CURRENT ISSUES
Validation of Fall Risk Assessment Specific to the Inpatient Rehabilitation Facility Setting Dan Thomas1, MSN, RN, CRRN, Andrea Pavic1, MS, OTR/L, Erin Bisaccia1, PT, DPT, NCS & Jonathan Grotts1, MA 1 Cottage Rehabilitation Hospital, Santa Barbara, CA, USA
Keywords
Abstract
Fall prevention; safety issues; evidence-based practice; rehabilitation. Correspondence Dan Thomas, Cottage Rehabilitation Hospital, PO Box 689, Santa Barbara, CA 93102. E-mail: [email protected] Accepted January 21, 2015. doi: 10.1002/rnj.211
Purpose: To evaluate and compare the Morse Fall Scale (MFS) and the Casa Colina Fall Risk Assessment Scale (CCFRA) for identification of patients at risk for falling in an acute inpatient rehabilitation facility. The primary objective of this study was to perform a retrospective validation study of the CCFRAS, specifically for use in the inpatient rehabilitation facility (IRF) setting. Design: Retrospective validation study. Method: The study was approved under expedited review by the local Institutional Review Board. Data were collected on all patients admitted to Cottage Rehabiliation Hospital (CRH), a 38-bed acute inpatient rehabilitation hospital, from March 2012 to August 2013. Patients were excluded from the study if they had a length of stay less than 3 days or age less than 18. The area under the receiver operating characteristic curve (AUC) and the diagnostic odds ratio were used to examine the differences between the MFS and CCFRAS. AUC between fall scales was compared using the DeLong Test. Findings: There were 931 patients included in the study with 62 (6.7%) patient falls. The average age of the population was 68.8 with 503 males (51.2%). The AUC was 0.595 and 0.713 for the MFS and CCFRAS, respectively (0.006). The diagnostic odds ratio of the MFS was 2.0 and 3.6 for the CCFRAS using the recommended cutoffs of 45 for the MFS and 80 for the CCFRAS. Conclusion: The CCFRAS appears to be a better tool in detecting fallers vs. nonfallers specific to the IRF setting. Clinical Relevance: The assessment and identification of patients at high risk for falling is important to implement specific precautions and care for these patients to reduce their risk of falling. The CCFRAS is more clinically relevant in identifying patients at high risk for falling in the IRF setting compared to other fall risk assessments. Implementation of this scale may lead to a reduction in fall rate and injuries from falls as it more appropriately identifies patients at high risk for falling.
Introduction The inpatient rehabilitation facility (IRF) setting presents unique challenges in screening patients for fall risk when compared to acute inpatient hospitals. These challenges are largely due to the lack of appropriate screening tools developed for IRF patients. While inpatient hospitals care © 2015 Association of Rehabilitation Nurses Rehabilitation Nursing 2015, 0, 1–7
for a wide variety of patients, IRFs generally admit patients recovering from conditions resulting in mobility, sensory, and cognitive impairments. These conditions are known to be risk factors for hospital falls (Evans, Hodgkinson, Lambert, & Wood, 2001). Previous studies have indicated fall rates vary between 3 and 6 falls per 1,000 patient days in the acute hospital setting, compared to
1
Validation of Fall Assessment in IRF
