Professional Documents
Culture Documents
Background. Acute kidney injury can be defined by a Results. Acute kidney injury developed in 1,061 of
fall in urine output, and urine output criteria may be 4,195 patients (25%). Urine output had moderate discrim-
more sensitive in identifying acute kidney injury than ination in predicting subsequent acute kidney injury
traditional serum creatinine criteria. However, as (C statistic [ .637 ± .054). Lower urine output and longer
pointed out in the Kidney Disease Improving Global duration of low urine output were associated with greater
Outcome guidelines, the association of urine output odds of developing acute kidney injury and death.
with subsequent creatinine elevations and death is Conclusions. We found that there is similar accuracy in
poorly characterized. The purpose of this study was to using urine output corrected for actual, ideal, or adjusted
determine what degrees of reduced urine output are weight to discriminate future acute kidney injury by
associated with subsequent creatinine elevation and creatinine elevation and recommend using actual weight
death. for its simplicity. We also found that low urine output is
Methods. This was a retrospective cohort study of adult associated with subsequent acute kidney injury and that
patients (age ‡18 years) cared for in a cardiovascular the association is greater for lower urine output and for
intensive care unit after undergoing cardiac operations in low urine output of longer durations. Low urine output
a tertiary care university medical center. All adult pa- (<0.2 mL $ kg–1 $ h–1), even in the absence of acute
tients who underwent cardiac operations and were not kidney injury by creatinine elevation, is independently
receiving dialysis preoperatively were studied. The associated with mortality.
development of acute kidney injury was defined as an
increase in creatinine of more than 0.3 mg/dL or by more (Ann Thorac Surg 2016;-:-–-)
than 50% above baseline by postoperative day 3. Ó 2016 by The Society of Thoracic Surgeons
(e.g., average vs. persistent reduction for the period The occurrence of low UO (<0.5 mL $ kg–1 $ h–1) was
specified) [9].” investigated by a member of the ICU team. If low UO was
Although there is overlap, a discordance frequently thought to be from hypovolemia, based on physical ex-
exists between patients diagnosed with AKI-Cr and AKI- amination and central and systemic blood pressures, a
UO. In these situations, whether AKI-Cr and AKI-UO fluid bolus of albumin or crystalloid was given. Hetas-
have different or similar prognoses is unclear [6, 10–12]. tarches were not used. If sufficient anemia (hematocrit of
Several recent studies have shown that AKI-UO is asso- w25% at the start of the study, decreasing to 20% by the
ciated with increased death [13–15]. However, two of end of the study) was present, red blood cell transfusion
these studies each only examined a single threshold of would be given. If the low UO was thought to be related
low UO (<7.2 mL $ kg–1 $ day–1 [13] and <500 mL/day to low cardiac output, a fluid bolus would be given if
[15]), and one required Cr exceeding 3.5 mg/dL to define there was associated hypovolemia, otherwise an inotrope
AKI [15] and were thus unable to establish a threshold or would be added or increased. If associated with hypo-
dose-dependency between AKI-UO and AKI-Cr and be- tension (mean arterial pressure <60 to 65 mm Hg) and a
tween AKI-UO and death. None of the studies evaluated normal or elevated cardiac output, then depending on
whether the effect of different types of weight (actual, presumed volume status, a fluid bolus or a vasopressor
ideal, or adjusted) should be used to adjust UO. Indeed, a would be used. Finally, if low UO was present despite
more comprehensive examination of the prognostic effect
of AKI-UO requires a large database with complete UO
and follow-up data. Table 2. Characteristics of Patients With and Without Acute
The purposes of this study were to determine (1) what Kidney Injurya
type of weight (actual, ideal, or adjusted) should be used
to adjust for hourly UO, (2) what, if any, level of UO is a AKI—Yes AKI—No
(n ¼ 1,061) (n ¼ 3,134)
harbinger of AKI-Cr, and (3) is AKI-UO associated with
death in patients after cardiac operations? Characteristic No. % No. % p Value
Fig 1. Urine output threshold and acute kidney injury. Heat map shows adjusted odds ratios of developing at least Kidney Disease Improving
Global Outcomes stage 1 acute kidney injury by creatinine criteria based on urine output meeting the hourly threshold averaged (A) over the several
hours (mean) or (B) below that value for all hours (consecutive). Each box shows the odds ratio (95% confidence interval) and the number of
patients who developed acute kidney injury by creatinine/the number of patients whose urine output met the threshold value for that many hours.
