Clinical Investigation|Articles in Press

Ok

# Comparative effectiveness and safety of bolus vs. continuous infusion of loop diuretics: Results from the MIMIC-III Database

Published:December 23, 2022

## Abstract

### Background

It is unclear whether fluid management goals are best achieved by bolus injection or continuous infusion of loop diuretics. In this study, we compared the effectiveness and safety of a continuous infusion with that of a bolus injection when an increased loop diuretic dosage is required in intensive care unit (ICU) patients.

### Methods

We obtained data from the MIMIC-III database for patients who were first-time ICU admissions and required an increased diuretic dosage. Patients were excluded if they had an estimated glomerular filtration rate <15 ml/min/1.73 m2, were receiving renal replacement therapy, had a baseline systolic blood pressure <80 mmHg, or required a furosemide dose <120 mg. The patients were divided into a continuous group and a bolus group. Propensity score matching was used to balance patients’ background characteristics.

### Results

The final dataset included 807 patients (continuous group, n = 409; bolus group, n = 398). After propensity score matching, there were 253 patients in the bolus group and 231 in the continuous group. The 24 h urine output per 40 mg of furosemide was significantly greater in the continuous group than in the bolus group (234.66 ml [95% confidence interval (CI) 152.13–317.18, p < 0.01]). There was no significant between-group difference in the incidence of acute kidney injury (odds ratio 0.96, 95% CI 0.66–1.41, p = 0.85).

### Conclusions

Our results indicate that a continuous infusion of loop diuretics may be more effective than a bolus injection and does not increase the risk of acute kidney injury in patients who need an increased diuretic dosage in the ICU.

## Introduction

Fluid overload is common in patients who are critically ill,
• Mehta R.L.
Fluid overload in the ICU: evaluation and management.
and several studies have demonstrated a correlation between fluid overload and adverse outcomes.
• Bouchard J.
• Soroko S.B.
• Chertow G.M.
• et al.
Fluid accumulation, survival and recovery of kidney function in critically ill patients with acute kidney injury.
,
• Boyd J.H.
• Forbes J.
• et al.
Fluid resuscitation in septic shock: a positive fluid balance and elevated central venous pressure are associated with increased mortality.
Conservative fluid management can improve the oxygenation index, increase the number of ventilator-free days, and shorten the stay in the intensive care unit (ICU).
• National Heart L.
• Wiedemann H.P.
• et al.
Blood Institute Acute Respiratory Distress Syndrome Clinical Trials N
Comparison of two fluid-management strategies in acute lung injury.
Loop diuretics administration is a well-established therapy that can decrease the incidence of fluid overload. Continuous infusion and bolus injection of diuretics are two methods that physicians can use to achieve their fluid management goals. However, there is still not enough evidence to indicate which strategy is better in terms of safety and efficiency. The randomized controlled Diuretic Optimization Strategies Evaluation in Acute Heart Failure (DOSE) study found no significant difference in patients’ global assessment of symptoms or in the change in renal function in a large group of patients with decompensated heart failure, when diuretics were administered by bolus as compared with continuous infusion.
• Felker G.M.
• Lee K.L.
• Bull D.A.
• et al.
Diuretic strategies in patients with acute decompensated heart failure.
Other randomized controlled trials have yielded conflicting results in patients with heart failure.
• Palazzuoli A.
• Pellegrini M.
• Ruocco G.
• et al.
Continuous versus bolus intermittent loop diuretic infusion in acutely decompensated heart failure: a prospective randomized trial.
• Llorens P.
• Miro O.
• Herrero P.
• et al.
Clinical effects and safety of different strategies for administering intravenous diuretics in acutely decompensated heart failure: a randomised clinical trial.
• Thomson M.R.
• Nappi J.M.
• Dunn S.P.
• et al.
Continuous versus intermittent infusion of furosemide in acute decompensated heart failure.
However, most of the studies only enrolled outpatients, who were very different from patients in the ICU. A recent meta-analysis reported that the beneficial effect on total urine output was greater when continuous furosemide was administered than when a bolus injection was delivered.
• Ng K.T.
• Velayit A.
• Khoo D.K.Y.
• et al.
Continuous infusion versus intermittent bolus injection of furosemide in critically ill patients: a systematic review and meta-analysis.
However, most of the randomized controlled trials included in that meta-analysis had small sample sizes.
Patients in the ICU have a high incidence of acute kidney injury (AKI) and are more vulnerable and more complicated to manage than the patients in a general medicine ward or outpatient. An appropriate diuretic strategy is needed to help physicians achieve their fluid management goals when a patient needs an increased dose of diuretics. The aim of this study was to compare the efficiency and safety of two diuretic strategies in a relatively large sample of patients who were admitted to the ICU based on the Medical Information Mart for Intensive Care (MIMIC)-III database.

