Geriatric Nursing 36 (2015) S3eS9
Contents lists available at ScienceDirect
Geriatric Nursing journal homepage: www.gnjournal.com
Feature Article
Length of hospital stay and discharge disposition in older trauma patients Rebecca A. Brotemarkle, PhD, RN a, *, Barbara Resnick, PhD, RN a, Kathleen Michaels, PhD, RN a, Patricia Morton, PhD, RN b, Chris Wells, PhD, PT c a
University of Maryland School of Nursing, 655 West Lombard Street, Baltimore, MD 21201, USA The University of Utah College of Nursing, 10 South 2000 East, Salt Lake City, UT 84112, USA c University of Maryland School of Medicine, Allied Health Building, 100 Penn Street, Baltimore, MD 21201, USA b
a b s t r a c t Keywords: Older adults Trauma Length of Stay Discharge Disposition Pain Therapy
As the number of older adults increases and life expectancies are increasing, more incidences of traumatic injury are expected in this population. In this study, the relationships between demographic variables, pain, days from admission to therapy evaluation, length of stay and discharge disposition were examined in 132 older adults who had experienced a traumatic event. Results showed that significant relationships existed between pain, age, comorbidities, injury severity and days from admission to therapy evaluation and length of stay; those with less pain, greater age and had more days between admission and when the first therapy evaluation occurred had longer lengths of stay. In addition, demographic variables, overall length of stay and pain intensity during therapy were associated with discharge location; for longer lengths of stay and higher pain, older trauma patients were less likely to be discharged to home. Ó 2015 Elsevier Inc. All rights reserved.
Discharge disposition and length of hospital stay in older trauma patients From 1900 to 1994, the number of Americans aged 65 and older increased from three million to 33 million and is expected to reach 80 million by 2050.1 As the population of older adults increases, the incidence of traumatic events in this population will also increase as older adults are remaining active longer.2 Falls and motor vehicle accidents are the most likely mechanisms of injury in those over the age of 65 and 25% of all trauma incidents were in those over the age of 65.3 Impact of trauma on the health care system Following a trauma admission, older adults commonly need and receive follow-up treatment in various settings.4e6 The appropriate discharge setting is decided upon by a multidisciplinary team, including the attending physicians, advanced practice nurses or physician assistants, physical, occupational and speech therapists, social workers, patient/family preferences and available resources and health insurance.7,8 The setting options include home or self
* Corresponding author. E-mail address: [email protected] (R.A. Brotemarkle). 0197-4572/$ e see front matter Ó 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.gerinurse.2015.02.016
care (this would include assisted living, boarding homes, senior housing etc), short term general hospital, skilled nursing facility, intermediate care facility, home under care of organized home health service, inpatient rehabilitation, long term care hospital, and hospice care.9 Matching the discharge disposition correctly to the older trauma patient’s needs and preferences has a major impact on the health care system, length of stay (LOS), patient function, appropriate resource allocation, and rate of readmissions.8 Length of stay for trauma patients varies widely and is dependent upon multiple factors. The impact of LOS on cost of care is significant10,11 thus the overall goal in trauma care is to decrease LOS in order to decrease cost. The obvious solution is to transition patients into lower levels of care as soon as possible post trauma.12 Factors that influence outcomes post trauma based on the SocialeEcological Model The SocialeEcological Model (SEM)13e17 suggests that intrapersonal factors, interpersonal factors, environment/community factors and policies/systems may influence outcomes. Intrapersonal factors are defined as those characteristics that are specific to the individual; interpersonal factors focus on the interactions between individuals that can influence outcomes; environmental factors include such things as the physical environment and why an
S4
R.A. Brotemarkle et al. / Geriatric Nursing 36 (2015) S3eS9
