This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.
Virtual care (VC) and remote patient monitoring programs were deployed widely during the COVID-19 pandemic. Deployments were heterogeneous and evolved as the pandemic progressed, complicating subsequent attempts to quantify their impact. The unique arrangement of the US Military Health System (MHS) enabled direct comparison between facilities that did and did not implement a standardized VC program. The VC program enrolled patients symptomatic for COVID-19 or at risk for severe disease. Patients’ vital signs were continuously monitored at home with a wearable device (Current Health). A central team monitored vital signs and conducted daily or twice-daily reviews (the nurse-to-patient ratio was 1:30).
Our goal was to describe the operational model of a VC program for COVID-19, evaluate its financial impact, and detail its clinical outcomes.
This was a retrospective difference-in-differences (DiD) evaluation that compared 8 military treatment facilities (MTFs) with and 39 MTFs without a VC program. Tricare Prime beneficiaries diagnosed with COVID-19 (Medicare Severity Diagnosis Related Group 177 or International Classification of Diseases–10 codes U07.1/07.2) who were eligible for care within the MHS and aged 21 years and or older between December 2020 and December 2021 were included. Primary outcomes were length of stay and associated cost savings; secondary outcomes were escalation to physical care from home, 30-day readmissions after VC discharge, adherence to the wearable, and alarms per patient-day.
A total of 1838 patients with COVID-19 were admitted to an MTF with a VC program of 3988 admitted to the MHS. Of these patients, 237 (13%) were enrolled in the VC program. The DiD analysis indicated that centers with the program had a 12% lower length of stay averaged across all COVID-19 patients, saving US $2047 per patient. The total cost of equipping, establishing, and staffing the VC program was estimated at US $3816 per day. Total net savings were estimated at US $2.3 million in the first year of the program across the MHS. The wearables were activated by 231 patients (97.5%) and were monitored through the Current Health platform for a total of 3474 (median 7.9, range 3.2-16.5) days. Wearable adherence was 85% (IQR 63%-94%). Patients triggered a median of 1.6 (IQR 0.7-5.2) vital sign alarms per patient per day; 203 (85.7%) were monitored at home and then directly discharged from VC; 27 (11.4%) were escalated to a physical hospital bed as part of their initial admission. There were no increases in 30-day readmissions or emergency department visits.
Monitored patients were adherent to the wearable device and triggered a manageable number of alarms/day for the monitoring–team-to-patient ratio. Despite only enrolling 13% of COVID-19 patients at centers where it was available, the program offered substantial savings averaged across all patients in those centers without adversely affecting clinical outcomes.
The COVID-19 pandemic forced health care systems around the world to rapidly innovate and adapt to unprecedented operational and clinical strain [
Early initiatives were heterogeneous, driven by local trends and available technologies [
In December 2020, the US Military Health System (MHS) Virtual Medical Center (VMC) implemented a VC program in 8 military treatment facilities (MTFs) across the United States for COVID-19 patients. The program was delivered virtually by a 24-hour dedicated nursing service using continuous remote patient monitoring. The implementation of a standardized VC program for the same condition in a subset of similar treatment facilities across the country made for an ideal natural experiment.
The VC program was available to any Tricare Prime beneficiary eligible for care within the MHS aged 21 years or older. Inclusion criteria were broad and included any patient who presented to hospital symptomatic for COVID-19 and those who, despite not requiring admission, had a high risk of exposure and were at risk for severe disease due to a comorbid state. The program was subsequently expanded to other use cases, including congestive heart failure, gestational hypertension, and postoperative monitoring for bariatric surgery; however, this analysis only pertains to COVID-19.
Referred patients were screened for eligibility and consented to participation in the VC program, and they were then issued the VC equipment and familiarized with it by specially trained nurses. Once home, patients were called to ensure kit setup was successful. Enrollment triage forms (
Patients were monitored using the Current Health (CH) VC platform (Current Health Inc), which included a US Food and Drug Administration (FDA) 510(k)-cleared wearable device worn on the upper arm, a blood pressure cuff and weighing scale, a tablet, and a network hub that operated via home Wi-Fi or roaming cellular signal, enabling access for patients without home internet. Continuous vital signs measured were pulse rate, respiratory rate, oxygen saturation, temperature, and motion. Blood pressure and weight were monitored intermittently. The tablet collected customizable patient-reported outcome measures, including symptom burden, and could be used to asynchronously request direct assistance from the on-call nursing team. All data were processed via cloud computing and displayed on a web dashboard for the clinical team. If any vital sign exceeded a predetermined threshold, an alert would trigger on the dashboard and send notifications to the appropriate staff. The team monitored patients across the 8 participating MTFs with a nurse-to-patient ratio of 1:30. Each MTF had designated on-call physicians available for on-demand support if care escalation was required.
