Abstract
Background
Spatial neglect (SN) is a risk factor of in-hospital falls among stroke survivors. Our prior study showed that receiving more sessions of prism adaptation treatment (PAT) in inpatient rehabilitation facilities (IRFs) predicted greater SN reduction and functional improvement.
Objective
To identify circumstances of falls specific to SN and explore whether increasing PAT sessions may reduce fall incidence.
Methods
The present study was a retrospective analysis of fall-related documentation, clinical data and notes regarding SN assessment, and treatment as part of standard care. Records of 3020 patients admitted to an IRF after stroke were reviewed, and 1489 (49%) had SN based on the Catherine Bergego Scale (CBS).
Results
A total of 276 patients (9% of all patients) fell at least once during their IRF stay, and 173 fallers (67% of all fallers) had SN. SN increased fall incidence (incident rate ratio [IRR] = 1.44, P = .005) after controlling for 5 covariates including age, sex, cognitive and motor functional level at admission, and length of stay. While independent of type or cause of falls, SN was associated with location of falls—SN increased fall incidence in hospital rooms (IRR = 1.55, P = .024), after controlling for the 5 covariates. 62 (36%) of fallers with SN received a median of 4.5 PAT sessions (range = 1-11; interquartile range = 2-10). Increased PAT sessions were associated with fewer falls after PAT (IRR = 0.82, P = .022), controlling for the 5 covariates and 2 additional factors including CBS and number of falls before PAT.
Conclusions
Treatment for SN such as PAT should be considered to reduce the risk of falls in these patients. Future research is needed to determine fall prevention measures for stroke survivors with SN, especially in their hospital rooms.
Keywords
Introduction
Stroke is a leading cause of disabilities in adults. Spatial neglect 1 (SN), a common neuropsychological disorder 2 after stroke, is caused by damage to neural networks critical to spatial processing and attention orientation,3-5 resulting in abnormalities in multiple perceptual modalities6-9 with the visual modality being most observable and investigated, and resulting in impairment in multiple cognitive and motor functions.10-13 SN is characterized by difficulties with reporting, responding, or orienting to information presented or mentally represented on the contralesional side of space,2,14 accompanied by anosognosia. 15 SN impedes rehabilitation progress16-19 and exacerbates disabilities.20-23
The estimated incidence of SN varies, depending on diagnostic methods and time post stroke. 24 Our previous study 25 showed that 58% of stroke patients (N = 3645) in inpatient rehabilitation facilities (IRFs) had SN, based on an ecological assessment, that is, the Catherine Bergego Scale (CBS). 26 In the United States, individuals are usually admitted to an IRF within the first 1 to 2 weeks post stroke. Patients receive 3 hours of rehabilitative therapies 5 days per week over 2 to 3 weeks, and the therapies include physical, occupational, and speech therapies. Patients with SN often have poorer rehabilitation outcome and stay in IRFs longer than patients without SN.16,27
Importantly, SN is a risk factor of in-hospital falls among stroke survivors.16,28 No fall prevention measure is yet decisively effective. 29 If certain circumstances related to a fall are specific to patients with SN, targeted measures (eg, modified environment, increased staff awareness) may be implemented to prevent falls. Thus, the first objective of the present study was to identify fall circumstances specific to patients with SN.
In addition, treatment for SN should be provided. Prism adaptation treatment (PAT) may have beneficial impacts on falls. PAT is an implicit sensorimotor training, requiring no explicit learning of strategies or relearning of skills, and the therapeutic effects are multi-modal and across domains30,31 with changes in neural activation and brain connectivity between the cerebral cortices and cerebellum.32,33 PAT was shown to improve postural balance in patients with SN,34,35 which might lead to fall prevention. Our previous study showed that receiving a greater number of PAT sessions was associated with greater SN reduction and functional improvement. 36 The second objective of the present study was to explore whether PAT could reduce the risk of in-hospital falls.
Methods
Data Extraction
Four years (2019-2022) of clinical data from 3 campuses of an IRF were reviewed. Patients whose diagnosis was coded as stroke and whose CBS score was available were included in the analysis. A fall was defined, following the U.S. Centers for Medicare and Medicaid Services (CMS), as “an unintentional change in position coming to rest on the ground, floor, or onto the next lower surface (eg, onto a bed, chair, or bedside mat). An intercepted fall was considered a fall.”