individual might or might not be able to go for walk (e.g., no flat open areas; unsafe neighborhoods); and policies, culture and systems can influence what patients are allowed to do on the unit and whether or not, for example, they need to be tethered to pulse oximetry throughout the hospital stay. The levels in the SEM are interactive. Intrapersonal factors and discharge disposition and LOS Demographic variables including increased age, race/ethnicity, and gender have all been noted to be non-modifiable factors that influence LOS. Older trauma patients, when compared with younger trauma patients, have worse outcomes including: more admissions to the hospital,18 longer hospital stays,18,19 and a greater need for post-acute rehabilitation and/or long term care.4,10,20 The following patient factors have been associated with worse outcomes for older trauma patients: increased age, increased pain, being uninsured, having more comorbidities, having higher injury severity, and type of injury (traumatic brain injury, hip fracture, multi-trauma).21e23 The effects of gender24e27 and race/ ethnicity28e30 on discharge outcomes have been mixed. Some studies have shown that women were more likely to be discharged to nursing homes,4,27,31e34 some showing that women were more likely to be discharged to home24,25,28 and others showing no differences.35,36 With regard to the impact of race and ethnicity study results have not been consistent. In some studies African Americans and Hispanic patients had higher mortality28,37 and nonwhite patients were more likely to be discharged to home.4,34 In other studies there were no differences between African American, Hispanic patients and Caucasian patients.30,38 Comorbidity, defined as the presence of disease conditions in addition to the condition for which the patient was admitted to the hospital,39 influences outcomes post trauma.10,40e43 Comorbidities are generally non-modifiable factors; however, signs and symptoms of comorbidities in some cases are able to be controlled (e.g. blood sugar levels, dizziness, urinary frequency). Although the findings are not always consistent, comorbidities are associated with worse outcomes in older adults who have experienced trauma.34,44,45 Specifically, comorbidities influence LOS with more comorbid conditions being related to longer LOS.46e48 Pain post trauma has also been reported to influence LOS and location of discharge and is a potentially modifiable intrapersonal factor. Prior research has shown that when older adults have acute pain it interrupts physical therapy in approximately a third of the patients.49 If pain interferes with a patient’s ability or desire to participate in therapy, slower recovery times and increased length of hospital stays are likely to occur.50e52 Pain that is not well controlled limits an individual’s functional performance and may result in a lower level of care (i.e., a long term care facilities versus a skilled post-acute rehabilitation stay).
The goal of discharge planning is to assure that patients receive needed services in the least restrictive setting. In addition, an appropriate discharge plan can decrease LOS and prevent readmissions to the hospital. Knowing what influences discharge disposition and LOS may help guide in improving the discharge planning process and decreasing LOS for older patients. The purpose of this study, therefore, was to comprehensively consider the factors associated with discharge disposition and length of stay among older adults hospitalized for traumatic injury. Specifically, it was hypothesized that demographic variables (gender, age), health related factors (comorbidities; injury severity; admitting diagnosis), pain and initiation of therapy would be directly and indirectly associated with LOS and discharge disposition (Fig. 1). Methods Design This study was a secondary data analysis using data from inpatient rehabilitation services provided to patients during inpatient stays in a Level 1 trauma center. Data collection was done using a retrospective chart review from Powerchart, an electronic medical record used in the participating trauma center. Specifically, the data sources included computerized therapy (physical, occupational, and speech) notes for each exposure to therapy during the patients’ hospital stay or up to 30 days post admission. Sample Eligibility was based on the participants being 65 years or older and having been admitted to an urban Level I Trauma Center following a traumatic event. A sample size of 137 adults was randomly chosen from 1387 admissions over a two-year period from 2010 to 2011. Five of the participants were excluded as three
Interpersonal factors and discharge disposition and LOS A potentially modifiable interpersonal factor is the type and timing of physical and occupational therapy. Physical and occupational therapy are an important part of a trauma patient’s recovery process. Older adults are known to have accelerated functional decline after a traumatic event.53e56 To prevent functional decline and assist older adults in returning to pre-injury functional levels, there has been a recent focus on early treatment and mobilization.57e61 Early physical rehabilitation programs have shown a length of stay reduction of 22% for patients in intensive care units and 19% for patients on general units.62
Fig. 1. Hypothesized Model.