2.72–17.8 falls per 1,000 patient days in an IRF (Ang, Mordiffi, Wong, Devi, & Evans, 2007; Gilewski, Roberts, Hirata, & Riggs, 2007). Current research has focused on acute care and skilled nursing facilities while little research has focused on fall risk and fall assessment tools in the IRF setting (Kwan, Kaplan, Hudson-McKinney, Redman-Bentley, & Rosario, 2012). In general, the prominent fall risk assessment tools used in health care, such as the Morse Fall Scale (MFS), were developed for use in acute inpatient hospitals (Morse, 1986). Several studies show that the MFS lacks ability to appropriately identify high or low fall risk patients and cannot differentiate between fallers and nonfallers in the IRF setting (Chow et al., 2007; Forrest, Chen, Huss, & Giesler, 2013; Kwan et al., 2012). The MFS was developed to identify patients at high fall risk using data from 100 patients who fell and 100 randomly selected patients who did not fall at an acute care facility (Morse, 1996). The MFS assessment tool includes information on patient’s mental status, mobility, incontinence, IV access, and history of falls and secondary diagnosis (Morse, 1996). However, in the IRF setting specific cognitive deficits, assistance needed for toileting, and activities of daily living may provide valid information to assess risk of falls (Kwan et al., 2012). Fall risk assessments are important for allocating appropriate resources to patients who are truly at high risk of falling. Use of an ineffective fall risk assessment can lead to complications in implementing successful fall risk precautions. Given the inadequate performance of current fall screening tools in the IRF setting, the rehabilitation clinician must rely on experience in combination with fall screening tools to guide fall risk assessments. Fall risk assessments in many IRFs include a 24-hour fall risk analysis by nursing staff and an additional evaluation of Functional Independence Measures (FIM) by physical therapy, occupational therapy, and speech therapy. FIM scores are an important component of the fall risk workup because of their strong correlation with risk of falling (Kwan et al., 2012). Interdisciplinary communication and evaluation of patient history are currently the best method for assessing fall risk given the lack of fall risk assessment tools validated in the IRF setting. The current literature attempting to validate fall risk assessment tools in the IRF is inconclusive (Chow et al., 2007; Forrest et al., 2013; Kwan et al., 2012; Lee, Geller, & Strasser, 2013; Smith, Forster, & Young, 2006). The majority of tools used in these studies were developed outside of the IRF setting. A new tool, the Casa Colina Fall
2
D. Thomas et al.
Risk Assessment Scale (CCFRAS), was developed specifically for the rehabilitation population and has not been included in prior validation studies (Rosario, Kaplan, Khonsari, & Patterson, 2014). The CCFRAS is unique in that it targets the diagnoses specific to the inpatient rehabilitation setting that are most at risk for falling including stroke, brain injuries, and amputations (Rabadi, Rabadi, & Peterson, 2008; Rosario et al., 2014; Table 1). The purpose of this study was to perform a retrospective validation study of the CCFRAS, specifically for use in the IRF setting. The researchers attempted to validate the CCFRAS by evaluating and comparing the MFS’ and CCFRAS’ ability to predict patient falls. Methods To validate the CCFRAS, a retrospective chart review was conducted on patients admitted to CRH. The project was approved by the local Institutional Review Board and granted a waiver of consent. A literature search was performed to find fall screening tools that were effective in the community rehabilitation setting. After identifying CCFRAS and the MFS as appropriate tools, a retrospective review was used to validate the tools’ effectiveness. The MFS items were identified in this study using elements as defined in the existing references. (Morse, 1986). Data Collection Data were retrospectively collected on all patients admitted from March 2012 through August 2013. Patients were excluded from the study if they had a length of stay less Table 1 CCFRAS* (If yes to tetraplegia, low risk, do not continue to score)
Diagnosis RCVA TBI All amputees Tetraplegia ASIA A, B, or C FIM score Toileting score 1, 2 Bed transfer 1, 2 Tub/shower transfer 0, 1 Stairs 0 Total
If Yes, Patient Receives the Following Score 20 50 40 __Yes __No 30 20 20 60
*Used with permission.
