Abstract
Background
Acute kidney injury (AKI) diagnosis requires ascertainment of change from a known baseline. Although pre-admission serum creatinine (SCr) is recommended, to date, all studies of AKI in acute stroke have used the first SCr on admission.
Methods
All patients admitted with an acute stroke to an emergency hospital were recruited. We compared use of pre-admission SCr with admission SCr to diagnose AKI. Regression analyses were used to identify risk factors for 30-day and 1-year mortality, respectively.
Results
A total of 1354 patients were recruited from December 2012 to September 2015. Incidence of AKI was 18.7 and 19.9% using pre-admission SCr and admission SCr, respectively. Diagnosis of AKI was associated with significantly increased 30-day and 1-year mortality. Diagnosis of AKI using pre-admission SCr had a stronger relationship with both 30-day and 1-year mortality. In 443 patients with a pre-admission SCr and at least two SCr during admission, AKI diagnosed using pre-admission SCr had a stronger relationship than AKI diagnosed using admission SCr with 30-day mortality [odds ratio (OR) = 2.64; 95% confidence interval (CI) 1.36–5.12; P = 0.004 versus OR = 2.10; 95% CI 1.09–4.03; P = 0.026] and 1-year mortality [hazard ratio (HR) = 1.90, 95% CI 1.32–2.76; P = 0.001 versus HR = 1.47; 95% CI 1.01–2.15; P = 0.046] in fully adjusted models.
Conclusions
AKI after stroke is common and is associated with increased 30-day and 1-year mortality. Using first SCr on admission gives a comparable AKI incidence to pre-admission SCr, but underestimates 30-day and 1-year mortality risk.
Keywords: acute kidney injury, mortality, stroke
INTRODUCTION
There is wide variation in the reported incidence of acute kidney injury (AKI) depending on definitions and populations studied: from 5.4% to 18.3% [1, 2] for hospitalized patients, to 40% in patients requiring intensive care [3, 4]. Having an episode of AKI is associated with increased mortality, inpatient length of stay and healthcare costs [5], and confers a higher risk of developing chronic kidney disease (CKD) [6] with its own sequelae.
Currently, AKI diagnosis relies on the ascertainment of a peak increase in SCr from a known ‘baseline’ value [7]. The Acute Kidney Injury Network (AKIN) [8, 9] and European Renal Best Practice (ERBP) guideline [10] recommend the use of first SCr on admission. Kidney Disease: Improving Global Outcomes (KDIGO) recommends that AKI diagnosis requires knowledge of change in SCr from a known ‘baseline’ [7]. A developing consensus favours the use of an average SCr 7–365 days up to admission [8, 9, 11–15].
Stroke is the second most common cause of death and the leading cause of neurological disability worldwide [16]. Patients suffering a stroke are typically older, have significant co-morbidities and often have associated CKD [17, 18]. All of these features are recognized to be associated with AKI [3]. However, a recent systematic review and meta-analysis highlighted the fact that risk factors for AKI after a stroke have not been extensively investigated, and most studies have confined themselves to known generic risk factors for AKI [19]. Only two studies examined the association between stroke severity and AKI and only one study investigating the relationship between radiological contrast exposure and development of AKI [19]. Furthermore, no study has examined the association between thrombolysis, angiographic procedures or vascular interventions after an ischaemic stroke [19]. With an ageing population and increasing prevalence of CKD, together with an increasing use of interventional procedures, it is important to establish risk factors for, and the true incidence of, AKI after a stroke in order to design studies to potentially improve outcomes.
To date, all studies investigating AKI in acute stroke have used the first SCr result on admission [20–22]. No study has investigated the use of different methods to classify AKI and their relationship to outcomes in acute stroke. To address this, we sought to determine in a population of patients admitted with an acute stroke:
the incidence of AKI determined using pre-admission SCr compared with first SCr on admission;
the risk factors for AKI in stroke patients; and
the impact of AKI on mortality at 30 days and 1 year using different methods for AKI diagnosis.
MATERIALS AND METHODS
Study design and population
This was a prospective cohort study of patients presenting with an acute stroke (acute ischaemic stroke or intracranial haemorrhage) between December 2012 and September 2015 to an acute hospital in UK. Readmissions and patients with end-stage renal disease (ESRD) were excluded. A stroke physician assessed each case by history, neurological examination and brain imaging (computerized tomography and/or magnetic resonance imaging). We followed the STrengthening the Reporting of OBservational studies in Epidemiology guidelines [23].
Data collection and follow-up period
We used data from the Sentinel Stroke National Audit Programme [24]. Demographic data including age, sex, ethnicity, Index of Multiple Deprivation score, comorbidities [hypertension, diabetes mellitus, congestive heart failure, previous stroke or transient ischaemic attack (TIA) and atrial fibrillation (AF)] and stroke type were extracted from the database. Degree of disability was determined using the modified Rankin scale [25] and stroke severity using the National Institutes of Health Stroke Scale score [26]. Mortality data up to 1 year were collected from the Office of National Statistics [27]. Pre-admission SCr was calculated as the mean of all pre-admission SCr values 7–365 days before admission [15].
Impaired renal function was defined as an estimated glomerular filtration rate (eGFR) of <60 mL/min/1.73 m2 using the Chronic Kidney Disease Epidemiology Collaboration and subdivided into GFR categories as per KDIGO CKD Guidelines [28]. For patients with no pre-admission SCr, impaired renal function was defined as an eGFR <60 mL/min/1.73 m2 using the first SCr value on admission. Anaemia was defined as a haemoglobin level <135 g/L for males and <115 g/L for females as per laboratory reference ranges.