R.A. Brotemarkle et al. / Geriatric Nursing 36 (2015) S3eS9
had primary diagnoses that were not trauma diagnoses and two died before they could participate in therapy. The final sample size was 132. The study was approved by a University Institutional Review Board. Data collection and measures Data were collected by registered nurses with acute care experience and expertise in the use of Powerchart. These individuals had not provided direct care to the study participants. The data collectors were instructed how to access the necessary data and how to complete the data collection forms for each patient. Therapy notes included preadmission function, admission day and length of stay, rehabilitation order day, day of first/initial therapy evaluation, and subsequent visits. Baseline assessments by all of the therapists include cognitive function and pain. In addition, physical and occupational therapists evaluated patients’ activities of daily living, instrumental activities of daily living, range of motion, manual muscle testing, and activity tolerance. Speech language pathologists included assessment of language skills and swallowing capabilities. Entries are not forced and therapists completed assessment components based on their expertise and clinical decision-making. Descriptive information about each rehabilitation encounter was also obtained, including the reason for the visit, what occurred during that visit, and documented reasons why a visit did not occur (e.g., patient refused). Demographic information about patients was obtained, including age, gender, race/ethnicity, comorbidities, living location prior to the traumatic event, and location after discharge. Length of stay was measured as the day of admission until the day of discharge, however, the day of discharge was not included in the count of days. For those patients admitted and discharged on the same day, the LOS was counted as one day. Pain was conceptualized as pain experienced prior to, during, and immediately following therapy (Fig. 1). The computerized system allows therapists to choose one of three pain assessment tools: the Verbal Rating Scale (VRS),63 Verbal Descriptor Scale (VDS)63,64 or the Checklist of Nonverbal Pain Indicators (CNPI).65 The VRS was the most frequently used among the three measures and involved a rating of zero to ten, where zero is no pain and ten is the worst imaginable pain. For the purpose of this study, the VRS data were used (n ¼ 88) as the measure of pain since the other measures are not on a numeric scale or do not measure pain intensity. Among the 46 participants in which the VRS was not completed, 30 (65%) had no assessment of pain identified and 16 (35%) used one of the other measurement options. As shown in Fig. 1, the measurement model of pain which included the prior to therapy, during and immediately following reports of pain by the patient fit the data with paths significant at the .001 level (pretherapy pain path was .89; during therapy path was .72; and post therapy path was .98). Injury severity was determined based on anatomic factors (the number of injured body areas) and the physiologic method of measuring severity of injury (shock or coma).66 Severe injury was based on having more than one trauma diagnosis and having a diagnosis of shock or a head injury with coma. Otherwise, if there was only one trauma diagnosis and there was no diagnosis of shock or coma, then the injury severity was considered to be mild/ moderate. Comorbidities were counted as the number of conditions not related to the admitting diagnosis that are checked off a list of diagnoses recorded in the electronic medical record. Some of the most frequent conditions in the sample were diabetes, dementia, cancer, cardiac disease, hypertension, stroke, arthritis, seizure, and a psychiatric diagnosis. Age was based on the patient’s age in years
S5
at the time of injury and hospitalization. Gender and race/ethnicity were based on what was recorded in the medical record. Primary admitting trauma diagnosis was separated into two possibilities: the patient was hospitalized for either a fracture (e.g., fracture of arm, leg, or ribs) or a head injury (e.g., skull fracture, subdural hematoma, or brain laceration). Additional admitting diagnoses among three participants did not fit into these two categories. These additional diagnoses included infection, syncope, and seizure and the three individuals with these diagnoses were excluded from the analysis. Discharge disposition was based on recommendations of the physical and/or occupational therapists in conjunction with case management and patient/caregiver input and the available resources. The discharge disposition options were: home, home with home health services, inpatient rehabilitation, acute rehabilitation, chronic rehabilitation, traumatic brain injury rehabilitation, stroke rehabilitation, spinal cord injury rehabilitation, subacute rehabilitation, or skilled nursing facility. For the purposes of data analysis, these options were collapsed into two categories, either to home (with or without home health services) or to a facility (rehabilitation, skilled nursing, or long term care). Data analysis Descriptive statistics were used to describe the sample characteristics. Frequency and percentages were used for categorical level data; mean and standard deviation were used for continuous level data. Normality and outliers were examined to determine if variables met assumptions for the parametric statistical tests (e.g., correlations, t-tests, ANOVA) that were used. All variables except for length of stay met the assumption of normality. Length of stay was transformed using the natural log and then the assumption of normality was met. Logistic regression was used to determine the most parsimonious model for identifying the predictors of discharge disposition (nominal level of data as the outcome). Data were analyzed using SPSS 21.0. Structural equation modeling was used to test the hypothesized model as shown in Fig. 1. Model testing was done using AMOS 21.0 and maximum likelihood estimation. Missing data were accounted for by using the ’estimate means and intercepts’ function in AMOS. AMOS handles missing data through the full information maximum likelihood (FIML) method in which model parameters and standard errors are estimated directly from the available data. As previously noted, pain data was available in 88 cases. With 46 cases missing, a missing values analysis was performed, which showed that the data were missing completely at random using Little’s MCAR test (Chi-square ¼ 47.737, df ¼ 38, p ¼ .134). Model fit was estimated using the chi square statistic, the normed fit index (NFI) and Steigers Root Mean Square Error of Approximation (RMSEA). The larger the p value associated with the chi square, the better the model fits the data.67 The NFI, an incremental fit index, tests a hypothesized model against a baseline or one that does not fit the data. Ideally, the NFI should be as close to 1 with .95 as the recommended cut-off value to indicate a good fit.68 The RMSEA is an absolute fit index which explains how well the model would fit the population covariance matrix.69 The RMSEA is sensitive to the number of parameters in the model and a cut-off value less than .06 indicates a good model fit (Hu & Bentler, 1999). The Critical Ratio (CR), the parameter estimate divided by an estimate of the standard error, is used to determine path significance. A CR that is greater than the absolute value of two is considered significant (Pedhazur & Schmelkin, 1991). A p < .05 level of significance was used for all analyses.
