% due to rounding of decimal points.

^bSynchronous mode (video and/or audio calla): 2-way interactive medium where both patients and clinicians exchange information instantaneously during the same time period using either video or audio calls.

^cAsynchronous mode (no video or audio calls): e-visits not in real time, eg, text messaging or applications/links to educational materials.

^dMixed modes: both synchronous and asynchronous mediums are used during the same episode of care.

Discussion

Our study examined a very large patient cohort contributed by multitudes of frontline clinicians across 50 US states investigating rehabilitation care models administering TR compared with in-person visits during the COVID-19 pandemic. The national database included hundreds of thousands of complete episodes of rehabilitation care across a breadth of Rural-Urban Commuting Area classification categories^35,36 representing diverse and robust patient demographics, conditions treated, as well as medical and other health-related variables.

Major Findings

The study’s major findings were: (1) only 6% of episodes of care in our sample incorporated some level of TR provided by 37% of clinicians; (2) TR was more likely to be administered during the second quarter 2020 (10%) compared with the third quarter 2020 (5%); (3) meaningful differences in some patient health and demographic characteristics were observed between TR and no TR subgroups; (4) TR frequency levels varied from 55%, 20%, and 25% for few, most, or all visits, respectively; (5) any TR use was equally administered across orthopedic body parts, with lower use for conditions of stroke, upper or lower quadrant edema, and vestibular dysfunction; and (6) percentages per TR technology modes were 60%, 21%, and 19% for synchronous, asynchronous, and mixed modes, respectively.

Implications for Practice

The low TR administration rate by service providers in our study contrasts with recent studies consistently recommending a rapid adoption and implementation of TR in replacement of or in addition to in-person rehabilitation clinic visits since the start of the COVID-19 pandemic.^2,4,20,37,38 The recommendations supporting TR’s outcome effectiveness, reduction in patient indirect costs, and acceptability by providers and patients were based on evidence published prior to the COVID-19 pandemic. Since the pandemic onset, we are aware of only 2 studies examining actual TR feasibility and outcomes.^2,39 For instance, Negrini et al³⁹ demonstrated that a complete shift from traditional in-person clinic care to telehealth was feasible and acceptable during the COVID-19 pandemic. In another recent study, the authors observed 4548 physical therapy sessions provided by 40 therapists, of which 85% were administered using telehealth, and all participating physical therapists conducted at least 1 telehealth session indicating 100% adoption.² Both of these studies, however, were limited to 1 medical setting and their results may not be generalizable to other medical contexts.

Many of the patient health and demographic characteristics between the TR and no TR subgroups were meaningfully different, which is consistent with previous research suggesting that clinicians prefer to use TR for certain patients but not others. Previous research, however, did not use standardized difference analyses to determine if reported inequalities or differences observed between patient variables in TR and comparison groups were clinically meaningful.^2,40 For example, Miller et al² using P-value statistics reported that during their study’s TR implementation phase a greater proportion of patients were younger, primarily English speaking, and had fewer medical comorbidities versus the comparison phase. As previously reported, we also observed important differences in patient characteristics between those receiving and not receiving TR. For instance, patients receiving TR were younger, more likely to be commercially insured versus Medicare, and to reside in large metropolitan areas. The standardized differences for these variables were greater than 0.1 suggesting groups were not similar enough to allow comparisons of outcome of care that would take into account the probability of different patients to receive TR. Therefore, when interpreting associations between patients receiving TR versus no TR and health outcomes, rigorous statistical methods have been recommended.^4,32 For instance, propensity score matching has been identified as an important analytic approach to balance or control for differences between the comparison groups (eg, with or without TR) that may be a result of their different probabilities of receiving the treatment under investigation (eg, TR), which may confound the outcome of interest.^4,41 Future studies applying such adjustments are needed to identify the treatment effect of TR and specific subgroups of patients that might or might not benefit from specific technology modes, dosage, and frequency levels of TR administration.