## Methods

### Data source and ethical approval

The MIMIC-III database is a publicly available dataset developed by the Massachusetts Institute of Technology Lab for Computational Physiology and includes de-identified health data for approximately 40,000 critical care patients. It includes demographics, vital signs, laboratory tests, medications and other data
• Johnson A.E.
• Pollard T.J.
• Shen L.
• et al.
MIMIC-III, a freely accessible critical care database.
(https://mimic.physionet.org/). The MIMIC database is approved by the Massachusetts Institute of Technology (Cambridge, MA) and Beth Israel Deaconess Medical Center (Boston, MA), and patient consent is obtained when the data are first collected. Therefore, the need for ethical approval and informed consent was waived for this study. The investigators also completed the required training course before requesting access to the database. The study was performed in accordance with the Declaration of Helsinki.

### Study population and grouping

The study population comprised 6830 adult patients with data in the MIMIC-III database who had been treated with diuretics during a first ICU stay. The time of diuretic use was divided into 12 h time units, and the total amounts of diuretics used in each time unit were calculated. After comparing the amounts of diuretics used in each 12 h unit with that in adjacent 12 h units, we defined a specific 12 h period as the “start unit”, namely, when the amounts of diuretics used were more than that in the previous time unit and the amounts in 24 h were more than 120 mg of furosemide (or more than 60 mg of torsemide or 3 mg of bumetanide). The beginning of this time unit was defined as the “start time”. Patients whose start unit occurred during their first ICU stay were subjected to further analysis. Patients who had received renal replacement therapy for end-stage renal disease (estimated glomerular filtration rate ≤15 ml/min/m2) and those with a baseline systolic blood pressure less than 80 mmHg before the start unit were excluded. In total, 878 subjects were enrolled. After exclusion of those with missing data and outliers, 807 subjects were included in the final dataset. Subjects who received continuous intravenous diuretics in “start unit” were assigned into the continuous infusion group, the others were assigned into the bolus group. For those who received both administrations of diuretics, we compared the amounts of diuretics between bolus injection and continuous infusion. And these patients were included into continuous infusion group for the diuretics were dominantly administrated by continuous infusion. A flow chart summarizing the enrollment process is shown in Fig. 1.

### Data collection

#### Demographic characteristics

We collected demographic information, including age, gender, and the last body weight recorded before the start unit. The type of ICU was also recorded.

#### Vital signs

Vital signs included the most recent systolic blood pressure (SBP) and diastolic blood pressure (DBP) recorded before the start unit.

#### Laboratory examinations

The most recent serum creatinine, blood urea nitrogen (BUN), blood sugar, and serum sodium level recorded before the start unit were obtained. The maximum serum creatinine level recorded in the 48h after the start time was also obtained to detect changes in kidney function. Estimated glomerular filtration rate (eGFR) was calculated by Modification of Diet in Renal Disease (MDRD) equation.
• Poggio E.D.
• Wang X.
• Greene T.
• et al.
Performance of the modification of diet in renal disease and Cockcroft-Gault equations in the estimation of GFR in health and in chronic kidney disease.

#### Past medical history and risk factors

Past medical history was obtained based on the admitting diagnosis and history of hypertension, diabetes mellitus, chronic kidney disease, and chronic heart failure (CHF). International Classification of Diseases codes were used rather than information on past medical history from the notes. Severity of illness was evaluated using the Simplified Acute Physiology Score (SAPS) III value, which was based on the MIMIC-III GitHub code.
• Metnitz P.G.
• Moreno R.P.
• Almeida E.
• et al.
SAPS 3–From evaluation of the patient to evaluation of the intensive care unit. Part 1: objectives, methods and cohort description.
,
• Moreno R.P.
• Metnitz P.G.
• Almeida E.
• et al.
SAPS 3–From evaluation of the patient to evaluation of the intensive care unit. Part 2: development of a prognostic model for hospital mortality at ICU admission.