We extracted data related to CBS, falls, and PAT. We also extracted demographic information (age, sex, race, Hispanic origin), administrative information (days post stroke at admission, length of stay), and selected items from the Continuity Assessment Record and Evaluation (CARE). All personal health information was removed. Due to the nature of retrospective analyses on de-identified data, the study was exempt from the review of the Institutional Review Board at Kessler Foundation.
SN Assessment
CBS was part of standard care in the IRF, used in all neurological patients including stroke patients, as a method to detect SN and measure SN severity. CBS was administered by occupational therapists (OTs) following Kessler Foundation Neglect Assessment Process (KF-NAP®) 26 about 5 days (median) after IRF admission. KF-NAP provides standardized guidance and instructions to make observations and assign scores on 10 CBS items, including gaze orientation, limb awareness, auditory attention, personal belongings, dressing, grooming, navigation, collisions, meals, and cleaning after meals. Each item is rated from 0 (no neglect) to 3 (severe neglect). When an item is not observed and thus not rated, the final CBS score is prorated (ie, average score of all rated items, multiplied by 10). 37 The range of CBS scores is 0 to 30, greater values indicating greater SN severity. For the purpose of the present study, we categorized patients into 2 groups: no SN if CBS = 0 (SN−) and having SN if CBS > 0 (SN+).
Activities of Daily Living Assessment
CARE is developed by the CMS, replacing the Functional Independence Measure (FIM) as the gold standard of the rehabilitation functional outcome measure. CARE is administered by OTs, physical therapists, speech therapists, and nurses. For the purpose of the present study, we created an activities of daily living (ADL) score with 7 self-care items (eating, oral hygiene, toileting hygiene, shower/bathe self, upper body dressing, lower body dressing, and putting on/taking off footwear) and 8 mobility items (roll left to right, sit to lying, lying to sitting on side of bed, sit to stand, chair/bed-to-chair transfer, toilet transfer, car transfer, and walk 10 feet), with each item rated from 1 (dependent) to 6 (independent). Assessment of these items was administered at IRF admission and discharge, and thus ADL scores were obtained for those 2 time points. When the rating of an item was unattainable, the ADL score was prorated (ie, average score of all rated items, multiplied by 15), and the range of scores was from 15 to 90, greater values indicating higher function.
Cognitive Status Assessment
The Brief Interview for Mental Status (BIMS) score was extracted from CARE as an index for cognitive status. BIMS assesses the ability to repeat 3 words (sock, blue, and bed), orient to the year, month, and day of the week, and recall the 3 words. The range of BIMS scores is 0 to 15, greater values indicating better performance. BIMS was administered at IRF admission. For the purpose of the present study, the BIMS score was assigned 0 in patients unable to be tested.
Prism Adaptation Treatment
PAT equipment was available in the IRF, and OTs were trained to use it as part of their standard care. Specifically, OTs provided PAT to patients with SN, following a standardized protocol, Kessler Foundation Prism Adaption Treatment (KF-PAT®) 38 and using the KF-PAT Portable Kit that included all the apparatuses and materials necessary for delivering PAT. Eligibility for PAT includes no history of vertigo, no lesion or functional impairment in the cerebellum, ability to make controlled arm movement of the less-affected upper limb, not being blind in both eyes, and ability to engage in PAT. In a PAT session, patients completed 60 arm reaching movements from their chest to a target on the table while wearing 20-diopter prism lenses, which shifted their visual field laterally by 11.4 degrees of visual angle toward the non-neglected side of space, and a board that blocked visual feedback of the first half to two-thirds of arm reaching trajectory (for an illustration, see Goedert et al 39 ) Prism aftereffects were measured using 2 pointing tasks immediately before putting on prisms and after taking them off, resulting in changes of pointing errors. An expected aftereffect was an error change being more toward the neglected side of space. If no aftereffect was detected in the first 3 consecutive PAT sessions, OTs may have discontinued PAT per KF-PAT protocol. 38 A PAT session took 10 to 20 minutes within a 45-minute occupational therapy.
