Predicting Domain-Specific Health-related Quality of Life using Acute Infarct Volume

Abstract

Background and Purpose

Limited data exist regarding the relationship between acute infarct volume and health-related quality of life (HRQOL) measures after ischemic stroke. We evaluated whether acute infarct volume predicts domain-specific Neuro-Quality of Life (Neuro-QOL) scores at 3 months after stroke.

Methods

Between 2012 and 2014, we prospectively enrolled consecutive patients with ischemic stroke and calculated infarct volume. Outcome scores at 3 months included modified Rankin score (mRS) and Neuro-QOL T-scores. We evaluated whether volume organized by quartiles predicted mRS and HRQOL scores at 3 months using logistic and linear regression as appropriate, adjusting for relevant covariates. We calculated variance accounted for (R²) overall and by volume for each domain of HRQOL.

Results

Among 490 patients (mean age 64.2 ± 15.86 years; 51.2% male; 63.3% Caucasian) included for analysis, 58 (11.8%) were disabled (mRS score >2) at 3 months. In unadjusted analysis, the highest volume quartile remained a significant predictor of one HRQOL domain, applied cognition-general concerns (R² 0.06, p<0.001). Our fully-adjusted prediction model explained 32–51% of the variance in HRQOL: upper extremity (R² 0.32), lower extremity (R² 0.51), executive function (R² 0.45), and general concerns (R² 0.34).

Conclusions

Acute infarct volume is a poor predictor of HRQOL domains after ischemic stroke, with the exception of the cognitive domain. Overall, clinical and imaging variables explained <50% of the variance in HRQOL outcomes at 3 months. Our data imply that a broad range of factors, some known and others undiscovered, may better predict post-stroke HRQOL than what is currently available.

Introduction

Acute infarct volume may be a useful biomarker of long-term outcomes after ischemic stroke^{1, 2}. However, the data evaluating the relationship between acute infarct volume measured by diffusion weighted imaging (DWI) long-term impairment and disability have been inconsistent and conflicting^3–5. Increasingly, patient-reported outcomes are utilized in clinical practice and research. The impact of acute infarct volume on health-related quality of life (HRQOL) scales has been poorly studied^{6, 7}. Standard clinical tools often lack adequate assessment of day-to-day functioning, made even more complicated by cognitive deficiencies and functional decline⁷. The National Institute of Health has created the Neuro-Quality of Life (Neuro-QOL) scale in part to address these limitations and needs⁸. No prior studies have assessed whether acute infarct volume predicts domain-specific HRQOL⁹. We, therefore, sought to evaluate whether acute infarct volume predicts domain-specific Neuro-QOL scores at 3 months and the variance in outcomes explained by acute infarct volume.

Methods

The local Institutional Review Board approved the study. Between August 2012 and July 2014, we prospectively enrolled consecutive patients presenting with imaging-verified ischemic stroke. Written informed consent was obtained from the patient or a legally authorized representative.

Ischemic stroke was defined as sudden onset of neurologic deficits and confirmed with corresponding hyperintense and hypointense lesions on DWI and acquired diffusion coefficient (ADC) sequences, respectively. Demographics, baseline functioning, insurance status, risk factors, hospital course, discharge disposition, interval rehabilitation services, and recurrent stroke¹⁰ were prospectively recorded. History of hypertension, diabetes mellitus, dyslipidemia, prior stroke, atrial fibrillation or flutter, and cardiac disease (history or angina, myocardial infarction, coronary bypass or intervention, or congestive heart failure) were defined by documented history, medications, or findings at presentation. Brain imaging was independently reviewed by two investigators (S.P. and C.L.) for presence of acute infarction along with location and vascular territory, blinded to outcome data. Acute infarct volume was calculated by an automated algorithm using thresholds for ADC and DWI. By consensus, clinical and radiographic data were prospectively reviewed to determine Trial of Org 10172 in Acute Stroke Treatment (TOAST) subtype for each confirmed case by consensus¹¹. The National Institutes of Health Stroke Scale (NIHSS) score was recorded on admission by certified clinician raters¹².