Patients were divided into three groups depending on the availability of SCr data (Figure 1): Group A: all patients with at least one pre-admission SCr available in the 7–365 days preceding admission; Group B: all patients with at least two SCr available after admission; and Group C: all patients that that fulfilled criteria for Groups A and B. Patients could therefore be assigned to more than one group.
FIGURE 1.
Study flow diagram. SSNAP, Sentinel Stroke National Audit Programme.
Ethical approval was granted for this study (East of England–Essex Research Ethics Committee 16/EE/0166).
Definition of AKI
The rate of AKI using pre-admission SCr (termed AKIpre) was compared with the surrogate, first SCr on hospital admission (termed AKIadm). AKI was defined as per KDIGO guidelines [7]. Urine output criteria were not used since electronic records of urine output were incomplete.
Statistical analysis
Analyses were performed using SPSS 24.0 (SPSS Inc., Chicago, IL, USA). Continuous variables are expressed as mean ±standard deviation (SD) for normally distributed variables or median and interquartile range for non-normally distributed variables and compared using the t-test or Mann–Whitney U test. Categorical variables are expressed as proportions and compared using the Chi-squared test or McNemar test for dichotomous variables [29]. All variables used in the analysis had <5% of values missing and were therefore treated as missing completely at random with case-wise deletion.
AKIpre and AKIadm methods were compared using the Bland–Altman method [30]. Sensitivity and specificity were reported for each method with the kappa value and 95% confidence intervals (CIs) used to denote the level of agreement between different methods. We calculated misclassification rates as the proportion of patients incorrectly assigned as having AKI as compared with AKIpre and compared correctly classified and misclassified AKI using the McNemar test.
Kaplan–Meier survival curves were drawn to assess group differences for time-to-event data and compared using the Log rank test. Logistic regression was used to assess the relationship between outcomes and parameters under investigation, expressed as an odds ratio (OR) with 95% CI. Time-to-event analysis for cumulative 1-year mortality was performed using the Cox proportional hazards model and results expressed as a hazard ratio (HR) with 95% CI. Variables found to be associated with the outcome under investigation in the univariable analysis were included in the multivariable models. A P-value threshold of <0.15 was selected in order to retain all potential risk factors and minimize the chance of type II errors [31].
RESULTS
In total, 1440 hospital admissions with acute stroke occurred within the study period (Figure 1). From these, 52 patients who were readmitted over the same period were excluded, as well as 13 duplicates and 21 patients with ESRD. The remaining 1354 patients were included for analysis. Among these patients, 725 (53.5%) fulfilled criteria for inclusion in Group A, 808 (59.7%) for Group B and 443 (32.7%) for Group C.
Characteristics of patients with and without pre-admission SCr
The baseline characteristics of the study population with and without pre-admission SCr data are shown in Table 1. In multivariable analysis, patients with pre-admission SCr values were more likely to be older, of non-Black ethnicity, have a concurrent diagnosis of diabetes mellitus, AF, previous stroke or TIA, and have higher disability on admission (Supplementary data, Appendix S1).
Table 1.
Baseline characteristics according to presence of pre-admission SCr
| Parameters | With pre-admission SCr (n = 725) | Without pre-admission SCr (n = 629) |
|---|---|---|
| Age, years, mean (SD) | 74.83 (14.09) | 69.04 (16.33) |
| Male (%) | 366 (50.5) | 362 (57.6) |
| Ischaemic stroke (%) | 643 (88.7) | 553 (87.9) |
| NIHSS score on admission, median (IQR) | 3.0 (8) | 4.0 (8) |
| NIHSS level of consciousness (0–3), median (IQR) | 0 (0) | 0 (0) |
| IMD score, mean (SD) | 30.51 (15.37) | 30.60 (17.82) |
| Ethnic group (%) | ||
| White | 625 (86.2) | 505 (80.3) |
| Asian/Asian British | 68 (9.4) | 61 (9.7) |
| Black/Black British | 18 (2.5) | 32 (5.1) |
| Mixed/other/unknown | 14 (1.9) | 31 (4.9) |
| Pre-admission SCr, mean (SD) (µmol/L) | 91.32 (31.40) | – |
| Pre-admission eGFR, mean (SD) (mL/min/1.73 m2) | 68.45 (22.72) | – |
| Admission SCr, mean (SD) (µmol/L) | 91.15 (41.55) | 84.14 (24.78) |
| Admission eGFR, mean (SD) (mL/min/1.73 m2) | 67.90 (22.71) | 75.84 (22.35) |
| Pre-admission eGFR <60 mL/min/1.73 m2 (%) | 278 (38.3) | – |
| Admission eGFR <60 mL/min/1.73 m2 (%) | 275 (37.9) | 159 (25.3) |
| Hypertension (%) | 378 (52.1) | 273 (43.4) |
| Diabetes mellitus (%) | 175 (24.1) | 105 (16.7) |
| CHF (%) | 35 (4.8) | 17 (2.7) |
| Previous stroke/TIA (%) | 218 (30.1) | 122 (19.4) |
| AF (%) | 179 (24.7) | 75 (11.9) |
| Thrombolysis (%) | 64 (8.8) | 105 (16.7) |
| Thrombectomy (%) | 14 (1.9) | 38 (6.0) |
| Any iodinated contrast exposure (%) | 130 (17.9) | 151 (24.0) |
Data are presented as mean±SD, median (IQR) or n (%).
