key: cord-0729517-vj4v69pg
authors: Tevald, Michael A.; Clancy, Malachy J.; Butler, Kelly; Drollinger, Megan; Adler, Joe; Malone, Daniel
title: AM-PAC “6-Clicks” for the Prediction of Short Term Clinical Outcomes in Individuals Hospitalized with COVID-19: A Retrospective Cohort Study
date: 2021-09-05
journal: Arch Phys Med Rehabil
DOI: 10.1016/j.apmr.2021.08.006
sha: 76417ead6b4a79fdbe017656036fc1b7b7dc1f0d
doc_id: 729517
cord_uid: vj4v69pg

OBJECTIVE To determine the ability of the Activity Measure for Post-Acute Care (AM-PAC) assessments of mobility and activity to predict key clinical outcomes in patients hospitalized with COVID-19. DESIGN Retrospective cohort study SETTING An academic health system in the United States consisting of five inpatient hospitals. PARTICIPANTS Adult patients (N=1486) urgently or emergently admitted, tested positive for COVID-19, and had at least 1 AM-PAC assessment. INTERVENTIONS Not applicable MAIN OUTCOME MEASURES Discharge destination, hospital length of stay, in-hospital mortality, and readmission. RESULTS 1486 admission records were included in the analysis. After controlling for covariates, initial and final mobility (OR = 0.867 and 0.833, respectively) and activity scores (OR = 0.892 and 0.862, respectively) were both independent predictors of discharge destination with a high accuracy of prediction (AUC = 0.819-0.847). Using a threshold score of 17.5, sensitivity ranged from 0.72-0.79, while specificity ranged from 0.74-0.83. Both initial AM-PAC mobility and activity scores were independent predictors of mortality (OR = 0.885 and 0.877, respectively). Initial mobility, but not activity, scores were predictive of prolonged LOS (OR = 0.957 and 0.980, respectively). However, the accuracy of prediction for both outcomes was weak (AUC = 0.659-0.679). AM-PAC scores did not predict re-hospitalization. CONCLUSION Functional status as measured by the AM-PAC “6-clicks” measures of mobility and activity are independent predictors of key clinical outcomes individual hospitalized with COVID-19.

The outbreak of the novel SARS-Cov-2 has had a significant impact on healthcare delivery throughout the world, with over 175 million cases and nearly 4 million deaths worldwide. 1 Infection with SARS-Cov-2 leads to COVID-19, which is highly variable in terms of clinical presentation and severity. 2 While some will present with no or minimal symptoms, 20% of infected individuals develop severe COVID-19, and a significant fraction will require prolonged hospitalization, placement in a post-acute care facility, or readmission to the hospital. 3, 4 Early identification of individuals at risk for such outcomes may help reduce the burden imposed on healthcare systems.

The risks of infection, hospitalization, and death due to COVID-19 are higher in the elderly, those with medical comorbidities, 5 and in the Black and Hispanic-Latino populations. 6 However, other factors may be necessary to identify individuals who require additional services in the acute and early post-acute phase. For example, three recent studies have reported significant associations between impairments in physical function and important outcomes like discharge destination 7 , hospital readmission, 8 and mortality 9 in this population. These results are consistent with a small but growing number of reports highlighting prognostic significance of physical function in hospitalized patients, 10 suggesting that routine assessment of physical function in the hospital may help guide the delivery of care, particularly during a pandemic when resources are stretched.

The Activity Measure for Post-Acute Care (AM-PAC) "6-Clicks'' assessments of basic mobility and activities of daily living are specific measures of patient functioning that have been adopted by numerous healthcare systems. The tools are quick and easy to implement and predict discharge destination in a variety of hospitalized populations. [11] [12] [13] The association RUNNING HEAD: Physical function and outcomes in COVID between AM-PAC scores and clinical outcomes in COVID-19, however, have not been fully investigated. Therefore, the purpose of this study was to test the hypothesis that AM-PAC "6clicks" activity and mobility scores would predict discharge disposition, hospital length of stay (LOS), in-hospital mortality and, and 30-day readmission in individuals hospitalized with COVID- 19 .

