Primary Care Telehealth Initiation and Engagement Among Veterans at High Risk, 2019-2022

Key Points Question How did patients at high risk (≥75th percentile for 90-day risk of hospitalization) engage with primary care telehealth modalities (telephone, video visits, or secure messaging) within the Veterans Health Administration between 2019 and 2022? Findings In this cohort study including 1 383 070 patients at high risk, among those newly engaged in telehealth with the onset of the COVID-19 pandemic, 38% remained regular telehealth users the following year. Patients had distinct patterns of primary care telehealth use during and after the COVID-19 pandemic. Meaning This study suggests that access barriers may limit initial telehealth engagement among some patients at high risk of hospitalization or mortality, although factors associated with uptake and sustainment vary; these patterns can inform future resource allocation.


Introduction
During the COVID-19 pandemic, the Veterans Health Administration (VHA) rapidly expanded telehealth services, including secure messaging and video and telephone visits. 1,2Telehealth outreach, such as development of a VHA loaned tablet program, particularly targeted veterans at high risk of adverse events given known access barriers. 2However, evaluations of telehealth use among patients at high risk have been limited to early or prepandemic findings. 3,4After the pandemic, telehealth infrastructure has gained permanence, with nearly 50% of visits in primary care being done via video or telephone modalities. 5,6Understanding primary care use and telehealth engagement among high-risk patients after these changes in access could inform resource allocation and policy to optimize primary care staffing, care delivery, and alignment with patient preferences among a high-need, high-cost population. 7,8

Study Design
This cohort study examined primary care use by modality among patients in the VHA at high risk of adverse hospitalization or mortality during and after the COVID-19 pandemic from March 11, 2019, to   March 10, 2022.We describe primary care utilization and changes by year among high-risk patients

Data Sources
Patient and clinic characteristics were from the Veterans Affairs (VA) Corporate Data Warehouse (CDW), except as described. 11Primary care empanelment was determined using the Primary Care Management Module for active assignment in the period of interest.Patient death data in the CDW are updated daily and aggregate data from VHA facilities, death certificates, and National Cemetery Administration records.Race and ethnicity data were included given the potential for inequitable access to telehealth modalities by race and ethnicity. 12,13These data use a previously developed algorithm that aggregates values across multiple VA medical record and demographic datasets, prioritizing sources of patient self-report; categories of "Other" race are included from historic response options in some datasets and include "declined," "multiple," or "unknown." 14Serious mental illness is an indicator based on diagnoses codes and functional assessments, established by the VA Program Evaluation Resource Center, and includes schizophrenia, schizoaffective disorder, bipolar disorder, major depressive disorder with psychosis, or posttraumatic stress disorder in a patient receiving antipsychotic medications. 15Veterans Affairs Priority Groups are categories of military service-connected disability or income.Chronic condition counts include 28 health conditions listed within the Gagne comorbidity index, 16  and <100 MB/s; upload speed, Ն5 and <100 MB/s), and optimal (download and upload speeds, Ն100 MB/s). 18Geographic location (rurality) was from the VA Planning Systems Support Group (PSSG) geocoded data, linked to patient zip codes from the US Census Bureau.Driving distance is also from PSSG data, using patient travel time to the nearest VHA primary care clinic with distances from the VA Site Tracking System.Distances are categorized as less than 64.4 km or more, consistent with definitions for care referral parameters. 19terans Health Administration Primary Care staffing ratios were included as covariates indicating clinic-level measures of support staff per primary care clinicians, obtained from the Patient Aligned Care Team Compass module within the CDW.Clinics were categorized as community affiliated or hospital affiliated per the VA Site Tracking System.Care modalities and type were defined by VA Managerial Cost Accounting procedural and encounter stop codes (eTable 1 in Supplement 1).

Patient Population
We included veterans with an estimated risk of hospitalization or mortality in the 75th percentile or greater within 90 days (ie, high risk for this study) enrolled in VHA primary care before the start of the COVID-19 pandemic (March 10, 2020). 20Estimated risk refers to a percentile of probabilities using the previously validated Care Assessment Need score, a VA-specific model based on patient clinical and demographic characteristics that estimates the probability of future hospitalization or mortality in a period of time. 21,22