CHF, congestive heart failure; IMD, Index of Multiple Deprivation; IQR, interquartile range; NIHSS, National Institutes of Health Stroke Scale.
AKI diagnosis
Rates of AKI using AKIpre and AKIadm were compared in patients in Group C. There was no difference in the rates of AKI using AKIpre or AKIadm (18.7% versus 19.9%; P = 0.63). The rates of AKI and agreement between methods are summarized in Supplementary data, Appendix S2. The overall misclassification rate was 17.4% (Kappa statistic 0.46 with 9.7% of cases overclassified and 7.7% of cases underclassified as having AKI) (Supplementary data, Appendix S3). A Bland–Altman plot (Supplementary data, Appendix S4) showed the majority of values fall within the limits of agreement.
Comparison of pre-admission SCr and admission SCr
In Group C, 25.1% had a first admission SCr that was ≥110% above the pre-admission SCr (Supplementary data, Appendix S5). A greater proportion of patients classified as having AKIpre experienced this pattern compared with patients classified as having AKIadm (48.2 and 20.5%, respectively). Conversely, 23.9% had a first admission SCr that was ≤90% of the pre-admission SCr with a greater proportion of patients with AKIadm compared with AKIpre experiencing this pattern (33.0 and 12.0%, respectively). Similar results were obtained in Group A (Supplementary data, Appendix S6).
Factors associated with AKI
In Group C, factors associated with AKIpre and AKIadm in multivariable analyses (Supplementary data, Appendix S7) were the presence of an eGFR <60 mL/min/1.73 m2 (OR = 2.78; 95% CI 1.69–4.57; P <0.001 and OR = 2.60; 95% CI 1.60–4.23; P < 0.001, respectively) and anaemia (OR = 1.78; 95% CI 1.06–2.97; P = 0.03 and OR = 1.96; 95% CI 1.19–3.24; P = 0.009, respectively). Similar associations were found in Group A for AKIpre (Supplementary data, Appendix S8) and in Group B for AKIadm (Supplementary data, Appendix S9).
AKI and 30-day mortality
In Group A, 30-day mortality was significantly higher in patients with AKIpre compared with patients without AKI [(24/88) 27.3% versus (66/637) 10.5%; P <0.001]. In Group B, the mortality rate was also higher in the AKIadm group [(27/134) 20.1% versus (82/674) 12.2%; P = 0.01]. In Group C, mortality was higher in the AKI group when both AKIpre [(24/83) 28.9% versus (50/360) 13.9%; P = 0.001] and AKIadm [(21/88) 23.9% versus (53/355) 14.9%; P = 0.04] were used to identify AKI compared with patients without AKI.
The full univariable and multivariable associations with 30-day mortality in Groups A and B are shown in Supplementary data, Appendices S10 and S11. In Group A, AKIpre was associated with 30-day mortality in multivariable analysis (OR = 2.66; 95% CI 1.40–5.05; P = 0.003; Table 2). In Group B, AKIadm was associated with 30-day mortality in multivariable analysis (OR = 1.79; 95% CI 1.04–3.08; P = 0.04; Table 2).
Table 2.
Logistic univariable and multivariable associations of AKI calculated using pre-admission SCr (AKIpre) or admission creatinine (AKIadm) with 30-day mortality
| Models | Group A (AKIpre) |
Group B (AKIadm) |
Group C (AKIpre) |
Group C (AKIadm) |
||||
|---|---|---|---|---|---|---|---|---|
| OR | P-value | OR | P-value | OR | P-value | OR | P-value | |
| Model 1 | 3.21 | <0.001 | 1.82 | 0.015 | 2.52 | 0.001 | 1.79 | 0.046 |
| (1.88–5.48) | (1.13–2.95) | (1.44–4.42) | (1.01–3.16) | |||||
| Model 2 | 2.86 | <0.001 | 1.58 (0.96–2.60) | 0.07 | 2.37 | 0.003 | 1.71 (0.95–3.07) | 0.07 |
| (1.65–4.98) | (1.33–4.22) | |||||||
| Model 3 | 2.66 | 0.003 | 1.79 | 0.04 | 2.64 | 0.004 | 2.10 | 0.03 |
| (1.40–5.05) | (1.04–3.08) | (1.36–5.12) | (1.09–4.03) | |||||
| Model 4 | 2.66 | 0.003 | 1.79 | 0.04 | 2.64 | 0.004 | 2.10 | 0.03 |
| (1.40–5.05) | (1.04–3.08) | (1.36–5.12) | (1.09–4.03) | |||||
| Model 5 | 2.66 | 0.003 | 1.79 | 0.04 | 2.64 | 0.004 | 2.10 | 0.03 |
| (1.40–5.05) | (1.04–3.08) | (1.36–5.12) | (1.09–4.03) | |||||
| Model 6 | – | – | – | – | 2.52 | 0.001 | ||
| (1.44–4.42) | ||||||||
| Model 7 | – | – | – | – | 2.37 | 0.003 | ||
| (1.33–4.22) | ||||||||
| Model 8 | – | – | – | – | 2.64 | 0.004 | ||
| (1.36–5.12) | ||||||||
| Model 9 | – | – | – | – | 2.64 | 0.004 | ||
| (1.36–5.12) | ||||||||
| Model 10 | – | – | – | – | 2.64 | 0.004 | ||
