INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

There have been multiple retrospective and prospective studies of the epidemiology of acute renal failure (ARF) (1). Most of them rely on arbitrary biochemical definitions of ARF (2). Some studies focus on patients outside of the intensive care unit (ICU) (4, 5) and are, therefore, of limited relevance to critical care physicians. Only a few studies provide information about the overall population incidence of this condition (4). Still fewer studies (6) focus, even in part, on the epidemiology of ARF requiring dialysis within the ICU (severe acute renal failure of critical illness). This lack of information is unfortunate, because the need for both dialysis and ICU care of defines a specific group of critically ill patients who may have a particularly poor prognosis (6, 9, 10) and who consume vast amounts of resources (11). Information on the overall incidence, style of management, and patient outcome would be useful in assessing current therapeutic approaches, research and resource allocation, and future therapies. Furthermore, much controversy exists about the type of renal support that should be used in critically ill patients with severe ARF (12). Intermittent hemodialysis, prescribed by the nephrology team and applied by a dialysis nurse, has been and still is the main form of acute renal replacement therapy in many countries. More recently, however, this practice has been challenged by the use of continuous renal replacement therapies (13). These therapies are often prescribed by the critical care physician and applied by the intensive care nurse. They are prescribed within a care delivery system commonly referred to as the "closed" ICU model. In such a setting, the critical care team is in direct control of therapy and input from the nephrology unit is often minimal or absent. In the State of Victoria, Australia, such a system has been operating for more than a decade providing an ideal environment for the assessment of its achievements and shortcomings. Such information has important public health implications related to the overall model of delivery of critical care, patient outcomes, its costs, and physician and nurse remuneration. Accordingly, we have conducted a prospective, multicenter study to (1) define the epidemiology of ARF requiring acute renal replacement therapy in the ICU, (2) study its style of management in a "closed" ICU system, and (3) describe patient and organ outcomes in such a system of care.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study was conducted over a 3-mo period (September 1, 1996- November 30, 1996) in 24 intensive care units across the State of Victoria. Because the study did not involve any intervention and because of its confidentiality and anonymity, the investigation was considered an audit of statewide practices. As is the case for auditing activities solely involving access to information contained in medical records and requiring no patient or next-of-kin contact, Ethics Committee approval was not required.

We enrolled all patients with severe ARF admitted to the participating hospitals during the study period. Severe ARF was defined as any degree of ARF, which, in the opinion of the treating physician, required the commencement of renal replacement therapy.

We divided patients into three groups: (1) those requiring acute renal replacement therapy outside the intensive care unit, (2) those requiring such therapy within the intensive care unit, and (3) those who required such therapy within the intensive care unit but who already were dialysis dependent.

Patients from groups 1 and 3 were identified and counted for epidemiological purposes but not studied further. Detailed information was prospectively obtained by the investigators in the various centers on patients from group 2.

A case report form was developed for the purpose of the study and the following information was obtained: Name of hospital, age of patient, sex, date of admission to hospital and ICU, diagnostic category (from an Australian and New Zealand Intensive Care Society-modified APACHE III diagnostic category list), premorbid renal function, and most recent serum creatinine before hospital admission. In the first 24 h of ICU admission, Acute Physiology and Chronic Health Evaluation (APACHE II) (14) score, Simplified Acute Physiology Score (SAPS II) (15), and serum creatinine value were obtained. The main cause of acute renal failure was indicated from a group of seven possible choices (hypotension/ischemia, sepsis, septic shock, rhabdomyolysis, nephrotoxic, radiocontrast, other) according to the judgment of the clinician. Information was obtained prior to the start of renal replacement of therapy. Such information included the use of mechanical ventilation, the use, type, and amount of inotropic/vasopressor drugs, the presence of sepsis (ACCP consensus criteria), the serum creatinine and bilirubin values, the Glasgow coma score, the presence of persistent hypotension, and urinary output for the preceding 24 h. Such information was used to calculate a previously described and validated ARF-specific prognostic score (16).

We obtained specific information regarding the following items: (1) starting date of acute renal replacement therapy, (2) the reason for its initiation, (3) the type of renal replacement therapy used, (4) the site and type of vascular access, (5) the type of filtering or dialyzing membrane (6) the type of anticoagulation and the type of complications associated with renal replacement therapy, and (7) the peak serum creatinine concentration during renal replacement therapy. Information was also obtained on the type of physician in charge of renal replacement therapy and patient management, on which nurse was responsible for the set-up and running of renal replacement therapy, and whether the nephrology unit was consulted to assist in the care of the patient. Data were obtained on ICU patient outcome, number of days of renal replacement therapy, organ outcome at ICU discharge, time on the ventilator, and duration of ICU stay. Finally, duration of hospital stay and survival to hospital discharge were also recorded.

