Improvement in ADLs During a Nursing Home Stay and One Year Mortality among Older Adults with Sepsis

Abstract

Background/objective:

To describe the recovery of activities of daily living (ADLs) during a skilled nursing facility (SNF) stay and the association with one-year mortality after SNF discharge among Medicare beneficiaries treated in intensive care for sepsis.

Design:

Retrospective cohort study.

Setting:

Skilled nursing facilities in the United States.

Participants:

Medicare fee-for-service beneficiaries admitted to a SNF within 3-days of discharge from a hospitalization that included an intensive care unit (ICU) stay for sepsis between January 1, 2013 and September 30, 2015 (N=59,383).

Measurements:

Data from the Minimum Data Set (MDS) were used to calculate a total score for seven ADLs. Improvement was determined by comparing the total ADL scores from the first and last MDS assessments of the SNF stay. Proportional hazard models were used to estimate the association between improvement in ADL function and one-year mortality after SNF discharge.

Results:

Approximately 58% of SNF residents had any improvement in ADL function. Residents who had improvement in ADL function had 0.72 (95% CI=0.69-0.74) lower risk for mortality following SNF discharge than residents who did not improve. Residents who improved 1-3 points (HR=0.82, 95% CI=0.79-0.84) and four or more points (HR=0.57, 95% CI=0.55-0.60) in ADL function had significantly lower mortality risk than residents who did not improve.

Conclusions:

Older adults treated in an ICU with sepsis can improve in ADL function during a SNF stay. This improvement is associated with lower one-year mortality risk after SNF discharge. These findings provide evidence that ADL recovery during a SNF stay is associated with better health outcomes for older adults who have survived an ICU stay for sepsis.

Introduction

Sepsis is a systemic inflammatory response to an infection that can lead to widespread organ failure.¹ Over half of hospitalized patients with sepsis are over the age of 65.² Older adults with sepsis often receive care in an intensive care unit (ICU).³ ICU patients can experience long periods of inactivity,⁴ disrupted sleep,⁵ and delirium.⁶ Consequently, many older adults with sepsis develop muscle weakness and physical disability while in the ICU.⁷

Sepsis has a high mortality rate, especially for older adults.⁸ The Centers for Disease Control and Prevention estimates that 75% of adults in which sepsis is the recorded cause of death are 65 or older.⁹ However, several studies have reported a decreasing trend in hospital mortality rates.^{10, 11} This trend in hospital mortality has coincided with an increase in the number of survivors discharged to post-acute care.^{12, 13}

Skilled nursing facilities (SNFs) are the most common site of post-acute care for older adults who have survived a hospitalization for sepsis.¹⁴ Older adults admitted to a SNF after a hospitalization for sepsis have considerable limitations in activities of daily living (ADL). An analysis of 135,370 Medicare beneficiaries discharged to a SNF following a hospitalization for severe sepsis found that nearly 40% of older adults were dependent in four or more ADLs at SNF admission.¹⁵ This level of ADL dependence is substantially higher than other SNF populations.¹⁶

The increasing use of SNF care by older adults who have survived sepsis, the substantial ADL disability, and high mortality rates in this population make it important to investigate the recovery of ADL function during a SNF stay and determine if this improvement is associated with reduced mortality after SNF discharge. However, the functional recovery of older adult sepsis survivors discharged to post-acute care has not been described. Additionally, ADL limitations have been associated with increased mortality among older adults with sepsis,¹⁵ but it is not known if improvement in ADL function during a post-acute care stay is associated with decreased mortality.

We used data from a national sample of Medicare beneficiaries who were discharged to a SNF for post-acute care following an ICU stay with sepsis. Our first objective was to identify resident characteristics at SNF admission that are associated with the recovery of ADL function during a SNF stay. Our second objective was to determine if the improvement in ADL function during a SNF stay is associated with lower one-year mortality risk after SNF discharge.