The red outline shows boxes where the odds ratio is statistically significant.
adequate hemodynamics and volume status, a loop inotropes, and hourly UO was no longer necessary.
diuretic (furosemide, 20 to 80 mg intravenous push, or Catheters and hourly UOs were collected for at least 24
bumetanide, 1 to 2 mg intravenous push) would be given. hours in 69% of patients and for at least 48 hours in 25%
of patients. When clinical symptoms were suggestive of
urinary retention, such as prior urinary clots and a
Data Collection sudden decrease in UO, catheters were flushed and irri-
Patient demographics, preoperative comorbidities, and gated, as necessary.
type of cardiac operation were recorded. Laboratory data
were collected from the preoperative period through
postoperative day 7 and accessed from the data ware- Outcomes
house, an electronic repository of patient information. In AKI was explored using both Cr-based and UO-based
the ICU, UO was routinely measured hourly until the criteria. AKI-Cr was defined using KDIGO criteria
bladder catheter was removed and recorded in the elec- (stage 1: >0.3 mg/dL increase in Cr or >50% increase,
tronic medical record (Centricity; GE Healthcare IT, stage 2: >100% increase, stage 3: >200% increase) using
Barrington, IL). Bladder catheters were maintained until the highest Cr value by the morning of postoperative day
the patient was liberated from mechanical ventilation, 3. The last preoperative Cr value was used as the baseline.
hemodynamically stable without vasopressors or For multivariable analyses, the different stages of AKI-Cr
4 ENGOREN ET AL Ann Thorac Surg
URINE OUTPUT AND CREATININE ELEVATION 2016;-:-–-
Fig 1. (Continued).
were collapsed into one group as the dependent variable. and Acute Kidney Injury Network (AKIN) definitions all
UOs were summed for each day for all possible hourly used thresholds to define AKI-UO [9, 16, 17]. To answer
intervals (1, 2, 3, . , 11, 12) and adjusted by weight. To the second question of how urine volume criteria should
answer the first question of which weight to use, we be applied (eg, average vs persistent reduction for the
applied three different patient weights (kg): (1) actual; (2) period specified), UOs were determined two different
ideal, calculated as height (cm) [0.9 – 88], if male, or ways: by consecutive hours (all UOs for those hours were
height (cm) [0.9 – 92], if female; and (3) adjusted, less than the specified threshold) and mean (although the
calculated as ideal body weight þ 0.4 [actual weight – hourly UO could be greater than the threshold, the mean
ideal weight]. We used nine different volume thresholds UO over the hours was less than the specified threshold).
for UO (0.1, 0.2, ., 0.9 mL $ kg–1 $ h–1). This produced A patient who had at least one episode of UO below the
108 UO-time interval combinations for each of the three weight-adjusted threshold for the time period was
weight definitions. deemed to have met the threshold for that day. If the time
Death was defined using The Society of Thoracic Sur- period extended across midnight, the partial time period
geons definition of death within 30 days of the operation before midnight was analyzed as part of the prior day,
or longer if the patient was still hospitalized from the and the total time period was analyzed as part of the
index operation. succeeding day.
