Original Investigation Association Between Kidney Function, Rehabilitation Outcome, and Survival in Older Patients Discharged From Inpatient Rehabilitation Edward M. Doyle,1 Joanne M. Sloan, BM, BCh,2 James A. Goodbrand, MSc,1 Marion E.T. McMurdo, MD,1 Peter T. Donnan, PhD,3 Mark M. McGilchrist, PhD,3 Helen Frost, PhD,4 and Miles D. Witham, PhD1 Background: Chronic kidney disease (CKD) is common in older people, but it is unclear if it affects survival and rehabilitation outcomes independent of comorbid conditions and physical function in this population. Study Design: Cohort analysis of prospective, routinely collected, linked clinical data sets. Setting & Participants: Patients discharged from a single inpatient geriatric rehabilitation center over a 12-year period. Predictors: Admission estimated glomerular filtration rate (eGFR) category as a predictor of improvement in the 20-point Barthel score (activities of daily living measure) during rehabilitation; discharge eGFR category and Barthel score as predictors of survival postdischarge. Outcomes: Survival postdischarge was modeled using Cox regression analyses, unadjusted and adjusted for age, sex, morbidities (ischemic heart disease, chronic obstructive pulmonary disease, stroke, diabetes, and heart failure), Barthel score and eGFR category on discharge, and serum calcium, hemoglobin, and albumin levels. The effect of admission eGFR category on change in Barthel score during admission was modeled using analysis of covariance, adjusted for admission, Barthel score, and comorbid conditions. Results: 3,012 patients were included; mean age, 84 years. 2,394 patients died during a mean follow-up of 8.3 years. Compared with patients with eGFR of 60 to 89 mL/min/1.73 m2, adjusted HRs for death were 1.26 (95% CI, 1.13-1.40), 1.45 (95% CI, 1.29-1.63), and 1.68 (95% CI, 1.42-1.99) for eGFR categories of 45 to 59, 30 to 44, and ,30 mL/min/1.73 m2, respectively. The relationship between discharge Barthel score and survival was similar within each discharge eGFR category (HRs of 0.95, 0.93, 0.92, 0.95, and 0.90 per Barthel score point within eGFR categories of $90, 60-89, 45-59, 30-44, and ,30 mL/min/1.73 m2; P for interaction 5 0.2). Similar improvements in Barthel score between admission and discharge were seen for each admission eGFR category. Limitations: Single-center study using routinely collected clinical data. Conclusions: eGFR category and Barthel score are independent risk markers for survival in older rehabilitation patients, but advanced CKD does not preclude successful rehabilitation. Am J Kidney Dis. -(-):---. ª 2015 by the National Kidney Foundation, Inc. INDEX WORDS: Chronic kidney disease (CKD); reduced renal function; older adults; oldest old; geriatric; Barthel score; activities of daily living (ADLs); mortality; physical function; functional impairment; frailty; rehabilitation.

C

hronic kidney disease (CKD) increases in prevalence with advancing age. Between 31% and 45% of people older than 70 years have CKD stage 3a (estimated glomerular filtration rate [eGFR] of 45-59 mL/min/1.73 m2) or worse.1,2 This decreased kidney function is associated with increased all-cause and cardiovascular mortality even in people older than 75 years.3 Although impaired physical function appears to be a risk marker for mortality in younger patients with CKD,4 it is less clear in older patients whether decreased kidney function and impaired physical function are independent risk markers for mortality or whether low GFR merely reflects frailty or impaired physical function, themselves both powerful markers for mortality in older people.5 Clarifying this issue is important, first in deciding whether labeling very old patients as having CKD serves a useful purpose, and second in considering whether intervening to ameliorate declines in kidney Am J Kidney Dis. 2015;-(-):---

function in the oldest old is likely to translate into clinically relevant benefits. This is of particular importance given that chronic kidney failure is a comparatively less common mode of death in older patients with CKD.3