S6
R.A. Brotemarkle et al. / Geriatric Nursing 36 (2015) S3eS9
Results
Table 2 Sample characteristics (N ¼ 132).
A total of 132 patients were included in the analysis. The group was almost half male (46.9%, n ¼ 62) and half female (53.1%, n ¼ 70). The average age of the sample was 78.3 (S.D. ¼ 9.7) years. The sample was 84.1% (n ¼ 111) Caucasian and 15.9% (n ¼ 21) other races/ethnicities (Table 2). The average number of comorbidities per patient was 2.29 (S.D. ¼ 1.96). Injury severity showed that about a third (34.1%, n ¼ 45) of the patients had mild or moderate injuries and the remaining two thirds (65.9%, n ¼ 87) had severe injuries. Approximately one third (40.2%, n ¼ 53) of the patients had an admitting diagnosis of head injury and almost two thirds of the patients (59.8%, n ¼ 79) had an admitting diagnosis of fracture. The average length of stay was 4.3 (S.D. ¼ 4.0) days. Discharge disposition showed that 42.4% (n ¼ 56) were discharged to home and 57.6% (n ¼ 76) were discharged to a facility (inpatient rehabilitation, acute rehabilitation, chronic rehabilitation, traumatic brain injury rehabilitation, stroke rehabilitation, spinal cord injury rehabilitation, subacute rehabilitation, or skilled nursing facility). Two patients died during the hospital stay. Model testing for length of stay Of the 21 hypothesized paths (Fig. 1), 12 were significant (bolded paths). Specifically, younger age, more comorbidities, greater injury severity, and admitting diagnosis were directly associated with pain and accounted for 13.7% of the variance in pain. In addition, these variables were indirectly associated with LOS through pain. Number of comorbidities and injury severity were directly associated with days between admission until first therapy evaluation and accounted for 11.5% of the variance in the days from admission to evaluation. Controlling for comorbidities, admitting diagnosis, and injury severity, days from admission to evaluation, age, and pain were directly associated with LOS and accounted for 37% of the variance in LOS (Table 1). Those with less pain, who were older, and had a greater number of days from admission until first therapy evaluation were noted to have a longer hospital LOS. The full hypothesized model did not fit the data with a Chisquare of 89.782 (degrees of freedom ¼ 34, [c2]/df ¼ 2.641, p < .001). The NFI was .776 and the RMSEA was .110. In order to test a more parsimonious model, the model was re-specified by
Table 1 Standardized regression weights for full model. Variables
Estimate
p
LOS ) Age LOS ) Comorbidities LOS ) Severity LOS ) Admitting diagnosis LOS ) Race LOS ) Days to eval LOS ) Pain LOS ) Gender Days to eval ) Comorbidities Days to eval ) Severity Days to eval ) Admitting diagnosis Pain ) Age Pain ) Comorbidities Pain ) Severity Pain ) Admitting diagnosis Pain ) Race Pain ) Days to eval Pain ) Gender Pre pain ) Pain During pain ) Pain Post pain ) Pain
L.163 .197 .252 .129 .102 .386 L.250 .024 .270 .201 .032 L.246 .258 .016 .108 .045 .052 .024 .894 .719 .978
.019 .007
Contents lists available at ScienceDirect
Geriatric Nursing journal homepage: www.gnjournal.com
Feature Article
Length of hospital stay and discharge disposition in older trauma patients Rebecca A. Brotemarkle, PhD, RN a, *, Barbara Resnick, PhD, RN a, Kathleen Michaels, PhD, RN a, Patricia Morton, PhD, RN b, Chris Wells, PhD, PT c a
University of Maryland School of Nursing, 655 West Lombard Street, Baltimore, MD 21201, USA The University of Utah College of Nursing, 10 South 2000 East, Salt Lake City, UT 84112, USA c University of Maryland School of Medicine, Allied Health Building, 100 Penn Street, Baltimore, MD 21201, USA b
a b s t r a c t Keywords: Older adults Trauma Length of Stay Discharge Disposition Pain Therapy
As the number of older adults increases and life expectancies are increasing, more incidences of traumatic injury are expected in this population. In this study, the relationships between demographic variables, pain, days from admission to therapy evaluation, length of stay and discharge disposition were examined in 132 older adults who had experienced a traumatic event. Results showed that significant relationships existed between pain, age, comorbidities, injury severity and days from admission to therapy evaluation and length of stay; those with less pain, greater age and had more days between admission and when the first therapy evaluation occurred had longer lengths of stay. In addition, demographic variables, overall length of stay and pain intensity during therapy were associated with discharge location; for longer lengths of stay and higher pain, older trauma patients were less likely to be discharged to home. Ó 2015 Elsevier Inc. All rights reserved.