This retrospective observational cohort study utilized an inpatient COVID-19 registry comprised of medical record data of patients admitted to five hospitals within the University of Pennsylvania Health System (UPHS). The UPHS includes both community and academic medical centers in urban and suburban areas across the greater metropolitan Philadelphia region. The records from individuals 17 years of age or older who were urgently or emergently admitted to one of the five hospitals from March 1st to July 31st, 2020, who tested positive for COVID-19 during their admission, and who had at least 1 AM-PAC "6 Clicks" rating were included in the analysis. Hospice or psychiatric admission records were excluded. The study protocol was approved by the University of Pennsylvania Institutional Review Board (protocol #843920).

Outcome variables included discharge destination (home vs. facility, including long term acute hospitals, skilled nursing facilities, and inpatient rehabilitation facilities), hospital LOS (dichotomized by 75 th percentile; 11 days), in-hospital mortality, and 30-day readmission. The analysis of discharge destination only included those who were discharged from the hospital alive. Readmission data was retrieved more than 30 days after the end of the observation period, and included individuals who were readmitted to one of the five hospitals in the system within 30 days of discharge from the index admission, excluding hospice admissions or patients admitted within six hours from discharge.

Our predictor variables were AM-PAC "6-clicks" mobility and activity scores. Each assessment consists of 6 items scored on a 1-4 scale, producing an overall score of 6-24, with lower scores indicating poorer performance. The mobility assessment includes the following tasks; 1) turning over in bed, 2) sitting down and standing up from a chair, 3) moving from lying to sitting, 4) moving to and from a bed to a chair, 5) walk in room, and 6) climb 3-5 steps. The activity assessment includes items related to activities of daily living: 1) putting on and taking off lower body clothing, 2) bathing, 3) toileting, 4) putting on and taking off upper body clothing, 5) grooming, and 6) eating meals. Previous studies have demonstrated the validity of the assessments in the hospital environment 14 , and that they can be assessed by multiple disciplines with excellent reliability. 15, 16 In the UPHS system, nurses, physical therapists and occupational therapists are all trained to use the assessments, and assessments from all three disciplines were included in this dataset. The patient's initial AM-PAC score was used to predict mortality and hospital LOS, while the final score was used to predict readmission. Both initial and final scores were used to predict discharge destination in separate analyses.

Demographic variables considered as potential covariates include age (> or <= 75 years) 17 , sex, race (black, white, other) 18 , ethnicity (Hispanic/Latino vs. not), and insurance status (uninsured, Medicaid, Medicare, private, other). The admitting hospital was also included to account for variation across facilities. 19 Clinical variables included primary/admitting diagnosis (COVID-19 vs. other), admission to the intensive care unit (ICU; yes vs. no), RUNNING HEAD: Physical function and outcomes in COVID mechanical ventilation (yes vs. no), and hospital LOS (days). The Charlson Comorbidity Index 20 was included, as the presence of medical comorbidities increase the risk and severity of infection with COVID-19. 5 Preliminary analysis suggested a significant impact of date of admission on several variables, including mortality; 21 as a result, the sample was divided into two "cohorts" based on whether their admission date was in the first (3/1 to 5/17) or second (5/18 to 7/31) half of the study time period. Due to the variability in the timing of the AM-PAC assessments, the number of days between admission and initial AM-PAC score was included as a covariate for the prediction of mortality, hospital LOS, and discharge destination. Similarly, the number of days between the final AM-PAC assessment and discharge was used as a covariate in the prediction of discharge destination and readmission.

Continuous variables are expressed as median and interquartile range, as the distribution of most variables is skewed, while categorical variables are expressed as frequencies and percentages. Subject characteristics were compared between subsets (COVID as admitting/primary diagnosis vs. COVD as secondary diagnosis, first cohort vs. second cohort, included vs. excluded) using the Mann-Whitney U test for continuous variables and Chi-square for categorical variables. Preliminary bivariate analysis demonstrated that all potential covariates were significantly associated with at least two of the four outcome measures, so all were included in the multivariate analyses. The assumption of linearity of each of the continuous predictor variables with the logit was assessed using the Box-Tidwell procedure with Bonferroni correction. 22