Population details for the State of Victoria were obtained from the latest census figures (1996) by the Australian Federal Government. The State of Victoria is the second most populous state in Australia. It is located in the southeastern part of the country and has Melbourne as its capital. Its size is approximately half that of California. Of a population of 4.75 million people, 3.75 million live in Melbourne and its suburbs. All patients with ARF requiring dialysis are either treated in larger regional centers or referred to hospitals in Melbourne. The provision of medical care in Australia is regulated by the Federal Government and operates within a mixed system of freely available national health care and private insurance-financed care. The unique features of this system have been recently reviewed in the American literature (17).

Calculations and Statistics

From the SAPS II scores of each patient, individual predicted risk of death was calculated. With the clinical information obtained prior to the start of the renal replacement therapy, an acute renal failure-specific prognostic index was calculated for each patient with its corresponding estimated risk of death (4). Descriptive statistics and comparisons were performed using the statistical analysis package of the spreadsheet software (Microsoft Excel 1997; Microsoft Corporation, Redmond, WA). Comparisons between survivors and nonsurvivors were performed using either the chi-square test for nominal variables or Student's t test for numerical variables. Data are presented either as actual numbers, with percentages within parentheses, or as means with 95% confidence interval.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

One hundred sixteen adult critically ill patients developed severe acute renal failure requiring acute renal replacement therapy during the 3-mo study period. In addition, seven patients with dialysis-dependent renal failure who were admitted to the ICU for advanced life support also received acute renal replacement therapy. Six of these patients were treated with continuous renal replacement therapy while in the ICU and four survived to discharge from the hospital. In all of these patients, vascular access was by means of double-lumen catheters and, in five, management was under the combined supervision of the nephrology unit and the intensive care team. These patients were excluded from further analysis.

At the same time, 38 patients received renal replacement therapy for acute renal failure outside the ICU. All renal replacement therapy outside of the ICU was in the form of intermittent hemodialysis. However, 19 of these patients were actually continuing renal replacement, which had been started while in ICU. Therefore, only 19 ward patients with severe ARF were "new" and had never received ICU care. Thus, a total of 135 patients developed severe acute renal failure over a 3-mo period. Given an adult (older than 18 yr of age) population of 3.45 million, the incidence of adult ARF requiring dialysis in the state of Victoria is 13.4 cases/100,000 people/yr.

The clinical features of the 116 patients admitted to the ICU with severe acute renal failure are summarized in Table 1. This table highlights the high level of illness severity in this cohort and the need for mechanical ventilation and inotrope/ vasopressor support in the vast majority of patients. A list of the main diagnoses, which led to ICU admission and subsequent or associated severe acute renal failure, is presented in Table 2. Such a list demonstrates that the three main triggers for ICU admission were severe sepsis/septic shock, cardiothoracic surgery, and myocardial pump failure.

Admission to the ICU occurred, on average, after 4.1 (95% CI 2.8-5.4) d in the hospital. Acute renal replacement therapy was started at a mean of 2.5 d (95% CI 1.7-3.3) after ICU admission. According to the treating clinician, the major but not the only reason for starting renal replacement therapy was severe and persistent oliguria or anuria in 72 (62%) patients, uncontrolled uremia in 26 patients (22.4%), fluid overload in seven patients (6%), hyperkalemia in four patients (3.5%), severe acidosis in three patients (2.6%), and multifactorial in four patients (3.5%).

A double-lumen catheter was used for vascular access in all cases. The femoral position was chosen in 80 (69%) patients, the subclavian position in 29 (25%) patients, and the jugular in seven patients (6%). The AN69 polyacrylonitrile membrane (Hospal, Lyon, France) was used in 107 (92.2%) patients and the FH88 polyamide membrane (Gambro, Lund, Sweden) was used in nine (7.8%) patients. The approach to circuit anticoagulation was quite variable: 36 (31%) patients received systemic anticoagulation with heparin, 20 (17.2%) patients received regional heparinization (prefilter heparin with postfilter protamine), 14 (12%) patients were treated with low-dose heparin (< 10 IU/kg/h), 24 (20.7%) patients received no anticoagulation, and the remaining 22 (19.1%) patients received other forms of anticoagulation (low-molecular-weight heparin, heparinoids, and/or prostacyclin).