| (1.36–5.12) | ||||||||
| Model 11 | – | – | – | – | AKIadm 1.13 (0.57–2.25) | 0.73 | – | – |
| AKIpre 2.37 (1.21–4.62) | 0.01 | |||||||
| Model 12 | – | – | – | – | AKIadm 1.09 (0.54–2.22) | 0.81 | – | – |
| AKIpre 2.323 (1.160–4.650) | 0.02 | |||||||
| Model 13 | – | – | – | – | AKIadm 1.35 (0.61–3.01) | 0.46 | – | – |
| AKIpre 2.20 (1.00–4.84) | 0.051 | |||||||
| Model 14 | – | – | – | – | AKIadm 1.38 (0.62–3.10) | 0.43 | – | – |
| AKIpre 2.22(1.00–4.91) | 0.049 | |||||||
| Model 15 | – | – | – | – | AKIadm 1.35 (0.60–3.02) | 0.47 | – | – |
| AKIpre 2.25 (1.02–4.97) | 0.046 | |||||||
Model 1: unadjusted. Model 2: adjusted for age and sex. Model 3: adjusted for age, sex, presence of eGFR <60 mL/min/1.73 m2 and AF, stroke type, disability score on admission, stroke severity and anaemia (all factors associated in univariable analysis with 30-day mortality in Group A). Model 4: adjusted for age, sex, ethnicity, presence of eGFR <60 mL/min/1.73 m2, AF and hypertension, disability score on admission, stroke severity and anaemia (all factors associated in univariable analysis with 30-day mortality in Group B). Model 5: adjusted for age, sex, ethnicity, presence of eGFR <60 mL/min/1.73 m2, AF and hypertension, stroke type, disability score on admission, stroke severity and anaemia (all factors associated in univariable analysis with 30-day mortality in Groups A and B). Model 6: Model 1 + AKIpre and AKIadm. Model 7: Model 2 + AKIpre and AKIadm. Model 8: Model 3 + AKIpre and AKIadm. Model 9: Model 4 + AKIpre and AKIadm. Model 10: Model 5 + AKIpre and AKIadm. Model 11: Model 6 with AKIadm forced into model. Model 12: Model 7 with AKIadm forced into model. Model 13: Model 8 with AKIadm forced into model. Model 14: Model 9 with AKIadm forced into model. Model 15: Model 10 with AKIadm forced into model.
The full univariable and multivariable associations with 30-day mortality in Group C are shown in Supplementary data, Appendix S12. Both AKIpre (OR = 2.64; 95% CI 1.36–5.12; P = 0.004) and AKIadm (OR = 2.10; 95% CI 1.09–4.03; P = 0.03) remained associated with 30-day mortality in the adjusted models (Table 2). Given that AKIpre appeared to have a stronger relationship with 30-day mortality than AKIadm, we constructed further models adjusting for factors associated with 30-day mortality when either AKIpre or AKIadm were used in the multivariable model. The OR for mortality remained higher using AKIpre than AKIadm in all models (Models 3–5; Table 2). We created additional models by entering both AKIpre and AKIadm. In all models, only AKIpre was retained (Models 6–10; Table 2), suggesting that AKIpre does indeed have a stronger relationship with 30-day mortality than AKIadm. We further explored this relationship by forcing AKIadm into the multivariable models and then adding AKIpre. In all cases, AKIpre was also retained in the models, suggesting that AKIpre carries further information than that provided by AKIadm alone (Models 11–15; Table 2).
AKI and 1-year mortality
In Group A, 1-year mortality was significantly higher in patients with AKIpre compared with patients without AKI [(42/88) 47.7% versus (131/637) 20.8%; P <0.001]. In Group B, the mortality rate was higher in the AKIadm group [(51/134) 38.1% versus (156/674) 23.1%; P <0.001]. In Group C, mortality was higher in the AKI group when both AKIpre [(42/83) 50.6% versus (97/360) 26.9%; P <0.001] and AKIadm [(38/88) 43.2% versus (101/355) 28.5%; P <0.001] were used to identify AKI compared with patients without AKI (Figure 2).
FIGURE 2.
(A) Kaplan–Meier survival curve at 365 days for Group C using AKIpre to diagnose AKI rates. Log rank test = 19.73; P < 0.001. (B) Kaplan–Meier survival curve for Group C at 365 days using AKIadm to diagnose AKI. Log rank test = 6.93; P = 0.008.
The full univariable and multivariable associations with 1-year mortality in Groups A and B are shown in Supplementary data, Appendices S13 and S14. In Group A, AKIpre was associated with 1-year mortality in multivariable analysis (HR = 2.00; 95% CI 1.40–2.86; P <0.001; Table 3). In Group B, AKIadm was associated with 1-year mortality in multivariable analysis (HR = 1.50; 95% CI 1.10–2.07; P = 0.01; Table 3).
Table 3.