From 1Medical Research Institute, University of Dundee; Department of Renal Medicine, NHS Tayside; 3Population Health Sciences, University of Dundee, Dundee, United Kingdom; and 4Scottish Collaboration for Public Health Research and Policy, University of Edinburgh, Edinburgh, Scotland. Received November 13, 2014. Accepted in revised form April 21, 2015. Address correspondence to Miles D. Witham, PhD, Ageing and Health, Medical Research Institute, Mailbox 1, Ninewells Hospital, Dundee DD1 9SY, United Kingdom. E-mail: m.witham@dundee. ac.uk  2015 by the National Kidney Foundation, Inc. 0272-6386 http://dx.doi.org/10.1053/j.ajkd.2015.04.041 2

1

Doyle et al

Poor physical function and decreased exercise capacity have been found to exist in patients with CKD, and recent studies suggest that patients with CKD are more likely both to be frail4 and to undergo more rapid decline in physical function.6 This close relationship between CKD and physical function suggests that incorporating measures of physical function into analyses of outcome for older patients with CKD thus is important. It also raises the question of how the presence of CKD affects the ability of older people to rehabilitate following illness. We therefore undertook an analysis to examine the effect of CKD on mortality in very old patients, accounting for the effect of physical function. We also aimed to examine the effect of CKD stage on the ability of older people to successfully rehabilitate after illness.

METHODS Study Population The study cohort was derived from a group of older patients who underwent inpatient rehabilitation over a 12-year period (January 1999 through December 2011) in the Dundee Medicine for the Elderly service; details of the service and a subset of the cohort have been published previously.7 Patients were transferred to the rehabilitation unit after recovery from acute illness including stroke, fractures, and general medical or general surgical illness. Data for routinely collected rehabilitation outcomes for the cohort include the 20-point Barthel score (a measure of activities of daily living [ADLs])8 at admission and discharge. These data were linked to data held by the University of Dundee Health Informatics Centre (HIC) on a range of other routinely collected health care measures.9 Information for biochemistry and hematology results, hospitalization data, and diagnoses (Scottish Morbidity Record [SMR] 01) coded using International Classification of Diseases, Tenth Revision (ICD-10) codes were accessible in the linked data set. Dates of death, derived from the Scottish Government Records Office, which records all deaths registered within Scotland, were also held by HIC in the linked data set. For this analysis, data from the first admission to the rehabilitation service were used. Analyses of improvement in rehabilitation were confined to those who had an admission and discharge Barthel score recorded; survival analyses were confined to those alive at discharge from rehabilitation who had a discharge Barthel score recorded. A flow chart depicting the derivation of the analysis cohort is given in Fig 1. Data Protection Office (Caldicott Guardian) approval was obtained prior to data linkage and analysis, but the need for institutional review board approval was waived in view of the routinely collected nature of the data.

Exposure and Outcome Measures and Covariates Time to death was calculated as the time between the date of discharge from rehabilitation facility to the date of death as recorded in Scottish General Records Office death records. The Barthel score was recorded as part of routine care by the multidisciplinary rehabilitation team. As a measure of basic ADLs, the Barthel score records what a person is capable of doing across 10 activities: walking, transfers, feeding, bathing, grooming, dressing, bowel function, bladder function, toilet use, and stair use8 (score range, 0-20; 20 5 highest level of independence). It is widely used as a measure of basic ADLs in older people, and previous work in this cohort has shown that a 1-point difference between admission 2