Image acquisition

Images were acquired on a 1.5T or 3T scanner (Siemens Medical Systems, Erlangen, Germany). Axial T1 spin echo images were used for reference. Pre-contrast images were preferentially used if available, otherwise post-contrast T1 images were used (typical sequence parameters: TR/TE 1500/45, 13–15 slices, in-plane resolution 1.15×1.15m, slice thickness 5mm, FOV 22cm, matrix 192×192, flip angle 30). Diffusion weighted images (TR/TE 4000/88, in-plane resolution 1.39×1.39mm, slice thickness 5mm, matrix 192×192) were acquired at b-values of 0 and 1000 and ADC maps were generated in-line on the scanner.

ADC lesion volume calculation: Before calculating the ADC lesion volume, anatomical reference images were used to mask out any non-brain matter. To do this, first the reference T1 image co-registered to the ADC image. Next, the T1 image was segmented into gray, white, and cerebrospinal fluid probability maps using a unified segmentation model¹³. The brain tissue mask was then defined as any voxel with >80% probability of being gray or white matter. All co-registration and segmentation was performed using SPM (SPM8; Wellcome Department of Cognitive Neurology, London, England). The final steps in lesion definition were performed in Matlab (The Mathworks Inc, Natick, United States). After masking out all non-brain tissue, a threshold approach to delineate the ischemic core was used¹⁴. Diffusion restricted voxels were defined as any voxel with ADC less than 680 ×10⁻⁶ mm²/s. To minimize noise, a minimum cluster size threshold of 0.145 mL was then applied to arrive at the final infarct volumes.

Outcome measures

For functional outcomes, the mRS was obtained using a validated telephone questionnaire at 3 months¹⁵. The mRS is scored from 0 (no symptoms) to 6 (dead), dichotomized between good (mRS score ≤2) and poor outcome (mRS score >2). We used the Neuro-QOL questionnaire, a health-related quality of life assessment tool (HRQOL) for common adult neurological disorders including stroke⁸. At 3 months, patients or their proxies completed 4 domain-specific assessments using the Neuro-QOL instrument (SF Version 1.0): upper extremity function (fine motor activity, reaching activities), lower extremity function (mobility including walking on stairs or uneven surfaces), applied cognition–executive function (planning, organizing, calculating, working with memory and learning), and applied cognition– general concerns (perceived difficulties with abilities such as memory, attention, and decision-making). Proxy assessments have been validated in previous studies featuring stroke populations^{10, 16}. Neuro-QOL results were expressed as T scores referenced to general United States population (with enrichment from a clinical neurologic sample) demographics with mean 50 and standard deviation (SD) 10^{17, 18}; additional information available at www.neuroqol.org⁸.

Statistical analysis

We calculated the means (SD) and medians (interquartile range [IQR]) for continuous variables including acute infarct volumes and HRQOL T-scores in any of the 4 Neuro-QOL domains and percentages for categorical variables. Mean T-scores were compared across baseline demographic, clinical, and hospital course variables using t-tests. Multiple linear regression analysis was used to examine associations between acute infarct volumes and 4 domains of Neuro-QOL using T-scores measured 3-months post-stroke. We developed multivariable models by sequentially adding variables that showed univariate association (p <0.10) in at least one domain. Covariates included in our model were: age, sex, race, payor status (no insurance, government insurance (Medicaid/Medicare), or private insurance), baseline mRS prior to stroke, current smoking within last 6 months, past medical history of atrial fibrillation, hypertension, dyslipidemia, type 2 diabetes mellitus, prior stroke, initial NIHSS score, acute infarct volume, TOAST subtype, proxy vs. patient reporting of HRQOL, discharge destination, recurrent stroke, and any rehabilitation therapy after index hospitalization. Variables known to be associated with QOL outcomes (eg. initial NIHSS, age etc) based on prior work¹⁰ were included in the final model regardless of univariate association. Three models were built, with each successive model repeating the adjustments of the previous model, as follows: model 1 (unadjusted), model 2 (adjustment for volume groups and baseline demographics: age, sex, race, smoking, baseline mRS, past medical history, payor status), and model 3 (model 2 plus adjustment for stroke presentation: initial NIHSS score, TOAST, proxy report, MCA territory involvement, discharge destination, ambulation at discharge, any rehabilitation after stroke, and recurrent stroke). The effect of acute infarct volume on the primary outcome (Neuro-QOL 4 domains) was evaluated with multiple linear regression of T-scores at 3 months post-stroke to determine the variance accounted for (VAF, R²) by quartile of acute infarct volume.