Statistical Analysis
We For each year, we created a binary indicator of regular telehealth use for primary care encounters, defined as 1 or more video visit or secure message or greater than median per-year proportion of telephone visits to other primary care modalities.We examined use of general primary care (including women's health) vs specialized primary care (home-based primary care, geriatrics, homeless care) for descriptive trends.
We classified cohort patients still living and active in primary care at the end of year 2 into 5 subgroups according to retroactive telehealth use as a binary factor per year across the 3-year period: never users (no telehealth use in any year); transient users (telehealth use in pandemic year 1 only); new persistent users (telehealth use in pandemic years 1 and 2); consistent users (telehealth use in all 3 years); and remaining as all others.Variation in patient-level characteristics among these subgroups for 3-year patterns of telehealth use were examined using exploratory multinomial logistic regression models with clinic-level random effects to control for differences in infrastructure and facility-level telehealth ability.Except as described above, covariates for models were from the most recent quarter in the prepandemic year.Models were also adjusted for clinic staffing ratios, community-or hospital-level affiliation, and baseline patient primary care visit use.Outcomes are presented as adjusted relative risk ratios (ARRs).Never users were set as the reference group.Dataset preparation and descriptive statistics used SAS EG, version 7.1 (SAS Institute Inc). 23atistical modeling was conducted using R, version 4.3.1 (R Project for Statistical Computing), 24 with extension mclogit or mblogit version 0.9.6.25 For all analyses, significance was determined using a 2-tailed type I error rate of P < .05.

Results
Before the pandemic, of  2).Those who were never users were the most likely to be men with the fewest chronic conditions and lowest disability burden at baseline.
Consistent users were more likely to be women and had the highest comorbidity and disability

JAMA Network Open | Health Policy
Primary Care Telehealth Initiation and Engagement Among Veterans at High Risk burden.Both transient users and new persistent users were more often located in urban settings.
New persistent users were proportionately those more often identifying as Black and Hispanic, women, low-income individuals, and those who had greater comorbidity burdens than transient users.Transient users had fewer chronic conditions and lower disability burdens than all other subgroups except never users.Age, marital status, serious mental illness, substance use, and internet adequacy at baseline were less characterizing of differences in ongoing telehealth use between subgroups (eFigure in Supplement 1).

Discussion
Patterns of telehealth use in primary care for high-risk patients changed throughout the COVID-19 pandemic.Our study has 3 main takeaways relevant to health systems.First, we found that while telehealth primary care use increased dramatically with pandemic onset, telehealth use declined subsequently as use of in-person care rebounded.Despite growing interest in virtual care, this finding underscores the continued need for resources to support in-person care among high-risk populations. 26cond, we found that general primary care, as opposed to more specialized primary care teams, remained the front line for high-risk patients during and after the pandemic.Despite expected attrition due to mortality and disengagement, [27][28][29] high-risk patients who remained in care throughout the pandemic remained at high complexity with a higher disability burden despite a slight lessening of psychosocial factors.Change in military service-connected disability is not insignificant for complexity; veterans with a high disability rating have mortality rates 2.5 times those with lower levels. 30This finding highlights the ongoing importance of general primary care after the pandemic for high-complexity, high-need populations, consistent with prepandemic literature.

Limitations
This study has some limitations, including caution when generalizing beyond veteran populations and potential misclassification due to administrative data.Results are descriptive of care use among high-risk patients who remained active and alive during the period of study and are not indicative of outcomes for all patients.Patients missing continuous covariates were excluded from models (n = 2201 [0.2%]).Models also adjusted for primary care staffing ratio and community vs hospital affiliation.
a Combined levels due to small cell sizes.

for 3
time periods (before the pandemic [March 11, 2019, to March 10, 2020], pandemic year 1 [March 11, 2020, to March 10, 2021], and pandemic year 2 [March 11, 2021, to March 10, 2022]).Among the patients living and active in VHA primary care during all 3 years, we explored factors associated with telehealth initiation and sustained engagement among demographic and geographic subgroups of high-risk patients.This study was designated nonresearch quality improvement by the VHA Office of Primary Care in accordance with the national VHA Office of Research and Development policy 9 and not subject to institutional review board review or exemption.This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. 10 examined 3 time periods for descriptive trends: before the pandemic (March 11, 2019, to March 10, 2020), pandemic year 1 (March 11, 2020, to March 10, 2021), and pandemic year 2 (March 11, 2021, to March 10, 2022).

Figure 2 .
Figure 2. Telehealth Use Among Veterans at High Risk Engaged in Primary Care Throughout 2019 to 2022 Compared With Never Users

8
Figure 1.Primary Care by Visit Modality Among Veterans at High Risk of Adverse Events Primary Care Telehealth Initiation and Engagement Among Veterans at High Risk clinics with higher proportions of patients with characteristics consistent with never users might anticipate that staffing and resource allocation should go toward ongoing support of in-person visits.
Abbreviations: CBOC, community-based outpatient clinic; VHA, Veterans Health Administration.a May not sum to 100%; sex missing not shown.b Other and unknown races and ethnicities are not shown; this category includes multiple races or response declined.c Veterans Affairs priority group is a military service-related disability or income determination.d All-cause acute hospitalizations (Ն1) in 12 months prior.e Includes unknown geography.JAMA Network Open | Health Policy JAMA Network Open.2024;7(7):e2424921.doi:10.1001/jamanetworkopen.2024.24921(Reprinted) July 31, 2024 6/10 Downloaded from jamanetwork.comby guest on 08/04/2024