Cox univariable and multivariable associations of AKI calculated using pre-admission SCr (AKIpre) or admission creatinine (AKIadm) with 1 year mortality
| Models | Group A (AKIpre) |
Group B (AKIadm) |
Group C (AKIpre) |
Group C (AKIadm) |
||||
|---|---|---|---|---|---|---|---|---|
| HR | P-value | HR | P-value | HR | P-value | HR | P-value | |
| Model 1 | 2.74 | <0.001 | 1.77 | <0.001 | 2.22 | <0.001 | 1.64 | 0.009 |
| (1.937–3.886) | (1.29–2.42) | (1.55–3.19) | (1.13–2.38) | |||||
| Model 2 | 2.43 | <0.001 | 1.52 | 0.009 | 2.08 | <0.001 | 1.55 | 0.02 |
| (1.72–3.45) | (1.11–2.09) | (1.44–2.99) | (1.07–2.26) | |||||
| Model 3 | 2.00 | <0.001 | 1.50 | 0.01 | 1.98 | <0.001 | 1.53 | 0.03 |
| (1.40–2.86) | (1.10–2.07) | (1.37–2.85) | (1.05–2.23) | |||||
| Model 4 | 2.00 | <0.001 | 1.50 | 0.01 | 1.98 | <0.001 | 1.53 | 0.03 |
| (1.40–2.86) | (1.10–2.07) | (1.37–2.85) | (1.05–2.23) | |||||
| Model 5 | 2.00 | <0.001 | 1.50 | 0.01 | 1.90 | 0.001 | 1.47 | 0.05 |
| (1.40–2.86) | (1.10–2.07) | (1.32–2.76) | (1.01–2.15) | |||||
| Model 6 | – | – | – | – | 2.22 | <0.001 | ||
| (1.55–3.19) | ||||||||
| Model 7 | – | – | – | – | 2.08 | <0.001 | ||
| (1.44–2.99) | ||||||||
| Model 8 | – | – | – | – | 1.98 | <0.001 | ||
| (1.37–2.85) | ||||||||
| Model 9 | – | – | – | – | 1.98 | <0.001 | ||
| (1.37–2.85) | ||||||||
| Model 10 | – | – | – | – | 1.90 | 0.001 | ||
| (1.32–2.76) | ||||||||
| Model 11 | – | – | – | – | AKIadm 1.11 (0.71–1.74) | 0.65 | – | – |
| AKIpre 2.10 (1.36–3.25) | 0.001 | |||||||
| Model 12 | – | – | – | – | AKIadm 1.07 (0.68–1.66) | 0.78 | – | – |
| AKIpre 2.06 (1.33–3.18) | 0.001 | |||||||
| Model 13 | – | – | – | – | AKIadm 1.09 (0.69–1.72) | 0.70 | – | – |
| AKIpre 2.02 (1.29–3.17) | 0.002 | |||||||
| Model 14 | – | – | – | – | AKIadm 1.07 (0.68–1.70) | 0.76 | – | – |
| AKIpre 1.96 (1.25–3.08) | 0.004 | |||||||
| Model 15 | – | – | – | – | AKIadm 1.06 (0.67–1.68) | 0.81 | – | – |
| AKIpre 2.00 (1.27–3.16) | 0.003 | |||||||
Model 1: unadjusted. Model 2: adjusted for age and sex. Model 3: adjusted for age, sex, presence of eGFR <60 mL/min/1.73 m2, AF and CHF, stroke type, disability score on admission, stroke severity and anaemia (all factors associated in univariable analysis with 1-year mortality in Group A). Model 4: adjusted for age, sex, presence of eGFR <60 mL/min/1.73 m2 and AF, disability score on admission, stroke severity and anaemia (all factors associated in univariable analysis with 1-year mortality in Group B). Model 5: Model 3 plus ethnicity. Model 6: Model 1 + AKIpre and AKIadm. Model 7: Model 2 + AKIpre and AKIadm. Model 8: Model 3 + AKIpre and AKIadm. Model 9: Model 4 + AKIpre and AKIadm. Model 10: Model 5 + AKIpre and AKIadm. Model 11: Model 6 with AKIadm forced into model. Model 12: Model 7 with AKIadm forced into model. Model 13: Model 8 with AKIadm forced into model. Model 14: Model 9 with AKIadm forced into model. Model 15: Model 10 with AKIadm forced into model.
The full univariable and multivariable associations with 1-year mortality in Group C are shown in Supplementary data, Appendix S15. Both AKIpre (HR = 1.90; 95% CI 1.32–2.76; P = 0.001) and AKIadm (HR = 1.47; 95% CI 1.01–2.15; P = 0.05) remained associated with 1-year mortality in the adjusted models (Table 3). Similar to 30-day mortality, AKIpre appeared to have a stronger relationship with 1-year mortality than AKIadm, and further models were constructed adjusting for factors associated with 1-year mortality when either AKIpre or AKIadm were used in the multivariable model (Table 3). The HR for mortality remained higher using AKIpre than AKIadm in all models (Models 3–5; Table 3). We created additional models by entering both AKIpre and AKIadm. In all models, only AKIpre was retained (Models 6–10; Table 3), suggesting AKIpre does indeed have a stronger relationship with 30-day mortality than AKIadm. We further explored this relationship by forcing AKIadm into the multivariable models and then adding AKIpre. In all cases, AKIpre was also retained in the models, suggesting that AKIpre carries further information than that provided by AKIadm (Models 11–15; Table 3).
DISCUSSION
To the best of our knowledge, this is the first study to investigate the rate of AKI in acute stroke utilizing two different methods. We show that AKI in unselected patients hospitalized after a stroke is high at ∼20% and this rate does not appear to be dependent on the method used to diagnose AKI. We also show that mortality after AKI is high but consistently higher in AKIpre, with 27.3 and 47.7% of patients being dead at 30 days and 1 year, respectively. Use of AKIadm consistently underestimates mortality risk at both time points.