Discharge disposition and length of hospital stay in older trauma patients From 1900 to 1994, the number of Americans aged 65 and older increased from three million to 33 million and is expected to reach 80 million by 2050.1 As the population of older adults increases, the incidence of traumatic events in this population will also increase as older adults are remaining active longer.2 Falls and motor vehicle accidents are the most likely mechanisms of injury in those over the age of 65 and 25% of all trauma incidents were in those over the age of 65.3 Impact of trauma on the health care system Following a trauma admission, older adults commonly need and receive follow-up treatment in various settings.4e6 The appropriate discharge setting is decided upon by a multidisciplinary team, including the attending physicians, advanced practice nurses or physician assistants, physical, occupational and speech therapists, social workers, patient/family preferences and available resources and health insurance.7,8 The setting options include home or self
* Corresponding author. E-mail address: [email protected] (R.A. Brotemarkle). 0197-4572/$ e see front matter Ó 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.gerinurse.2015.02.016
care (this would include assisted living, boarding homes, senior housing etc), short term general hospital, skilled nursing facility, intermediate care facility, home under care of organized home health service, inpatient rehabilitation, long term care hospital, and hospice care.9 Matching the discharge disposition correctly to the older trauma patient’s needs and preferences has a major impact on the health care system, length of stay (LOS), patient function, appropriate resource allocation, and rate of readmissions.8 Length of stay for trauma patients varies widely and is dependent upon multiple factors. The impact of LOS on cost of care is significant10,11 thus the overall goal in trauma care is to decrease LOS in order to decrease cost. The obvious solution is to transition patients into lower levels of care as soon as possible post trauma.12 Factors that influence outcomes post trauma based on the SocialeEcological Model The SocialeEcological Model (SEM)13e17 suggests that intrapersonal factors, interpersonal factors, environment/community factors and policies/systems may influence outcomes. Intrapersonal factors are defined as those characteristics that are specific to the individual; interpersonal factors focus on the interactions between individuals that can influence outcomes; environmental factors include such things as the physical environment and why an
S4
R.A. Brotemarkle et al. / Geriatric Nursing 36 (2015) S3eS9
individual might or might not be able to go for walk (e.g., no flat open areas; unsafe neighborhoods); and policies, culture and systems can influence what patients are allowed to do on the unit and whether or not, for example, they need to be tethered to pulse oximetry throughout the hospital stay. The levels in the SEM are interactive. Intrapersonal factors and discharge disposition and LOS Demographic variables including increased age, race/ethnicity, and gender have all been noted to be non-modifiable factors that influence LOS. Older trauma patients, when compared with younger trauma patients, have worse outcomes including: more admissions to the hospital,18 longer hospital stays,18,19 and a greater need for post-acute rehabilitation and/or long term care.4,10,20 The following patient factors have been associated with worse outcomes for older trauma patients: increased age, increased pain, being uninsured, having more comorbidities, having higher injury severity, and type of injury (traumatic brain injury, hip fracture, multi-trauma).21e23 The effects of gender24e27 and race/ ethnicity28e30 