Patients were disenrolled from VC at their physicians’ discretion or if escalated back from VC to inpatient care due to clinical decompensation. CH coordinated kit return, sanitization, service, and repackaging and returned the kits to their original MTF. This ensured that each MTF maintained a consistent stock of equipment to enroll new patients. The infrastructure, personnel, and fiscal resources for the program were directly funded by the MHS. A total of 200 CH kits were available for distribution across the 8 participating MTFs, dynamically divided based on utilization and demand. There were always sufficient kits available at each location.
The difference-in-differences (DiD) model used ordinary least squares regression, regressing the outcome on an indicator for whether the hospital was included in the VC program, an indicator for whether the patient’s date of admission was after the earliest implementation of the VC program, and an interaction variable of the 2 indicators. The coefficient of this interaction can be interpreted as the effect of being admitted to a hospital that had an active VC program regardless of enrollment in the program. The model controlled for age, gender, marital status, COVID-19 pneumonia diagnosis during the index admission, and Elixhauser comorbidity score [
The sample was constructed using all Tricare beneficiaries admitted to an MTF with COVID-19 from December 7, 2020, to December 6, 2021. Only the patient’s first admission was included, with patients admitted either directly by their physician or through an emergency department. Patients with any Medicare Severity Diagnosis Related Groups other than 177 were excluded, along with any who were not discharged to their home. The final sample included 3988 index admissions: 1838 patients who were admitted into an MTF with an VC program and 2150 who were admitted to an MTF without a program. Of the 1838 patients admitted into an MTF with a VC program, 237 (13%) were enrolled in VC. The average cost of VC was calculated per day based on the capital expenditure and ongoing monitoring contract, costs of nursing labor, and program management support. While VC program initiation varied at the MTF level, the Defense Health Agency (DHA) paid for the VC centrally with a single, centralized monitoring hub. This makes sense given that VC is a high fixed-cost investment that can be managed from a single location, allowing for economies of scale. However, this means that costs can only be calculated at the system level and not at the MTF level.
Outcome data for the 237 patients enrolled in the program were obtained from MHS Mart, known as “M2,” which is a queryable data repository for the DHA. Vital-sign and alarm data were obtained from CH. A manual chart review of patients’ electronic medical records (EMRs) was conducted for the subgroup of patients at Brooke Army Medical Center (BAMC) between December 7, 2020, and June 7, 2021. The review was limited to the first 6 months of the program due to the availability of clinicians to conduct the review. That review focused on comorbidities that increased risk for severe COVID-19, including smoking, diabetes, being immunocompromised, chronic kidney disease, and hypertension, as well as validated scores for severity and readmission: the Quick COVID-19 Severity Index (qCSI), Quick Sepsis Related Organ Failure Assessment (qSOFA), and the HOSPITAL score (an abbreviation that represents “hemoglobin at discharge, discharge from an oncology service, sodium level at discharge, procedure during the index admission, index type of admission, number of admissions during the last 12 months, and length of stay”) [
The study complied with the Declaration of Helsinki. This was a retrospective study of data collected for clinical rather than research purposes, so prior informed consent was not sought and no compensation was offered. An exemption for retrospective analysis was granted by the BAMC Institutional Review Board (reference number C.2021.103e). The data were deidentified for analysis.
During the study period, 3988 patients were admitted to the MHS with COVID-19.
The first column in
The DiD provided an average effect, but conceptually the program should have impacted those that would have stayed in the hospital for monitoring.
Patient covariates at military treatment facilities with a virtual care program (treatment), and without a program (control) (N=3988).
Characteristics | Treatment (n=1838) | Control (n=2150) | % Difference (treatment minus control) | |
Mean age (years) | 55.98 | 56.27 | –0.29 | .58 |
Female, n (%) | 717 (39) | 774 (36) | 3 | .08 |
Married, n (%) | 1489 (81) | 1613 (75) | 6 | <.001 |
Pneumonia, n (%) | 1011 (55) | 1226 (57) | –2 | .21 |
Mean Elixhauser score | 1.96 | 1.98 | –0.02 | .78 |
Results of difference-in-differences analysis (N=3988). All regressions include the full set of control variables and fixed effects.