We also observed that clinicians administered TR across different frequency levels for either a few, most, or all visits during the episode of care. Some level of TR was used in addition to or replacement of in-person clinic visits in 75% and 25% of episodes, respectively. This suggests that TR use is not a one-size-fits-all approach, but delivery of TR is flexible and can be tailored to individual patients’ needs.

Prior studies investigating TR use in clinical practice examined patient samples with specific orthopedic conditions, mainly low back, knee, or hip complaints^12,17 and to a lesser extent upper limb and ankle foot impairments.^5,15,42 It is not known if clinicians during routine practice administer TR to all patients referred to physical therapy outpatient services regardless of condition type. Some authors speculate that only specific patient populations, for example, acute versus chronic spinal pain, benefit from TR services.⁸ Our findings suggest that TR was administered equally across a wide variety of typical orthopedic conditions commonly managed by clinicians in outpatient rehabilitation settings. In addition, we also observed that TR was administered to other care types such as vertigo, stroke, and edema, but these sample sizes were small, thereby limiting our interpretations regarding the descriptive results for nonorthopedic impairments. Our findings support the feasibility to administer at least some level of TR during everyday outpatient services to patients treated with conditions assessed in this study.

We observed that synchronous modes (2-way real-time communication) were used in 60% of episodes with TR, 21% administered TR asynchronously (1-way E-visit communication not in real time), and 19% used both modes during the episode of care. Although clinicians primarily used synchronous TR modes, asynchronous mode was used by many physical therapists during the patient’s episode of care. Medicare B as well as certain private payers pay clinicians for TR care administered using either mode separately or together.⁴³ Although the percentage of asynchronous use was lower compared with the synchronous mode, asynchronous TR modes were reported useful for clinicians to follow up with patients via recorded videos, secure messages, or with written materials to maximize efficiency and outcomes.⁴⁴ There is also some evidence that asynchronous mode can be effective for patients’ status post total joint replacement surgery.⁴⁵

For synchronous mode, video call (71%) was the preferred communication medium. This is consistent with previous findings reporting physical therapists favored video technologies rather than other mediums such as the telephone to deliver care.^46,47 Despite the preference for communication using video mediums to deliver TR, prior studies reported that audio modes such as telephone-supported or telephone coaching programs may be a more practical alternative to real-time video telecommunication.^40,45 Recent evidence showed little differences between these 2 modalities in terms of patient outcomes, and more research was recommended to determine the circumstances under which video medium is superior to telephone as a telehealth modality.⁴⁸ In another recent study, the authors reported that 75% of patients aged 65 or over operated TR telecommunication mediums independently and concluded that TR was feasible in adults of all ages.³⁸ We recommend selecting TR communication technologies that are most user-friendly and capable of expanding access of rehabilitation services to all of our patients. We recommend future studies examining the optimal interactions between TR delivery modes and frequency levels to achieve best patient outcomes.

We did not anticipate the low adoption and implementation of TR by providers and the sharp reduction in TR administration between the second and third quarters of 2020, given recent reports that telehealth is rapidly being implemented in light of the COVID-19 pandemic.⁴⁴ We speculate that the reduction in TR use observed may be primarily explained by: (1) the easing of stay-at-home and mandatory restrictions during the third quarter 2020, and (2) subsequent easing of pandemic restrictions decreased some of the motivation of both patients and providers to continue using alternative TR care services, favoring the familiarity of an in-person approach. One possible explanation for the decrease in TR administration may be patient dissatisfaction with TR compared with in-person care. The unadjusted results indicate that it is plausible that patients were less likely to be very satisfied with TR and that patients with specific conditions may have had less improvement than those treated in-person. However, due to the nature of this descriptive study and no adjustment as to the probability of being treated using TR, further research is needed to study associations between TR use and patient satisfaction or other patient outcomes while controlling for potential confounders of the outcomes assessed. Implementation of TR faces many challenges, which may also explain the overall low TR adoption by clinicians in our study.^17,49 For instance, findings suggest that clinicians have been reluctant to consider TR because of the impossibility of using palpation as well as certain manual techniques and diagnostic tests, and the lack of rehabilitation equipment commonly used in outpatient clinics.¹⁷

