Impact of the Veterans Affairs National Telestroke Program’s Subacute Telestroke Service on Interhospital Transfers

Abstract

Background and Purpose: Telestroke improves access to acute ischemic stroke (AIS) expertise, aids in decision-making, and reduces interhospital transfers. Few studies have examined subacute inpatient telestroke services, which focus on inpatient stroke workup and management. Methods: In this retrospective cohort study of patients with emergency department (ED) diagnosis of AIS from 10/2021-6/2024, we sought to determine the impact of a novel subacute telestroke program on ED transfer rates at participating hospitals. For intervention sites (that implemented the subacute teleconsult program), the period prior to subacute consult “go-live” date was considered the pre-implementation period. Control sites (without the subacute program) were considered preimplementation prior to 5/22/2023 (when approximately half of intervention sites had initiated the subacute program). Logistic regression with generalized estimating equations evaluated the association between implementation time-period and odds of ED transfer in intervention and control sites, adjusting for age, NIHSS, sex, race, and an intervention by time-period interaction term. Results: 1266 patients met eligibility criteria (N = 544 patients from 11 control sites and N = 722 patients from 11 intervention sites). The ED transfer rate was lower within intervention sites post-implementation (pre: 25.7% to post: 22.5%) and higher in control sites (pre: 25.9% to post: 27.1%). These comparisons were statistically nonsignificant in the multivariable analysis. Conclusions: There was a reduction in interhospital transfers after implementation of a subacute telestroke consultation service, but results were nonsignificant in adjusted models. Future analyses should evaluate whether receipt of a subacute telestroke encounter at the patient-level is associated with reduced interhospital transfer for stroke.

Introduction

Telestroke connects acute ischemic stroke (AIS) experts with under-resourced hospitals using videoconferencing technology. Existing telestroke research has largely focused on consultations to aid in acute AIS decision-making (thrombolysis and endovascular thrombectomy). Another potential benefit of telestroke is reduction in interhospital transfer. ¹ A prior Veterans Health Administration (VHA) study demonstrated that implementation of an acute national telestroke program (NTSP) led to 60% reduction in odds of interhospital transfer. ² Few studies have examined subacute telestroke services, ³ which focus on inpatient stroke workup and management, rather than acute management.

In 2022, the VHA NTSP instituted a novel subacute telestroke consult service to extend beyond acute stroke management to provide detailed recommendations in AIS workup and management, including diagnostic evaluation of stroke mechanism and appropriate secondary prevention. We sought to determine the impact of this subacute telestroke program on emergency department (ED) transfer rates at participating hospitals. We hypothesized that the subacute telestroke program would be associated with a decrease in interhospital transfer.

Methods

This analysis was part of a Rapid Response Project within the VA EXTEND QUERI (Quality Enhancement Research Initiative), to evaluate the subacute telestroke consultation service before disseminating and implementing throughout the remaining NTSP sites. The EXTEND QUERI work was reviewed and determined by the local Roudebush VAMC/Indiana University IRB to be non-research.

The NTSP subacute consult service was designed for Veterans with suspected or confirmed stroke who were not candidates for acute treatments (thrombolysis and endovascular thrombectomy), to provide standard of care workup, treatment recommendations and education as well as subsequent follow-up and management. We obtained Central Data Warehouse (CDW) data for Veterans with an ED discharge AIS diagnosis from 10/2021-6/2024 at NTSP subacute consult intervention sites and similar control NTSP sites (without the subacute consult service). Control sites were selected if they had expressed interest in the subacute consult but had not initiated the implementation process. These control sites were considered by the program for future implementation based on their stroke volume, ED/inpatient neurology availability, and documented need/interest in the program. The CDW data included the VA Emergency Department Integration Software, which includes discharge disposition from the VA ED. For intervention sites, the period prior to subacute consult “go-live” date (unique to each site) was considered the pre-implementation period. Control sites were considered pre-implementation prior to 5/22/2023 (the date when approximately half of intervention sites had initiated the subacute program).

Patient-level data included demographics and National Institutes of Health (NIHSS) Score, which was abstracted from Computerized Patient Record System (CPRS) notes using SQL text mining and previously validated algorithms. ⁴ Receipt of a NTSP subacute telestroke consult at intervention sites was identified via note titles and NTSP patient logs.

Statistical Analysis

A logistic regression model was fit to the outcome of ED transfer (yes/no) with effects in the model including time period (post vs pre), site type (intervention vs control), and a time period*site type interaction to test whether the effect of time period varies between intervention and control sites. Generalized estimating equation (GEE) methodology was used to account for clustering of patients within facilities. The interaction term allowed us to use contrasts from the model to estimate the odds ratio for ED transfer for post-implementation vs pre-implementation within the intervention and control sites. Other effects in the model to account for potential confounding included age, NIHSS, sex, and race (black, white, other/unknown).

