The impact of apraxia and neglect on early rehabilitation outcome after stroke

This study followed the principles of the Declaration of Helsinki in its current version from October 2013 (Fortaleza, Brasilia). Data were analyzed retrospectively from the medical records compiled between November 2017 and October 2020 at the Department of Neurology of the University Hospital Cologne. All patients were enrolled in the Neurological-Neurosurgical Early Rehabilitation program for acute to early subacute stroke patients established in Germany [42]. Only patients with an Early Rehabilitation Barthel Index [41] of less than 30 are eligible for this program. The patients receive at least 300 min of therapy per day, including specialized nursing, physical, occupational, and speech therapy. On average, patients are included in the early rehabilitation program for 2–3 weeks before being transferred to standard in-patient or out-patient rehabilitation centers, their nursing homes, or their home. The usual duration of the early rehabilitation in our department is two weeks for patients without invasive interventions and three weeks for patients after invasive interventions (e.g., thrombectomy). There was some variation regarding the duration of the early rehabilitation program due to holidays, weekends, and extra days compensating for medically necessary pauses. In our sample of 150 acute to sub-acute stroke patients, 110 patients were enrolled in the 2-week program and 40 patients were enrolled into the 3-week program. Thus, most patients stayed in the program for 14 days only, where the behavioural and neuropsychological assessments were usually performed on day 2 and the last 2–3 days of the program.

In total, the medical records of 515 patients were screened. An initial neuropsychological assessment was available for 379 patients since these patients had sufficient command of German, could follow instructions, did not refuse the assessment, and were not under legal guardianship (see Fig. 1).

A further 32 patients were excluded because a neurosurgical intervention or an (intermittent or persistent) delirium interfered with the neuropsychological follow-up assessments. Of the remaining 347 patients who underwent (at least in part) the three neuropsychological assessments at admission, at mid-term, and at discharge, 150 patients suffered from a hemispheric ischemic or hemorrhagic stroke and did not present with any of the following exclusion criteria: age > 90 years, additional contralateral stroke, a clinically relevant stroke of the cerebellum, brain stem, pons, or medulla oblongata, subarachnoid hemorrhage, other neurological diseases affecting the central nervous system, clinically relevant psychiatric diseases (e.g., psychosis, addiction, major depression), and dementia. Data of patients with previous strokes of the same hemisphere or an ipsilateral cerebellar stroke without motor symptoms (ataxia and intentional tremor) of the hand involved in the testing were not excluded, as well as patients suffering post-stroke depression. Table 1 shows the demographic and clinical features of the final sample (n = 150).

Handedness was assessed by the Edinburgh Handedness Questionnaire [38].

Three different parameters for functional outcome after stroke were employed: The National Institute of Health Stroke Scale, NIHSS), the German Early Rehabilitation Barthel Index (Frühreha-Barthel-Index, FRBI), and the composite score of the Functional Independence Measure (FIM) and Functional Assessment Measure (FAM). The NIHSS is a standard evaluator-based score to describe the severity of stroke symptoms [19]. It contains 11 items (each 0–max. 4 points); higher scores represent more severe impairments (total score 0–max. 42). The FRBI is based on the Barthel Index (BI) comprising 10 items (total score 0–max. 100, [34]) plus seven additional items with possible negative scores 0 to − 325 [41]. The FIM estimates disability in terms of caregivers’ burden considering motor and cognitive functions/deficits [20]. It covers 18 items (total score 0–max. 126). Commonly, the FIM is used together with the FAM [21], contributing additional information about psychosocial functions. The FAM consists of 13 items (total score 0–max. 42).

The cognitive status of the stroke patients was tested using the Cologne Neuropsychological Screening for Stroke Patients (KöpSS) [28], specifically with the KöpSS versions A and B at admission and discharge, respectively. The KöpSS can be performed even by severely impaired patients and examines the general cognitive performance level and seven relevant cognitive domains. Multiple subtasks assess each domain. Cut-off values exist for each domain and subtask. Overall, the cut-off value indicating cognitive impairment is set at ≤ 98 (of max. 108). In the current study, we applied a modified version of the KöpSS to avoid duplicate testing of cognitive functions by the KöpSS and the below-described apraxia and neglect assessments. Furthermore, KöpSS-items that required writing with the right hand or bimanual movements could not be performed by the right-handed stroke patients with motor impairments. The modified KöpSS (total score 0–max. 70) still encompassed 5 domains (orientation, language without the writing subtask, calculation, memory, attention and executive functions).