One hundred ten patients were treated with continuous renal replacement therapy. Of these, 51 (46.4%) received convective therapy as continuous venovenous hemofiltration (CVVH) and 59 (53.6%) received mostly diffusive therapy either as continuous venovenous hemodiafiltration or as continuous venovenous hemodialysis (CVVHDF or CVVHD). There were no differences in outcomes between these two modes of continuous renal replacement therapy. Three patients received intermittent venovenous hemofiltration (8 h/d) and three other patients received intermittent hemodialysis. Four patients who were treated with CVVHD for most of their ICU stay were switched to intermittent hemodialysis in the last few days of their ICU admission to facilitate patient mobilization.

Twenty-three patients (19.8%) experienced complications. In 13 patients (11.2%), there was therapy-induced hypotension (one of the three patients treated with intermittent hemodialysis, 12 among the rest of the patients). In six (5.1%) patients, vasopressor drugs had to be increased at the time of start of renal replacement therapy. Four (3.4%) patients experienced bleeding requiring transfusion. This bleeding was attributed to the anticoagulant therapy used during renal replacement therapy. There were no reported episodes of line-related sepsis associated with renal replacement therapy.

All patients were treated in "closed" ICUs. The critical care physician was in charge of patient care in all cases and the critical care nurse set up and operated the renal replacement therapy circuit in all but three patients. The nephrologist was consulted in only 30 (26%) cases. There was no difference in outcome between the patients who received a nephrology unit consult and the 86 who did not.

The major outcome variables under study included duration of renal replacement therapy, duration of mechanical ventilation, ICU and hospital length of stay, recovery of renal function to dialysis independence, and ICU and hospital survival. The findings related to such outcomes are summarized in Table 3. Using the SAPS II score, the predicted in-hospital mortality was 59.6%. Using a recently validated, acute renal failure-specific (16) prognostic score (Liano's score), mean predicted in-hospital mortality was 59%. However, 59 patients survived to hospital discharge, representing an actual mortality of 49.2%.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

There have been many studies of the epidemiology of acute renal failure (1). However, in most cases, the definition of acute renal failure rests on arbitrary biochemical cut-off points (2). These biochemical dividing lines vary from study to study making comparisons difficult. Biochemical subdivisions also have no clear association with outcome. On the other hand, the need to start renal replacement therapy (severe ARF) defines as a quantum leap in illness severity, cost of care, and complexity of treatment, which has been demonstrated to be associated with a poor prognosis (7, 9). Furthermore, the combined need for dialysis and ICU care defines a group of patients with an even poorer prognosis (6, 10, 18). These patients are of particular interest to the critical care physician because they pose major management challenges, use vast resources, and require expensive technology. It is highly controversial who should be in charge of the care of these patients and who should prescribe and conduct renal replacement therapy. It has been argued that such patients should be cared for by an intensivist in a "closed" ICU system (19). On the other hand, it has also been argued that nephrological management is essential for safe and effective care (20). Unfortunately, to date, there has not been a prospective study of the epidemiology of severe ARF in the ICU and of its management and outcome in a "closed" ICU system. Such information has powerful public health, medical training, and the therapeutic implications. Accordingly, we conducted such a study in the state of Victoria, where all ICUs operate within a "closed" system.

First, this investigation sought to address the epidemiology of severe ARF by prospectively studying its incidence in the State of Victoria, Australia. The study's findings highlight several previously unreported or underemphasized facts. First, in a developed country, severe ARF is no longer a disease of nephrology wards: it is now a disease, which, because of its association with multiorgan dysfunction, is seen and managed in the intensive care unit. In our study, 90 patients were treated solely in the ICU, 19 in the ICU and nephrology ward, and 19 in a nephrology ward only. These findings are in keeping with a recent prospective study (21) and demonstrate a major epidemiological shift. This redistribution from the wards to the ICU has important implications regarding physician and nurse workload, resource allocation, physician training, control of overall patient management, and expertise development.

Second, severe ARF in Victoria has an incidence of 13.4 cases/100,000 adults/yr. This incidence is greater than that recently reported for acute respiratory distress syndrome (ARDS) (20). ARDS, however, attracts more research and publications in the intensive care literature. The knowledge that severe acute renal failure is a disease of critically ill patients, is more common than ARDS, has a similar mortality, and is treated by intensivists may change this imbalance.

Third, many patients who develop severe ARF have premorbid renal dysfunction. These patients are at particular risk of remaining dialysis dependent (9 of 11 patients who survived and remained dialysis dependent had premorbid renal dysfunction). This insight demands that a particular strategy of careful and aggressive monitoring and renal protection be put in place. Such strategy applies in particular to patients with impaired renal function who undergo surgery, develop a septic state, or are exposed to any potential threat to renal homeostasis.

