The study was approved by the Johns Hopkins Institutional Review Board, and all patients gave written informed consent. Inclusion criteria were admission to the hospital for acute RIS with diffusion weighted imaging (DWI) and clinically indicated TTE admitted at two academic hospitals (Johns Hopkins Hospital or Johns Hopkins Bayview Hospital). Patients were excluded if they had a cerebral hemorrhage, were non-English speaking, had history of dementia or other neurological disease affecting the brain, uncorrected hearing or visual acuity deficits, ongoing sedation, or were unable to undergo MRI imaging. Patients were approached for written consent while inpatients in the hospital and were evaluated for neglect testing within a time window of 24 h to 5 days from time of RIS. This inclusion criteria reflects that of a parent study, which is an ongoing longitudinal, prospective parent study of the history of neglect in patients with RIS.

Neglect was determined using four, well-validated tests: horizontal line bisection, line cancellation, Ogden copy scene and gap detection and scored without knowledge of the study hypothesis, infarct volumes, or TTE findings (10). Four different tests to determine neglect were administered as multiple tests are more sensitive to neglect than a single test and might reveal different types of neglect (a focus of the parent study). The horizontal line bisection test entails drawing a horizontal line at the middle of the vertical line. Line cancellation asks patients to cross out all 28 vertical lines on the page. Ogden copy scene involves copying the “Ogden scene,” with the picture including a house, a fence and trees with the total number of components to the picture being 16, yielding a percent error. Finally, in gap detection, the patient is presented with 30 circles and is asked to circle complete circles and cross out circles with the gap, with 10 having a gap on the right, 10 on the left and 10 having no gap. Each test, while possibly revealing different types of neglect, is sufficient to diagnose a global state of neglect (yes/no). Some patients did not complete all of the tests, usually due to time constraints due to other clinical assessments, rehabilitation, visitors, etc.

Z-scores for neglect were defined based on normative data from 58 individuals of similar age who were admitted with TIA or were preoperative coronary artery bypass graft surgical candidates. These patients were chosen as norms due to the fact that they were felt to best represent the vascular risk factor profile that might influence the left ventricle and its function, but did not have an ischemic stroke. Although, using this population for norms might underestimate neglect (if controls showed neglect due to diastolic dysfunction, for example), it would not overestimate neglect or effect the correlations evaluated in this study. Neglect was considered as both a binary and a categorical variable. Neglect was defined as having a Z-score of ≥2 on one of any of the four tests. Neglect severity was then categorized for those who completed at least two tests as: no neglect, neglect on 1 or neglect on ≥2 or more tests, as previously reported (11). The data are available from the first author upon reasonable request.

TTE was obtained during admission at the discretion of the inpatient treatment team, but is fairly routine at the recruiting institution. All TTEs were performed by accredited technicians, with reports generated by board-certified cardiologists, and using the Philips IE33 ultrasound in standard 2D-analysis. The following variables were recorded from each patient's TTE report according to the pre-specified analysis: LV ejection fraction (LVEF, %), LV ventricular diastolic diameter (cm), LV systolic diameter (cm), Left atrial (LA) diameter (cm) and mitral E to A ratio. Mitral E and A point velocities are components of the mitral inflow signal to assess diastolic filling.

All patients underwent a brain 3-T MRI with contrast. Infarct volume was defined using DWI with manual lesion tracing performed using either ImageJ^© software or MRICron^©. Such techniques have been well-validated for calculating infarct volumes (12). Technicians measuring infarct volume were masked to patient characteristics, neglect scores and TTE variables.

Primary dependent variables were presence of neglect (vs. none) or severity of neglect as defined above. Pre-specified TTE variables (primary independent variables) were each evaluated descriptively and in univariate analyses in association with the outcome of interest using crude logistic regression models. Subsequent models then explored the association between presence of neglect (multivariable logistic regression), and then severity of neglect (multinomial logistic regression) with each specified TTE variable in separate analyses, with the same nested adjustment models. Each of the LV variables were considered as separate independent variables, in separate models and were pre-specified prior to the analysis. Covariates were obtained at the time of consent and were chosen based on the possibility of confounding the association between neglect and the TTE variables. In a sensitivity analysis, the Bonferroni correction was applied to adjust for the potential for multiple comparisons.

The nested adjustment models were as follows: Model 1 adjusted for sex, race, and age; Model 2 for Model 1 plus atrial fibrillation, and Model 3 for Model 2 plus education and DWI volume. Age was centered at 65 years old and considered in increments of 10 years. DWI volume was centered at the mean and considered in increments of 10 cm³. Education was dichotomized into 12th grade and over vs. below 12th grade. LVEF is reported per 10% increase.

All statistical analyses were performed using Stata v14.1. Two-sided p < 0.05 was considered statistically significant.

Two hundred and eighteen patients met inclusion criteria (Table 1) and 212 patients completed at least 2/4 of the assessments for neglect. One hundred and sixty patients did not have any evidence of neglect on any assessment, while 58 participants had evidence of neglect on at least one test. Among those who had neglect, 27 (47%) completed all four tests, while 81 (51%) without neglect completed all four tests. There was a significant difference in age (≥65 vs. <65 yo) among those who had neglect, vs. those who did not have neglect (χ = 6.25, P-value 0.01).

There was no significant association between any of the TTE variables of interest and presence of neglect (Table 2). When considering the relative risk of more severe neglect, a 1 cm increase in LV systolic diameter was associated with a higher relative risk of having neglect on two tests compared to those with no neglect for a participant 65 yo and similar risk factors (RRR = 1.83, 95% CI 1.01–3.32, Model 2, Table 3), but not after accounting for education and DWI volume (RRR = 1.68, 95% CI 0.89–3.19, Model 2). A 1 cm larger LA diameter was associated with approximately twice the risk of having neglect on two tests vs. no neglect (RRR = 2.23, 95% CI 1.04–4.77, Model 2), but again lost significance in the final model (RRR = 1.73, 95% CI 0.76–3.94). When considering the results after adjustment for the potential for multiple comparisons, the effect estimates were no longer statistically significant (p-value ≤ 0.01).