Methods

Data Sources

We used 100% National Medicare data files from January 01, 2013 through December 31, 2015. These data files included the Master Beneficiary Summary File, Medicare Provider Analysis and Review (MedPAR) files, and Minimum Data Set (MDS) version 3.0. The MDS assesses the health, function, and treatments needs of nursing home residents.¹⁷ The MDS has undergone multiple revisions, most recently in 2010 (MDS 3.0). The major revisions include the addition of questionnaires for mood, pain, and cognition and changes to existing sections (e.g., nutrition, falls, balance) that have improved the MDS.¹⁸ We accessed the Medicare files after review by our University Institutional Review Board and submitting a Data Use Agreement approved by the Centers for Medicare and Medicaid Services.

Sample Selection

The sample selection criteria are shown in Figure 1. We identified 953,118 Medicare beneficiaries admitted to SNF within 3-days of hospital discharge after receiving care in an ICU and who survived the SNF stay. Approximately nine percent (n=81,785) of these beneficiaries had a diagnostic related group (DRG) code for septicemia in the hospital claim (DRG 870, 871, or 872). We first excluded SNF residents whose first MDS assessment was coded as a re-entry assessment. Next, we excluded residents who did not have complete information for the variables of interest at their first MDS assessments. Of the 5,154 residents with missing data, 5,043 residents were missing data for one or more variables to determine cognitive status. The greatest differences between residents with and without missing data were for hours of physical therapy (PT) and occupational therapy (OT) during the first 7 days of the SNF stay (Supplemental Table S1). Nearly 74% of residents with missing data received four hours or less of PT and 77% received four hours or less of OT compared to 35.2% and 41.5% of residents with no missing data, respectively. As shown in Supplementary Table S1, the majority (60.3%) of residents with missing data were deceased within one-year of SNF discharge compared to 46.9% of residents with no missing data (hazard ratio [HR]=1.66, 95% CI=1.59-1.73). Residents with missing data were also at greater risk for one-year mortality compared to residents included in the final sample (HR=1.47, 95% CI=1.42-1.53).

Figure 1:

The selection of skilled nursing facility residents discharged from the hospital after an intensive care unit stay with sepsis.

We also excluded residents who did not have two or more MDS assessments during their Part-A covered SNF stay. Residents who did not have at least two MDS assessments were in poorer health and were more likely to be deceased within one-year of SNF discharge (HR=1.51, 95% CI=1.45-1.56) than residents with two or more assessments (Supplemental Table S2) and when compared to residents in the final sample (HR=1.57, 95% CI=1.51-1.62). We excluded residents who had a length of stay of less than 10-days to account for residents who did not have a SNF stay that was long enough to improve in ADL function. Patients whose last MDS assessment did not have complete information for seven ADLs were also excluded. Finally, we excluded 139 residents who were recorded as independent on all seven ADL items on their first MDS assessment because these residents could not improve in ADL function. The final sample included 59,383 residents.

Resident Characteristics

We controlled for resident characteristics that are associated with ADL recovery and mortality.^{19, 20} The Master Beneficiary Summary File was used for information on residents’ age, sex, race/ethnicity, and Medicaid eligibility. The MedPAR file was used to determine the number of days in the ICU, hospital length of stay, the number of hospitalizations in the past year, if the resident had an ICD-9-CM code for delirium in their hospital claim,²¹ SNF length of stay, and SNF discharge status according to the SNF claim. ICU length of stay included days spent in a coronary care unit. SNF length of stay was grouped into quartiles. Discharge status was categorized as home (with or without home health), inpatient hospital, still a resident in the facility, and other.