on discharge outcomes have been mixed. Some studies have shown that women were more likely to be discharged to nursing homes,4,27,31e34 some showing that women were more likely to be discharged to home24,25,28 and others showing no differences.35,36 With regard to the impact of race and ethnicity study results have not been consistent. In some studies African Americans and Hispanic patients had higher mortality28,37 and nonwhite patients were more likely to be discharged to home.4,34 In other studies there were no differences between African American, Hispanic patients and Caucasian patients.30,38 Comorbidity, defined as the presence of disease conditions in addition to the condition for which the patient was admitted to the hospital,39 influences outcomes post trauma.10,40e43 Comorbidities are generally non-modifiable factors; however, signs and symptoms of comorbidities in some cases are able to be controlled (e.g. blood sugar levels, dizziness, urinary frequency). Although the findings are not always consistent, comorbidities are associated with worse outcomes in older adults who have experienced trauma.34,44,45 Specifically, comorbidities influence LOS with more comorbid conditions being related to longer LOS.46e48 Pain post trauma has also been reported to influence LOS and location of discharge and is a potentially modifiable intrapersonal factor. Prior research has shown that when older adults have acute pain it interrupts physical therapy in approximately a third of the patients.49 If pain interferes with a patient’s ability or desire to participate in therapy, slower recovery times and increased length of hospital stays are likely to occur.50e52 Pain that is not well controlled limits an individual’s functional performance and may result in a lower level of care (i.e., a long term care facilities versus a skilled post-acute rehabilitation stay).
The goal of discharge planning is to assure that patients receive needed services in the least restrictive setting. In addition, an appropriate discharge plan can decrease LOS and prevent readmissions to the hospital. Knowing what influences discharge disposition and LOS may help guide in improving the discharge planning process and decreasing LOS for older patients. The purpose of this study, therefore, was to comprehensively consider the factors associated with discharge disposition and length of stay among older adults hospitalized for traumatic injury. Specifically, it was hypothesized that demographic variables (gender, age), health related factors (comorbidities; injury severity; admitting diagnosis), pain and initiation of therapy would be directly and indirectly associated with LOS and discharge disposition (Fig. 1). Methods Design This study was a secondary data analysis using data from inpatient rehabilitation services provided to patients during inpatient stays in a Level 1 trauma center. Data collection was done using a retrospective chart review from Powerchart, an electronic medical record used in the participating trauma center. Specifically, the data sources included computerized therapy (physical, occupational, and speech) notes for each exposure to therapy during the patients’ hospital stay or up to 30 days post admission. Sample Eligibility was based on the participants being 65 years or older and having been admitted to an urban Level I Trauma Center following a traumatic event. A sample size of 137 adults was randomly chosen from 1387 admissions over a two-year period from 2010 to 2011. Five of the participants were excluded as three
Interpersonal factors and discharge disposition and LOS A potentially modifiable interpersonal factor is the type and timing of physical and occupational therapy. Physical and occupational therapy are an important part of a trauma patient’s recovery process. Older adults are known to have accelerated functional decline after a traumatic event.53e56 To prevent functional decline and assist older adults in returning to pre-injury functional levels, there has been a recent focus on early treatment and mobilization.57e61 Early physical rehabilitation programs have shown a length of stay reduction of 22% for patients in intensive care units and 19% for patients on general units.62
Fig. 1. Hypothesized Model.