© 2015 Association of Rehabilitation Nurses Rehabilitation Nursing 2015, 0, 1–7
D. Thomas et al.
than 3 days or age less than 18 years old. Patients were excluded if they had a length of hospital stay less than 3 days because FIM scores are completed on the third day of hospital admissions. Scores for the fall risk assessment tools were calculated based on data captured in the electronic medical record and every patient in the study received a score for the MFS and CCFRAS. All fall risk assessments used in the analysis were done using the assessment performed on hospital admission. Falls were queried from the risk management team at the hospital. Only the first fall from a patient admission was used in summarizing the effectiveness of the screening tools. For this study, a fall was considered a sudden, uncontrolled, unintentional, nonpurposeful, downward displacement of the body to the floor/ground, or hitting another object like a chair or a stair. Statistical Methods and Analysis Data were summarized using the mean with standard deviation in parentheses for continuous data and number in group with percent of group in parentheses for discrete variables. The performance of the fall risk scales were summarized using sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and diagnostic odds ratio. The area under the receiver operating characteristic curve (AUC) was used to summarize overall performance of the fall risk scale. In the context of this study, the AUC can be defined as the probability that a classifier will rank a randomly chosen patient that fell higher than a randomly chosen patient that did not fall. AUC between scales were compared using the DeLong Test (DeLong, DeLong, & Clarke-Pearson, 1988). A value of 80 for the CCFRAS and a value of 45 for the MFS were used to determine high risk of fall. An attempt was made to find optimal cutoffs for determining high-risk fall patients using the Youden Index for receiver operating curve (ROC) analysis (Youden, 1950). The R Statistical Programming Environment (R Core Team) was used to conduct statistical analysis. Results There were 931 patients included in the study with 62 (6.7%) patient falls. The mean length of stay was 14.1 days (Table 2). The majority of patients were admitted with a weak or impaired gait (94.1%), secondary diagnosis (94.8%), needing assistance with toileting (75.6%), and unable to negotiate stairs (75.4%). Accord© 2015 Association of Rehabilitation Nurses Rehabilitation Nursing 2015, 0, 1–7
Validation of Fall Assessment in IRF
Table 2 Patient characteristics (n = 931) *
Age Number of males LOS* Morse Fall Scale components History of falls Use of ambulatory aid IV therapy Weak/impaired gait Altered mental state Secondary diagnosis CCFRAS components Right CVA TBI All amputees Toileting score of 1 or 2 Bed transfer score of 1 or 2 Tub/shower transfer score of 0 or 1 Stairs score of 0 CCFRAS scale* Pts at high risk according to CCFRAS scale Morse Fall Scale* Pts at high risk according to Morse Fall Scale Falls Discharge disposition Home Rehab or SNF setting Another hospital Unknown
68.7 (16.2) 479 (51.5%) 14.1 (10.3) 373 275 195 876 171 883
(40.1%) (29.5%) (20.9%) (94.1%) (18.4%) (94.8%)
94 (10.1%) 62 (6.7%) 20 (2.1%) 704 (75.6%) 305 (32.8%) 186 (20.0%) 702 (75.4%) 84.7 (42.8%) 622 (66.8%) 48 (18) 540 (58.0%) 62 (6.7%) 756 (81.1%) 111 (11.9%) 60 (6.4%) 5 (0.5%)
*Mean (standard deviation). Table 3 Fall characteristics (n = 62) Days from admission to fall Fall type Other/unknown Fall ambulating (w or w/o assist) Transferring to chair, bed, etc. Reaching for item Changing position Fall outcome Required monitoring or intervention No apparent injury
9 (11.3)* 26 (41.9%) 8 (12.9%) 8 (12.9%) 3 (4.8%) 3 (4.8%) 7 (11.3%) 55 (88.7%)
*Mean (standard deviation).
ing to the CCFRAS, 66.8% of patients were considered high risk while 58.0% of patients were high risk according to the MFS. The mean difference between admission date and occurrence of a fall was 9 days with 88.7% of the falls causing no apparent injury (Table 3).
3
Validation of Fall Assessment in IRF
D. Thomas et al.
Using the recommended cutoffs for classifying highrisk fall patients, the CCFRAS had a superior sensitivity (0.871) compared to the MFS (0.726), but the MFS had a higher specificity (0.430) than the CCFRAS (0.346). The difference between AUC and the diagnostic odds ratio were statistically significant between the fall scales with AFRAS having a diagnostic odds ratio of 3.6 while the MFS had a diagnostic odds ratio of 2.4 (Table 4). Figure 1 presents the AUC for both fall scales. The underlying components of each fall scale are compared between patients that fell and those that did not in Table 5. The only component of the MFS that occurred at a significantly greater rate in fall patients compared to nonfall patients was altered mental state (41.9% vs. 16.7%, respectively, p < .001). Primary diagnosis of right cerebrovascular accident, needing assistance with toileting, needing assistance with bed transfer, needing assistance or
Table 4 Fall scale performance
AUC Sensitivity Specificity PPV NPV Accuracy Diagnostic odds ratio
Morse Fall Scale
CCFRAS Scale
0.599 0.726 0.430 0.083 0.957 0.450 2
0.710 0.871 0.346 0.087 0.974 0.381 3.6
p-Value .010
Sensitivity, Specificity, PPV, and NPV calculated using suggested high risk cutoffs for fall scales (Morse ≥45 and CCFRAS ≥80).