Apraxic deficits were assessed with the Cologne Apraxia Screening (KAS) [48] and the finger imitation test by Goldenberg [18]. The KAS (20 items, total score 0–max. 80, cut-off ≤ 76) comprises tasks that assess pantomiming the use of objects and imitation and include bucco-facial and arm/hand gestures resulting in four subtests (bucco-facial pantomime, arm/hand pantomime, bucco-facial imitation, arm/hand imitation). Patients are instructed to perform the pantomime corresponding to an object or imitate the presented gesture. All stimuli (objects, gestures to be imitated) are presented using photos. Patients with RH stroke were assessed with the KAS-R, a shorter version of the KAS (12 items, total score 0–max. 48, cut-off ≤ 46) with mirror-inverted stimuli that facilitate the spatial perception of the presented material [52]. For reasons of comparability, the raw scores of both KAS versions were transformed to a relative score for analysis (relative score = raw score * 100/maximally possible score). The Goldenberg Finger Imitation Test consists of 10 finger configurations that the patient should imitate with the ipsilesional hand in a mirror-like fashion after a demonstration by the examiner. For each item, 2 points are allocated for an immediate correct response, 1 point if the second attempt is successful, and 0 points if the patient fails on both attempts (total score 0–max. 20, cut-off-score ≤ 16).

The neglect assessment was based on two subtests of the Neglect-Test (NET) [15]—the German version of the Behavioral Inattention Assessment (BIT) [51]. The line bisection test (0–max. 9 points, cut-off ≤ 7) was used to assess a putative spatial perception bias, while the star cancellation test provided a score for visual exploration and allowed calculating a laterality quotient (LQ). Here, the absolute value of the LQ was considered to reflect spatial biases in either direction (LQ =|(hits contralesional-hits ipsilesional)/(hits contralesional + hits ipsilesional)|, range LQ: 0–1, cut-off ≥|0.2|; [14]). Figure 2 shows a conceptual schema of the dependent and independent variables.

Based on a screening of 515 hospitalized stroke patients from the University Hospital Cologne’s early rehabilitation ward, 150 stroke patients (75 left-hemispheric strokes, 75 right hemispheric strokes) fulfilled the inclusion criteria and were enrolled in this observational, longitudinal study. The cognitive status was documented at admission using the KöpSS Parallel-Version A and at discharge using the KöpSS Parallel-Version B. We also used the scores of the functional outcome scales at admission to and discharge from the Neurological-Neurosurgical Early Rehabilitation program (FRBI, NIHSS, and FIM/FAM). The assessments of apraxia (KAS, Goldenberg Imitation Test) and neglect (NET-line bisection, NET-star cancellation) took place at midterm.

To allow for the maximal exploitation of the current data set by using more accurate methods we treated the data (e.g., FIM/FAM scores) like interval-scaled data in the current statistical evaluation. Furthermore, since we used the same method for all statistical evaluations, the current approach ensures comparability both within the study and with previous studies [8, 31, 36, 46].

To delineate the inherent structure of apraxia and neglect tests and reduce the number of independent variables, the two apraxia tests and the two neglect tests were evaluated using a principal component analysis (PCA). Two components reaching the Kaiser’s Criterion (eigenvalue > 1) were obtained and underwent a varimax rotation. Individual component scores were calculated by the regression method based on the two components.

Furthermore, we applied the wild bootstrap method, which is applicable if normality assumptions are not fulfilled (i.e., in case of exceeding skewness and heteroscedasticity). Moreover, based on 2000 samples, bias corrected and accelerated (BCa) confidence intervals were applied.

We applied four multiple regression analyses to identify relevant predictors of the cognitive and functional outcome scores at discharge. The dependent variables were the KöpSS score, the FRBI, NIHSS, and the FIM/FAM Composite Score at discharge (T2). Age, the Neglect and the Apraxia Component, and the score of the respective scales at admission (T1) were used as independent variables. All analyses were carried out using SPSS Version 27 (SPSS Inc., Chicago, Illinois).

Finally, to control potential influences of the affected hemisphere, four supplementary moderator analyses were performed using the PROCESS macro for SPSS (Version 3.5) [24]. The variables were used in analogy to the multiple regression analyses. In detail, the Neglect and the Apraxia Components were entered as independent variables, the remaining independent variables were classified as covariates, and the affected hemisphere (RH or LH) was added as a moderator.

Due to missing values, the number of cases entering the respective analyses differed and will be mentioned separately for each analysis. Note that the minimal number of cases was 106.