R.A. Brotemarkle et al. / Geriatric Nursing 36 (2015) S3eS9
had primary diagnoses that were not trauma diagnoses and two died before they could participate in therapy. The final sample size was 132. The study was approved by a University Institutional Review Board. Data collection and measures Data were collected by registered nurses with acute care experience and expertise in the use of Powerchart. These individuals had not provided direct care to the study participants. The data collectors were instructed how to access the necessary data and how to complete the data collection forms for each patient. Therapy notes included preadmission function, admission day and length of stay, rehabilitation order day, day of first/initial therapy evaluation, and subsequent visits. Baseline assessments by all of the therapists include cognitive function and pain. In addition, physical and occupational therapists evaluated patients’ activities of daily living, instrumental activities of daily living, range of motion, manual muscle testing, and activity tolerance. Speech language pathologists included assessment of language skills and swallowing capabilities. Entries are not forced and therapists completed assessment components based on their expertise and clinical decision-making. Descriptive information about each rehabilitation encounter was also obtained, including the reason for the visit, what occurred during that visit, and documented reasons why a visit did not occur (e.g., patient refused). Demographic information about patients was obtained, including age, gender, race/ethnicity, comorbidities, living location prior to the traumatic event, and location after discharge. Length of stay was measured as the day of admission until the day of discharge, however, the day of discharge was not included in the count of days. For those patients admitted and discharged on the same day, the LOS was counted as one day. Pain was conceptualized as pain experienced prior to, during, and immediately following therapy (Fig. 1). The computerized system allows therapists to choose one of three pain assessment tools: the Verbal Rating Scale (VRS),63 Verbal Descriptor Scale (VDS)63,64 or the Checklist of Nonverbal Pain Indicators (CNPI).65 The VRS was the most frequently used among the three measures and involved a rating of zero to ten, where zero is no pain and ten is the worst imaginable pain. For the purpose of this study, the VRS data were used (n ¼ 88) as the measure of pain since the other measures are not on a numeric scale or do not measure pain intensity. Among the 46 participants in which the VRS was not completed, 30 (65%) had no assessment of pain identified and 16 (35%) used one of the other measurement options. As shown in Fig. 1, the measurement model of pain which included the prior to therapy, during and immediately following reports of pain by the patient fit the data with paths significant at the .001 level (pretherapy pain path was .89; during therapy path was .72; and post therapy path was .98). Injury severity was determined based on anatomic factors (the number of injured body areas) and the physiologic method of measuring severity of injury (shock or coma).66 Severe injury was based on having more than one trauma diagnosis and having a diagnosis of shock or a head injury with coma. Otherwise, if there was only one trauma diagnosis and there was no diagnosis of shock or coma, then the injury severity was considered to be mild/ moderate. Comorbidities were counted as the number of conditions not related to the admitting diagnosis that are checked off a list of diagnoses recorded in the electronic medical record. Some of the most frequent conditions in the sample were diabetes, dementia, cancer, cardiac disease, hypertension, stroke, arthritis, seizure, and a psychiatric diagnosis. Age was based on the patient’s age in years
S5
at the time of injury and hospitalization. Gender and race/ethnicity were based on what was recorded in the medical record. Primary admitting trauma diagnosis was separated into two possibilities: the patient was hospitalized for either a fracture (e.g., fracture of arm, leg, or ribs) or a head injury (e.g., skull fracture, subdural hematoma, or brain laceration). Additional admitting diagnoses among three participants did not fit into these two categories. These additional diagnoses included infection, syncope, and seizure and the three individuals with these diagnoses were excluded from the analysis. Discharge disposition was based on recommendations of the physical and/or occupational therapists in conjunction with case management and patient/caregiver input and the available resources. The discharge disposition options were: home, home with home health services, inpatient rehabilitation, acute rehabilitation, chronic rehabilitation, traumatic brain injury rehabilitation, stroke rehabilitation, spinal cord injury rehabilitation, subacute rehabilitation, or skilled nursing facility. For the purposes of data analysis, these options were collapsed into two categories, either to home (with or without home health services) or to a facility (rehabilitation, skilled nursing, or long term care). Data analysis Descriptive statistics were used to describe the sample characteristics. Frequency and percentages were used for categorical level data; mean and standard deviation were used for continuous level data. Normality and outliers were examined to determine if variables met assumptions for the parametric statistical tests (e.g., correlations, t-tests, ANOVA) that were used. All variables except for length of stay met the assumption of normality. Length of stay was transformed using the natural log and then the assumption of normality was met. Logistic regression was used to determine the most parsimonious model for identifying the predictors of discharge disposition (nominal level of data as the outcome). Data were analyzed using SPSS 21.0. Structural equation modeling was used to test the hypothesized model as shown in Fig. 1. Model testing was done using AMOS 21.0 and maximum likelihood estimation. Missing data were accounted for by using the ’estimate means and intercepts’ function in AMOS. AMOS handles missing data through the full information maximum likelihood (FIML) method in which model parameters and standard errors are estimated directly from the available data. As previously noted, pain data was available in 88 cases. With 46 cases missing, a missing values analysis was performed, which showed that the data were missing completely at random using Little’s MCAR test (Chi-square ¼ 47.737, df ¼ 38, p ¼ .134). Model fit was estimated using the chi square statistic, the normed fit index (NFI) and Steigers Root Mean Square Error of Approximation (RMSEA). The larger the p value associated with the chi square, the better the model fits the data.67 The NFI, an incremental fit index, tests a hypothesized model against a baseline or one that does not fit the data. Ideally, the NFI should be as close to 1 with .95 as the recommended cut-off value to indicate a good fit.68 The RMSEA is an absolute fit index which explains how well the model would fit the population covariance matrix.69 The RMSEA is sensitive to the number of parameters in the model and a cut-off value less than .06 indicates a good model fit (Hu & Bentler, 1999). The Critical Ratio (CR), the parameter estimate divided by an estimate of the standard error, is used to determine path significance. A CR that is greater than the absolute value of two is considered significant (Pedhazur & Schmelkin, 1991). A p < .05 level of significance was used for all analyses.