To check linearity, associations between outcomes and predictors were plotted, and the significance of adding a quadratic term for continuous variables was assessed for each regression model. Log-transforming predictors or outcomes did not achieve linearity due to the right skewness of volume. We used acute infarct volume based on quartile (Q) groups (Q1: ≤0.30 mL; Q2: 0.31–1.20 mL; Q3: 1.20 mL-5.28 mL; Q4: >5.28 mL) in all analyses for ease of interpretation and since all linear regression assumptions were not met using continuous volume measurements. Normality and constant variance of residuals were checked. Multicollinearity among predictors was checked and all predictors in the model 3 have variance inflation factor less than 10.

Multivariable logistic regression was used to assess the association between acute infarct volume and other predictors with mRS at 3 months (poor 0–2 vs. good 3–5) reporting the odds ratios (OR) and corresponding 95% confidence intervals. The 3 models were built similarly to the linear regression models for Neuro-QOL domains. All statistically analyses were performed using SAS 9.4 (Cary, NC). P-values were two-tailed. Statistical significance was considered to be a p-value <0.05.

Results

Among 490 patients (mean age 64.2 ± 15.86 years; 51.2% male; 63.3% Caucasian) included for analysis (Figure 1), 21 (4.3%) had recurrent stroke within 3 months and 58 (11.8%) were disabled (mRS score >2) at 3 months. Median acute infarct volume was 1.24 mL (IQR 0.31–5.08 mL). Proxy reporting of 3 month HRQOL data was 16.9% (Table 1). The overall mean T-scores for each HRQOL domain in each quartile of infarct volume is shown in Supplemental Table I. Mean T-scores decrease in all domains across ascending quartiles of infarct volume.

Figure 1.

Flowchart for study inclusion

Table 1.

Patient characteristics (n=490)

	Overall baseline demographics
Age (SD), years	64.2 (15.9)
Male, n (%)	251 (51.2%)
Acute infarct volume groups, n (%)
Quartile 1 (≤0.30 mL)	122 (24.9%)
Quartile 2 (0.31–1.20 mL)	122 (24.9%)
Quartile 3 (1.21–5.28 mL)	127 (25.9%)
Quartile 4 (>5.28 mL)	119 (24.3%)
Race, n (%)
Caucasian	310 (63.3%)
Non-Caucasian	180 (36.7%)
Payor Status, n (%)
No insurance	64 (13.1%)
Government insurance (Medicare/Medicaid)	246 (50.2%)
Private insurance	180 (36.7%)
Discharge destination, n (%)
Home	316 (64.5%)
Acute inpatient rehabilitation	167 (29.4%)
Skilled nursing facility and other	39 (8.0%)
Ambulating at discharge, n (%)
Independently with or without device	354 (72.2%)
With assistance from another person	115 (23.5%)
Unable to ambulate	21 (4.3%)
mRS at baseline, median (IQR)	0 (0–0)
Comorbid medical conditions, n (%)
Diabetes mellitus	131 (26.7%)
Dyslipidemia	329 (67.1%)
Atrial fibrillation	59 (12.0%)
Coronary artery disease	87 (17.8%)
Prior stroke	88 (18.0%)
Current smoker (<6 months)	87 (17.8%)
NIHSS score, median (IQR)	2 (1–6)
Toast subtype, n (%)
Cardioembolic	100 (20.4%)
Large artery atherosclerosis	83 (16.7%)
Small artery disease	67 (13.7%)
Other determined	79 (16.1%)
Cryptogenic	161 (32.9%)
3-month data
Proxy report, n (%)	81 (16.5%)
Recurrent stroke, n (%)	21 (4.3%)
mRS score 0–2, n (%)	385 (78.6%)
mRS score 3–5, n (%)	105 (21.4%)
Mean Neuro-QOL T-score (SD)
Upper extremity function (n=490)	49.1 (8.6)
Lower extremity function (n=490)	46.5 (9.8)
General concerns (n=486)	54.9 (6.8)
Executive function (n=486)	52.0 (9.5)