Separate multivariate logistic regression models were generated for each combination of AM-PAC (mobility, activity) and outcome variable (mortality, LOS, discharge destination, readmission). All variables were entered using the forced-entry method in two steps; the first step included just the covariates, while the second included covariates and the AM-PAC score of interest (e.g., mobility or activity, initial or final). Receiver operating characteristic (ROC) curves were created for each step using the predicted probabilities from each model, and change in the area under the curve (AUC) was used to evaluate the impact of the addition of the AM-PAC score on the overall prediction of the model. Additionally, separate ROC curves were created using only the raw AM-PAC scores to determine the AUC for prediction and identify threshold scores for predicting each of the outcome variables. Stratified analyses, using identical regression models, were performed to further explore the impact of admitting/primary diagnosis (COVID vs. other), date of admission (first vs. second cohort), admission to the ICU (yes vs. no), medical comorbidity (dichotomized by the median), and race (Black vs. White + Other). The impact of selection bias on the relationship between the covariates and the dependent variables was explored using separate regression models including only covariates as predictors for included and excluded participants. All analyses were conducted using SPSS v26 (IBM Inc., Armonk, NY), and the level of significance was set at p <0.05.

Participant flow is shown in Figure 1 , and demographics are shown in Table 1 . Sixty-two percent of adult urgent/emergent admissions had at least one AM-PAC mobility score, while RUNNING HEAD: Physical function and outcomes in COVID 51% had at least one AM-PAC activity score. Physical and/or Occupational Therapy was consulted in 861 of the 1456 records with at least one AM-PAC mobility score. 74% of those with a Physical Therapy consult had at least one AM-PAC Mobility score entered by a Physical Therapist, while 68% of those with an Occupational Therapy consult had at least one AM-PAC Activity score entered by an Occupational Therapist. Compared to participants who had at least 1 AM-PAC score in their record, excluded participants were younger, had fewer comorbidities, shorter LOS, were less likely to be admitted to the intensive care unit, more likely to be discharged to home, but also more likely to die in the hospital (see Supplemental Table 1 ).

COVID-19 was the admitting or primary diagnosis for 48.6% of included participants.

Primary diagnoses varied for the remainder of the sample, with the most common being infection/sepsis (27% of the total), pneumonia (1.6%), gastro-intestinal issues (1.4%), trauma/fracture (1.3%). Those with COVID as an admitting or primary diagnosis were more likely to be discharged home, less likely to be admitted to the ICU or require mechanical ventilation, and had higher AM-PAC scores than those with COVID as a secondary diagnosis. As seen in Supplemental Table 2 , participants in the first cohort were more likely to have COVID as an admitting or primary diagnosis, to require mechanical ventilation, have a longer hospital LOS, die in the hospital, and be discharged to a post-acute care facility that those in the second cohort.

The results of the step-wise logistic regression analyses are presented in Table 2 , and the result of the stratified analyses are displayed in Tables 3-5 . For discharge destination, the predictive ability of initial and final AM-PAC scores was assessed in separate regression models. RUNNING HEAD: Physical function and outcomes in COVID After controlling for the influence of covariates, AM-PAC mobility and activity scores were both independent predictors of discharge destination (see Table 3 ). Furthermore, their addition significantly improved the accuracy of prediction of the model, as shown by the change in AUC (see Table 2 ). Each one-point decrease in initial AM-PAC score increased the odds of discharge to a facility by 1.15 (1.12, 1.19) and 1.12 (1.08, 1.12) fold for mobility and activity, respectively. no), the number of medical comorbidities, and race (see Supplemental Table 3 ). In addition, the comparison of regression models including only covariates for included and excluded individuals suggest similar relationships between each of the covariates and discharge destination in the two groups (see Supplemental Table 4 ).