Fourth, sepsis and septic shock are the major cause of severe ARF in the ICU. This information highlights the fact that any successful way of decreasing the incidence of severe ARF is likely to be based on the development of more effective therapies for the prevention or rapid treatment of sepsis. Strategies that do not confront this issue are likely to fail.

Finally, isolated severe ARF is uncommon in the ICU. More than 80% of patients with severe ARF of critical illness have associated respiratory and circulatory failure (multiorgan dysfunction syndrome). It is unlikely that therapies exclusively directed at ARF will have a major impact on patient outcome, because they do not address the disease process of which severe ARF is but one clinical expression.

An important finding of our study concerns the style of management of severe ARF. In this regard, we found that in Victoria, critical care physicians and critical care nurses fully control the management and renal replacement prescription and application for the majority of patients with severe ARF while in the ICU. They almost completely rely on continuous renal replacement therapy. Nephrology unit opinion is sought in only a small group of patients. Victorian intensivists initiate renal replacement therapy early (within 2.5 d of ICU admission and with a mean serum creatinine concentration of 311 µmol/L). In more than 70% of cases, severe persistent oliguria is the major trigger for starting artificial renal support. Such behavior suggests a strong desire to prevent, rather than simply treat, fluid overload and excessive uremia. It is also in keeping with recent suggestions about the need to change the indications for dialysis in ICU patients (22). Once continuous renal replacement therapy is started, tight uremic control is pursued, as evidenced by a mean peak creatinine concentration of only 396 µmol/L during treatment. The potential benefits of tight uremic control in patients with acute renal failure have recently been highlighted (23, 24).

The third aim of our study was to describe the outcome of severe ARF in a "closed" system. The comparison of actual to predicted outcome was particularly important because the style and system of patient care ("closed" ICU, intensivist-controlled ARF management) practiced in Victoria are different from many centers in North America and, therefore, likely to come under close scrutiny. Critics of the Victorian (and indeed Australian) system will argue that an approach to ARF care that excluded nephrology unit input would inevitably result in increased morbidity and mortality. The findings of our study, however, do not support this contention. On the contrary, they suggest that satisfactory patient outcomes can be achieved with the "closed" ICU approach to the care of ARF patients. Nonetheless, we recognize that a randomized study would be needed to fully test the proposed advantages or disadvantages of a "closed" system. Such a study, however, would be extraordinarily difficult to conduct because "systems" cannot be readily changed across an entire state. In-hospital mortality in our study, at 49.2%, is among the lowest reported in the English literature for a cohort of ICU patients requiring dialysis in the last decade. Shaefer and coworkers (8) reported a 56.7% ICU mortality for their patients, Cosentino and coworkers (18) a 79% ICU mortality, and Schwilk and coworkers (26) a 58% ICU mortality (no in-hospital mortality available). The ICU mortality for our cohort was only 41.4%. The in-hospital mortality for ICU patients requiring dialysis has been the focus of only two studies during the last decade. Chertow and coworkers (6) reported an in-hospital 70% mortality and Douma and colleagues (10) reported an in-hospital mortality of 76%. In the first study, illness severity was less than in our population according to APACHE II scores. Recently, Brivet and colleagues (7) studied acute renal failure in the ICU in order to develop specific prognostic markers. Although their study did not specifically focus on severe ARF requiring dialysis, they reported that the overall mortality for those ARF patients who had been treated with renal replacement therapy was 64%. Equally, a recent study by Liano and coworkers (21) obtained prospective data on ARF requiring dialysis. It was found that 66.2% of cases of such severe ARF occurred in ICU patients and that mortality in these patients was 79%. In contrast to the above findings, the in-hospital mortality in our study was only 49.2%. Interestingly, only two studies have reported a mortality of less than 50% for critically ill patients with severe ARF this decade. Both are series reports from single institutions using continuous renal replacement therapy in a "closed" ICU system (27, 28).

Another important measure of quality care is "renal outcome" (the percentage of survivors who were discharged from hospital free of dialysis). In our cohort, only 18.6% of survivors were dialysis dependent at discharge. This finding is in contrast to a 34.3% incidence of dialysis dependence in Cosentino's patients (18) and a 33% incidence in Chertow's group (6). It is important to note that the patients in these two studies had almost exclusively been treated with intermittent hemodialysis (91.4%). When the cumulative proportion of survivors recovering dialysis independence in these two groups is compared to our results, the difference is significant (z value = 2.107, p = 0.036). These considerations support the notion that continuous renal replacement therapy may facilitate renal recovery when compared to intermittent hemodialysis. They also reinforce the contention that intensivist-controlled management of severe ARF within a "closed" ICU system is safe.