#### Medications

Information on use of spironolactone, thiazide, tolvaptan, norepinephrine, milrinone, and infusions of albumin and plasma during the start unit was recorded. The amounts of loop diuretics used in the previous unit (defined as “previous diuretics”) and in the 24 h after the start time were calculated.

#### Endpoints

Urine output in the 24 h after the start time was extracted from the database. Urine output per 40 mg of furosemide was calculated using the following formula:
$urineoutput(ml)amountofdiuretics(mg)×40$

AKI was defined as an increase in the creatinine level of ≥0.3 mg/dl within 48 h in accordance with the 2012 Kidney Disease: Improving Global Outcomes criteria.

J.A. Kellum, N. Lameire, P. Aspelin, et al. Kidney disease: improving global outcomes (KDIGO) acute kidney injury work group. KDIGO clinical practice guideline for acute kidney injury. 2012;

### Propensity score matching and statistical analysis

We selected variables known to influence the outcome (urine output) in clinical practice, namely, patient age, weight, and sex, SBP, serum creatinine, BUN, and sodium levels, history of chronic kidney disease, history of CHF, SAPS III value, use of spironolactone, thiazide, norepinephrine, and milrinone, infusion of albumin or plasma, type of ICU, previous diuretics and amounts of diuretics in 24h for propensity score matching. We chose 1:1 matching using nearest neighbor matching with a caliper width of 0.05 standard deviation of the logit of the propensity scores without replacement. The propensity score distribution before and after matching was shown in Fig. 2. The value of the C-statistic was 0.78, indicating that the model was of good quality. After propensity score matching, there were 231 subjects in the continuous infusion group and 253 in the bolus group.
Data that were normally distributed were shown as the mean ± standard deviation and data that were not normally distributed were shown as the median (interquartile range). Proportions were compared using the chi-squared test and means were compared using the t-test when data were normally distributed and the Wilcoxon test when they were not. Both linear regression and logistic regression were used in the final model. All statistical analyses were performed using STATA 12.0 (StataCorp LLC, College Station, TX, USA) and EmpowerStats software (Greenwood Village, CO). A p-value <0.05 was considered statistically significant.

