Trends in One-Year Outcomes of Dialysis-Requiring Acute Kidney Injury in Denmark 2005-2012: A Population-Based Nationwide Study

Background Dialysis-requiring acute kidney injury (AKI) is associated with substantial mortality and risk of end-stage renal disease (ESRD). Despite considerable growth in incidence of severe AKI, information pertaining to trends in outcomes remains limited. We evaluated time trends in one year risks of ESRD and death in patients with dialysis-requiring AKI over an eight year period in Denmark. Methods In a retrospective nationwide study based on national registers, all adults requiring acute renal replacement therapy between 2005 and 2012 were identified. Patients with preceding ESRD were excluded. Through individual-level cross-referencing of administrative registries, information pertaining to comorbidity, preceding surgical interventions, and concurrent other organ failure and sepsis was ascertained. Comparisons of period-specific one year odds ratios for ESRD and death were calculated in a multiple logistic regression model. Results A total of 13,819 patients with dialysis-requiring AKI were included in the study. Within one year, 1,017 (7.4%) patients were registered with ESRD, and 7,908 (57.2%) patients died. The one-year rate of ESRD decreased from 9.0% between 2005 and 2006 to 6.1% between 2011 and 2012. Simultaneously, the one-year mortality rate decreased from 58.2% between 2005 and 2006 to 57.5% between 2011 and 2012. Consequently, the adjusted odds ratios for the period 2011–2012 (with the period 2005–2006 as reference) were 0.75 (0.60–0.95, p = 0.015) and 0.87 (95% CI 0.78–0.97, p = 0.010) for ESRD and death, respectively. Conclusions In a nationwide retrospective study on time trends in one year outcomes following dialysis-requiring AKI, risk of all-cause mortality and ESRD decreased over a period of 8 years.


Introduction
Acute kidney injury (AKI) continues to be a severe disease associated with poor prognosis [1,2]. AKI is independently associated with diminished short-and long-term survival [3][4][5], and outcomes are proportional to AKI severity [6,7]. Additionally, AKI is associated with substantially increased propensity for chronic kidney disease (CKD) including end-stage renal disease (ESRD) [8][9][10][11]. Crude incidence rates of dialysis-requiring AKI have increased substantially throughout the last two decades [12,13]; however, Danish incidence rates have remained stable at approximately 350 per million since 2006 [14], and information pertaining to temporal change in outcomes following AKI continues to be limited [15,16]. Development of AKI is predominantly characterized by concurrence of acute and chronic disease [17,18]. Consequently, risk prediction in AKI is dependent on an accurate comprehension of the interplay between potential risk factors for adverse outcomes. The clinical implications associated with changing employment of acute renal replacement therapy (RRT) however remain undetermined. Accordingly, the aim of the present study was to investigate temporal change in outcomes following dialysis-requiring AKI. Consequently, we assessed one-year risk for ESRD and all-cause mortality following dialysis-requiring AKI between 2005 and 2012 in a nationwide population-based cohort.

Data sources
Danish residents are issued a civil registration number at birth or time of immigration. As health services are tax-funded; all Danish citizens receive comprehensive medical coverage through the public health care system. Treatment is registered according to the civil registration number, and consequently, cross-referencing of registers is possible. Information concerning public health care is recorded in multiple national administrative registers. All hospitalizations are recorded in the National Patient Registry, and reimbursement is contingent on accurate registration by departments. Diagnostic coding is registered in accordance with the 10 th edition of the International Classification of Diseases (ICD), and procedural coding is registered in accordance with the Nordic Medico-Statistical Committee Classification of Surgical Procedures [19,20]. The predicative value of diagnostic codes included in the Charlson Index Score has recently been validated with excellent results [21]. Additionally, chronic dialysis treatment and renal transplantations are recorded in the validated Danish National Registry on Regular Dialysis and Transplantation [22]. Dispensation of prescription medication by Danish pharmacies is recorded in the Danish Register of Medicinal Product Statistics [23]. The register records the Anatomical Therapeutic Chemical Classification System (ATC) code, quantity, strength and date of all prescriptions dispensed in Danish pharmacies. Prescription medication is partially reimbursed by the public healthcare system, and the register has Funding from commercial sources, i.e., Bristol-Myers Squibb did not alter the authors' adherence to PLOS ONE policies on sharing data and materials. The opinions, results, and conclusions in this paper are solely representative of the authors, and are fully independent from funding sources, and no conflicts of interest exist. Additionally, the results presented in this paper have not been published previously in whole or part, except in abstract format. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript, and the opinions, results, and conclusions in this paper are solely representative of the authors, and are fully independent from funding sources.
been validated with excellent results [23,24]. Limited biochemistry was available for a fraction of the study population; results were accessible from multiple laboratories across Denmark. Finally, information relating mortality and causes of death was determined via the National Registry of Causes of Death [25].