Qualitative Data Coding
De-identified documentation of falls was reviewed by 2 authors who were blind to patients’ SN diagnosis. These 2 authors were practicing OTs, working in one of the IRF campuses for at least 6 years and familiar with abbreviations and short-hand notes. They identified a set of fall circumstances (ie, codes) and definitions which emerged from the first review of 30 falls conducted together. They then independently coded the rest by a set of 30 to 50 falls. A third author compiled the codes and facilitated discussions, leading to consensus with the original coders regarding discrepancies. During the 7-month process, the agreement rate increased from 90% to 96% before each consensus was reached. Four categories of falls (ie, themes) were then identified based on 22 fall circumstances (Table 1).
Categories and Circumstances of Fall Incidents.
Covariates
Older age, poorer cognitive and motor abilities at admission, and longer length of stay have been shown as potential risk factors of in-hospital falls.16,40 Thus, age, BIMS, ADL at admission, and length of stay were included in the analysis as a priori covariates. In addition, a preliminary multivariate Poisson regress analysis was conducted to explore whether other factors (sex, race, Hispanic origin, and time post stroke at admission) would be considered a covariate too. Based on predetermined criteria (incident rate ratio [IRR] >1.005 or <0.995), we included sex as a covariate as well. These 5 a priori covariates were included in the analysis.
Statistical Analyses
All the continuous variables were skewed. Thus, summaries were described using medians and interquartile ranges (IQRs), and group comparisons were performed using the Mann–Whitney U test. Categorical variables were described using counts and percentages, and group comparisons of categorical variables were conducted using the χ2 test. Outcome variables, which were rates of falls in specific context, were examined using Poisson regression models. The alpha level of each statistical test was set at .05. All analyses were performed using STATA-SE 16.1.
Results
Demographic and Clinical Characteristics
A total of 3020 patients met the criteria for being included in the analysis. A total of 1489 (49%) patients had SN (CBS >0; median = 5.56, IQR = 2.22-11.67). Table 2 summarizes the information and the results of comparisons between the SN− and SN+ groups. Demographic information reflected the characteristics of the cohort: In comparison with the SN− group, the SN+ group was 1 year younger and had a lower ratio of White individuals (62.1% vs 71.7%). The 2 groups did not differ statistically in sex or Hispanic origin. Clinical characteristics relevant to SN (as reported in prior studies)16-19,27,28,41 were observed: The SN+ group was admitted to IRF 2 days later post stroke, had a lower cognitive status indicated by 2 points on BIMS at admission, had lower ADL scores at admission and discharge by 12.8 and 14.6 points, respectively, and stayed 6 days longer in IRF than the SN− group.
Patient Characteristics and Fall-related Information.
Abbreviations: SN, spatial neglect; CBS, Catherine Bergego Scale; BIMS, Brief Interview for Mental Status; ADL, activities of daily living.
P values are results from comparisons between patients without spatial neglect (SN−) and patients with spatial neglect (SN+), and the test used for each comparison is noted. Categorical variables are presented with counts and percentages in parenthesis. Continuous variables are presented with medians and interquartile ranges in parenthesis.
Falls
276 patients (9%) had at least 1 fall. Among them, 173 (63%) were patients with SN (Table 2). The SN+ group was more likely to fall than the SN− group (IRR = 1.87, standard error [SE] = 0.22, 95% confidence interval [CI] [1.28, 2.36], P < .001). A multivariate Poisson regression further examined the IRR of falls of the SN+ over SN− group with the a priori covariates. The results are summarized in Table 3, showing that the SN+ group was associated with a higher fall rate (IRR = 1.44, SE = 0.19, 95% CI [1.12, 1.86], P = .005), and so were being younger, male, having better ADL function at admission, and staying longer in IRF.
Rate Ratio of Falls Predicted by Spatial Neglect.
Abbreviations: BIMS, Brief Interview for Mental Status; ADL, activities of daily living; IRR, incident rate ratio; SE, standard error; CI, confidence interval.
Another multivariate Poisson regression was conducted to examine whether SN severity based on the CBS score predicted falls after considering the same covariates. Summarized in Table 4, results indicated that patients with mild and moderate SN had higher incident rates than patients with no SN (IRR = 1.43, P = .008; IRR = 1.52, P = .027, respectively). There was no statistical difference between patients with no SN and those with severe SN (IRR = 1.42, P = .203).