SCr is an imperfect measurement of dynamic glomerular filtration rate [32] and is modified by age, sex and race as well as nutritional state, muscle mass and hydration [3]. Consequently, novel biomarkers for earlier and more accurate detection of AKI are under investigation, but to date, these have not progressed to use in routine medical practice [33–36]. Therefore, SCr remains the only routinely used laboratory test for the diagnosis of AKI [37–39].
Use of admission SCr, as recommended by AKIN [8, 9] and ERBP guidelines [10], to diagnose AKI has low sensitivity and may therefore fail to detect both community-acquired and hospital-acquired AKI [40]. Use of a baseline SCr taken 7–365 days up to admission [8, 9, 11–15] is not available for many patients because of the absence of a pre-admission value. In our cohort, 46.5% of patients did not have a pre-admission SCr, consistent with published data [41], and we found that patients with a pre-admission SCr were more likely to be older and have more comorbidities. Although we have performed several analyses to exclude this as a major source of confounding, it remains possible that some residual confounding persists and our findings require confirmation in other studies. Patients of black ethnicity were less likely to have a pre-admission SCr, consistent with reports that this ethnic group is less likely to have access to health systems both in the USA and the UK [42–44]. This has partly been attributed to socioeconomic status [45]. However, in this study, socioeconomic deprivation did not appear to be associated with having a pre-admission SCr. It is uncertain whether this is a factor of the UK’s social system of healthcare or whether it would be replicated in other healthcare systems [44].
Using admission SCr to diagnose AKI in our cohort produced a similar rate of AKI to pre-admission SCr. These rates are comparable to other published studies that utilize SCr values rather than coding to diagnose AKI, with rates ranging from 15% to 27% [20–22]. However, although the absolute rates are similar, AKIadm when compared with AKIpre had low sensitivity and a Kappa value indicating only moderate agreement. We also found that individual patients were just as likely to have a SCr on admission that was 10% higher or 10% lower than the pre-admission SCr. This could potentially suggest a random effect consistent with day-to-day biological fluctuations in SCr [46, 47] or analytical variability in the laboratory [48]. Consistent with this, we found that patients with a higher admission SCr were more likely to be diagnosed with AKIpre, whereas patients with a lower admission SCr were more likely to be diagnosed with AKIadm. Despite the similar rates of AKI diagnosed with either method, we found that AKIpre was consistently associated with higher 30-day and 1-year mortality than AKIadm. Although we cannot prove this conclusively, our results would suggest that using AKIpre correctly identifies more patients with ‘true’ AKI. Furthermore, our finding that AKIpre was associated with a higher mortality risk than AKIadm indicates that AKI-associated mortality in acute stroke, as reported to date, probably underestimates the association, given that all of the studies that used SCr to define AKI used first SCr on admission [20–22]. These findings have major implications for clinical management and further research. Future epidemiological, or indeed interventional studies, should make every effort to use pre-admission SCr as the baseline for AKI diagnosis.
In some clinical situations, where the first SCr on admission is taken as the baseline value, cases of community-acquired AKI may be missed, thus underestimating AKI incidence [8]. Conversely, in older and frailer patients with an acute illness, the admission SCr may be lower than the outpatient ‘baseline’, leading to overestimation of AKI rates [14, 49]. Inclusion of patients without true AKI in the analysis may introduce further bias. This was emphasized by a recent US study which reported that 45% of all hospitalized patients had a first SCr on admission <90% of the pre-admission SCr [14]. However, we feel that these are unlikely to have been significant factors in our study. Acute stroke, by its very nature, happens suddenly and patients are usually admitted to hospital almost immediately, giving little time to develop community-acquired AKI. Patients presenting with a first acute stroke might also be less likely to have a gradually worsening health state preceding the admission, and therefore less likely to have a lower SCr as might be expected in patients with more chronic conditions.
Our study has some limitations to acknowledge. A significant proportion of the cohort had no pre-admission SCr. This may have caused confounding since patients with available pre-admission SCr likely represent a more comorbid group. To address this, we selected a core cohort of patients (Group C) who had both a pre-admission SCr value and at least two blood tests during admission and used multiple adjustments in our analysis. However, it is possible that residual confounding remains. We do not have data on the length of time between the onset of the acute stroke and the drawing of blood for measurement of SCr. However, patients suffering an acute stroke are generally admitted to hospital immediately and blood drawn shortly after arrival in the emergency department. Therefore, the interval is unlikely to be ever more than a few hours and it is doubtful that this would have a significant effect on the findings. We do not have data on cause of death. There is surprisingly little data on the cause of death after AKI. A recent population-based study has shown that the most common causes of death after AKI are cancer or cardiovascular events [50]. We would therefore predict that cardiovascular events were the most common cause of death after AKI in stroke patients, but further work is required to prove this. Our study utilizes data from a cohort of stroke patients and therefore AKI incidence, risk factors and outcomes may be different to an unselected hospital population with AKI. Finally, we were limited by the observational nature of this study and were unable to address causality.
CONCLUSIONS
In summary, AKI after stroke is common and associated with increased mortality. First creatinine on admission gives a comparable incidence of AKI to pre-admission creatinine but with low sensitivity. First creatinine on admission to classify AKI consistently underestimates 30-day and 1-year mortality compared with pre-admission SCr. Further studies are required to help inform future AKI prevention trials that are urgently needed to ultimately improve clinical outcomes.