and discharge on the Barthel score relates to a 5% to 10% difference in mortality.7 All biochemical indexes were taken from routinely collected clinical data, held by HIC. All analyses were performed in a single laboratory (Biochemical Medicine, NHS Tayside). The measurement taken closest to the date of admission to rehabilitation was used in analyses of the effect of eGFR category on change in Barthel score during rehabilitation. For postdischarge survival analyses, the measurement taken closest to the date of discharge from rehabilitation was used. Creatinine was measured using a compensated rateblanked method on a Roche multichannel analyzer; the methodology was traceable to isotope-dilution mass spectrometry. In February 2007, a new algorithm for creatinine calculation was introduced locally to align results with UK National External Quality Assessment Service (UK NEQAS) results. Results in our database prior to this point were corrected in our analyses using the equation: Creatinine 5 (old creatinine 3 1.1358) 2 26. eGFR was derived from creatinine measures using the CKD-EPI (CKD Epidemiology Collaboration) creatinine equation.10 Kidney function was then categorized on the basis of KDIGO (Kidney Disease: Improving Global Outcomes) GFR categories in CKD11 ($90, 60-89, 45-59, 3044, 15-29, and ,15 mL/min/1.73 m2). Data for proteinuria were not available and hence were not included in analyses. Only a small proportion of patients had phosphate or bicarbonate measurements performed because these analytes were not part of standard biochemistry panels; these analytes therefore are not included in the analyses. Specific morbidities (ischemic heart disease, stroke, heart failure, or chronic obstructive pulmonary disease) were coded from ICD-10 codes on the basis of previous hospitalizations for myocardial infarction, stroke, and heart failure. Diabetes status (present or absent) was taken from linked data from the Scottish Care Information– Diabetes Collaboration database,12 also held by HIC. Cancer diagnosis in the 5 years prior to rehabilitation admission was taken from the Scottish Cancer Registry (SMR06) records, held by HIC.

Data Analysis Descriptive statistics were generated for each category of eGFR; values in each category were compared to the reference group (eGFR $ 90 mL/min/1.73 m2) using analysis of variance (ANOVA) for continuous variables and either Pearson c2 or Fisher exact test for categorical variables. Two main analyses were undertaken. The first compared kidney function category with the improvement in Barthel score from admission to discharge. The group with eGFRs of 60 to 89 mL/min/1.73 m2 was used as reference, and each category of kidney function was compared Number of rehabilita on admissions 1/1/1999 to 31/12/2011 (n=5650)

First admissions only (n=4382)

Alive at discharge from rehabilita on (n=4009)

Discharge Barthel score and all covariates available (n=3012)

Admission and discharge Barthel score, plus all covariates available (n=2868)

Survival analyses (n=3012)

Analysis of rehabilita on efficacy (n=2868)

Figure 1. Flow chart shows selection of cohort for analysis. Am J Kidney Dis. 2015;-(-):---

Kidney Function, Rehabilitation, and Survival with the index group using ANOVA, both unadjusted and adjusted for a range of covariates. The second analysis used Cox regression analysis to compare the risk of death between different categories of kidney function. Both unadjusted models and models adjusted for a range of covariates were developed; adjusted models all used forced entry of covariates, which were selected a priori on the basis of biological plausibility. Covariates used in the analysis were age; sex; previous hospital admission for myocardial infarction, stroke, heart failure, or chronic obstructive pulmonary disease; presence of diabetes mellitus; admission Barthel score; previous cancer diagnosis; and serum calcium, serum albumin, and hemoglobin levels. Follow-up was censored at July 23, 2012. To formally test for interaction between eGFR and discharge Barthel score on survival, both terms were entered into the adjusted Cox regression model as continuous variables, along with an interaction term. All statistical methods were performed with SPSS, version 22 (IBM). A 2-sided P , 0.05 was considered significant for all analyses.

the eGFR category of 15 to 29 mL/min/1.73 m2. At discharge from rehabilitation, 3,012 patients were alive and had data for all covariates, including discharge Barthel score; mean follow-up from date of discharge to the censor date for this group was 8.3 6 3.4 (standard deviation) years, with 2,394 (79.5%) patients dying during this follow-up period. Table 1 gives baseline data for this analyzed study cohort. No patient in the study was undergoing renal replacement therapy during rehabilitation. During rehabilitation, eGFR category did not change in 1,966 (65.3%), worsened in 406 (13.5%), and improved in 640 (21.2%) of 3,012 patients.

RESULTS

Table 2 gives unadjusted and adjusted results of improvement in Barthel score during rehabilitation for each category of kidney function at admission, using the 2,868 individuals with all covariates including admission and discharge Barthel scores. (Baseline characteristics by level of eGFR at admission are provided in Table S1, available as online supplementary material.) Patients with admission eGFRs $ 90 mL/ min/1.73 m2 had slightly lower admission Barthel scores than the reference group and improved slightly more during rehabilitation.