## Results

Table 1 showed the characteristics of the two study groups before and after propensity score matching. Before matching, there were significant between-group differences in body weight, SAPS III value, SBP, serum sodium, BUN, and creatinine levels, previous diuretics, amounts of diuretics in 24h, history of chronic kidney disease, history of CHF, use of spironolactone and thiazides, type of ICU, use of norepinephrine and milrinone, and albumin level. After propensity score matching, the only significant between-group difference was in type of ICU.
Table 1Characteristics of two groups before and after propensity matching.
Before PS matchingAfter PS matching
VariablesBolusContinuouspBolusContinuousp
n398409253231
Demographic characteristics
Age69.95±14.8469.35±14.340.5670.05±13.8569.97±14.650.95
Weight85.37±26.2390.76±30.27<0.0186.84±28.6086.50±27.000.90
Gender0.470.288
Male203(51.01%)219 (53.55%)117(46.25%)118(51.08%)
Vital sighs
SBP119.00±14.36112.34±13.68<0.01116.85±14.04115.94±13.290.47
DBP59.15±8.9759.05±9.170.8858.59±9.1359.85±9.750.14
Laboratory test
Scr1.20(0.80–1.60)1.40(1.00–2.10)<0.011.20(0.80–1.70)1.30(0.90–1.90)0.16
MAX Scr1.20(0.90–1.60)1.60(1.10–2.20)<0.011.30(0.90–1.90)1.40(0.90–2.00)0.54
BUN26.50(20.00–39.75)35.00(23.00–57.00)<0.0129.00(19.00–44.00)32.00(20.00–44.50)0.22
Na138.91±4.76138.11±5.270.03138.58±4.97138.91±4.820.45
GLU139.97±52.82145.90±55.770.12141.16±54.76141.81±54.210.90
Past history and risk factor
Hypertension166 (41.71%)161 (39.36%)0.498105(41.50%)90(38.96%)0.57
DM177 (44.47%)191 (46.70%)0.525108(42.69%)98(42.42%)0.95
CKD122 (30.65%)144 (35.21%)0.1775(29.64%)74(32.03%)0.57
CHF239 (60.05%)288 (70.42%)<0.01161(63.64%)147(63.64%)0.79
SAPS III50.62±19.9154.37±19.53<0.0151.58±20.6452.27±18.830.70
Medications
Spironolactone17(4.27%)19(4.65%)0.809(3.56%)9(3.90%)0.84
Thiazide25(6.28%)44(10.76%)0.0219(7.51%)22(9.52%)0.43
Norepinephrine30(7.54%)61(14.91%)<0.0124(9.49%)27(11.69%)0.75
Milrinone26(6.53%)57(13.94%)<0.0121(8.30%)18(7.79%)0.84
Infusion albumin92(23.12%)73(17.85%)0.0663(24.90%)55(23.81%)0.78
Infusion plasma39(9.80%)33(8.07%)0.3919(7.51%)20(8.66%)0.64
Previous diuretics40.00(20.00–40.00)60.00(37.50–100.48)<0.0140.00(20.00–60.00)40.00(29.88–60.14)0.48
24h diuretics amount160.00(120.00–200.00)232.06(160.11–349.93)<0.01160.00(120.00–240.00)177.74(133.63–237.70)0.18
ICU type<0.010.03
Cardiac ICU86(21.61%)149 (36.43%)55(21.74%)68(29.44%)
Medical ICU105(26.38%)52 (12.71%)61(24.11%)39(16.88%)
Surgery ICU162(40.70%)149 (36.43%)110(43.48%)86(37.23%)
Cardiac surgery ICU29(7.29%)35(8.56%)17(6.72%)20(8.66%)
Trauma surgery ICU16(4.02%)24(5.87%)10(3.95%)18(7.79%)
Endpoints
Urine output 24h2685.00(1799.25–3503.75)3309.00(2130.00–4455.00)<0.012495.00(1746.00––3455.00)3485.00(2385.00–4840.00)<0.01
Urine output/40mg Lasix624.21(359.53–933.33)544.49(288.20–959.53)<0.01524.00(315.62–821.67)797.75(470.27––1165.98)<0.01
AKI113(28.39%)155(37.90%)<0.0185(33.60%)77(33.33%)0.95
AKI, acute kidney injury; BUN, blood urea nitrogen; CHF, chronic heart failure; CKD, chronic kidney disease; DBP, diastolic blood pressure; DM, diabetes mellitus; GLU, serum glucose; ICU, intensive care unit; Na, serum Natrium; Previous diuretics, the amount of diuretics used in the last unit before start unit; SAPS III, Simplified Acute Physiology Score III; SBP, systolic blood pressure; Scr, serum creatinine; 24h diuretics amount, the amount of diuretis used in 24h after start time.
Multiple linear regression was used to compare the urine output in 24h after the start time between the two groups. As shown in Table 2, after adjustment for type of ICU, the output in 24h was significantly greater in the continuous group than in the bolus group (by 930.71 ml, 95% confidence interval [CI] 655.51–1205.90, p < 0.01), as was urine output in 24h per 40 mg of furosemide (by 234.66 ml, 95% CI 152.13–317.18, p < 0.01). Multiple logistic regression showed no significant between-group difference in the incidence of AKI after adjustment for type of ICU (odds ratio 0.96, 95% CI 0.66–1.41, p = 0.85).
Table 2Multiple linear regression between urine output and inject type
: ICU type was adjusted in this regression model.
.
Variablesβ95%CIp
Urine output in 24h930.71655.51–1205.90<0.01
Urine output in 24h per 40mg furosemides234.66152.13–317.18<0.01
a : ICU type was adjusted in this regression model.