The first MDS assessment was used for information on marital status, cognitive status, total hours of PT and OT received in the past seven days, active diagnoses from section I of the MDS for heart failure, pneumonia, chronic obstructive pulmonary disease (COPD) or asthma, depression, diabetes, and urinary tract infection (UTI), and severe ADL limitations. Marital status was dichotomized (not married, married). Being married has been associated with lower mortality risk²² and lower likelihood to transition from a SNF to long-term care.²³ Cognitive status was categorized as cognitively unimpaired, mild impairment, moderate impairment, or severe impairment.²⁴ The total hours of PT and OT were calculated by summing the minutes of individual, concurrent, and group therapy. The total hours were categorized into quartiles. Finally, residents were classified as having severe functional limitations if their ADL total score was 23 points or higher.²⁰

Outcomes

The two outcomes were change in ADL function between the first and last MDS assessments during the Part-A covered SNF stay and one-year mortality after SNF discharge.

Improvement in ADL function

We used the following ADL measures from the first and last MDS assessments of the SNF stay: (1) dressing; (2) personal hygiene; (3) toileting; (4) locomotion on the unit; (5) transferring; (6) bed mobility; and (7) eating. Residents’ self-performance on each ADL is rated as independent (0-points), needing supervision (1 point), limited assistance (2 points), extensive assistance (3 points), and totally dependent (4 points). The MDS also allows staff to indicate if an ADL activity occurred once or twice or if the resident never did the ADL activity during the previous seven days. In these instances, we coded residents as totally dependent for that ADL. We calculated a total score for ADL function that ranged from 0 to 28 points with a higher score indicating poorer function.²⁵

We created three variables for improvement in ADL function. First, a dichotomous variable for any improvement in ADL function was created according to if the ADL total score from the first MDS was higher than the last MDS assessment. This variable is consistent with the quality measure for the percentage of short-stay nursing home residents who made any improvement in function between admission and discharge that was added to Nursing Home Compare in July, 2016.²⁶ Among residents who had any improvement in ADL function, the median improvement was 3 points. Thus, we created a second variable that categorized residents as having no improvement (stayed the same or decreased), improved 1-3 points, or improved four or more points in ADL function. This variable was created to determine if there was a dose-response relationship between ADL improvement and mortality risk, as well as to identify residents who had clinically meaningful improvement. A four point improvement is the difference between needing extensive assistance on all seven ADL tasks and needing limited assistance on the majority of ADLs. Finally, we calculated the improvement in ADL function per 10-days to account for differences in the number of days between the first and last MDS assessments and for easier interpretability than the change per day.

Mortality

Mortality was defined as being deceased within one-year of the discharge date in the SNF claim. Information for date of death came from the Master Beneficiary Summary File.

Statistical Analysis

Multivariable logistic regression was used to estimate the average marginal effects of the resident characteristics on any improvement in ADL function. The analysis controlled for demographic characteristics, ICU and hospital lengths of stay, the number of hospitalizations in the last year, and characteristics from the first MDS assessment. The model also included a facility fixed effect.

Cox proportional hazard regression models were used to examine the relationship between the three variables for improvement in ADL function and one-year mortality after SNF discharge. The models for any improvement in ADL function and improving 1-3 points or four or more points controlled for the same covariates as the logistic regression model, as well as SNF length of stay and SNF discharge status. Length of SNF stay is used in the calculation of the improvement per 10 days and was thus not included as a covariate in this proportional hazards model. No variables in the models violated the proportional-hazards assumption based on visual inspection of the scaled Schoenfeld residuals according to survival time.

Results

Resident characteristics

The characteristics of the 59,383 residents in the final sample according to improvement in ADL function are shown in Supplement Table S3. The average age of the sample was 81.1 years and residents were predominately female (58.2%), White (80.8%), and not married (64.5%). The average ADL total score at admission was 18.3 points, with 11.2% of residents being classified as having severe ADL impairment. A total of 34,262 residents (57.7%) had improvement in ADL function. Among residents who improved in ADL function, the average total ADL score at the first MDS assessment was 18.0 points (SD=4.11) and 13.4 points (SD=5.74) at the last MDS assessment.

