Acute kidney disease in hospitalized acute kidney injury patients

Study design, setting and population

The multi-center retrospective cohort study identified AKI patients from hospitalized patients aged more than 14 years in three affiliated hospitals of Central South University in China from January 1st, 2015 to December 31st, 2015. Study procedures were shown in Fig. 1. AKI was determined according to the 2012 Kidney Disease: Improving Global Outcomes Clinical Practice Guideline for Acute Kidney Injury (Kidney Disease: Improving Global Outcomes KDIGO, 2012) by the serum creatinine (SCr) level. Increase in SCr by ≥ 0.3 mg/dl within 48 h or increase in SCr to ≥ 1.5 times baseline within 7 days was used to identify AKI patients. Patients with CKD stage 5 (admission diagnosis including CKD stage 5, end stage renal disease or uremia), kidney transplantation (admission diagnosis including kidney transplantation or kidney transplant status), hospital stay <48 h, non-AKI or incomplete medical records were excluded. For patients with multiple hospitalizations, only the first hospitalization was included in the analysis. The Medical Ethics Committee of the Second Xiangya Hospital of Central South University approved the study protocol (2013-S061) and waived the patient consent. This project has been registered in Chinese Clinical Trial Registry (ChiCTR 1800019857).

Identification of AKD

AKD was defined as a condition in which AKI stage 1 or greater persists for more than 6 days and less than 90 days after an AKI initiating event according to the consensus report of the ADQI 16 Workgroup (Chawla et al., 2017). AKD stage 1 was defined as SCr being 1.5–1.9 times of the baseline level, AKD stage 2 as SCr being 2.0–2.9 times of the baseline level, and AKD stage 3 as SCr being ≥3.0 times of the baseline level or ongoing need for renal replacement therapy. Patients with SCr <1.5 times of baseline level were categorized as AKD stage 0 (non-AKD group). AKD and AKD stage were identified based on the SCr level on the 7th day or the day nearest to the 7th day but not more than the 10th day after the diagnosis of AKI. Baseline SCr was defined as the mean SCr value that was available between 7-365 days prior to admission (Siew et al., 2012). For patients without a reliable baseline SCr before admission and with no evidence of baseline CKD, a back-estimation of the baseline SCr was performed using the Modification of Diet in Renal Disease study equation assuming that the baseline eGFR is 75 ml/min per 1.73 m² (Bellomo et al., 2004).

Data collection and definition

Patients’ information was extracted from Hospital Information System, Laboratory Information System and medical records, which included age, sex, clinical departments, comorbidities, diagnosis on admission and discharge, surgeries, invasive procedures, all-cause in-hospital death, as well as laboratory test results and time. Identification and classification of AKI (2012), AKI stage, AKI type, comorbidity, Charlson comorbidity index (CCI) (Charlson et al., 1994), organ failure (Blanco et al., 2008) and sepsis (Bone et al., 1992) were evaluated by trained nephrologists through reviewing medical records and laboratory data. AKI stage was defined according to the 2012 KDIGO guide for AKI ( Kidney Disease: Improving Global Outcomes KDIGO, 2012) and the highest stage within the first 6 days of AKI diagnosis was used to determine the AKI stage. Community-acquired AKI was identified when patients met the KDIGO AKI definition according to SCr change on the first day of admission; hospital-acquired AKI was identified when patients who developed AKI did not meet community-acquired AKI criteria. Critically ill patients were identified as those who were admitted to intensive care units when AKI was diagnosed. Patients who could not be classified as critically ill patients were regarded as non-critically ill patients. Presence of comorbidities was determined by the diagnosis codes at admission and discharge. The burden of comorbidities was evaluated by CCI. Organ failure (cardiovascular system dysfunction, the arterial systolic blood pressure ≤ 90 mm Hg or mean arterial pressure ≤ 70 mm Hg or use of vasopressors; kidney dysfunction, urine output <0.5 ml/kg /h despite adequate blood volume, or creatinine >1.9 mg/dl; respiratory system dysfunction, hypoxemia with PaO2 <60 mm Hg or mechanical ventilation; hematologic dysfunction, the platelet count <80,000/mm³; liver dysfunction, bilirubin ≥ 3 mg/dl; nervous system dysfunction, use of sedative), and diagnosis of anemia (hemoglobin ≤ 10 g/dl), hypoalbuminemia (serum albumin <30 g/l), proteinuria (dipstick urinalysis protein positive), and hyperuricemia (serum uric acid ≥7.0 mg/dl in men and 6.0 mg/dl in women) were determined by the data within seven days prior to AKD diagnosis. The worst result was used if there were multiple results for the same test. Data on hospital operation were determined by patients’ medical orders information and collected within three days prior to AKD diagnosis.