## Discussion

Fluid management is an essential component of health care in critically ill patients. Fluid overload, which can lead to cognitive impairment, systolic and diastolic dysfunction, reduced gas exchange, reduced renal blood flow, impaired absorption and synthesis and is correlated with a poor outcome in these patients.
• Mehta R.L.
Fluid overload in the ICU: evaluation and management.
A study by Bouchard et al. demonstrated that patients with AKI and fluid overload had increased 30-day and 60-day mortality rates.
• Bouchard J.
• Soroko S.B.
• Chertow G.M.
• et al.
Fluid accumulation, survival and recovery of kidney function in critically ill patients with acute kidney injury.
Wiedemann et al. found that conservative fluid management could improve the oxygenation index and lung injury score, increase the number of ventilator-free days, and shorten the length of ICU stay in patients with acute lung injury.
• National Heart L.
• Wiedemann H.P.
• et al.
Blood Institute Acute Respiratory Distress Syndrome Clinical Trials N
Comparison of two fluid-management strategies in acute lung injury.
Moreover, the Vasopressin in Septic Shock Trial demonstrated that a more positive fluid balance was associated with an decreased risk of mortality in patients with septic shock.
• Boyd J.H.
• Forbes J.
• et al.
Fluid resuscitation in septic shock: a positive fluid balance and elevated central venous pressure are associated with increased mortality.
Administration of diuretic agents, in particular loop diuretics, is the strategy most commonly used to manage fluid balance. The main side effects of diuretics are hypokalemia, hypotension, and increased serum BUN and creatinine levels. In theory, compared with bolus injection, continuous intravenous infusion may be more effective because it maintains a consistent rate of drug excretion and inhibits reabsorption of sodium chloride in the loop of Henle over time.
• Martin S.J.
• Danziger LH.
Continuous infusion of loop diuretics in the critically ill: a review of the literature.
Several studies have compared the efficiency of these two strategies but no consistent conclusions have been reached. The DOSE study included 308 patients with acute decompensated heart failure and focused on co-primary endpoints of patient-reported global assessment of symptoms, quantified as the area under the curve (AUC) for the visual analog scale score, and the change in serum creatinine level between baseline and 72h. The investigators found no significant difference in patient-reported symptoms (mean AUC, 4236 ± 1440 vs 4373 ± 1404, p = 0.47) or in the mean change in creatinine level (0.05 ± 0.3 mg/dl vs 0.07 ± 0.3 mg/dl, p = 0.45) between bolus injection and continuous infusion group.
• Felker G.M.
• Lee K.L.
• Bull D.A.
• et al.
Diuretic strategies in patients with acute decompensated heart failure.
A study by Makhoul et al. found that urine output was higher in patients in the ICU who received diuretics via continuous infusion than in those who received the same amount of diuretics via bolus injection.
• Makhoul N.
• Friedstrom S.
• et al.
Frusemide in pulmonary oedema: continuous versus intermittent.
Furthermore, in a study by Ostermann et al., a significantly higher dose of furosemide was needed to achieve target diuresis in a bolus group than in a continuous infusion group (24.1 mg/h vs. 9.2 mg/h, p = 0.0002) with no difference in the change in serum creatinine level.
• Ostermann M.
• Alvarez G.
• Sharpe M.D.
• et al.
Frusemide administration in critically ill patients by continuous compared to bolus therapy.
Shah et al. investigated 90 patients in a cardiac ICU and documented significantly greater diuresis in the first 24h (p = 0.002) and a significantly shorter hospital stay (p = 0.023) in the bolus group without any significant difference in renal function.
• Shah R.A.
• Subban V.
• Lakshmanan A.
• et al.
A prospective, randomized study to evaluate the efficacy of various diuretic strategies in acute decompensated heart failure.
Another study in patients with chronic heart failure and high-risk acute decompensation found that continuous infusion of intravenous furosemide was associated with a better decongestive effect.
• Frea S.
• Pidello S.
• Volpe A.
• et al.
Diuretic treatment in high-risk acute decompensation of advanced chronic heart failure-bolus intermittent vs. continuous infusion of furosemide: a randomized controlled trial.
Moreover, a meta-analysis that included nine studies for a total of 464 ICU patients concluded that continuous infusion of furosemide was associated with a significantly greater total urine output than bolus therapy with no statistically significant change in the creatinine level.
• Ng K.T.
• Velayit A.
• Khoo D.K.Y.
• et al.
Continuous infusion versus intermittent bolus injection of furosemide in critically ill patients: a systematic review and meta-analysis.
However, a more recent meta-analysis found no significant difference in all-cause mortality, length of hospital stay, or electrolyte disturbance between patients with acute decompensated heart failure who received furosemide by continuous infusion and those who received it by bolus injection.
• Ng K.T.
• Yap J.L.L.
Continuous infusion vs. intermittent bolus injection of furosemide in acute decompensated heart failure: systematic review and meta-analysis of randomised controlled trials.
Therefore, it is still unclear which strategy is better. One explanation for the inconsistent findings to date is that the target populations have differed from study to study. Therefore, more clinical research is needed to compare these two strategies to guide the diuretics use.