Patient characteristics associated with improvement in ADL function

The multivariable logistic regression model identified several resident characteristics that were significantly associated with improvement in ADL function (Figure 2). The probability of improvement in ADL function decreased with more hospitalizations in the prior year, older age, and more severe cognitive impairment at SNF admission. Conversely, more hours of PT and OT during the first seven days were associated with significantly higher probability for improvement in ADL function. Severe ADL limitations at the first MDS assessment were also associated with significantly higher probability to have improvement in ADL function during the SNF stay.

Figure 2:

The average marginal effects for the association between skilled nursing facility (SNF) resident characteristics at admission and any improvement in activities of daily living during a SNF stay.

Association between improvement in ADL function and mortality

Residents who had improvement in ADL function had 0.72 (95% CI =0.69-0.74) lower risk for one-year mortality following SNF discharge than patients who did not improve (Table 1). Severe ADL limitations and more severe cognitive impairment were associated with significantly higher mortality risk whereas more hours of PT and OT received were associated with significantly lower mortality risk (Table 1).

Table 1:

Association between improvement in ADL function and one-year mortality after discharge from a skilled nursing facility, adjusting for patient characteristics.

Characteristic	Hazard Ratio (95% CI)	p-value
Improved in ADL function (ref no)	0.72 (0.69, 0.74)	<0.001
Age (ref 66-70)
71-75	1.06 (1.00, 1.13)	0.04
76-80	1.17 (1.11, 1.24)	<0.001
≥81	1.49 (1.42, 1.57)	<0.001
Women (ref men)	0.80 (0.78, 0.83)	<0.001
Married (ref not married)	1.03 (1.00, 1.06)	0.04
Race / ethnicity (ref White)
Black	0.86 (0.82, 0.90)	<0.001
Hispanic	0.88 (0.83, 0.94)	<0.001
Other	0.84 (0.78, 0.91)	<0.001
Hospitalizations in last year (ref 0 )
1-2	1.25 (1.21, 1.29)	<0.001
3+	1.63 (1.55, 1.71)	<0.001
ICU length of stay (ref 1-2)
3-4 days	0.97 (0.93, 1.01)	012
≥5 days	0.97 (0.93, 1.00)	0.08
Hospital length of stay (ref 1-4)
5-6 days	1.11 (1.06, 1.15)	<0.001
7-10 days	1.17 (1.12, 1.22)	<0.001
≥11 days	1.18 (1.13, 1.24)	<0.001
Delirium in hospital (ref no)	0.97 (0.94, 1.00)	0.04
SNF LOS (ref 10-17)
18-26 days	1.02 (0.98, 1.06)	0.34
27-41 days	1.00 (0.96, 1.04)	0.84
≥42 days	0.92 (0.88, 0.95)	<0.001
Severe ADL limitations (ref no)	1.26 (1.21, 1.31)	<0.001
Cognitive status (ref cognitively unimpaired)
Mild impairment	1.21 (1.17, 1.25)	<0.001
Moderate impairment	1.39 (1.34, 1.45)	<0.001
Severe impairment	1.56 (1.46, 1.65)	<0.001
Heart failure (ref no)	1.21 (1.17, 1.24)	<0.001
Pneumonia (ref no)	1.06 (1.03, 1.09)	<0.001
COPD / asthma (ref no)	1.13 (1.09, 1.16)	<0.001
Depression (ref no)	0.91 (0.88, 0.94)	<0.001
Diabetes (ref no)	0.93 (0.90, 0.96)	<0.001
UTI (ref no)	0.94 (0.91, 0.97)	<0.001
Hrs of physical therapy (ref < 3.92 hrs)
Quartile 2 (≥3.92-4.73 hrs)	0.93 (0.89, 0.97)	<0.001
Quartile 3 (≥4.73-5.88 hrs)	0.89 (0.85, 0.93)	<0.001
Quartile 4 (≥5.88 hrs)	0.84 (0.80, 0.89)	<0.001
Hrs of occupational therapy (ref < 3.63 hrs)
Quartile 2 (≥3.63-4.46 hrs)	0.97 (0.93, 1.01)	0.14
Quartile 3 (≥4.46-5.50 hrs)	0.95 (0.91, 0.99)	0.01
Quartile 4 (≥5.50 hrs)	0.91 (0.86, 0.95)	<0.001
Discharge disposition (ref home)
Hospital	2.78 (2.68, 2.87)	<0.001
Still patient	1.41 (1.33, 1.49)	<0.001
Other	1.39 (1.33, 1.45)	<0.001