Endpoints and follow-up

The primary endpoint was the proportion of patients who met one or more criteria for Major Adverse Kidney Events within 30 days (MAKE30) (Kellum, Zarbock & Nadim, 2017; Semler et al., 2016) after diagnosis of AKI which consists of all-cause mortality (hereafter referred to as mortality), new receipt of renal replacement therapy, or persistent renal dysfunction (defined as the creatinine value ≥ 200% of the baseline value). All events were censored at hospital discharge or 30 days after the diagnosis of AKI, whichever came first. The second endpoint was the recipient of chronic dialysis or mortality one year after the diagnosis of AKI. MAKE30 and chronic dialysis in one year were identified through reviewing all relevant medical records (Hospital Information System, Laboratory Information System and out-patients records), making phone calls and sending messages. Dialysis performed during hospitalization was determined based on patients’ medical orders information which included ‘hemodialysis’ or ‘peritoneal dialysis’. Chronic dialysis was further determined by referring to the Chinese National Renal Data System. Death was confirmed through data linkage to the Chinese Center for Disease Control and Prevention cause-of-death reporting system which is a national administrative registry responsible for collection and management of death information from all provinces in China.

Statistical analysis

Patients’ baseline characteristics and outcomes were compared between AKD stages using One-way ANOVA test or Kruskal-Wallis test for continuous variables and chi-square test for categorical variables as appropriate. Multiple comparisons between AKD stages were made by Student-Newman-Keuls test for continuous variables and Scheffe’s confidence interval method for categorical variables as appropriate. Multivariable logistic regression analysis was used to estimate the impact of AKD stage on persistent renal dysfunction and new receipt of renal replacement therapy in 30 days, MAKE30, one-year chronic dialysis as well as composite adverse outcomes in one year. Cox proportional hazard regression model was used to estimate the hazard ratio of AKD on 30-day and one-year mortality. Variables that were considered clinically relevant or associated with outcomes were adjusted in the multivariable models. Proportional hazards assumption was assessed by the curves of log[-logS(t)] to t test for each covariable. Multivariable logistic regression analysis was used to determine the predictors of AKD stage 2–3 and establish the risk prediction model. Baseline variables that were considered clinically relevant or that showed a univariable relationship with outcome were entered into the model and selected by the forward selection method. Discrimination and calibration of the model were tested by area under the receiver operating characteristic curve (AUROC) and Hosmer and Lemeshow test.

Statistical analyses were performed using SPSS 18.0 (IBM Corporation). P < 0.05 was considered statistically significant.

Incidence of AKD

The overall incidences of AKI and AKD in hospitalized patients were 8.31% (2884/34709) and 3.34% (1161/34709) respectively. 2,556 AKI patients meeting the inclusion criteria were enrolled in our analysis. AKD occurred in 45.4% (1161/2556) of all AKI patients, 14.5% (141/971) of AKI stage 1 patients, 44.6% (308/691) of AKI stage 2 patients and 79.6% (712/894) of AKI stage 3 patients. Characteristics of the study population stratified by AKD stages were presented in Table 1. The prevalences of AKD stage 1 and stage 2–3 were 16.9% and 28.5% respectively. AKD stage 2–3 tended to occur in patients who were older and accompanied with more comorbidities. A larger percentage of patients with AKI stage 3, intrinsic renal AKI as well as mechanical ventilation were also observed in AKD stage 2–3.