Out of the 139 patients who could be assessed with the Edinburgh Handedness Questionnaire, 133 (95.7%) were right-handed. The average time post-stroke at admission to the Neurological-Neurosurgical Early Rehabilitation program was 5.0 days (SD = 3.7). The interval between the assessments at admission and discharge was, on average, 12.2 days (SD = 3.9, range 5–23). The interval between the midterm and discharge assessments was 7.9 days (SD = 4.4, range 1–19). Visuospatial neglect was revealed in 45 (33.8%) of 133 patients by the line bisection test (mean = 7.14, SD = 2.82, range 0–9) and in 26 (19.1%) of 136 patients by the star cancellation test (mean = 0.16, SD = 0.33, range 0–1). Twenty-four (18.9%) of 127 patients who performed both the line bisection test and the star cancellation test showed impaired performance in one test and 20 patients (15.8%) in both neglect tests. Apraxia was diagnosed in 79 (54.1%) of 146 patients by the Goldenberg Finger imitation test (mean = 14.25, SD = 5.54, range 0–20) and for 100 (70.9%) of 141 patients by the KAS (mean = 81.12, SD = 21.6, range 0–100). Forty-two patients of 136 patients (30.9%) who performed both the Goldenberg Finger imitation test and the KAS scored below the cut-off in one of the apraxia tests, 63 (46.3%) patients were impaired in both apraxia tests.

Following previous studies [30, 50], we diagnosed apraxia and neglect when patients scored below the cut-off in at least one test. Accordingly, 105 (of 134 patients, 77.2%) patients were apraxic, and 44 (of 127 patients, 34.6%) suffered from neglect.

A significant improvement was observed for the cognitive and all functional scales during the rehabilitation period (Table 2).

Entering the scores of the two apraxia tests and the two neglect tests into a Principal Component Analysis (PCA) revealed two components that fulfilled the Kaiser criterion (eigenvalue > 1), explaining 79.8% of the variance (Table 3). A subsequent varimax rotation revealed that the first component (hereafter: Neglect Component) primarily represented the scores of the neglect tests, while the second component reflected the apraxia test scores (hereafter: Apraxia Component). Individual component scores for further calculations were obtained using the regression method. Higher scores indicate better performance (i.e., less severe apraxia or neglect).

For all scales at discharge, the multiple regression analysis yielded a strong predictive value of the respective scale at admission. The KöpSS score at admission predicted the KöpSS score at discharge (b = 0.425, t = 6.714, p < 0.001, n = 100, Table 4). The initial assessment of FIM/FAM predicted the final variance of the FIM/FAM compound score (b = 1.187, 16.324, p < 0.001, n = 113). Likewise, the initial NIHSS score predicted the final NIHSS score (b = 0.699, t = 9.075, p < 0.001, n = 106). Finally, the initial FRBI score predicted the final FRBI measure (b = 0.55, t = 5.858, p < 0.001, n = 121).

Besides the dominant effects of the initial behavioral scores, we found significant effects on cognitive and rehabilitation outcomes for the components reflecting apraxia and neglect severity. In particular, the Apraxia Component accounted for the KöpSS score variability (b = 5.084, t = 5.378, p < 0.001; see Fig. 3A). In contrast, the Neglect Component showed an effect (b = 5.668, t = 2.34, p < 0.05; see Fig. 3B) on the discharge FRBI, with the latter effect being independent of the affected hemisphere. A significant negative impact of age (b = −0.166, t = −3.398, p < 0.01) and education (b = 0.373, t = 2.011, p < 0.05) was found only for the KöpSS score at discharge.

Supplementary moderator analysis revealed a significant differential effect of the affected hemisphere on the Neglect Component when predicting the KöpSS score at discharge. As illustrated in Fig. 3C, more severe neglect (as indicated by a lower Neglect Component score) predicted 2.9% of the variance of the final KöpSS score (F(1,93) = 9.45, p < 0.01, n = 100) for the patients suffering from a RH but not LH stroke. No difference between the affected hemispheres was found for the Apraxia Component (or any other scale/ score).

Apraxia and neglect are multi-faceted syndromes. Because of limited statistical power, this study could not account for potential differences between the different facets of apraxia, e.g., bucco-facial versus limb apraxia [32], or neglect, e.g., peripersonal versus extra-personal neglect [49]. The current statistical power also precluded an analysis of subscales of the cognitive and functional outcome scores. Finally, further studies are warranted to investigate other common cognitive sequelae of LH stroke like aphasia [50] or RH stroke like anosognosia [47]. A further limitation of our study is that data about the long-term outcome of the current stroke patient sample are not available. Future studies are warranted that investigate the important relationship between improvements in the early rehabilitation after stroke (and their predictors) and the functional and cognitive long-term outcome after stroke. Concerning the statistical methods applied, we know that the FIM/FAM is an instrument with an ordinal Likert scale like many clinical questionnaires. However, to maximally exploit the current data set and to ensure the comparability with previous studies, we treated the FIM/FAM data like interval-scaled data in the current statistical evaluation.