ADL (activities of daily living); ICU (intensive care unit); SNF (skilled nursing facility); UTI (urinary tract infection); MDS (Minimum Data Set)

Severe ADL limitations defined as a total ADL score of 23 points or higher.

Mortality risk also decreased with greater improvement in ADL function. Residents who improved 1-3 points in ADL function had nearly 20% lower risk (hazard ratio [HR]=0.82, 95% CI=0.79-0.84) and residents who improved four or more points had 43% lower risk (HR=0.57, 95% CI=0.55-0.60) for one-year mortality after SNF discharge compared to residents with no improvement. Finally, a one-point improvement per 10-days in ADL function was associated with 0.88 (95% CI=0.86-0.88) times lower risk for mortality.

Discussion

ADL limitations are an important risk factor for mortality for community dwelling older adults^{27, 28} and for older adults receiving SNF care.^{20, 29} Using a national sample of Medicare beneficiaries with sepsis who were treated in intensive care, we found that the majority of patients who survived a SNF stay improved in ADL function.

In our sample, the average total ADL score at the first MDS assessment was 18.3 points. This is equivalent to needing extensive assistance to complete each of the seven ADLs included in our analysis. Despite the low independence, 26.8% of all residents improved in ADL function by four or more points. The average ADL total score at the last MDS assessment for residents who had any improvement was 13.4 points. This is comparable to needing supervision or limited assistance when performing an ADL.

Our study gives new evidence that functional improvement during a SNF stay can be important to reducing mortality risk for older adults who have survived sepsis. Several factors may contribute to this finding. First, the recovery of ADL function may reflect the improvement in overall health during the SNF stay. Second, ADL disability among SNF residents has been associated with increased risk for poor outcomes, such as rehospitalizations.³⁰ Improvement in ADL function may decrease the risk for these and other outcomes that are associated with mortality.³¹

Nearly half of the residents in our sample were classified as having mild, moderate, or severe cognitive impairment at SNF admission. The probability for improvement in ADL function decreased substantially with more severe cognitive impairment whereas the risk for mortality increased. Sepsis can severely impair brain function³² and people with sepsis often show rapid cognitive decline while in the hospital.³³ Survivors can regain some cognitive functioning,³⁴ but many experience prolonged cognitive impairment.³⁵ Cognitive impairment has been found in other SNF populations to negatively impact physical recovery^{19, 36} and to be associated with a higher risk for long-term ADL disability³⁷ and mortality.³⁸

The high prevalence of cognitive impairment has important implications given the growing number of older adults who are discharged to post-acute care after a hospitalization for sepsis.^{12, 13} Existing guidelines for treating older adults with sepsis focus on reducing mortality and information on ways to prevent or minimize cognitive impairment is limited.³⁹ There is some evidence from small randomized trials that at-home interventions that include a combination of cognitive and physical rehabilitation therapies can improve cognition after an ICU stay.^{40, 41} Continued research is needed to determine if such interventions are feasible and effective in SNFs. Existing rehabilitation strategies commonly used in SNFs may also need to be modified according to a person’s level of cognitive impairment.⁴²

Greater hours of PT and OT during the first seven days of the SNF stay were associated with significantly higher probability for improvement in ADL function as well as lower mortality risk. These associations may be due in part to residents with better ADL function at SNF admission being able engage in more therapy. Early rehabilitation and mobilization are recommended strategies to mitigate muscle atrophy and weakness in the ICU.⁴³ Hospitalized patients who are able to receive PT and OT while in the ICU can improve in ADL function.⁴⁴ Early rehabilitation therapy while in the ICU may enable older adults to participate in a greater number of hours of PT and OT earlier in their SNF stay. Greater therapy intensity is associated with shorter length of stay, more improvement in mobility and ADL function, and greater likelihood for community discharge.^{45, 46} Rehabilitation therapy after ICU discharge has also been associated with lower mortality for sepsis survivors.⁴⁷