Study Population
Dialysis-requiring AKI was defined as treatment with acute RRT in patients without previously registered chronic RRT. All patients requiring acute RRT between 1 st January 2005 and 31 st December 2012 were identified in the Danish National Patient Registry. Patients <18 years old, patients migrating within one year of treatment, patients previously treated with acute RRT, and patients previously registered with chronic RRT were excluded. All patients were followed until an endpoint or the end of the study. A flow chart depicting study design is shown in Fig 1.

Study Outcomes
Primary outcomes were defined as all-cause mortality and ESRD at one year. Secondary outcomes were defined as estimated glomerular filtration rate (eGFR) (plasma creatinine measurements recorded 52-104 weeks following index) in survivors with preserved renal function at one year, duration of dialysis-requirement, and duration of hospital admission.

Study Covariates
Comorbidities were identified based on diagnostic codes found in the National Patient Registry. Sensitivity for hypertension, and diabetes was augmented through the inclusion of information pertaining to outpatient medication. Preceding eGFRs were calculated as means all available plasma creatinine measurements recorded 52 to 2 weeks prior to index in accordance with the Chronic Kidney Disease Epidemiology Collaboration equation [26]. Azotemic parameters registered one day prior to index were included where available. Major cardiothoracic, gastric and orthopedic surgical interventions were identified based on procedural codes found in the National Patient Registry. Post-procedural dialysis-requirement was defined as acute RRT within 14 days of a major surgical intervention. Dialysis modalities were identified based on procedural codes found in the National Patient Registry as continuous RRT (CRRT), acute intermittent hemodialysis, or acute peritoneal dialysis. Similarly, requirements of mechanical ventilation and/or circulatory support were also identified based on procedural codes found in the National Patient Registry. Outpatient medications as baseline were identified in the prescription database. A comprehensive list of the medications, relevant codes, and algorithms are provided in S1 Appendix.

Statistical Analysis
Outcomes following dialysis-requiring AKI were compared between the periods; for known risk factors associated with prognosis. As such, the models were adjusted for patient age, gender, dialysis modality, comorbidity, surgery, sepsis, mechanical ventilation, and circulatory support. Comorbidities included in the model were; non-insulin dependent and insulindependent diabetes, congestive heart failure, ischemic heart disease, cardiac arrhythmia, valvular heart disease, peripheral vascular disease, stroke, chronic obstructive pulmonary disease, solid and non-solid cancer, and chronic kidney disease. Preceding surgeries included in the model were any major cardiothoracic, gastric or orthopedic procedures. Furthermore, odds ratios for one-year mortality and ESRD were also computed in various selected subpopulations. These included; gender-specific strata, age-specific strata, modalityspecific strata, non-surgery-related and surgery-related strata, non-intensive care requiring and intensive care requiring strata, and Charlson Index Score stratified (<3, 3-6, and >6). In additional analyses, odds ratios were calculated for 90-day mortality and 90-day renal survival using multiple logistic regression analyses analogous to the principal analyses. As incidence of dialysis-requiring AKI in the general population was lowest in 2005, sensitivity analyses with exclusion of patients from the aforementioned year were performed to confirm the principal results; no differences were discernable. Finally, residual renal function was assessed in renal survivors by comparing the proportionate number of patients with significant (>20%) decrease in eGFR 90-365 days following index. Retrospective register based studies do not need ethical approval in Denmark; however, the Danish Data Protection Agency has approved use of data (ref. 2007-58-015 / I-suite nr 00916 GEH-2010-001).

Results
A total of 13,819 patients with dialysis-requiring AKI were identified between January 1 st 2005 and December 31 st 2012. Although patient characteristics principally remained stable, use of CRRT and mechanical ventilation increased over time. Period-specific baseline characteristics are shown in Table 1.