Abbreviations: IQR=interquartile range; mRS=modified Rankin Scale; NIHSS=National Institutes of Health Stroke Scale; QOL=quality of life; TOAST=Trial of Org 10172 in Acute Stroke Treatment

In the unadjusted analysis (Model 1), the highest quartile (>5.28 mL) volume group compared to the lowest quartile (≤0.30 mL), was associated with lower T-scores for each QOL domain: upper extremity (ΔT-score: −4.27, p<0.0001), lower extremity (ΔT-score: −3.66, p=0.0034), executive function (ΔT-score: −3.36, p=0.0061), and general concerns (ΔT-score: −3.48, p<0.0001). Acute infarct volume explained only 2–4% of the variances in T-scores by domain, with highest being 4% for applied cognition-general concerns (Table 2, Model 1). Univariate analysis is shown in Supplemental Table III.

Table 2.

Final fully adjusted prediction model of Neuro-QOL measures at 3 months

Predictors		Upper Extremity Function N=490			Lower Extremity Function N=490			Executive Function N=486			General concerns N=486			mRS at 3 month (Poor 3–5 vs Good 0–2) N=490
		B	SE	P-value	B	SE	P-value	B	SE	P-value	B	SE	P-value	OR	(95% CI)	P-value
Volume Groups	Q2 (0.31–1.20 mL)	0.44	0.97	0.65	1.63	0.91	0.07	0.45	0.92	0.63	0.31	0.72	0.67	2.21	(0.66, 7.34)	0.20
(reference=Q1)	Q3 (1.21–5.28mL)	−0.05	0.98	0.96	0.51	0.93	0.58	−0.34	0.94	0.71	−0.22	0.73	0.76	3.57	(1.12, 11.38)	0.03
	Q4 (>5.28 mL)	−0.95	1.08	0.38	−0.35	1.02	0.73	−0.97	1.03	0.35	−1.86	0.80	0.02	6.73	(2.00, 22.59)	0.002
Age		0.06	0.03	0.04	−0.05	0.03	0.10	−0.04	0.03	0.13	0.005	0.02	0.84	0.98	(0.95, 1.01)	0.27
Sex	Male	−0.36	0.69	0.60	−1.00	0.65	0.13	−0.56	0.66	0.40	−0.64	0.51	0.21	1.60	(0.79, 3.25)	0.19
Race	Caucasian	−0.75	0.75	0.32	0.02	0.71	0.98	0.65	0.72	0.37	−0.21	0.56	0.70	0.58	(0.26, 1.28)	0.18
Baseline mRS	2–5 vs 0–1	−4.08	1.56	0.01	−4.33	1.48	0.004	−4.18	1.49	0.005	−1.49	1.16	0.20	5.85	(1.65, 20.76)	0.006
Comorbid conditions	Hypertension	−0.26	0.88	0.77	−0.54	0.84	0.51	−1.17	0.84	0.16	−0.015	0.66	0.98	1.23	(0.50, 3.01)	0.66
	Diabetes mellitus	0.33	0.83	0.69	0.31	0.79	0.69	−0.05	0.80	0.95	−0.57	0.62	0.36	0.67	(0.29, 1.54)	0.34
	Hyperlipidemia	0.46	0.79	0.56	0.51	0.75	0.50	−0.58	0.76	0.45	−0.09	0.59	0.88	2.01	(0.89, 4.56)	0.10
	Atrial fibrillation	−1.05	1.25	0.40	0.30	1.19	0.80	−0.60	1.19	0.62	1.21	0.93	0.19	1.40	(0.46, 4.24)	0.56
	Coronary artery disease	−1.77	0.94	0.06	−1.71	0.89	0.06	−1.14	0.90	0.21	−1.06	0.70	0.13	1.07	(0.44, 2.58)	0.88
	Any prior stroke	0.61	0.91	0.51	0.75	0.86	0.38	−1.63	0.87	0.06	−0.80	0.68	0.24	0.54	(0.20, 1.45)	0.22