SUPPLEMENTARY DATA
Supplementary data are available at ckj online.
Supplementary Material
ACKNOWLEDGEMENTS
We would like to acknowledge and thank James Hodson and Peter Nightingale from the statistical department of University Hospitals Birmingham NHS Foundation Trust for their statistical input.
CONFLICT OF INTEREST STATEMENT
None declared.
REFERENCES
- 1. Selby NM, Crowley L, Fluck RJ. et al. Use of electronic results reporting to diagnose and monitor AKI in hospitalized patients. Clin J Am Soc Nephrol 2012; 7: 533–540 [DOI] [PubMed] [Google Scholar]
- 2. Zeng X, McMahon GM, Brunelli SM. et al. Incidence, outcomes, and comparisons across definitions of AKI in hospitalized individuals. Clin J Am Soc Nephrol 2014; 9: 12–20 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Waikar SS, Liu KD, Chertow GM.. Diagnosis, epidemiology and outcomes of acute kidney injury. Clin J Am Soc Nephrol 2008; 3: 844–861 [DOI] [PubMed] [Google Scholar]
- 4. Nisula S, Kaukonen K-M, Vaara ST. et al. Incidence, risk factors and 90-day mortality of patients with acute kidney injury in Finnish intensive care units: the FINNAKI study. Intensive Care Med 2013; 39: 420–428 [DOI] [PubMed] [Google Scholar]
- 5. Chertow GM, Burdick E, Honour M. et al. Acute kidney injury, mortality, length of stay, and costs in hospitalized patients. J Am Soc Nephrol 2005; 16: 3365–3370 [DOI] [PubMed] [Google Scholar]
- 6. Coca SG, Singanamala S, Parikh CR.. Chronic kidney disease after acute kidney injury: a systematic review and meta-analysis. Kidney Int 2012; 81: 442–448 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kidney Disease: Improving Global Outcomes (KDIGO) Acute Kidney Injury Work Group. KDIGO Clinical Practice Guideline for Acute Kidney Injury. Kidney Int 2012; 2(Suppl): 1–138 [Google Scholar]
- 8. Siew ED, Matheny ME, Ikizler TA. et al. Commonly used surrogates for baseline renal function affect the classification and prognosis of acute kidney injury. Kidney Int 2010; 77: 536–542 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Siew ED, Ikizler TA, Matheny ME. et al. Estimating baseline kidney function in hospitalized patients with impaired kidney function. Clin J Am Soc Nephrol 2012; 7: 712–719 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Ad-hoc working group of E, Fliser D, Laville M. et al. A European Renal Best Practice (ERBP) position statement on the Kidney Disease Improving Global Outcomes (KDIGO) clinical practice guidelines on acute kidney injury: part 1: definitions, conservative management and contrast-induced nephropathy. Nephrol Dial Transplant 2012; 27: 4263–4272 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Go AS, Parikh CR, Ikizler TA. et al. The assessment, serial evaluation, and subsequent sequelae of acute kidney injury (ASSESS-AKI) study: design and methods. BMC Nephrol 2010; 11: 22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Matheny ME, Peterson JF, Eden SK. et al. Laboratory test surveillance following acute kidney injury. PLoS One 2014; 9: e103746. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Leaf DE, Srivastava A, Zeng X. et al. Excessive diagnostic testing in acute kidney injury. BMC Nephrol 2016; 17: 9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Liu KD, Hsu C-Y, Yang J. et al. Acute kidney injury ascertainment is affected by the use of first inpatient versus outpatient baseline serum creatinine. Kidney Int Rep 2018; 3: 211–215 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Siew ED, Matheny ME.. Choice of reference serum creatinine in defining acute kidney injury. Nephron 2015; 131: 107–112 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Prabhakaran S, Ruff I, Bernstein RA.. Acute stroke intervention: a systematic review. JAMA 2015; 313: 1451–1462 [DOI] [PubMed] [Google Scholar]
- 17. Masson P, Webster AC, Hong M. et al. Chronic kidney disease and the risk of stroke: a systematic review and meta-analysis. Nephrol Dial Transplant 2015; 17: 1162–1169 [DOI] [PubMed] [Google Scholar]
- 18. Arnold J, Sims D, Ferro CJ.. Modulation of stroke risk in chronic kidney disease. Clin Kidney J 2016; 9: 29–38 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Arnold J, Ng KP, Sims D. et al. Incidence and impact on outcomes of acute kidney injury after a stroke: a systematic review and meta-analysis. BMC Nephrol 2018; 19: 283. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Covic A, Schiller N-G, Mardare L. et al. The impact of acute kidney injury on short-term survival in an Eastern European population with stroke. Nephrol Dial Transplant 2008; 23: 2228–2234 [DOI] [PubMed] [Google Scholar]
- 21. Tsagalis G, Akrivos T, Alevizaki M. et al. Long-term prognosis of acute kidney injury after first acute stroke. Clin J Am Soc Nephrol 2009; 4: 616–622 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Khatri M, Himmelfarb J, Adams D. et al. Acute kidney injury is associated with increased hospital mortality after stroke. J Stroke Cerebrovasc Dis 2014; 23: 25–30 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. von Elm E, Altman DG, Egger M. et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. PLoS Med 2007; 4: e296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Morris S, Ramsay AIG, Boaden RJ. et al. Impact and sustainability of centralising acute stroke services in English metropolitan areas: retrospective analysis of hospital episode statistics and stroke national audit data. BMJ 2019; 364: l1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. van Swieten JC, Koudstaal PJ, Visser MC. et al. Interobserver agreement for the assessment of handicap in stroke patients. Stroke 1988; 19: 604–607 [DOI] [PubMed] [Google Scholar]