Since NIHSS score was missing for 23% (292/1266) of the cohort, multiple-imputation was conducted in which the NIHSS score was imputed using predictive mean matching with fully conditional specification including variables of age, sex, and race. This method imputes NIHSS scores from observed values. ⁵ The primary model was fit to both the complete-case data and the imputed data. Reported odds ratios and 95% confidence intervals are reported for both.

As the cut point used to identify the pre vs post time period for the control sites was arbitrarily selected, a sensitivity analysis was conducted by fitting the same model with the pre- and post-implementation time period for controls defined differently. Only observations that occurred prior to the first intervention site going live with the subacute program were considered as ‘pre’ for control sites and observations that occurred after all sites went live were considered as ‘post’. This sensitivity analysis removed 179 observations from the control sites that occurred when only some of the intervention sites had initiated the subacute program.

Results

A total of 1266 Veterans met eligibility criteria (N = 544 patients from 11 control sites and N = 722 patients from 11 intervention sites). Descriptive statistics of the site-level characteristics are provided in Supplemental Table 1. The rate of ED transfer was 24.2% (175/722) in intervention sites and 26.5% (144/544) in control sites. Within intervention sites, the ED transfer rate was slightly lower after the subacute program began with 25.7% (102/397) prior to the subacute program and 22.5% (73/325) after, whereas control sites had a slightly higher ED transfer rate in the later time period relative to the earlier time period: ED transfer was 25.9% (73/282) vs 27.1% (71/262), respectively (Table 1, Figure 1).

Table 1.

Demographic Characteristics and Outcome by Assignment and Pre/post Implementation

		Control*		Intervention†
		Pre N = 282	Post N = 262	Pre N = 397	Post N = 325
Age (years): Mean (SD)/median (IQR)		70.9 (10.5)/72 (65, 77)	70.0 (10.4)/72 (63, 77)	72.8 (10.6)/74 (67, 80)	72.3 (10.0)/74 (68, 78)
NIHSS score: Mean (SD)/median (IQR)		3.4 (4.6)/2 (0, 5)	2.9 (3.2)/2 (1, 4)	4.2 (5.5)/3 (1, 5)	3.6 (4.4)/3 (1, 5)
Sex	Female	13 (4.6%)	13 (5.0%)	17 (4.3%)	14 (4.3%)
	Male	269 (95.4%)	249 (95.0%)	380 (95.7%)	311 (95.7%)
Race	Black	41 (14.5%)	50 (19.1%)	28 (7.1%)	28 (8.6%)
	White	226 (80.1%)	198 (75.6%)	336 (84.6%)	272 (83.7%)
	Other or unknown	15 (5.3%)	14 (5.3%)	33 (8.3%)	25 (7.7%)
Transferred	No	209 (74.1%)	191 (72.9%)	295 (74.3%)	252 (77.5%)
	Yes	73 (25.9%)	71 (27.1%)	102 (25.7%)	73 (22.5%)

*Control site pre/post: before/after 5/22/2023.

^†Intervention site pre/post: before/after individual site subacute go live date.

Figure 1.

Emergency Department (ED) Transfer Rates Pre- and Post-implementation of the Subacute Telestroke Consult Service

Based on the model which incorporated multiple imputation, the interaction of time period and site type was not statistically significant (Table 2, P = .17). The odds of ED transfer after the subacute program began in intervention sites was lower relative to prior to the program (OR 0.83, 95% CI 0.57, 1.21) but was not statistically significant (P = .34) (Table 2). The odds of ED transfer in control sites increased in the later time period relative to the earlier (OR 1.14, 95% CI 0.89, 1.45) but was non-significant (P = .30). Younger age and higher NIHSS were associated with increased odds of transfer in all models (Table 2).

Table 2.

Generalized Estimating Equations (GEE) Logistic Regression Results for the Unadjusted, Complete Case, and Multiple-Imputation Models