Discussion
In this nationwide retrospective study on outcomes following dialysis-requiring AKI, prognosis was observed to improve incrementally during an eight year study period. Specifically, the odds ratio of both all-cause mortality and ESRD was observed to decrease by~20% between 2005 and 2012. Simultaneously, durations of dialysis-requirement and hospitalization decreased. Irrespectively, prognosis following dialysis-requiring AKI remains unfavorable, with only approximately one third of patients surviving without requiring chronic RRT. AKI is associated with significant mortality, with one-year mortality rates fluctuating from~35% tõ 70% dependent on population and AKI severity [4,[27][28][29]. However, mortality rates have been reported to be decreasing [13,15,16,30,31], although the attributable risk for death associated with AKI may be increasing [32]. Additionally, AKI survivors remain at increased risk of ESRD. Reported rates of ESRD range from 2% to 20% [3,5,8,15,33,34]; however, risk of ESRD following dialysis-requiring AKI-as is the case with mortality-could be decreasing [15]. Regardless, our results indicate incremental improvement in prognosis following dialysisrequiring AKI from 2005 till 2012.
Although initiation of acute RRT in the context of imminent indications is associated with poorer outcomes [35], evidence supporting early initiation of acute RRT in AKI remains wholly insufficient [36]. Accordingly, initiation of acute RRT continues to be based on imminent indications with no apparent shift towards earlier initiation of therapy [37,38]. Our results corroborate this observation, as we were unable to document any substantial change in pre-dialysis azotemia or serum lactate. Furthermore, incidence rates of non-dialysis-requiring AKI have increased throughout the last decade, while incidence rates of dialysis-requiring AKI have remained stable [14]. As such, no existing evidence clearly substantiates changing thresholds for initiation of acute RRT.
Although no novel interventions pertaining to dialysis-requiring AKI have demonstrated genuine benefit throughout the last decade [39][40][41][42][43], our results nonetheless provide further evidence of increments in survival and renal recovery following severe AKI. Use of CRRT increased throughout the study period. Dialysis-modality has previously been demonstrated to be associated with outcomes in observational data [38]; however, the association is plausibly driven by selection biases, and no evidence currently supports any benefit associated with a specific modality [44][45][46][47][48]. Improvements in prognoses could alternately represent the beneficial effects related to implementation of evidence-based goal-directed therapy in critical illness [49,50]. Outcomes following sepsis and septic shock are known to be improving [51,52], possibly due to the benefits associated with the surviving sepsis guidelines [53], and outcomes following acute respiratory failure may also be improving [54]; albeit, thresholds for use of mechanical ventilation could be changing [55]. As such, the observed improvement of outcomes following dialysis-requiring AKI remains plausible, although possibly driven by interventions targeting supportive and non-kidney related care.
Our results confirm the detrimental consequences associated with severe AKI. Initial survivors of dialysis-requiring AKI remain at substantially increased risk of ESRD. Tentatively, our results indicate an improvement in residual eGFR in non-ESRD survivors; however, as preceding and succeeding eGFRs were only available in a small fraction of our population, the analysis was both underpowered and plausibly subject to certain biases. While outpatient care holds potential in identifying CKD in time to mitigate complications, only a fraction of patients with dialysis-requiring AKI are currently referred to a nephrologist following discharge [56,57]. Notably, referral to a nephrologist following discharge is associated with interventions [58], and improved outcomes in retrospective cohorts [59].
Our study had numerous strengths. First, due to the structure of public health care in Denmark, national registries record comprehensive information pertaining to medical care of all Danish citizens. Follow-up is generous and essentially unflawed, and national registries are extensively validated. Second, the availability of reliable data from the Danish National Registry on Regular Dialysis and Transplantation ensures an accurate dissemination between dialysis-requiring AKI and acute RRT in ESRD. Additionally, the accuracy of CRRT in the intensive care setting has previously been validated with excellent results [60]. However, our study also had a number of limitations. First, correlation is not causation; due to the observational design, our results do not provide definite answers related to cause. Second, the absence of systematic and universal laboratory and clinical data is regrettable; consequently, a number of possible confounders remain unaddressed. Specifically, the limited data on biochemical and clinical indications for initiating acute RRT remains unfortunate; particularly as the availability of data pertaining to baseline azotemia was underpowered for definite conclusions regarding clinical implications. Furthermore our algorithm for identification of dialysis-requiring AKI does not rule out alternate indications for acute RRT to AKI. Additionally, prevalence of comorbidity may be underestimated due to a reliance on diagnostic and procedural coding. Finally, due to the particular demographics of Denmark, results cannot necessarily be extrapolated to other populations.

Conclusions
In a nationwide retrospective study on outcomes within the first year following dialysis-requiring AKI, we observed incremental decrease in odds ratios for all-cause mortality and ESRD between 2005 and 2012. Although mortality of patients with greatest vulnerability and women remained unchanged, an overall improvement in outcomes following dialysis-requiring AKI of all patients was observed. Although interpretation of our results is subject to limitations pertaining to residual confounding, dialysis-requiring AKI nonetheless continues to be associated with disheartening prognoses, with current one-year mortality and ESRD rates exceeding 50% and 10%, respectively.