- 26. Brott T, Adams HP, Olinger CP. et al. Measurements of acute cerebral infarction: a clinical examination scale. Stroke 1989; 20: 864–870 [DOI] [PubMed] [Google Scholar]
- 27. Arnold JJ, Hayer M, Sharif A. et al. Acute Care QUAliTy in chronic Kidney disease (ACQUATIK): a prospective cohort study exploring outcomes of patients with chronic kidney disease. BMJ Open 2015; 5: e006987. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Kidney Disease: Improving Global Outcomes (KDIGO) CKD Work Group. KDIGO 2012 clinical practice guideline for the evaluation and management of chronic kidney disease. Kidney Int 2013; 3: 1–150 [DOI] [PubMed] [Google Scholar]
- 29. Hoffman JI. The incorrect use of Chi-square analysis for paired data. Clin Exp Immunol 1976; 24: 227–229 [PMC free article] [PubMed] [Google Scholar]
- 30. Bland JM, Altman DG.. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1: 307–310 [PubMed] [Google Scholar]
- 31. Bursac Z, Gauss CH, Williams DK. et al. Purposeful selection of variables in logistic regression. Source Code Biol Med 2008; 3: 17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Uchino S. Creatinine. Curr Opin Crit Care 2010; 16: 562–567 [DOI] [PubMed] [Google Scholar]
- 33. Murray PT, Mehta RL, Shaw A. et al. Potential use of biomarkers in acute kidney injury: report and summary of recommendations from the 10th Acute Dialysis Quality Initiative consensus conference. Kidney Int 2014; 85: 513–521 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Alge JL, Arthur JM.. Biomarkers of AKI: a review of mechanistic relevance and potential therapeutic implications. Clin J Am Soc Nephrol 2015; 10: 147–155 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Zhang A, Cai Y, Wang PF. et al. Diagnosis and prognosis of neutrophil gelatinase-associated lipocalin for acute kidney injury with sepsis: a systematic review and meta-analysis. Crit Care 2016; 20: 41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Su LJ, Li YM, Kellum JA. et al. Predictive value of cell cycle arrest biomarkers for cardiac surgery-associated acute kidney injury: a meta-analysis. Br J Anaesth 2018; 121: 350–357 [DOI] [PubMed] [Google Scholar]
- 37. Pickering JW, Endre ZH.. The definition and detection of acute kidney injury. J Renal Inj Prev 2014; 3: 21–25 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Thomas ME, Blaine C, Dawnay A. et al. The definition of acute kidney injury and its use in practice. Kidney Int 2015; 87: 62–73 [DOI] [PubMed] [Google Scholar]
- 39. Malhotra R, Siew ED.. Biomarkers for the early detection and prognosis of acute kidney injury. Clin J Am Soc Nephrol 2017; 12: 149–173 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Broce JC, Price LL, Liangos O. et al. Hospital-acquired acute kidney injury: an analysis of nadir-to-peak serum creatinine increments stratified by baseline estimated GFR. Clin J Am Soc Nephrol 2011; 6: 1556–1565 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Bernier-Jean A, Beaubien-Souligny W, Goupil R. et al. Diagnosis and outcomes of acute kidney injury using surrogate and imputation methods for missing preadmission creatinine values. BMC Nephrol 2017; 18: 141. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Gornick ME, Eggers PW, Reilly TW. et al. Effects of race and income on mortality and use of services among Medicare beneficiaries. N Engl J Med 1996; 335: 791–799 [DOI] [PubMed] [Google Scholar]
- 43. Fiscella K, Franks P, Gold MR. et al. Inequality in quality: addressing socioeconomic, racial, and ethnic disparities in health care. JAMA 2000; 283: 2579–2584 [DOI] [PubMed] [Google Scholar]
- 44. Szczepura A. Access to health care for ethnic minority populations. Postgrad Med J 2005; 81: 141–147 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Nazroo JY. The structuring of ethnic inequalities in health: economic position, racial discrimination, and racism. Am J Public Health 2003; 93: 277–284 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Toffaletti JG, McDonnell EH.. Variation of serum creatinine, cystatin C, and creatinine clearance tests in persons with normal renal function. Clin Chim Acta 2008; 395: 115–119 [DOI] [PubMed] [Google Scholar]
- 47. Reinhard M, Erlandsen EJ, Randers E.. Biological variation of cystatin C and creatinine. Scand J Clin Lab Invest 2009; 69: 831–836 [DOI] [PubMed] [Google Scholar]
- 48. Delanaye P, Cavalier E, Pottel H.. Serum creatinine: not so simple! Nephron 2017; 136: 302–308 [DOI] [PubMed] [Google Scholar]
- 49. Doi K, Yuen PST, Eisner C. et al. Reduced production of creatinine limits its use as marker of kidney injury in sepsis. J Am Soc Nephrol 2009; 20: 1217–1221 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Silver SA, Harel Z, McArthur E. et al. Causes of death after a hospitalization with AKI. J Am Soc Nephrol 2018; 29: 1001–1010 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.