Our findings have potential implications that are relevant to the new reimbursement model for SNFs. Since 1998, SNFs were reimbursed according to the number of therapy minutes a resident received over a seven day period. This reimbursement model was replaced in October 2019 with the Patient Driven Payment Model, which bases reimbursement on the medical complexity of residents and includes financial incentives for shorter SNF stays and less therapy.⁴⁸ A potential unintended consequence is that some SNFs may be more willing to admit highly complex residents but not provide sufficient therapy for these residents to improve in function.^{48, 49} Continued research will be needed to determine if the Patient Driven Payment Model leads to changes in the medical characteristics of residents admitted to SNFs, the amount of therapy that residents receive, and outcomes of SNF care for highly complex residents.

Our study has limitations. First, determining the change in ADL function required our analysis to only include residents who had at least two MDS assessments. Patients who were excluded spent more days in in the hospital and ICU, and were in generally worse health than patients who had two or more MDS assessments. Also, investigating survival after SNF discharge required us to restrict our analysis to older adults who survived a SNF stay. Consequently, our cohort represents a healthier subset of residents who were discharged to a SNF following an ICU stay with sepsis. Second, our analysis only assessed ADL recovery during a Part-A rehabilitative SNF stay. ADL limitations among older adults who have survived a hospitalization for sepsis can persist for several years.⁵⁰ Future research should investigate the changes in ADL function that occur after SNF discharge. This can include older adults who are discharged to home with home health and those who remain in a nursing home as long-term care residents. The introduction of uniform resident assessments as part of the Improving Medicare Post-Acute Care Transformation Act will allow for research on changes in ADL function as older adults transition across different post-acute care settings. Third, our analysis was limited to Medicare fee-for-service beneficiaries. This is important to consider because Medicare Advantage beneficiaries have been shown to have shorter lengths of SNF stays and receive fewer total hours of rehabilitation therapy but are more likely to be discharged to the community when compared to fee-for-service beneficiaries. Finally, we did not have information on patients’ ADL function before their ICU stay or changes in ADL function that may have occurred while in the ICU. Patients’ ADL function and other health characteristics before and during an ICU stay for sepsis may contribute to the likelihood for ADL recovery during a SNF stay.

Conclusions and Implications

In summary, the majority of Medicare beneficiaries who were treated in intensive care for sepsis and who survive a SNF stay improve in ADL function. Improvement in ADL function during the SNF stay was associated with significantly lower mortality risk after SNF discharge. Important factors associated with improvement in ADL function and mortality risk included cognitive status at SNF admission and the total hours of PT and OT received during the first seven days of a SNF stay. These findings are evidence that older adults who have been discharged from the ICU for sepsis can have meaningful improvement in ADL function during a SNF stay and that this improvement is associated with better survival after SNF discharge. These findings can also assist with making prognoses and goal setting for rehabilitation outcomes during a SNF stay as well as inform discharge planning from a SNF.

Impact statement: We certify that this work makes a novel contribution to existing research on the rehabilitation and outcomes of older adults who have survived an intensive care unit stay for sepsis. Prior research has focused on short-term mortality for older adults with sepsis. Using national Medicare claims files, we focus our analysis on older adults who have been discharged to a skilled nursing facility (SNF) in order to investigate what proportion of older adults improve in activities of daily living (ADL) and if this improvement is associated with reduced mortality risk after SNF discharge. We found that 57.7% of older adults treated in intensive care for sepsis and who survived a skilled nursing facility stay improved in ADLs. Greater improvement in ADL function during a SNF stay was associated with significantly lower one-year mortality.