Use of Discharge Disposition to Determine Stroke Severity after TAVR

Despite reductions in procedure-related mortality and complications,¹ stroke remains a persistent complication after transcatheter aortic valve replacement (TAVR). However, only disabling strokes are associated with increased mortality post-TAVR, whereas non-disabling strokes are not.² Consequently, only disabling stroke has been included in the composite endpoint for recent TAVR clinical trials. The distinction between disabling and non-disabling stroke post-TAVR is further highlighted by the PROTECTED-TAVR trial, which demonstrated a potential benefit of cerebral embolic protection on disabling stroke but not for the primary endpoint of any stroke.³

Given these differences in outcomes and preventive strategies, it is important to understand the severity of post-TAVR strokes that occur in clinical practice. While formal neurologic adjudication is frequently used in clinical trials, this approach is impractical outside of the trial setting. The TVT registry only captures stroke in aggregate, administrative claims do not specify functional status after stroke, and attempts to collect more detailed information on stroke severity in other national registries have been unsuccessful owing to frequent missing data. As such, it is challenging to distinguish between disabling and non-disabling strokes after TAVR using real-world data.

Discharge location is a variable readily available in real-world data that could serve as a proxy for stroke severity, since patients who are more disabled may be less likely to be discharged home. We therefore examined the performance of discharge location in distinguishing disabling vs. non-disabling stroke among patients undergoing TAVR.

We pooled trial-level data from patients who experienced in-hospital stroke in 6 large TAVR clinical trials (PARTNER 2A, PARTNER 3, Sapien S3i Registry, SURTAVI, Evolut Low-Risk, and PROTECTED TAVR). All trials included systematic neurologic assessments before and after TAVR to screen for stroke. For each trial, the presence and severity of stroke was determined by an independent clinical events committee (CEC), which differentiated between disabling and non-disabling stroke according to VARC-2 criteria.⁴ We categorized discharge location for each patient as home versus not home, including deaths in the non-home discharge category. Discharge location was ascertained directly from trial data, except for the SURTAVI and Evolut Low-Risk trials, for which it was obtained through linkage with claims.⁵ For the primary analysis, we calculated test characteristics for discharge location to indicate stroke severity. We performed sensitivity analyses (1) including only the three trials conducted since 2017, given recent changes in operator technique and clinical pathways that have reduced the rate of discharge to a non-home location and are more reflective of current practice, and (2) excluding trials without direct discharge disposition information. This study was waived from Institutional Review Board review. Data will not be made available per data use agreements.

A total of 109 patients with adjudicated stroke after TAVR were included. Of the 40 patients with disabling stroke, 32 were discharged to a non-home location resulting a sensitivity of 80% (95% CI 64–91%) (Table). Of the 69 patients with non-disabling stroke, 46 were discharged home resulting in a specificity of 67% (95% CI 54–78%). Results were similar when trials without direct discharge disposition information were excluded. When the analysis was limited to studies conducted after 2017 (45 strokes), the sensitivity of non-home discharge location for identifying disabling stroke was 73% and specificity was 90%, with positive predictive value of 79% and negative predictive value of 87%. Thus, among patients with CEC-adjudicated stroke after TAVR, we found that discharge location was a strong predictor of whether the stroke was disabling or non-disabling, particularly in studies conducted with contemporary techniques and devices.

Table.

Assessment of Stroke Severity by Discharge Location in TAVR Clinical Trials

	Stroke severity			Sensitivity (195% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)
	Disabling	Non-disabling	Total
All studies (n=6)
Discharge to non-home	32	23	55	80.0% (64.4%, 91.0%)	66.7% (54.3%, 77.6%)	58.2% (49.1%, 66.8%)	85.2% (75.2%, 91.6%)
Discharge to home	8	46	54
Total	40	69	109
Studies post-2017 (n=3)
Discharge to non-home	11	3	14	73.3% (44.9%, 92.1%)	90.0% (73.5%, 97.9%)	78.6% (54.6%, 91.8%)	87.1% (74.3%, 94.0%)
Discharge to home	4	27	31
Total	15	30	45
Studies with direct discharge disposition information (n=4)
Discharge to non-home	25	15	40	78.1% (60.0%, 90.7%)	68.1% (52.9%, 80.9%)	62.5% (51.4%, 72.5%)	82.1% (69.8%, 90.1%)
Discharge to home	7	32	39
Total	32	47	79

PPV=positive predictive value, NPV=negative predictive value, CI=confidence interval

Sensitivity = (true positives)/(true positives + false negatives); specificity = (true negatives)/(true negatives + false positives); PPV = (true positives)/(true positives + false positives); NPV = (true negatives)/(true negatives + false negatives)

Home includes home with services. Non-home includes skilled nursing facility, rehabilitation, long-term care facility, other hospital, inpatient or home hospice, or death.

Discharge location was not available for 3 patients with disabling stroke in PROTECTED TAVR due to loss to follow up and 5 patients (2 disabling, 3 non-disabling) in SURTAVI and 6 patients (0 disabling, 6 non-disabling) in the Evolut Low Risk trial due to lack of linkage with Medicare claims. These patients were thus not included in our sample.

In addition to stroke, there are several reasons why a patient might be discharged to a non-home location after TAVR including living in a non-home location prior to TAVR or an in-hospital complication that precluded discharge to home. In contemporary practice, the most common TAVR complications that might necessitate higher level of care post-discharge include major vascular complications and disabling stroke. Over the past decade, development of lower-profile devices and minimalist transfemoral techniques has been associated with a reduction in the incidence of vascular complications after TAVR, yet the incidence of stroke has remained unchanged.¹ Thus, the specificity of discharge location for identifying disabling stroke has increased in post-2017 trials, which are more likely to reflect contemporary practice.

Despite use of clinical trial data, we were unable to determine whether a patient lived in a non-home location prior to TAVR. It is likely that exclusion of such patients would have further improved the specificity of discharge location for identification of disabling strokes. To overcome this limitation, future studies leveraging discharge location to distinguish between disabling and non-disabling stroke should account for pre-hospital location if possible. Finally, we used claims to obtain discharge disposition location for 2 trials, though overall results were similar when these trials were excluded.

This study suggests that discharge location is a potential proxy for distinguishing stroke severity after TAVR, particularly for applications where specificity is more important than sensitivity. In the future, this approach may be useful to track incidence and correlates of disabling stroke after TAVR using real-world data.