After controlling for the influence of covariates, initial AM-PAC mobility, but not activity, was a significant independent predictor of hospital LOS (see Table 4 ). However, it's addition to the model had only a modest effect on the accuracy of prediction (see Table 2 ). Each point decrease in mobility score increased the odds that the hospital LOS would be 11 days or longer by 1.04 (95% CI = 1.02,1.07) fold. Stratified analysis revealed similar results for AM-PAC mobility score regardless of primary/admitting diagnosis (COVID vs. other), and cohort (date of admission in first vs. second half of specified time period). However, mobility scores were no longer predictors when the analysis was restricted to those who were admitted to the ICU. RUNNING HEAD: Physical function and outcomes in COVID Because AM-PAC activity was not a predictor in the primary analysis, additional stratified analyses were not conducted.

After controlling for the influence of covariates, initial AM-PAC mobility and activity scores were both independent predictors of mortality after controlling for the influence of covariates (see Table 5 ). However, their addition did not significantly improve the overall prediction of the model (see Table 2 ). Each point decrease in the initial AM-PAC score increased the odds of in-hospital mortality by 1.13 (95% CI = 1.08, 1.18) or 1.14 (95% CI = 1.09-1. 20 

These results indicate that AM-PAC "6-clicks" mobility and activity scores are strong predictors of discharge destination in individuals hospitalized with COVID-19. Strengths of the study include a large sample size from a health system consisting of multiple centers serving urban and suburban areas, including community hospitals and academic medical centers. The results suggest that simple measures of physical function can help guide discharge planning to maximize the efficiency of care delivery in the setting of a global pandemic.

Our results confirm and extend those of other studies demonstrating a link between functional impairments and discharge destination in COVID-19, 7 as well those supporting the prognostic significance of AM-PAC scores in mixed 13,23 and diagnosis-specific patient groups. 11, 24 The similarity in the threshold scores and accuracy of prediction between this and other [11] [12] [13] demonstrate that the relationship between AM-PAC scores and discharge destination is robust, and applicable even in the setting of a global pandemic.

We are unaware of other studies assessing the association between physical function and LOS in COVID-19, but Laosa and colleagues 9 reported that poor function on admission predicted mortality in those hospitalized in the first month of the pandemic. While we found AM-PAC scores to be poor predictors of mortality overall, the strength of the association was substantially higher in the first cohort than the second. Thus, the improvements in medical management of COVID-19 appear to have blunted to association between physical function and mortality, which may contribute to the discrepancy. The lack of association between physical function and readmission differs from the results of Bowles and colleagues, 8 who studied individuals discharged to home following hospitalization for COVID-19. In contrast, nearly 40% RUNNING HEAD: Physical function and outcomes in COVID of the individuals in the current study were discharged to a post-acute care facility, where it is likely that their medical status was closely monitored, decreasing their risk of re-hospitalization.

In addition, only readmission to one of the five hospitals in the UPHS would be captured by our registry. Therefore, our data may underestimate the readmission rate, which may limit our ability to draw conclusions about the relationship between in-hospital physical function and readmission.

Overall, the results of this investigation suggest that AM-PAC scores provide important information that can guide discharge planning, which may reduce costly delays in the transitioning of patients to the appropriate next level of care 25 and the rate of hospital readmission. 26 Early assessment of mobility can also guide resource allocation decisions. For example, Johnson 27 and colleagues found that increased frequency of rehabilitation services improved function at the time of hospital discharge and increased the likelihood of discharge to home. The fact that AM-PAC scores were assessed by multiple disciplines, including nursing, suggests that mobility assessment can provide important information even when Physical or Occupational Therapy are not involved in a patient's care.

Several limitations of this study must be considered. First, all individuals with a positive COVID test were considered for eligibility, regardless of the severity of their COVID-specific symptoms. As a result, some individuals in our sample may have had no or minor symptoms specifically related to COVID-19. However, our sensitivity analysis demonstrated similar results when the analysis was restricted to those whose admitting or primary diagnosis was COVID-19, all of whom presumably had a severe presentation. Second, we did not have information on RUNNING HEAD: Physical function and outcomes in COVID other factors known to impact the outcomes of interest. Prior living and functional status, marital status, and baseline cognition are all known to impact discharge destination, but those data were not available in our registry. In addition, medical complications that did not require admission to the intensive care or mechanical ventilation were not captured in our dataset, and we cannot assess the impact of these factors on our results. Third, only including individuals who had an AM-PAC score documented in their medical record may limit the generalizability of our results to all individuals hospitalized with COVID-19. While it is unclear what factors might make a clinician more or less likely to implement the AM-PAC assessment in a given patient, the results of our analysis suggest that the potential for significant selection bias is small. Finally, our readmission rates may be underestimates, as our registry only included readmission data from hospitals in our system.