	Univariable analysis N = 1266		Multivariable analysis with available NIHSS N = 974		Multivariable analysis with multiple imputed NIHSS N = 1266
	OR (95%CI)	P value	OR (95%CI)	P value	OR (95%CI)	P value
Assignment: Intervention	0.80 (0.37 - 1.73)	0.57	0.90 (0.37 - 2.16)	0.81	0.94 (0.43 - 2.06)	0.87
Time: Post	0.92 (0.72 - 1.17)	0.48	1.05 (0.81 - 1.36)	0.72	1.14 (0.89 - 1.45)	0.30
Age	0.99 (0.98 - 1.00)	0.049	0.99 (0.97 - 1.00)	0.03	0.98 (0.97 - 1.00)	0.01
NIHSS	1.07 (1.05 - 1.10)	<.01	1.08 (1.05 - 1.10)	<.01	1.07 (1.05 - 1.10)	<.01
Sex: Female	1.02 (0.63 - 1.64)	0.94	0.73 (0.40 - 1.34)	0.32	0.91 (0.55 - 1.51)	0.72
Race: Black vs White	1.10 (0.71 - 1.71)	0.68	1.09 (0.69 - 1.72)	0.72	1.06 (0.68- 1.67)	0.79
Race: Other or Unknown vs White	1.15 (0.82 - 1.61)	0.42	1.13 (0.67 - 1.92)	0.64	1.18 (0.84 - 1.66)	0.33
Assignment*Time	--	--	--	0.24	--	0.17
Contrast Estimates
			OR (95%CI)	P value	OR (95%CI)	P value
Intervention: Post vs Pre	--	--	0.78 (0.52 - 1.19)	0.25	0.83 (0.57 - 1.21)	0.34
Control: Post vs Pre	--	--	1.05 (0.81 - 1.36)	0.72	1.14 (0.89 - 1.45)	0.30

Results were very similar in the sensitivity analysis where we defined the pre-implementation time period in the control site as that occurring prior to the first go-live date in intervention sites and the post-implementation period as occurring after the last site went live (Supplemental Table 2).

Discussion

A decrease in transfer rates of approximately 12% (absolute decrease of 3.2%) was observed in sites which implemented the NTSP subacute consult. Facilities that did not implement the subacute consult had a 4.6% increase in transfer rates (absolute increase of 1.2%) during this time period. Although possibly a clinically significant drop in transfer rates, when controlling for patient characteristics, this decrease was not statistically significant, with younger age and increased stroke severity being the primary factors associated with ED transfers.

Prior literature has revealed mixed findings on the association between telestroke implementation and interhospital transfer for AIS. ¹ A 2022 VHA study by Lyerly et al found a 14.4% decrease in transfer rates after implementation of the acute NTSP consult. ² Similarly, a study from the Medical University of South Carolina demonstrated a decreased transfer rate from 29.4% to 20.2% after the implementation of a telestroke program. ⁶ An analysis from Catalonia, Spain found that transfers were avoided in 46.8% of patients with AIS after implementation of a telestroke network, but this study was limited by lack of a pre-post design. ⁷ A differences-in-differences analysis using a sample of US hospitals published in 2021 found no effect of telestroke implementation on interhospital transfer. ⁸ Whereas prior studies focused on acute telestroke systems, our present study extends this literature given our focus on a subacute telestroke service implementation.

One potential reason that a lower effect size was seen in the present study for subacute telestroke implementation (as compared to that for acute telestroke implementation in the prior VHA study) ² is that the subacute telestroke intervention sites also had acute telestroke services available, which may have diluted the effect of the novel subacute telestroke service.

This study has several limitations. First, NIHSS score was a key covariate that was missing in 23% of the study sample. Though this missingness for NIHSS is comparable to that from large national stroke registries that operate outside the VHA system, such as the Get With the Guidelines-Stroke registry,^9,10 we attempted to account for this limitation by using multiple imputation with predictive mean matching. Reassuringly, results were similar in the complete case and multiple imputation models (Table 2). Second, this was an observational study to evaluate the implementation of a program at the site-level using patient-level data. This study design was necessary given the relatively recent implementation of the subacute telestroke service at certain sites, which precluded the ability to randomize sites (or even patients) to the subacute telestroke program. Because only 12.8% of Veterans with AIS discharged from a participating facility actually received a subacute consult, our sample size available for this program evaluation may not have been sufficient to detect a clinically meaningful effect of the intervention. This finding also highlights a major area for future quality improvement and further implementation efforts. Finally, there may be additional sources of bias towards the null, such as exposure misclassification issues related to our definition of the pre vs post study period. Future analyses should evaluate whether receipt of a subacute telestroke encounter at the patient-level is associated with reduced interhospital transfer for stroke.

ORCID iDs

Brian Stamm https://orcid.org/0000-0002-0862-9650

Linda S. Williams https://orcid.org/0000-0002-9228-9459

Ethical Considerations

The analysis was deemed quality improvement by the VHA QUERI program and was conducted at the request of the NTSP as part of a mixed-methods evaluation of program implementation prior to further expansion of the subacute consult program. The EXTEND QUERI work was also reviewed and determined at the level of the local IRB at the Roudebush VAMC to be non-research.

Data Availability Statement

It is not possible to share our data due to the sensitive/protected nature of the patient and hospital-level VHA data.