unable to transfer to tub/shower, and inability to negotiate stairs were all present at higher rates in patients that fell compared to those that did not and are components of the CCFRAS. An analysis of the best performing cutoffs for the fall risk scales yielded a cutoff of 47.5 for the MFS and 105 for the CCFRAS (Table 6). The diagnostic odds ratio for both fall risk scales increased using these cutoffs to 2.4 in the MFS and 3.9 in the CCFRAS. The CCFRAS improved from 0.381 to 0.639 in accuracy and from 0.346 to 0.652 in specificity from the recommended cutoff value to the theoretical best cutoffs. Discussion This study validates that the CCFRAS is an adequate fall risk screening tool for use in the inpatient rehabilitation hospital setting. When compared to the MFS in the IRF setting, the CCFRAS had a higher sensitivity and diagnostic odds ratio. The clinical relevance of the CCFRAS compared to the MFS for IRF setting is apparent when one compares the patient characteristics composing each scale. Altered mental status was the only condition in the MFS that was statistically significant while many of the components within the CCFRAS occurred at higher rates within the fallers. There were observed differences between a theoretical optimal cutoff for the fall risk assessments and the recommended cutoffs. The increase from the recommended cutoffs to the optimal cutoffs was 31% in the CCFRAS and 6% in the MFS. This could be explained by heterogeneity between the populations used to develop the fall risk
Figure 1 Plot of area under the ROC curve.
4
© 2015 Association of Rehabilitation Nurses Rehabilitation Nursing 2015, 0, 1–7
D. Thomas et al.
Validation of Fall Assessment in IRF
Table 5 Comparison of patient characteristic by fall incidence
*
Age Number of males LOS* Morse Fall Scale components History of falls Use of ambulatory aid IV therapy Weak/impaired gait Altered mental state Secondary diagnosis CCFRAS components Right CVA TBI All amputees Toileting score of 1 or 2 Bed transfer score of 1 or 2 Tub/shower transfer score of 0 or 1 Stairs score of 0 CCFRAS scale score* Morse Fall Scale score* Pts at high risk according to CCFRAS Pts at high risk according to Morse Fall Scale
No Fall (n = 869)
Fall (n = 62)
p-Value
68.8 (16.2) 444 (51%) 13.3 (9.2)
67.2 (16.2) 35 (56.5%) 24.9 (16.8)
.451 .409
Validation of Fall Risk Assessment Specific to the Inpatient Rehabilitation Facility Setting Dan Thomas1, MSN, RN, CRRN, Andrea Pavic1, MS, OTR/L, Erin Bisaccia1, PT, DPT, NCS & Jonathan Grotts1, MA 1 Cottage Rehabilitation Hospital, Santa Barbara, CA, USA
Keywords
Abstract
Fall prevention; safety issues; evidence-based practice; rehabilitation. Correspondence Dan Thomas, Cottage Rehabilitation Hospital, PO Box 689, Santa Barbara, CA 93102. E-mail: [email protected] Accepted January 21, 2015. doi: 10.1002/rnj.211
Purpose: To evaluate and compare the Morse Fall Scale (MFS) and the Casa Colina Fall Risk Assessment Scale (CCFRA) for identification of patients at risk for falling in an acute inpatient rehabilitation facility. The primary objective of this study was to perform a retrospective validation study of the CCFRAS, specifically for use in the inpatient rehabilitation facility (IRF) setting. Design: Retrospective validation study. Method: The study was approved under expedited review by the local Institutional Review Board. Data were collected on all patients admitted to Cottage Rehabiliation Hospital (CRH), a 38-bed acute inpatient rehabilitation hospital, from March 2012 to August 2013. Patients were excluded from the study if they had a length of stay less than 3 days or age less than 18. The area under the receiver operating characteristic curve (AUC) and the diagnostic odds ratio were used to examine the differences between the MFS and CCFRAS. AUC between fall scales was compared using the DeLong Test. Findings: There were 931 patients included in the study with 62 (6.7%) patient falls. The average age of the population was 68.8 with 503 males (51.2%). The AUC was 0.595 and 0.713 for the MFS and CCFRAS, respectively (0.006). The diagnostic odds ratio of the MFS was 2.0 and 3.6 for the CCFRAS using the recommended cutoffs of 45 for the MFS and 80 for the CCFRAS. Conclusion: The CCFRAS appears to be a better tool in detecting fallers vs. nonfallers specific to the IRF setting. Clinical Relevance: The assessment and identification of patients at high risk for falling is important to implement specific precautions and care for these patients to reduce their risk of falling. The CCFRAS is more clinically relevant in identifying patients at high risk for falling in the IRF setting compared to other fall risk assessments. Implementation of this scale may lead to a reduction in fall rate and injuries from falls as it more appropriately identifies patients at high risk for falling.