Outcomes of AKD patients

Outcomes of patients stratified by AKD stage were provided in Table 2. Kaplan–Meier survival curves for 30-day and one-year mortality among AKD stages were shown in Fig. 2. During the one-year follow-up, 26.8% (684/2556) of patients developed the composite endpoints of MAKE30, of whom 13.5% (92/684) came from the AKD stage 0 group, 11.0% (75/684) from the AKD stage 1 group and 75.6% (517/684) from the AKD stage 2–3 group. 19.0% (486/2556) of patients developed the composite endpoints of adverse outcomes (chronic dialysis and mortality) in one year, with 35.2% (171/486) from the AKD stage 0 group, 15.0% (73/486) from the AKD stage 1 group and 49.8% (242/486) from the AKD stage 2–3 group. A stepwise increase in the incidence of MAKE30 and one-year adverse outcomes was observed as the AKD stage got higher. Compared with AKD stage 0 and 1, patients with AKD stage 2–3 consistently showed significantly higher incidences of developing all adverse outcomes in 30 days and one year. AKD stage 1 only had higher incidences in persistent renal dysfunction and mortality in 30 days than AKD stage 0 (Table 2).

Association between AKD stage and 30-day as well as one-year adverse outcomes was shown in Table 3. Fully adjusted multivariable regression analysis indicated that AKD stage 1 mainly showed an increased risk of persistent renal function in 30 days, while AKD stage 2–3 conferred remarkably higher risks in receiving renal replacement therapy, developing persistent renal dysfunction and mortality in 30 days along with chronic dialysis and mortality in one year. OR/HR for covariables adjusted in the multivariable regression model were shown in Tables S1–S7.

Association between AKI stage, AKD stage and adverse outcomes in 30 days and one year

To rule out the impact of AKI stage on the association between AKD and 30-day as well as one-year adverse outcomes, the incidences of MAKE30 and one-year adverse outcomes in patients stratified by both AKI stage and AKD stage were analyzed and shown in Fig. 3. In each stratum of AKI, the incidences of MAKE30 and one-year adverse outcomes went up as AKD stage got higher and increased remarkably from AKD stage 1 to AKD stage 2–3. When stratified by AKD stage, the occurrence of MAKE30 increased with the increase of AKI stage while the occurrence of one-year adverse outcomes did not alter much among AKI stages. Notably, the incidences of MAKE30 and one-year adverse outcomes in patients with the lowest AKI stage and severe AKD (AKD stage 2–3) were nearly two times higher than those in patients with the most severe AKI and AKD stage 0 (MAKE30:20.50% vs 10.44%; one-year adverse outcomes: 33.33% vs 14.84%). Multivariable logistic regression analysis stratified by AKI stage showed that AKD stage 2–3 remained significantly higher risks in developing MAKE30 and one-year adverse outcomes among each stratum of AKI; however, AKD stage 1 almost conferred no extra risk in MAKE30 and one-year adverse outcomes than AKD stage 0 (Table 4).

Prediction of AKD stage 2–3

Multivariable logistic regression analysis showed that AKI stage, AKI type, AKI classification, proteinuria, anemia, hyperuricemia, organ failure as well as CCI were independent risk factors for the development of AKD stage 2–3 (Fig. 4). AKI stage was a major risk factor for AKD stage 2–3 with the highest odds ratio and predicted it in graded manner as the odds ratio got higher with the increase of AKI stage. This 8-variable model showed good discrimination and calibration in predicting AKD stage 2–3 with the AUROC being 0.85 (95% CI [0.83−0.87]) and the P value of the Hosmer and Lemeshow test being 0.78.

Sensitivity analysis

The association between AKD stage and adverse outcomes in 30 days and one year was further examined in critically ill and non-critically ill patients. The results were not significantly changed as AKD stage 2–3 was similarly associated with increased risk of developing adverse outcomes in 30 days and one year both in the critically ill and non-critically ill, while the association of AKD stage 1 and one-year adverse outcomes was not seen in critically ill patients or non-critically ill patients (Tables S8, S9).