	Current smoking (<6 months)	1.28	0.92	0.17	1.80	0.87	0.04	1.04	0.88	0.24	−0.25	0.68	0.72	0.47	(0.17, 1.35)	0.16
Insurance	Government	−1.47	1.21	0.22	−0.14	1.14	0.90	0.70	1.16	0.55	1.15	0.90	0.20	0.17	(0.04, 0.64)	0.01
(reference=no insurance)	Private	0.37	0.93	0.69	2.23	0.88	0.01	1.32	0.88	0.14	1.99	0.69	0.004	0.14	(0.05, 0.40)	0.0002
Initial NIHSS score		−0.35	0.08	<.0001	−0.19	0.08	0.02	−0.25	0.08	0.0021	−0.05	0.06	0.47	1.06	(0.99, 1.14)	0.12
TOAST subtype	Large artery	−1.37	1.25	0.27	0.09	1.18	0.94	0.10	1.19	0.93	1.85	0.93	0.05	1.37	(0.44, 4.22)	0.59
(reference=Cardioembolic)	Small vessel	−2.67	1.35	0.05	−0.61	1.28	0.64	0.46	1.29	0.72	1.32	1.01	0.19	0.58	(0.10, 3.23)	0.53
	Other	−0.38	1.29	0.77	0.24	1.22	0.84	1.23	1.23	0.32	1.99	0.96	0.04	2.48	(0.74, 8.32)	0.14
	Cryptogenic	−1.11	1.10	0.31	−0.38	1.04	0.72	−0.39	1.05	0.71	0.24	0.82	0.77	2.51	(0.91, 6.89)	0.08
MCA territory	Yes vs. No	−0.48	0.70	0.49	−0.43	0.66	0.52	−0.24	0.67	0.72	−0.72	0.52	0.17	0.85	(0.42, 1.75)	0.66
Proxy	Yes vs. No	−5.35	1.02	<.0001	−8.83	0.97	<.0001	−10.68	1.02	<.0001	−6.39	0.79	<.0001	8.41	(3.49, 20.26)	<.0001
Discharge destination	Acute inpatient rehab	−0.07	1.07	0.95	−3.23	1.01	0.002	−1.09	1.11	0.32	−2.24	0.86	0.009	2.36	(0.92, 6.09)	0.08
(reference=home)	SNF/other	−4.77	1.45	0.001	−4.24	1.37	0.002	−5.35	1.51	0.001	−3.89	1.18	0.001	4.37	(1.17, 16.29)	0.028
Ambulating at discharge	With assistance from another person	−2.10	1.02	0.04	−1.54	0.97	0.11	−0.21	1.01	0.84	−0.13	0.79	0.87	2.18	(0.96, 4.93)	0.061
(reference=independently with or without device)	Unable to ambulate	−10.48	1.93	<.0001	−8.56	1.84	<.0001	−2.94	1.90	0.12	−3.03	1.48	0.04	8.88	(1.45, 54.52)	0.018
Baseline ambulatory status	non-ambulatory vs. ambulatory	−2.10	2.95	0.48	−8.77	2.79	0.002	−1.13	2.85	0.69	0.06	2.22	0.98
Any rehabilitation after stroke		−2.15	0.93	0.02	−2.45	0.88	0.01	1.23	0.89	0.17	0.97	0.69	0.16	6.48	(2.24, 18.77)	0.0006
Recurrent Stroke	Yes vs. No	2.90	1.67	0.08	3.34	1.59	0.04	4.42	1.59	0.01	4.84	1.24	0.001	0.36	(0.06, 2.21)	0.27
Intercept		48.27	3.08	<.0001	50.80	2.91	<.0001	55.27	2.94	<.0001	51.86	2.29	<.0001
Adjusted R2	Model 1	3%			2%			1%			4%			6%
	Model 2	10%			24%			19%			13%			22%
	Model 3	32%			51%			45%			34%			44%
Volume Groups trend test p-value		0.435			0.038			0.495			0.256			0.001

Abbreviations: mRS=modified Rankin Scale; NIHSS=National Institutes of Health Stroke Scale; MCA=Middle cerebral artery