Association of Kidney Function With Rehabilitation Outcome

Study Participants The full cohort consisted of 4,382 patients undergoing their first admission for inpatient rehabilitation. Of these, 373 died during their inpatient stay and were excluded from the analyses. No participants were of African or African American ethnicity. Death rates during rehabilitation varied across eGFR category, being similar in groups with eGFRs $90, 60 to 89, 45 to 59, and 30 to 44 mL/min/1.73 m2 (death rates of 9.2%, 7.6%, 7.9%, and 8.7%, respectively), but higher for groups with eGFRs of 15 to 29 and ,15 mL/min/ 1.73 m2 (15.0% and 37.5%, respectively). Given the small numbers of patients with eGFRs , 15 mL/min/ 1.73 m2, these patients were aggregated with those in

Association of Kidney Function and Functional Status With Survival There were 3,012 individuals included in this analysis; data for one or more covariates were missing

Table 1. Baseline Details Stratified by Discharge eGFR Category

No. of patients Age (y) Male sex Barthel score Admissionb Discharge Previous admission for: Myocardial infarction Stroke CHF COPD Diabetes mellitus Serum albumin (g/L) Adjusted serum calcium (mmol/L) Hemoglobin (g/dL)

$90

60-89

45-59

30-44

,30

260 (9) 75.9 6 6.5a 164 (63)a

1,432 (48) 83.5 6 7.0 622 (43)

652 (22) 85.6 6 7.0a 231 (35)a

481 (16) 87.1 6 6.6a 167 (35)a

187 (6) 86.7 6 7.5a 58 (31)a

9.4 6 4.4a 13.4 6 6.0a

10.3 6 3.8 14.3 6 4.7

10.7 6 3.7c 14.6 6 4.6

11.0 6 3.6a 14.9 6 4.0c

11.0 6 3.5c 14.1 6 4.7

129 (20)a 51 (8) 77 (12)a 94 (14)

102 (21)a 29 (6) 71 (15)a 56 (12)

37 (20)a 14 (7) 36 (19)a 14 (7)c

116 (18) 37 6 4 2.41 6 0.14 11.9 6 1.8a

87 (18) 37 6 5 2.40 6 0.12 11.6 6 1.7a

46 (25)a 36 6 5c 2.41 6 0.13 11.2 6 1.8a

35 (13) 21 (8) 8 (3) 58 (22)a 42 (16) 35 6 6a 2.40 6 0.12 12.3 6 2.1

180 86 56 193

(13) (6) (4) (13)

231 (16) 37 6 5 2.40 6 0.11 12.3 6 1.9

Note: N 5 3,012. eGFRs expressed in mL/min/1.73 m2. Values for categorical variables are given as number (percentage); for continuous variables, as mean 6 standard deviation. Percentages for total numbers do not add to 100% due to rounding. Abbreviations: CHF, chronic heart failure; COPD, chronic obstructive pulmonary disease; GFR, estimated glomerular filtration rate. a P , 0.01 compared to eGFR of 60 to 89 mL/min/1.73 m2 category. b n 5 2,868. c P , 0.05 compared to eGFR of 60 to 89 mL/min/1.73 m2 category. Am J Kidney Dis. 2015;-(-):---

3

Doyle et al Table 2. Association Between eGFR at Admission and Rehabilitation Outcomes as Measured Using Barthel Score Increase in Barthel Score During Rehabilitation (95% CI) eGFR at Admission

No.

Admission Barthel Score

$90 mL/min/1.73 m2 60-89 mL/min/1.73 m2 45-59 mL/min/1.73 m2 30-44 mL/min/1.73 m2 ,30 mL/min/1.73 m2

256 1,272 583 512 245

9.5 6 4.4b 10.3 6 3.8 10.9 6 3.7b 10.8 6 3.7 10.7 6 3.8

Unadjusted

4.6 3.7 4.0 3.6 3.7

(4.2-5.1)b (3.5-3.9) (3.7-4.2) (3.3-3.9) (3.3-4.2)

Adjusteda

4.3 3.7 4.0 3.6 3.8

LOS (d)

(3.8-4.8)b (3.5-3.9) (3.8-4.3) (3.3-3.9) (3.4-4.2)