S6
R.A. Brotemarkle et al. / Geriatric Nursing 36 (2015) S3eS9
Results
Table 2 Sample characteristics (N ¼ 132).
A total of 132 patients were included in the analysis. The group was almost half male (46.9%, n ¼ 62) and half female (53.1%, n ¼ 70). The average age of the sample was 78.3 (S.D. ¼ 9.7) years. The sample was 84.1% (n ¼ 111) Caucasian and 15.9% (n ¼ 21) other races/ethnicities (Table 2). The average number of comorbidities per patient was 2.29 (S.D. ¼ 1.96). Injury severity showed that about a third (34.1%, n ¼ 45) of the patients had mild or moderate injuries and the remaining two thirds (65.9%, n ¼ 87) had severe injuries. Approximately one third (40.2%, n ¼ 53) of the patients had an admitting diagnosis of head injury and almost two thirds of the patients (59.8%, n ¼ 79) had an admitting diagnosis of fracture. The average length of stay was 4.3 (S.D. ¼ 4.0) days. Discharge disposition showed that 42.4% (n ¼ 56) were discharged to home and 57.6% (n ¼ 76) were discharged to a facility (inpatient rehabilitation, acute rehabilitation, chronic rehabilitation, traumatic brain injury rehabilitation, stroke rehabilitation, spinal cord injury rehabilitation, subacute rehabilitation, or skilled nursing facility). Two patients died during the hospital stay. Model testing for length of stay Of the 21 hypothesized paths (Fig. 1), 12 were significant (bolded paths). Specifically, younger age, more comorbidities, greater injury severity, and admitting diagnosis were directly associated with pain and accounted for 13.7% of the variance in pain. In addition, these variables were indirectly associated with LOS through pain. Number of comorbidities and injury severity were directly associated with days between admission until first therapy evaluation and accounted for 11.5% of the variance in the days from admission to evaluation. Controlling for comorbidities, admitting diagnosis, and injury severity, days from admission to evaluation, age, and pain were directly associated with LOS and accounted for 37% of the variance in LOS (Table 1). Those with less pain, who were older, and had a greater number of days from admission until first therapy evaluation were noted to have a longer hospital LOS. The full hypothesized model did not fit the data with a Chisquare of 89.782 (degrees of freedom ¼ 34, [c2]/df ¼ 2.641, p < .001). The NFI was .776 and the RMSEA was .110. In order to test a more parsimonious model, the model was re-specified by
Table 1 Standardized regression weights for full model. Variables
Estimate
p
LOS ) Age LOS ) Comorbidities LOS ) Severity LOS ) Admitting diagnosis LOS ) Race LOS ) Days to eval LOS ) Pain LOS ) Gender Days to eval ) Comorbidities Days to eval ) Severity Days to eval ) Admitting diagnosis Pain ) Age Pain ) Comorbidities Pain ) Severity Pain ) Admitting diagnosis Pain ) Race Pain ) Days to eval Pain ) Gender Pre pain ) Pain During pain ) Pain Post pain ) Pain
L.163 .197 .252 .129 .102 .386 L.250 .024 .270 .201 .032 L.246 .258 .016 .108 .045 .052 .024 .894 .719 .978
.019 .007