Patient treatment appeared to occur early in the course of severe ARF (serum creatinine just above 300 µmol/L) and to be triggered by severe oliguria/anuria. This approach to ARF management is often criticized. It is argued that patients who in fact have prerenal dysfunction are thus unnecessarily exposed to renal replacement therapy. If critical care physicians would just wait, several of their patients would improve and not require renal replacement. It is also argued that by including such patients, who do not have "true" severe ARF, outcomes are artificially improved. Both arguments are not supported by fact. All patients treated with CRRT in our study were oliguric (< 400 ml/d of urine output) or anuric for at least more than 3 d after initiation of therapy. No patients were placed on CRRT for fluid control only. Persistent oliguria/anuria occurred despite full, aggressive, and invasively monitored fluid and inotropic support. Close to 80% of patients were on vasoactive drugs and mechanical ventilation at the time of initiation of renal replacement therapy. Renal replacement was needed for an average period of 6 d. These patients clearly had "true" severe ARF. Furthermore, contrary to criticism, by treating such patients early, mortality may have actually been artificially inflated compared to a "waiting" strategy. A "waiting" strategy, in fact, would have allowed irretrievably ill patients to die (and thus not be counted) without dialysis, thereby decreasing mortality figures for dialyzed patients. Finally recent data in severe ARF associated with trauma support the view that early intervention with continuous therapy may improve survival (29).

It is possible that our patients had a particularly high incidence of chronic renal dysfunction, which could explain the difference between our outcome findings and those of others. However, premorbid renal dysfunction was present in 32.8% of our patients versus 39% of Liano's patients (21). Furthermore, premorbid renal dysfunction appeared to be associated with improved rather than worsened outcome (21). It is also possible that the illness severity scores used in our study provided an inaccurate assessment of predicted mortality. For this reason, we used two scores (SAPS II and the Madrid score), a general score and an ARF-specific score. They predicted a very similar mortality. Although the former score may be criticized, the latter has been repeatedly reported to perform very well (4, 10, 16, 21) in ARF patients. Using this score, Liano recently reported a standardized mortality ratio (observed/predicted mortality) of 1.12 for ICU patients with severe ARF in Madrid (21) with an overall mortality of 79%. Using the same scoring system, the standardized mortality ratio for our patients was 0.83 (overall mortality of 49.2%). From an epidemiological point of view, it is interesting to note that in the comprehensive study by the Madrid group, the incidence of severe ARF (ARF treated by dialysis) was 8.6 cases/100,000 people/ yr, a value similar to ours. However, only 5.7 were treated in ICU versus 9.8 in Victoria. In our view, this finding most likely reflects the difference in availability of ICU beds in Spain compared to Australia. Differences in health care rationing are likely to also explain the difference between the findings of Liano and coworkers and ourselves compared with those of Feest and coworkers (2) in the United Kingdom, who reported a yearly incidence of acute dialysis of only 2.2/100,000. The availability of ICU beds in the United Kingdom is currently approximately half that of Australia. No figures are available from North America.

It is also important to emphasize that in this study, the use of a "closed" ICU system approach to the management of severe ARF covaried with the use of continuous hemofiltration techniques. This covariance makes it impossible to discern whether the favorable results demonstrated in our investigation are due to one or the other or both. The goal of our study was to establish whether a system operating within the framework described previously is safe. Understanding its advantages and disadvantages will require further study. This investigation was also not designed to address the characteristics and dynamics of nephrological intervention. Thus, one of its limitations is related to its inability to define when nephrological consultation was obtained, what suggestions or management options were offered, whether such suggestions led to changes in management, or whether any "renal" diagnoses were missed entirely or identified late because of delayed or absent renal consultation. Although such information would be of great interest, it was outside of the aims of the study. Furthermore, it would not have applied to the vast majority of patients who, in fact, never had any nephrological input.

In conclusion, we found that treating severe ARF within a "closed" ICU system with near exclusive intensive care team control of management and renal replacement therapy was safe. Furthermore, this approach resulted in organ and patient outcomes, which compare favorably with those reported in the literature and are better than predicted by illness severity scores. Our findings raise the important issue of "system effect" and "practice style effect" on patient outcome. They also raise issues related to physician training and credentialling, resource allocation, research priorities, and the future role of critical care physicians and nurses in this important field of medicine.