We next analyzed UO based on thresholds, as KDIGO To ensure a directional association between UO and the
and prior Risk, Injury, Failure, Loss, End-Stage (RIFLE) subsequent development of AKI-Cr, once AKI-Cr
Ann Thorac Surg ENGOREN ET AL 5
2016;-:-–- URINE OUTPUT AND CREATININE ELEVATION
developed, all further UOs were censored and not prominent with longer durations. These general relation-
included in the models for AKI-Cr. The sensitivity and ships were observed when examining UO by consecutive
specificity of each UO-time interval combination were hours or mean hours (Fig 1).
determined, and receiver operator characteristic curves (C Patients with AKI were more likely to have longer times
statistic, a measure of discrimination) were created with of mechanical ventilation and ICU and hospital stays
KDIGO stage 1 or worse as the dependent variable of (Table 3). One hundred-twenty patients (2.8%) died. After
interest. adjustment for demographics, comorbidities, baseline
UO thresholds are the traditional way of defining AKI- creatinine, and type of operation, lower UO thresholds for
UO but may produce a false dichotomy at the threshold longer times were significantly associated with death
[8, 9, 16]. For example, if the hourly threshold is 0.5 mL/kg, (Fig 2). Results were similar for UO being continuously
then a patient whose UO is 0.45 mL/kg is considered the below the threshold for all hours in the time period and
same as a patient whose hourly UO is 0.1 mL/kg and for the UO averaged over the time period being below the
different from one whose hourly UO is 0.55 mL/kg. We threshold.
therefore also analyzed the association between AKI-UO When we analyzed the data, not by thresholds but by
and AKI-Cr with death using 30 different UO-time com- ranges, we found that lower UO and longer times of low
binations, comprising 5 hourly UO ranges (<0.2, 0.2 to UO were both associated with the subsequent develop-
< 0.4, 0.4 to < 0.6, 0.6 to < 0.8, and 0.8 to 0.9 mL/kg) and 6 ment of AKI-Cr. Even short durations (1 or 2 hours) were
time intervals (1 or 2, 3 or 4, 5 or 6, 7 or 8, 8 or 10, and 11 or associated with approximately a 35% to 120% increase in
12 hours). To maintain statistical power, all levels of AKI-Cr the ORs of AKI-Cr (Supplemental Fig 1). As the duration
were collapsed into one stage (KDIGO stage 1 or greater). of low UO progressively increased, the ORs of developing
The association between AKI and death (the third AKI-Cr progressively increased, particularly for UO of
question) was determined using binary logistic regres- less than 0.2 or 0.2 to 0.4 mL $ kg–1 $ h–1. Despite low UO
sion, adjusted for age, height, type of operation, ejection being associated with AKI-Cr, after adjustment for other
fraction, baseline creatinine, and comorbidities with 95% factors, only UO below 0.2 mL $ kg–1 $ h–1 was inde-
confidence intervals (CIs) that excluded 1 and a p of less pendently associated with increased risk of death, and
than 0.05 denoting statistical significance. No adjustment this risk increased with increasing duration of low UO
was made for multiple outcomes [18]. All statistics were (Supplemental Fig 2).
done with R software (R Foundation, Vienna, Austria).
Comment
Results
We have performed one of the largest studies to date
KDIGO stage 1 or greater AKI-Cr developed in 1,061 of
examining the prognostic effect of declines in UO and
4,195 patients (25%) within 3 days of their operation
relationship to AKI as defined by Cr-based criteria. Using
(Table 1). These patients were older, were more likely to
a large database of patients undergoing cardiac opera-
have comorbidities, and had lower ejection fractions
tions, we have described the clinical effect of several
(Table 2). The discriminatory ability of UO to predict AKI-
different approaches to applying AKI-UO definitions. In
Cr was similar between actual (C statistic ¼ .637 .054),
ideal (.623 .058), and adjusted (.631 .060) body weight.