42 [55] 38 [47] 32 [42]b 35 [42] 33 [41]b

Note: n 5 2,868. Unless otherwise indicated, values given as mean 6 standard deviation or median [interquartile range]. Abbreviations: CI, confidence interval; eGFR, estimated glomerular filtration rate; LOS, length of stay. a Adjusted for age, sex, myocardial infarction, stroke, heart failure, chronic obstructive pulmonary disease, diabetes, admission Barthel score, previous cancer diagnosis, and serum calcium, serum albumin, and hemoglobin levels. b P , 0.05 versus eGFR of 60 to 89 mL/min/1.73 m2 category.

for the excluded individuals. There were 2,394 (79.5%) who had died by the censor date of July 23, 2012. Unadjusted and adjusted hazard ratios for survival, calculated using Cox regression analysis, are given in Table 3. Worsening eGFR was associated with greater risk of death, as expected; within each eGFR category, higher discharge Barthel scores were associated with lower risk of death. The relationship between discharge Barthel score and death was similar across all eGFR categories. Interaction Between Barthel Score and eGFR on Survival Figure 2 shows the interaction between Barthel score and eGFR category, suggesting that Barthel score remained an independent predictor of outcome within each eGFR category. However, a formal test of interaction between eGFR and discharge Barthel score within the Cox regression analysis did not reach significance (P 5 0.2).

DISCUSSION Our results show that in very old functionally impaired people, both kidney function and ADLs (an indirect measure of physical and psychosocial function) are independently associated with risk of death,

with a clear relationship between both severity of CKD and risk of death and ADL impairment and risk of death. This relationship is evident even with relatively mild decreases in kidney function and is still evident after adjusting for functional status as measured by the Barthel score; the relationship between kidney function and mortality is more pronounced in those with worse Barthel scores, suggesting that both factors contribute independently to risk of mortality. We also found that in the subgroup of older people with advanced CKD still derived benefit from rehabilitation, with similar improvements seen to those without advanced CKD. These findings support the current KDIGO guidance encouraging exercise for patients with CKD across all ages and severity categories.11 The relationship between CKD and mortality is well described, and this relationship extends to older patients.13,14 Debate continues as to whether CKD, especially less severe CKD (eg, stage 3a), is a causal factor in older people or simply represents either normal aging or a marker of disease burden, frailty, and impaired homeostatic reserve.15,16 A weakness of previous studies is that few have been able to adjust for measures of functional status in mortality analysis; the fact that CKD remained as a risk marker for mortality in our study even after adjustment for

Table 3. Survival Postdischarge From the Rehabilitation Center, by eGFR Category at Discharge

eGFR on Discharge

No.

$89 mL/min/1.73 m2 60-89 mL/min/1.73 m2 45-59 mL/min/1.73 m2 30-44 mL/min/1.73 m2 ,30 mL/min/1.73 m2

260 1,432 652 481 187

Unadjusted HR (95% CI)

1.14 1.00 1.15 1.35 1.56

(0.98-1.33) (reference) (1.04-1.28) (1.21-1.51) (1.33-1.84)

Adjusted HRa (95% CI)

0.86 1.00 1.26 1.45 1.68

(0.74-1.01) (reference) (1.13-1.40) (1.29-1.63) (1.42-1.99)

HR per Unit Increase in Discharge Barthel Score (95% CI)a

0.95 0.93 0.92 0.95 0.90

(0.92-0.97) (0.92-0.94) (0.90-0.94) (0.93-0.97) (0.86-0.93)

Note: N 5 3,012. Abbreviations: CI, confidence interval; eGFR, estimated glomerular filtration rate; HR, hazard ratio. a Adjusted for age, sex, myocardial infarction, stroke, heart failure, chronic obstructive pulmonary disease, diabetes mellitus, previous cancer diagnosis, discharge Barthel score, and serum calcium, serum albumin, and hemoglobin levels. 4

Am J Kidney Dis. 2015;-(-):---

Kidney Function, Rehabilitation, and Survival 7 6.5 eGFR (mL/min/1.73 m2)

6

Hazard RaƟo for Death

5.5 5 4.5

>90 ≥90

60-90 60-89

45-59

30-44