Model 1 has only acute infarct volume groups

Model 2 adjusted for baseline demographics (age, sex, race, baseline mRS, comorbid conditions (past medical diagnoses, smoking), payor status) and acute infarct volume groups

Model 3 adjusted for Model 2 covariates plus stroke presentation (initial NIHSS, TOAST, MCA territory, proxy report, discharge destination, ambulation at discharge, any rehabilitation after stroke, recurrent stroke) and acute infarct volume groups

Our adjusted model featuring volume and clinical demographics (Supplemental Table IV), with covariates described in methods, explained 13–27% of the variance in HRQOL; upper extremity (R² 0.13), lower extremity (R² 0.27), executive function (R² 0.22), and general concerns (R² 0.16). The highest quartile (>5.28 mL) volume group compared to the lowest quartile (≤0.30 mL) was associated with lower T-scores for each QOL domain: upper extremity (ΔT-score: −4.27, p<0.0001), lower extremity (ΔT-score: −3.66, p=0.0034), executive function (ΔT-score: −3.36, p=0.0061), and general concerns (ΔT-score: −3.48, p<0.0001). Acute infarct volume explained only 3–7% of the variances in T-scores by domain, with the most significant contributions in applied cognition-general concerns (R² 0.07). The second and third quartiles were not associated with Neuro-QOL T-scores. In Model 2, besides volume, other significant baseline predictors were pre-stroke mRS, prior stroke, age, and payor status.

In our fully adjusted model (Table 2, Model 3), the highest volume quartile remained a significant predictor of only 1 HRQOL domain, applied cognition-general concerns (R² 0.06, p=0.303), compared to the other HRQOL domains: upper extremity (R² 0.03, p=0.455), lower extremity (R² 0.04, p=0.111), executive function (R² 0.010, p=0.556). In model 3, baseline mRS, initial NIHSS, CAD history, smoking, recurrent stroke, TOAST criteria, proxy reporting, and discharge destination, ambulation at discharge, recurrent stroke, and rehabilitation after stroke were significant predictors of Neuro-QOL domains. This model, including clinical demographics, volume, and stroke presentation, explained 32–51% of the variance in HRQOL; upper extremity (R² 0.32), lower extremity (R² 0.51), executive function concerns (R² 0.45), and general concerns (R² 0.34) (Table 2, Model 3).

In multivariable logistic regression analysis, acute infarct volume demonstrated a graded effect on poor vs. good outcome (mRS >2 vs. ≤2, respectively) in both models 2 and 3 (Table 2 and Supplemental Table IIII) with the highest quartile group showing a 7-fold increase in the likelihood of poor outcome compared to the lowest quartile (Q1 9.76%, Q2=17.07%, Q3=22.83%, and Q4 37.82%, p<0.0001).

Discussion

Acute infarct volume was associated with HRQOL domains in the unadjusted analysis. However, it only contributed modestly (33%) to prediction of a single HRQOL domain, applied cognition-general concerns, in the fully adjusted analysis. We observed that clinical and imaging factors explained only between 2132% and 51% of the variance in 3-month domain-specific HRQOL scores. In contrast, the largest acute infarct volume quartile showed a significant effect on 3-month disability by mRS independent of demographic and clinical factors. These data suggest that acute infarct volume may be a modest predictor of global function and disability (e.g., mRS) but provides poor discrimination of domain-specific HRQOL (e.g., upper limb). Lesion location specific to the function would likely be more significant, for instance the corticospinal tract for upper extremity function. Overall, more than half of the variance in specific domains of HRQOL is unexplained despite accounting for clinical, treatment, and imaging factors. Our data imply that other undiscovered biologic, psychosocial, clinical, and imaging factors may better predict post-stroke HRQOL than what is currently available.

There are conflicting data on the relationship between acute infarct volume and outcomes after stroke^{5, 7}. The largest previous study done (n=108) found no significant relationship between acute infarct volume and mRS at 3 months after stroke¹⁹. Other smaller studies have noted that acute infarct volume can predict long-term disability^{5, 20}. Schiemanck et al. found that infarct volume was a significant but modest predictor of long-term HRQOL (R²=−0.45, p<0.01) and mRS (R²=0.39, p<0.01)⁶. Our study results suggest that acute infarct volume is an independent predictor of global disability, as measured by mRS, but may not be as valuable in predicting domain-specific HRQOL.