Introduction The inpatient rehabilitation facility (IRF) setting presents unique challenges in screening patients for fall risk when compared to acute inpatient hospitals. These challenges are largely due to the lack of appropriate screening tools developed for IRF patients. While inpatient hospitals care © 2015 Association of Rehabilitation Nurses Rehabilitation Nursing 2015, 0, 1–7
for a wide variety of patients, IRFs generally admit patients recovering from conditions resulting in mobility, sensory, and cognitive impairments. These conditions are known to be risk factors for hospital falls (Evans, Hodgkinson, Lambert, & Wood, 2001). Previous studies have indicated fall rates vary between 3 and 6 falls per 1,000 patient days in the acute hospital setting, compared to
1
Validation of Fall Assessment in IRF
2.72–17.8 falls per 1,000 patient days in an IRF (Ang, Mordiffi, Wong, Devi, & Evans, 2007; Gilewski, Roberts, Hirata, & Riggs, 2007). Current research has focused on acute care and skilled nursing facilities while little research has focused on fall risk and fall assessment tools in the IRF setting (Kwan, Kaplan, Hudson-McKinney, Redman-Bentley, & Rosario, 2012). In general, the prominent fall risk assessment tools used in health care, such as the Morse Fall Scale (MFS), were developed for use in acute inpatient hospitals (Morse, 1986). Several studies show that the MFS lacks ability to appropriately identify high or low fall risk patients and cannot differentiate between fallers and nonfallers in the IRF setting (Chow et al., 2007; Forrest, Chen, Huss, & Giesler, 2013; Kwan et al., 2012). The MFS was developed to identify patients at high fall risk using data from 100 patients who fell and 100 randomly selected patients who did not fall at an acute care facility (Morse, 1996). The MFS assessment tool includes information on patient’s mental status, mobility, incontinence, IV access, and history of falls and secondary diagnosis (Morse, 1996). However, in the IRF setting specific cognitive deficits, assistance needed for toileting, and activities of daily living may provide valid information to assess risk of falls (Kwan et al., 2012). Fall risk assessments are important for allocating appropriate resources to patients who are truly at high risk of falling. Use of an ineffective fall risk assessment can lead to complications in implementing successful fall risk precautions. Given the inadequate performance of current fall screening tools in the IRF setting, the rehabilitation clinician must rely on experience in combination with fall screening tools to guide fall risk assessments. Fall risk assessments in many IRFs include a 24-hour fall risk analysis by nursing staff and an additional evaluation of Functional Independence Measures (FIM) by physical therapy, occupational therapy, and speech therapy. FIM scores are an important component of the fall risk workup because of their strong correlation with risk of falling (Kwan et al., 2012). Interdisciplinary communication and evaluation of patient history are currently the best method for assessing fall risk given the lack of fall risk assessment tools validated in the IRF setting. The current literature attempting to validate fall risk assessment tools in the IRF is inconclusive (Chow et al., 2007; Forrest et al., 2013; Kwan et al., 2012; Lee, Geller, & Strasser, 2013; Smith, Forster, & Young, 2006). The majority of tools used in these studies were developed outside of the IRF setting. A new tool, the Casa Colina Fall