Rate Ratio of Falls Predicted by Severity of Spatial Neglect.
Abbreviations: CBS, Catherine Bergego Scale; BIMS, Brief Interview for Mental Status; ADL, activities of daily living; IRR, incident rate ratio; SE, standard error; CI, confidence interval.
Regarding the number of patients who were injured after a fall, 239 (87%) of all 276 fallers met the CMS’s criteria for “no injury,” which are “no evidence of any injury is noted on physical assessment by the nurse or primary care clinician; no complaints of pain or injury by the patient; no change in the patient’s behavior is noted after the fall.” There was no statistical difference between the SN− and SN+ group in the ratio of injured fallers (Chi-squared = 2.34, P = .126; Table 2).
Fall Circumstances
A Poisson regression modeling analysis was performed to examine whether the SN+ group fell more often than the SN− group in each circumstance listed in Table 1. Results are summarized in Figure 1A, showing that in comparison to the SN− group, the SN+ group had a higher rate of falls from bed (IRR = 1.92, SE = 0.61, 95% CI [1.03, 3.58], P = .039), were found more often on the floor (IRR = 1.49, SE = 0.29, 95% CI [1.01, 2.19], P = .044), and fell in their hospital rooms more often (IRR = 1.52, SE = 0.28, 95% CI [1.06, 2.17], P = .022). One result that approached significance but did show a considerable effect size in the Cause of Falls category was circumstances related to equipment (IRR = 2.05, SE = 0.78, 95% CI [0.98, 4.31], P = .058). After adjusting for the a priori covariates (Figure 1B), the only circumstance that remained distinctive between the SN+ and SN− groups was patient’s room (IRR = 1.55, SE = 0.30, 95% CI [1.06, 2.26], P = .024).
Forest plots of incident rate ratios (IRRs) for falls in the SN+ group over those in the SN− group (A) when no covariate was included in the Poisson modeling and (B) adjusted after controlling for age, sex, cognitive status, activities of daily living at admission, and length of stay. One circumstance in Type of Falls (Self-reported Fall) and 3 in Location of Falls (Stair or Curb, Hallway, Monitored Space) were excluded due to the rates being too low to be analyzed using Poisson modeling.
Impact of PAT
In this cohort, 443 patients with SN (CBS: median = 11, IQR = 7.14-16.67) received PAT, ranging from 1 to 11 sessions (median = 5; IQR = 2-9). About 62 (14%) of them fell during their IRF stay, and the CBS of these fallers was 12.9 in median (IQR = 7.5-18.57).
To answer whether receiving more PAT sessions reduced the rate of falls, we included patients whose information regarding dates of PAT and falls was available in order to reveal the sequence of events (ie, falls before the first PAT session and falls after the last PAT session). The required information was available in 37 patients with SN who received PAT. A Poisson regression analysis, including no covariates, showed that receiving greater numbers of PAT sessions were associated with a reduced rate of falls after PAT (IRR = 0.77, SE = 0.07, 95% CI [0.65, 0.93], P = .006). To confirm the effect while considering potential confounding factors, we followed up with a multivariate Poisson regression including the a priori covariates and 2 additional covariates, which were CBS (indicating SN severity before PAT) and number of falls before PAT (Table 5). The result remained that more PAT sessions predicted a lower fall rate after PAT (IRR = 0.82, SE = 0.07, 95% CI [0.69, 0.97], P = .022).
Rate Ratio of Falls After PAT Predicted by Number of PAT Sessions.
Abbreviations: PAT, prism adaptation treatment; BIMS, Brief Interview for Mental Status; ADL, activities of daily living; CBS, Catherine Bergego Scale; IRR, incident rate ratio; SE, standard error; CI, confidence interval.
Lastly, to confirm whether our previous finding 36 on the association between PAT treatment intensity (ie, number of sessions) and functional improvement could be replicated in the present cohort, we conducted a linear regression analysis on ADL Gain (n = 1312), which was calculated by subtracting the ADL score at admission from that at discharge. After controlling for BIMS and ADL at admission, CBS, length of stay, age, and sex, the result confirmed that more PAT sessions predicted greater ADL Gain (b = 0.30, SE = 0.11, 95% CI [0.08, 0.52], P = .008).