Although discrimination improved with successive days Table 3. Outcomes
after the operation for all types of weights and both
continuous and mean UO, there was little difference AKI—Yes AKI—No
(n ¼ 1,061) (n ¼ 3,134)
when analyzed by weight type within body mass index
category (Supplemental Table 1). Outcome No. % No, % p Value
After adjustment for factors in Table 2, all UO thresholds
Prolonged ventilation 412 39 287 9 <0.001
(adjusted by actual weights) examined (up to 0.9 mL $ kg–1 $
Deep sternal wound 5 0.5 8 0.3 0.335
h–1) for any length of time were associated with increased infections
odds of developing AKI-Cr compared with patients whose
Reoperation for bleeding 63 6 45 1 <0.001
UO remained above 0.9 mL $ kg–1 $ h–1 (Fig 1). In general, a
Readmission to ICU 185 17 551 18 0.926
trend toward increasing risk for AKI was noted with
Readmission to hospital 163 15 429 14 0.185
increasing duration of UO within each threshold category
(horizontal left to right trend in Fig 1). However, this rela- Median IQR Median IQR
tionship was modest at higher UO thresholds (0.6 to 0.9 mL
$ kg–1 $ h–1) and became more pronounced only at lower Total ventilator 18 7–57 6 4–14 <0.001
time, h
UO thresholds. For example, the adjusted odds ratio (OR)
Length of stay
for AKI at a UO of 0.1 mL $ kg–1 $ h–1 ranged from
ICU, h 114 66–195 51 28–90 <0.001
2.355 (95% CI, 1.77 to 3.135) for a duration of 1 hour to
36.99 (95% CI, 4.38 to 312.24) for a duration of 12 hours. Postoperative, d 10 7–15 6 5–8 <0.001
Similarly, for any given duration, the risk of AKI generally Total hospital, d 12 7–21 7 5–11 <0.001
increased with lower UO threshold (vertical top to bottom AKI ¼ acute kidney injury; ICU ¼ intensive care unit; IQR ¼
trend in Fig 1). Once again, this relationship was most interquartile range.
6 ENGOREN ET AL Ann Thorac Surg
URINE OUTPUT AND CREATININE ELEVATION 2016;-:-–-
Fig 2. Urine output threshold and acute death. Heat map shows adjusted odds ratios of dying based on urine output meeting the hourly threshold
averaged (A) over the several hours (mean) or (B) below that value for all hours (consecutive). Each box shows the odds ratio (95% confidence
interval) and the number of patients who died/the number of patients whose urine output met the threshold value for that many hours. The red
outline shows boxes where the odds ratio is statistically significant.
general, declining UO was associated with developing threshold values may produce a false similarity between
AKI-Cr, but only extreme drops in UO (by threshold or dissimilar values (eg, 0.1 and 0.45) and suggest a false
by longer duration) were associated with increased death. separation between similar values (eg, 0.48 and 0.52). A 0.5
Our findings have several practical implications. mL $ kg–1 $ h–1 threshold includes 0.1 and 0.45 and their
Importantly, we found that there is similar discrimi- average outcome is compared with the average outcome
nation in using UO corrected for actual, ideal, or adjusted of all values exceeding 0.5. To avoid this problem, we
body weight. This did not vary when we stratified by repeated the analyses dividing UO into ranges instead of
three levels of obesity (Supplemental Table 1). For thresholds. This found a time-dependent association be-
simplicity, we therefore recommend using actual weight. tween UO and AKI-Cr: short times of profoundly low UO
Low UO was associated with subsequent rise in creati- were associated with AKI-Cr, whereas at higher levels of
nine levels. UO of less than 0.5 mL $ kg–1 $ h–1 for 6 or 12 UO, longer time periods were required to develop
hours was associated with an increased risk of developing AKI-Cr. UO below 0.2 or between 0.2 and 0.4 mL $ kg–1 $
AKI-Cr (Fig 1) but not of dying (Fig 2). However, as the h–1 was associated with increased odds of developing
UO threshold became lower (eg, <0.3 mL $ kg–1 $ h–1), AKI-Cr, with the ORs progressively increasing the longer
the association between UO and death became significant UO remained in these ranges. This suggests that low UO
and progressively increased the longer UO remained acts as an early predictor of AKI-Cr and that if in-
below that threshold. Although UO thresholds are part of terventions are beneficial in preventing AKI-Cr, the
the KDIGO, AKIN, and RIFLE definitions for AKI, physician has varying times to intervene depending on the
Ann Thorac Surg ENGOREN ET AL 7
2016;-:-–- URINE OUTPUT AND CREATININE ELEVATION
Fig 2. (Continued).