2
D. Thomas et al.
Risk Assessment Scale (CCFRAS), was developed specifically for the rehabilitation population and has not been included in prior validation studies (Rosario, Kaplan, Khonsari, & Patterson, 2014). The CCFRAS is unique in that it targets the diagnoses specific to the inpatient rehabilitation setting that are most at risk for falling including stroke, brain injuries, and amputations (Rabadi, Rabadi, & Peterson, 2008; Rosario et al., 2014; Table 1). The purpose of this study was to perform a retrospective validation study of the CCFRAS, specifically for use in the IRF setting. The researchers attempted to validate the CCFRAS by evaluating and comparing the MFS’ and CCFRAS’ ability to predict patient falls. Methods To validate the CCFRAS, a retrospective chart review was conducted on patients admitted to CRH. The project was approved by the local Institutional Review Board and granted a waiver of consent. A literature search was performed to find fall screening tools that were effective in the community rehabilitation setting. After identifying CCFRAS and the MFS as appropriate tools, a retrospective review was used to validate the tools’ effectiveness. The MFS items were identified in this study using elements as defined in the existing references. (Morse, 1986). Data Collection Data were retrospectively collected on all patients admitted from March 2012 through August 2013. Patients were excluded from the study if they had a length of stay less Table 1 CCFRAS* (If yes to tetraplegia, low risk, do not continue to score)
Diagnosis RCVA TBI All amputees Tetraplegia ASIA A, B, or C FIM score Toileting score 1, 2 Bed transfer 1, 2 Tub/shower transfer 0, 1 Stairs 0 Total
If Yes, Patient Receives the Following Score 20 50 40 __Yes __No 30 20 20 60
*Used with permission.
© 2015 Association of Rehabilitation Nurses Rehabilitation Nursing 2015, 0, 1–7
D. Thomas et al.
than 3 days or age less than 18 years old. Patients were excluded if they had a length of hospital stay less than 3 days because FIM scores are completed on the third day of hospital admissions. Scores for the fall risk assessment tools were calculated based on data captured in the electronic medical record and every patient in the study received a score for the MFS and CCFRAS. All fall risk assessments used in the analysis were done using the assessment performed on hospital admission. Falls were queried from the risk management team at the hospital. Only the first fall from a patient admission was used in summarizing the effectiveness of the screening tools. For this study, a fall was considered a sudden, uncontrolled, unintentional, nonpurposeful, downward displacement of the body to the floor/ground, or hitting another object like a chair or a stair. Statistical Methods and Analysis Data were summarized using the mean with standard deviation in parentheses for continuous data and number in group with percent of group in parentheses for discrete variables. The performance of the fall risk scales were summarized using sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and diagnostic odds ratio. The area under the receiver operating characteristic curve (AUC) was used to summarize overall performance of the fall risk scale. In the context of this study, the AUC can be defined as the probability that a classifier will rank a randomly chosen patient that fell higher than a randomly chosen patient that did not fall. AUC between scales were compared using the DeLong Test (DeLong, DeLong, & Clarke-Pearson, 1988). A value of 80 for the CCFRAS and a value of 45 for the MFS were used to determine high risk of fall. An attempt was made to find optimal cutoffs for determining high-risk fall patients using the Youden Index for receiver operating curve (ROC) analysis (Youden, 1950). The R Statistical Programming Environment (R Core Team) was used to conduct statistical analysis. Results There were 931 patients included in the study with 62 (6.7%) patient falls. The mean length of stay was 14.1 days (Table 2). The majority of patients were admitted with a weak or impaired gait (94.1%), secondary diagnosis (94.8%), needing assistance with toileting (75.6%), and unable to negotiate stairs (75.4%). Accord© 2015 Association of Rehabilitation Nurses Rehabilitation Nursing 2015, 0, 1–7
Validation of Fall Assessment in IRF
Table 2 Patient characteristics (n = 931) *
Age Number of males LOS* Morse Fall Scale components History of falls Use of ambulatory aid IV therapy Weak/impaired gait Altered mental state Secondary diagnosis CCFRAS components Right CVA TBI All amputees Toileting score of 1 or 2 Bed transfer score of 1 or 2 Tub/shower transfer score of 0 or 1 Stairs score of 0 CCFRAS scale* Pts at high risk according to CCFRAS scale Morse Fall Scale* Pts at high risk according to Morse Fall Scale Falls Discharge disposition Home Rehab or SNF setting Another hospital Unknown
68.7 (16.2) 479 (51.5%) 14.1 (10.3) 373 275 195 876 171 883
(40.1%) (29.5%) (20.9%) (94.1%) (18.4%) (94.8%)
94 (10.1%) 62 (6.7%) 20 (2.1%) 704 (75.6%) 305 (32.8%) 186 (20.0%) 702 (75.4%) 84.7 (42.8%) 622 (66.8%) 48 (18) 540 (58.0%) 62 (6.7%) 756 (81.1%) 111 (11.9%) 60 (6.4%) 5 (0.5%)
*Mean (standard deviation). Table 3 Fall characteristics (n = 62) Days from admission to fall Fall type Other/unknown Fall ambulating (w or w/o assist) Transferring to chair, bed, etc. Reaching for item Changing position Fall outcome Required monitoring or intervention No apparent injury
9 (11.3)* 26 (41.9%) 8 (12.9%) 8 (12.9%) 3 (4.8%) 3 (4.8%) 7 (11.3%) 55 (88.7%)
*Mean (standard deviation).
ing to the CCFRAS, 66.8% of patients were considered high risk while 58.0% of patients were high risk according to the MFS. The mean difference between admission date and occurrence of a fall was 9 days with 88.7% of the falls causing no apparent injury (Table 3).