Length of stay (n=3984) | Thirty-day readmission (n=3984) | Emergency department visit within 90 days (n=3984) | |
Panel A: linear (postimplementation), coefficient (SE) | –0.556 (0.256) | 0.007 (0.019) | –0.011 (0.018) |
Panel B: log transformation (postimplementation), coefficient (SE) | –0.135 (0.047) | N/Aa | N/A |
Panel C: alternative specification (postimplementation), coefficient (SE) | –0.115 (0.051)b | 0.094 (0.249)c | 0.45 (0.501)c |
Full sample mean | 4.78 | 0.08 | 0.02 |
aN/A: not applicable.
bAlternative specification: Poisson.
cAlternative specification: logit.
Difference in adjusted means of the treatment and control military treatment facilities before and after implementation of the virtual care program.
Residual versus predicted length of stay for those on virtual care, based on observable covariates and actual length of stay.
For the 237 patients enrolled in the program, mean age was 53 (SD 15.3) years, 100 (42%) were female, 137 (58%) were male, 231 (97.5%) activated their wearable, median activation time was 60 (IQR 11-186) minutes, and they were monitored through the CH platform for a total of 3474 (median 7.9, IQR 3.2-16.5, range 1-106) days. Wearable adherence was 85% (IQR 63%-94%). Patients triggered a median of 1.6 (IQR 0.7-5.2) physiological alarms per patient per day; 203 (85.7%) were monitored at home and directly discharged from VC; 27 (11.4%) were escalated to a physical hospital bed while on monitoring; and 1 (0.4%) was readmitted to the hospital within 30 days of discharge from VC. There were no deaths in the cohort. There were significant differences between those requiring escalation to physical care and those remaining at home throughout their time in the VC program. Those who required escalation activated their CH kit in a similar timeframe but were less adherent to using the wearable (median 63%, IQR 32%-83% vs 87%, IQR 70%-95%, respectively;
Charts from 39 patients (16% of the monitored cohort) from the first 6 months of the VC program at BAMC were hand-reviewed for COVID-19 risk factors. Their demographics and COVID-19 risk factors are presented in
Demographics and COVID risk factors of patients enrolled in the virtual care program at Brooke Army Medical Center (n=39).
Characteristics | Values | |
Age (years), mean (SD, range) | 59 (14.6, 31-86) | |
|
||
|
Female | 15 (38) |
|
Male | 24 (62) |
|
||
|
Asian | 1 (3) |
|
Black | 7 (18) |
|
Hispanic | 12 (31) |
|
Other or unspecified | 3 (8) |
|
Southeast Asian | 1 (3) |
|
White | 15 (38) |
|
||
|
High risk | 25 (64) |
|
Medium risk | 14 (36) |
|
||
|
Smoker, n (%) | 6 (15) |
|
Diabetes, n (%) | 12 (31) |
|
Immunocompromised, n (%) | 11 (28) |
|
Chronic kidney disease, n (%) | 4 (10) |
|
Hypertension, n (%) | 23 (59) |
|
Risk factors, median, (IQR, range) | 1 (0-2, 0-4) |
|
||
|
0 | 21 (54) |
|
1 | 17 (43) |
|
2 | 1 (3) |
HOSPITALa score, median (IQR, range) | 2 (1-2.5, 0-6) | |
|
||
|
No | 37 (95) |
|
Yes | 2 (5) |
|
||
|
No | 13 (33) |
|
Yes | 26 (67) |
aHOSPITAL score: hemoglobin at discharge, discharge from an oncology service, sodium level at discharge, procedure during the index admission, index type of admission, number of admissions during the last 12 months, and length of stay.
The MHS VC program was established during a time of acute national need, with patients offered round-the-clock remote care as an alternative to being in the hospital. Despite only enrolling 6% of patients with COVID-19 admitted to the MHS and only 13% of patients at centers where the program was available, the program had a disproportionately large impact. Overall length of stay was reduced by 12%, averaged across all COVID-19 patients at centers with availability, with an associated cost saving of $2047 per patient. Reassuringly, the 11% rate of escalation to physical care for patients enrolled in the program demonstrated that unwell patients were being identified and treated despite being at home, with no increase in emergency department attendance, 30-day readmissions, or deaths. It should be stressed that escalation to physical care was not a “readmission,” as the patient remained in their “initial admission” until discharged from the program. Indeed, escalation was desirable in those patients who warranted it, and any days spent at home rather than in the hospital increased inpatient capacity at the facilities while reducing exposure to COVID-19 for other patients and staff.