severity of the low UO. Although higher levels of UO were death are similar to two other studies [14, 15]. However,
associated with the subsequent development of AKI-Cr, our study differs by analyzing UO ranges and not just
the ORs were generally in the range of 1.5 to 2.0 and did thresholds. This better allows us to determine the effect of
not vary appreciably with duration. different levels of low UO.
Our findings that threshold values of UO are associated Current consensus AKI definitions use a UO threshold
with AKI-Cr is similar to that of Prowle and colleagues of 0.5 mL $ kg–1 $ h–1 for 6 hours to diagnose the earliest
[12], who in a study of 239 patients found that UO below stage of AKI. In our study, this threshold was associated
0.5 mL $ kg–1 $ h–1 was associated with subsequent AKI- with a nearly 80% increased risk for developing AKI
Cr, although they did not find an association between based on Cr criteria (adjusted OR, 1.79; 95% CI, 1.37 to
duration of oliguria and AKI-Cr. This may relate to 2.33) but not with a significant risk of death (adjusted OR,
different patient populations (after cardiac operations in 1.21; 95% CI, 0.63 to 2.33; Fig 1). Conversely, an observed
our study vs mixed medical and surgical ICU in theirs), UO of 0.1 mL $ kg–1 $ h–1 for just 2 hours was associated
different creatinine elevations to define AKI-Cr (0.3 with both a significant increased risk of subsequently
mg/dL or 50% rise vs 0.5 mg/dL), or their limited power meeting AKI by Cr-based criteria (adjusted OR, 3.22; 95%
with only 239 patients [12]. However, our findings extend CI, 3.44 to 4.45) and death (adjusted OR, 2.93; 95% CI, 1.37
theirs by analyzing the data by ranges instead of just to 6.28). These findings reinforce the sensitive nature of
thresholds. This allowed us to better relate the amount extreme drops in UO in identifying patients at risk for
and time of low UO to both AKI-Cr and to death. Our poor outcomes, even within only a few hours. Impor-
findings that low UO is independently associated with tantly, these early hours may be a critical time to
8 ENGOREN ET AL Ann Thorac Surg
URINE OUTPUT AND CREATININE ELEVATION 2016;-:-–-
13. Macedo E, Malhotra R, Bouchard J, Wynn SK, Mehta RL. failure—definition, outcome measures, animal models, fluid
Oliguria is an early predictor of higher mortality in critically therapy and information technology needs: the Second In-
ill patients. Kidney Int 2011;80:760–7. ternational Consensus Conference of the Acute Dialysis
14. Mandelbaum T, Lee J, Scott DJ, et al. Empirical relationships Quality Initiative (ADQI). Crit Care 2004;8:R204–12.
among oliguria, creatinine, mortality, and renal replacement 17. Perneger TV. What’s wrong with Bonferroni adjustments.
therapy in the critically ill. Intensive Care Med 2013;39:414–9. BMJ 1998;316:1236–8.
15. Teixeira C, Garzotto F, Piccinni P, et al. Fluid balance and 18. Chelsey LC. Renal excretion at low urine volumes and the
urine volume are independent predictors of mortality in mechanism of oliguria. J Clin Invest 1938;17:591–7.
acute kidney injury. Crit Care 2013;17:R14. 19. Hamada Y, Kawachi K, Tsunooka N, Nakamura Y, Takano S,
16. Bellomo R, Ronco C, Kellum JA, Mehta RL, Palevsky P, Imagawa H. Capillary leakage in cardiac surgery with cardio-
Acute Dialysis Quality Initiative Workgroup. Acute renal pulmonary bypass. Asian Cardiovasc Thorac Ann 2004;12:193–7.