3
Validation of Fall Assessment in IRF
D. Thomas et al.
Using the recommended cutoffs for classifying highrisk fall patients, the CCFRAS had a superior sensitivity (0.871) compared to the MFS (0.726), but the MFS had a higher specificity (0.430) than the CCFRAS (0.346). The difference between AUC and the diagnostic odds ratio were statistically significant between the fall scales with AFRAS having a diagnostic odds ratio of 3.6 while the MFS had a diagnostic odds ratio of 2.4 (Table 4). Figure 1 presents the AUC for both fall scales. The underlying components of each fall scale are compared between patients that fell and those that did not in Table 5. The only component of the MFS that occurred at a significantly greater rate in fall patients compared to nonfall patients was altered mental state (41.9% vs. 16.7%, respectively, p < .001). Primary diagnosis of right cerebrovascular accident, needing assistance with toileting, needing assistance with bed transfer, needing assistance or
Table 4 Fall scale performance
AUC Sensitivity Specificity PPV NPV Accuracy Diagnostic odds ratio
Morse Fall Scale
CCFRAS Scale
0.599 0.726 0.430 0.083 0.957 0.450 2
0.710 0.871 0.346 0.087 0.974 0.381 3.6
p-Value .010
Sensitivity, Specificity, PPV, and NPV calculated using suggested high risk cutoffs for fall scales (Morse ≥45 and CCFRAS ≥80).
unable to transfer to tub/shower, and inability to negotiate stairs were all present at higher rates in patients that fell compared to those that did not and are components of the CCFRAS. An analysis of the best performing cutoffs for the fall risk scales yielded a cutoff of 47.5 for the MFS and 105 for the CCFRAS (Table 6). The diagnostic odds ratio for both fall risk scales increased using these cutoffs to 2.4 in the MFS and 3.9 in the CCFRAS. The CCFRAS improved from 0.381 to 0.639 in accuracy and from 0.346 to 0.652 in specificity from the recommended cutoff value to the theoretical best cutoffs. Discussion This study validates that the CCFRAS is an adequate fall risk screening tool for use in the inpatient rehabilitation hospital setting. When compared to the MFS in the IRF setting, the CCFRAS had a higher sensitivity and diagnostic odds ratio. The clinical relevance of the CCFRAS compared to the MFS for IRF setting is apparent when one compares the patient characteristics composing each scale. Altered mental status was the only condition in the MFS that was statistically significant while many of the components within the CCFRAS occurred at higher rates within the fallers. There were observed differences between a theoretical optimal cutoff for the fall risk assessments and the recommended cutoffs. The increase from the recommended cutoffs to the optimal cutoffs was 31% in the CCFRAS and 6% in the MFS. This could be explained by heterogeneity between the populations used to develop the fall risk
Figure 1 Plot of area under the ROC curve.
4
© 2015 Association of Rehabilitation Nurses Rehabilitation Nursing 2015, 0, 1–7
D. Thomas et al.
Validation of Fall Assessment in IRF
Table 5 Comparison of patient characteristic by fall incidence
*
Age Number of males LOS* Morse Fall Scale components History of falls Use of ambulatory aid IV therapy Weak/impaired gait Altered mental state Secondary diagnosis CCFRAS components Right CVA TBI All amputees Toileting score of 1 or 2 Bed transfer score of 1 or 2 Tub/shower transfer score of 0 or 1 Stairs score of 0 CCFRAS scale score* Morse Fall Scale score* Pts at high risk according to CCFRAS Pts at high risk according to Morse Fall Scale
No Fall (n = 869)
Fall (n = 62)
p-Value
68.8 (16.2) 444 (51%) 13.3 (9.2)
67.2 (16.2) 35 (56.5%) 24.9 (16.8)
.451 .409