The median length of monitoring on VC was 7.9 days, and the overall adherence to wear was 85%. Adherence to wearables has typically been reported as the total number of days patients wore a device, rather than the consistency of wear during those days. In both postoperative and clinical trial contexts, median length of engagement has been reported at around 5 days [
Other previously identified barriers to VC adoption have included lack of connectivity (the so-called digital divide that disproportionately affects the elderly, those with low income, and rural populations) and concerns around privacy and usability [
The alarms needed to be specific as well as sensitive to avoid disturbing patients, bringing them into the hospital unnecessarily, or increasing the risk of alarm fatigue among the nursing team [
The facilitators and barriers to rolling out this intervention were the subject of a separate qualitative study, presently under review. However, to contextualize the findings of this paper, we noted that in common with similar interventions, clinician acceptance took time to establish, and more patients would have likely been enrolled if acceptance had come sooner [
However, the program was also hampered by the inability to coordinate community services or go into patients’ homes. The lack of an integrated inpatient/outpatient EMR, along with inexperience in the use of virtual care relative value units and Current Procedural Terminology–associated reimbursement also slowed revenue generation. Finally, the CH platform was FDA 510(k)-cleared only for patients older than 21 years, which made the proportion of the active-duty population aged between 18 and 21 ineligible for enrollment.
The unique strength of this study was its comparison of similar health facilities spread across the United States all attempting to treat an identical clinical condition concurrently. The DiD analysis compared a treatment group (centers with VC) to a control group (centers without VC) before and after the intervention (ie, VC), then estimated the divergence in outcome (ie, length of stay). The identifying assumption was that the treatment group would have followed parallel trends with the control group in the absence of the intervention. In other words, changes in the dominant COVID-19 variant, vaccination rates, treatment methodologies, or other factors would have impacted centers with and without VC equally. While this was an inherently untestable assumption,
Consequently, the DiD analysis could only estimate the effect that the presence of VC had on patients on average. It demonstrated that patients at centers with VC stayed in the hospital for less time than their counterparts in centers without VC. This may have been due to the specific patients being entered in the program but could also have been driven by hospitals and physicians having more ability to focus on those patients who remained in the hospital. The VC effect may also have been driven through creating additional capacity, as well as the care itself. To this point,
In conclusion, the unique structure of the MHS allowed comparison between MTFs that implemented a VC program for COVID-19 and those that did not. Despite the VC program enrolling only a small proportion of patients admitted with COVID-19, it offered substantial savings in centers where it was adopted. The program was effective in identifying suitable patients, escalating them appropriately to physical care, and discharging them once their illness was resolved. The program's military context may have aided its rapid rollout and adoption across multiple centers, and the single-payer nature of the DHA may have facilitated the economic justification of the initiative. However, the results are likely to be applicable to other large health systems that can support or engage a nurse monitoring service, particularly those systems that can reap the economic benefit of a cost-saving program.
Enrollment form and risk stratification for the Virtual Care program.
Brooke Army Medical Center
Current Health
Defense Health Agency
difference in differences
Diagnosis Related Group
electronic medical record
US Food and Drug Administration
hemoglobin at discharge, discharge from an oncology service, sodium level at discharge, procedure during the index admission, index type of admission, number of admissions during the last 12 months, and length of stay
Military Health System Mart
Military Health System
military treatment facility
Quick COVID-19 Severity Index
Quick Sepsis Related Organ Failure Assessment
virtual care
Virtual Medical Center
We acknowledge Dr Gary Legault, the Virtual Medical Center clinical team, Ms Jessie Lever-Taylor, Ms Hope Smith, and Ms Abenaa Boyce for their hard work in setting up the program and making it a success.
The data sets generated during and/or analyzed during the current study are not publicly available due to information governance restrictions within the Military Health System. Please make any requests for access to MJM (michael.j.morris34.civ@health.mil), who is chairman of the Brooke Army Medical Center Institutional Review Board.
MW, DY, NZ, JP, and AW are paid employees of Current Health and hold stock in Best Buy Inc. MJM is a US government employee and a paid speaker for Janssen.