In conclusion, AM-PAC "6-clicks" assessments of mobility and activity are easy to implement, can be completed by multiple disciplines, and provide important prognostic information. Consistent application of the AM-PAC assessments may help to maximize the efficiency of care delivery in this and future pandemics. Future research should assess the ability of other functional measures to predict short-and long term outcomes in individuals hospitalized with COVID-19. In addition, efforts should be focused on incorporating preexisting guidelines 28 for consistent outcome measure use and reporting in this population, which may ultimately improve patient outcomes. Results of the main (sample = All) and sub-analyses on different subsets of the cohort (e.g., only those with COVID as a primary or admitting diagnosis, etc.). Model p<0.001 for all analyses.

Nagelkerke R 2 ranged from 0.482 to 0.552. Results of the main (sample = All) and sub-analyses on different subsets of the cohort (e.g., only those with COVID as a primary or admitting diagnosis, etc.). Additional sub-analyses using AM-RUNNING HEAD: Physical function and outcomes in COVID PAC activity score were not conducted due to the lack of significance of the main analysis.

Model p<0.001 for all analyses. Nagelkerke R 2 ranged from 0.481 to 0.525. Results of the main (sample = All) and sub-analyses on different subsets of the cohort (e.g., only those with COVID as a primary or admitting diagnosis, etc.). Model p<0.001 for all analyses.

Nagelkerke R 2 ranged from 0.449 to 0.512. Accuracy of AM-PAC prediction of key outcomes using AM-PAC mobility and activity scores.

Sensitivity and specificity are for a cut-off score of 17.5. AUC = area under the curve.

RUNNING HEAD: Physical function and outcomes in COVID 

Characteristics of and Important Lessons From the Coronavirus Disease 2019 (COVID-19) Outbreak in China: Summary of a Report of 72314 Cases From the Chinese Center for Disease Control and Prevention

Presenting Characteristics, Comorbidities, and Outcomes Among 5700 Patients Hospitalized With COVID-19 in the New York City Area

Characterization of Patients Who Return to Hospital Following Discharge from Hospitalization for COVID-19

Centers for Disease Control and Prevention Risk for COVID-19 Infection, Hospitalization, and Death By Race

RUNNING HEAD: Physical function and outcomes in COVID

Identification of Functional Limitations and Discharge Destination in Patients With COVID-19

Surviving COVID-19 After Hospital Discharge: Symptom, Functional, and Adverse Outcomes of Home Health Recipients

Rapid Assessment at Hospital Admission of Mortality Risk From COVID-19: The Role of Functional Status

Utility of Functional Metrics Assessed During Acute Care on Hospital Outcomes: A Systematic Review

Use of AM-PAC "6 Click" socres to predict discharge location post-hospitalization in adults with cardiovascular disease: A retrospective cohort study

Association of AM-PAC "6-Clicks" Basic Mobility and Daily Activity Scores With Discharge Destination

AM-PAC "6-Clicks" functional assessment scores predict acute care hospital discharge destination

Physical function and outcomes in COVID 14

Validity of the AM-PAC "6-Clicks" inpatient daily activity and basic mobility short forms

Association of impaired functional status at hospital discharge and subsequent rehospitalization

Interrater Reliability of AM-PAC "6-Clicks" Basic Mobility and Daily Activity Short Forms

Age, Frailty, and Comorbidity as Prognostic Factors for Short-Term Outcomes in Patients With Coronavirus Disease 2019 in Geriatric Care

Racial and Ethnic Differences and Clinical Outcomes of COVID-19 Patients Presenting to the Emergency Department

Patient and Hospital Factors Associated With Differences in Mortality Rates Among Black and White US Medicare Beneficiaries Hospitalized With COVID-19 Infection

A new method of classifying prognostic comorbidity in longitudinal studies: development and validation