Besides acute infarct volume, there are other imaging biomarkers that may better predict outcomes after stroke. Small studies of axonal tract and structural integrity using diffusion tensor imaging suggest strong correlation with motor outcomes²¹; yet, HRQOL measures were not assessed. Infarct location has also shown a direct association with neurologic deficits after stroke²². Smaller subcortical, deep white matter, internal capsule and brainstem infarcts correlate with deficits out of proportion to the infarct volume than superficial cortical lesions^{3, 4, 7, 23}. However, specific QOL domains have not been previously reported. These imaging biomarkers, and additional ones such as white matter hyperintensity, silent infarcts, and microbleeds, could be incorporated to improve future predictive models of HRQOL and disability after stroke.

HRQOL scores provide additional context for patients with poor outcomes, assess day-to-day functioning, and may even overcome ceiling effects from traditional measures such as the mRS¹⁰. From previous studies, impairment in HRQOL, predicted from baseline demographics and clinical information, is common at 3 months after stroke, and can even be present despite no disability on mRS and minimal deficits by NIHSS score^{10, 24}. While we confirm these prior clinical predictors, we are the first to evaluate the explained variance in HRQOL accounted for by acute infarct volume.

While HRQOL represents an important patient-centered outcome measure, predicting HRQOL remains difficult. In adult cardiology, HRQOL data have been incorporated into heart failure clinical trials²⁵, but most have shown limited prediction of HRQOL outcomes using biological variables such as ejection fraction. In congenital heart disease, there is a high incidence of functional impairment²⁶. As short-term mortality has been reduced, research attention has turned to long-term morbidity and HRQOL, drawing parallels to advances in stroke care. In pediatric cardiology, HRQOL has been significantly linked with physical and psychosocial domains²⁶. These results reaffirm our conclusions that other psychosocial and biologic variables may exist that better explain HRQOL after stroke.

A major strength of our study is the large sample size with rigorous follow-up. There are, however, several limitations. First, this study was conducted in a single urban academic institution. The differences between institutions related to MRI acquisition and patient characteristics should not vary substantially, but our results may not be generalizable to all settings. Our cohort also included consecutive ischemic stroke patients, the majority of whom had mild stroke and small infarct volumes. Thus, our results may not be reflective of cohorts with large infarcts and more severe strokes. Second, besides vascular territory stratification, more specific location was not accounted for in this study. Third, Neuro-QOL measures are patient-reported outcomes. Objective measures of cognition and psychological function were not performed in this study. HRQOL assessments require active engagement with the patient. Results may be susceptible to inaccuracy if patients lose interest or tire during assessment. Fourth, not all domains of HRQOL were assessed including fatigue and depression. Fifth, the graded response we noted between infarct volume and mRS outcomes may be due to the small sample size of those with poor outcomes and small volumes. Finally, we did not assess pre-stroke measures of HRQOL and are unable to report changes in T-scores post-stroke. Pre-stroke measures of HRQOL are cumbersome to perform and not validated in hospitalized patients.

Our study is the first, to our knowledge, to incorporate acute infarct volume and clinical, demographic, and hospital course into a predictive model for domain-specific HRQOL measures using Neuro-QOL in a large prospective cohort. HRQOL impairments have been shown to correlate strongly with functional disability²³. Since HRQOL is relevant to both patients and caregivers, prediction of HRQOL after stroke warrants investigation. We found that infarct volume accounts for only a small fraction of explained variance in HRQOL measures at 3 months after ischemic stroke and is a significant predictor of only 1 domain, applied cognition-general concerns. Our data imply that a broad range of factors, some known and other undiscovered, may better predict post-stroke HRQOL than what is currently available. Future directions for imaging-based predictive models of HRQOL should incorporate infarct volume but also voxel- and network-based infarct location and integrity of affected tracts and structures²² in addition to psychosocial determinants.