Recent research is providing clear clinical guidance on best strategies necessary to successfully implement TR.^4,37,44,49 Emerging implementation strategies include a well-designed administrative plan within the organization backed with financial and technological support, provider education on choice of platform, legal and ethical considerations, implications for best treatment and examination choices, management processes, and timely recognition and feedback on adoption and implementation progress.^44,49,50 A systematic review and a qualitative study identified that the organization of the care model was the most important strategy in determining the value and implementation of TR service.^51,52 This organizational strategy was supported by 2 recent articles touting that the smooth transition to TR care observed within their facility followed a rigorous administrative and clinical TR plan of care, which we speculate accounted for their successful TR implementation.^2,39 Continued research exploring best methods and strategies to enhance widespread implementation of TR care across a broad spectrum of outpatient rehabilitation clinics from rural to metropolitan areas within the United States is warranted.

Limitations

Caution is recommended to avoid overinterpretation of our findings. First, due to the nature of this descriptive design and no risk adjustment, interpretation of PF, visits, and patient satisfaction outcome results for TR and no TR subgroups is limited. The focus of future research is required to: (1) study associations between TR use and patient satisfaction and improvement in physical function outcomes while controlling for potential confounders of the outcomes assessed, and (2) determine the value of TR care in physical therapy. Second, our descriptive data may not be generalizable to the overall outpatient rehabilitation population in the United States. Although this was not the purpose of this study, testing for the generalizability of the national database analyzed merits attention. Third, TR frequency level was based on patients’ recall at discharge. Validating patient-reported TR frequency levels using billing data, which were not available to us, is recommended for future research. Fourth, we analyzed data only from clinicians using FOTO to collect treatment outcomes, with no comparison with clinicians who either did not document treatment outcomes or did not use FOTO for outcome documentation. Therefore, we cannot rule out a potential selection bias.

Conclusion

Our results provide new knowledge regarding to whom and how TR is being administered during the COVID-19 pandemic in outpatient rehabilitation practices included in our study. The database assessed was found to be suitable for future studies on associations between TR use during the episode of care, and diverse outcome measures documented during routine rehabilitation outpatient practice, while controlling for a comprehensive set of patient characteristics. Studies on the generalizability of these findings into the “real-world” settings where TR is yet to be adopted are needed to advance best TR care models and promote patient outcomes during and after the COVID-19 pandemic.

Acknowledgments

This study was performed at Net Health Systems, Inc, Pittsburgh, Pennsylvania.

Author Contributions

Concept/idea/research design: M.W. Werneke, D. Deutscher, D. Hayes

Writing: M.W. Werneke, D. Deutscher

Data collection: J.E. Mioduski

Data interpretation: M.W. Werneke, D. Deutscher

Data analysis: D. Deutscher

Project administration: D. Hayes

Consultation (including review of manuscript before submitting): M.W. Werneke, D. Deutscher, D. Grigsby, C.A. Tucker, D. Hayes

Funding

There are no funders to report for this submission.

Ethics

This study was approved by Solutions IRB, a private institutional review board located in Yarnell, Arizona and given exempt status based on federal guidelines (IRB #: IORG0007116).

Disclosures

Mr. Werneke, Mr. Mioduski, and Drs. Deutscher, Tucker, and Hayes acknowledge that this research is part of their regular compensated work for Net Health Systems, Inc., the company that owns the Focus on Therapeutic Outcomes (FOTO) Patient Outcomes system that gathers and manages the data analyzed in this manuscript. Mr. Grigsby is an independent owner of a physical therapy clinic practice and has no affiliation with Net Health Systems, Inc. The authors also certify that they have no other affiliations with or financial involvement in any organization or entity with a direct financial interest in the subject matter or materials discussed in the article.