Discussion
The present study showed that SN was associated with an increased risk of falls among stroke survivors admitted to IRFs, especially in their hospital rooms. The present study also demonstrated that greater treatment intensity of PAT, that is, more sessions, predicted a lower rate of in-hospital falls in stroke survivors with SN. These findings suggest that patients need to be assessed for SN in IRF settings, and patients with SN should receive treatment such as PAT to reduce the risk of falls. These findings and clinical implications are the strength of the present study.
Another strength of the present study is that observations of prior research were replicated: The present study confirmed that SN is associated with a longer interval between stroke onset and inpatient rehabilitation admission (suggesting the need for more acute care), lower function level, increased in-hospital fall risk, prolonged length of inpatient stay, and slower rehabilitation progress.16-19,27,28,41 The high incidence of SN,24,25 for example, almost half of stroke patients in the present study, underscores the necessity for implementation of effective treatments, such as PAT.36,42-44 The present study also confirmed that a greater number of PAT sessions predicts greater functional improvement. 36 PAT when provided with appropriate intensity may help stroke survivors with SN reach the same rehabilitation outcome as those with no SN.
SN is a safety concern regarding falls16,18,28 as well as other aspects of everyday life (eg, potential traffic accidents). 45 The present study showed that SN was associated with falls occurring in patients’ rooms. Future research is required to determine effective fall prevention measures that can be implemented in patients’ rooms. For example, in addition to raising the bed rails when patients are in or on bed, a signage of SN can be posted in the room to notify all staff members, regardless of their roles in care and service provision.
In the present study, we included patients who sustained a stroke based on the coded rehabilitation impairment category used within the U.S. IRFs to classify patients by diagnosis. This led to a heterogenous sample because all stroke patients regardless of locations of brain damage, prior stroke history, or other neurological disorders were included. This fact may be considered a limitation or strength. The heterogeneity increases the noise in the data, but the result may be relatively generalizable to a larger pool of stroke patients instead of a selective group of stroke patients.
Factors not included in the study may have played a role in the current findings. These includes types or total hours of therapies, or treatment activities, other than PAT, within a therapy in which patients participated during their IRF stay. Whether or why patients received no or any number of PAT sessions were determined by various factors. 46 One factor was SN severity. While the median CBS score of all patients with SN was 5.56, that of patients who received PAT was 11, suggesting that OTs may have prioritized the use of PAT in patients whose SN was relatively severe. A limitation of retrospectively analyzing clinical data is the varied availability of information. For example, not all of the 62 fallers with SN who received PAT were included in the analysis seeking the association between the number of PAT sessions and fall risk reduction. This is because the dates of events including falls, first and last PAT sessions were required to be included. To explore the association between PAT and falls with a relatively small sample, we conservatively included multiple covariates (Table 5) and observed that more sessions of PAT received by patients predicted a reduced rate of falls.
In conclusion, the present study using a large sample of clinical data demonstrated that patients with SN had a 44% higher rate of falls than patients with no SN in IRF settings and identified patients’ hospital room as the location where patients with SN had a higher fall incidence than patients with no SN. With the information available, treatment for SN such as PAT, especially multiple sessions of PAT, should be considered to reduce the risk of falls in these patients. Further intervention development and research is needed to reduce the risk of falls in hospital rooms for patients with SN.
Footnotes
Acknowledgements
None.
Author Contributions
Peii Chen: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Project administration; Software; Supervision; Visualization; Writing - original draft; Writing - review & editing. Natalia Noce: Conceptualization; Formal analysis; Investigation; Resources; Writing - review & editing. Emily DeBel: Conceptualization; Data curation; Investigation; Project administration; Resources; Writing - review & editing. Jayme O'Connor: Data curation; Investigation; Resources; Validation; Writing - review & editing. John DeLuca: Investigation; Supervision; Writing - review & editing.
Declaration of Conflicting Interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: KF-NAP and KF-PAT are registered trademarks owned by Kessler Foundation. Four of the authors are employees of Kessler Foundation.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