Our present study analyzed data for ICU patients obtained from the MIMIC database and demonstrated that continuous infusion of diuretics was associated with a greater urine output and more urine output per 40 mg of furosemide without an increased risk of AKI. We divided the duration of diuretic therapy into 12 h units to include patients who needed an increased dose of diuretics. Our study population comprised critically ill patients who were already on relatively high doses of diuretics and needed an even higher dose in the ICU. Given that such patients are often encountered in clinical practice, we believe that the findings of this study could be used to assist physicians with clinical decision-making. We also considered that the use of propensity score matching in this study to balance covariates that could otherwise act as confounders improves the credibility of our results. Furthermore, we used the urine output per 40 mg of furosemide to balance the effect of administration of different amounts of diuretics on urine output, which allowed us to demonstrate the sensitivity to diuretics in the two groups.
Our study included patients from cardiac, medical, general surgery, cardiac surgery, and trauma surgery ICUs with the aim of finding a better strategy for administration of diuretics in a broad range of critically ill patients. These patients were all volume-loaded because they had already used relatively large doses of diuretics. We initially screened the main diagnosis for each patient and found that the diagnoses were too varied to be able to analyze the data on the basis of the type of disease. Therefore, we used the SAPS III value to evaluate the severity of disease in each individual patient; this could have attenuated the effect of type of disease on the endpoint to some extent. After propensity score matching, the SAPS III value was balanced between the two study groups. Mindful of the clustering effect, we included type of ICU as one of the covariates in propensity score matching and adjusted for it in the final regression model. We were unable to confirm the effect of different medical centers because the information in the MIMIC-III database is collected from a single center. Length of ICU stay and mortality were not included as endpoints in view of the large number of factors that could influence these variables, particularly the disease that prompted the ICU admission. We also calculated the incidence of hypotension and found no significant between-group difference.
In this study, we enrolled patients who had received more than 120 mg of furosemide because in clinical practice it is not usually necessary to use a continuous diuretic infusion in patients requiring a furosemide dose lower than 120 mg. Therefore, we were able to minimize the possibility of selection bias. The most important determinant of urine output is the amounts of diuretics administered. As Table 1 showed, the mean dose of previous diuretics was 40mg for bolus group and 60mg for continuous group and 24h diuretics was 160mg for bolus group and 232.06mg for continuous group. So, we included previous amounts and 24 h amounts of diuretics in propensity score model. And after propensity score matching, the difference of previous amounts and 24h amounts of diuretics between two groups were not significant, which reduced the impact to a minimum. The previous amounts of diuretics were much less than the 24h amounts, which made the carry over effect of the diuretics used prior to the start time could be ignored.
Diuretic response is defined as the capacity of inducing natriuresis or diuresis following diuretic administration and diuretic resistance means reduced sensitivity to diuretics because of the reduced natriuresis and diuresis.
• Mullens W.
• Damman K.
• Harjola V.P.
• et al.
The use of diuretics in heart failure with congestion - a position statement from the Heart Failure Association of the European Society of Cardiology.
The measurement of urinary sodium content was proven to be a good indicator for diuretic response.
• Testani J.M.
• Hanberg J.S.
• Cheng S.
• et al.
Rapid and highly accurate prediction of poor loop diuretic natriuretic response in patients with heart failure.
And low natriuresis after furosemide administration was an early marker of poor diuretic response and correlated with higher NT-proBNP and higher incidence of worsening renal function at 72 h.
• Galluzzo A.
• Frea S.
• Boretto P.
• et al.
Spot urinary sodium in acute decompensation of advanced heart failure and dilutional hyponatremia: insights from DRAIN trial.
In our study, we tried to include the data of natriuresis to reflect the diuretic response. However, the data of natriuresis was barely recorded in this database.
In this study, patients who received both bolus injection and continuous infusion of diuretics were included. In clinical practice, many of the physicians preferred to using bolus injection before continuous infusion, for rapidly increasing the blood concentration of diuretics to achieve a good diuretic effect. So, we included this part of patients in our analysis. And we also excluded this part of patients and re-do the analysis to investigate the influence of this part of patients. We found the conclusions remained consistent whether we excluded patients who received both diuretic administrations or not. For patients without receiving both diuretic administrations, the characteristics of two groups in patients before and after propensity matching and regression results were showed in Supplementary Tables 1 and 2.
This study had several limitations. First, we could not include every covariate that could potentially influence urine output. We attempted to extract information on other factors that could influence diuretic efficiency, such as plasma albumin, and use of dopamine, but found that the amount of missing data was too large for these additional parameters to be included. Furthermore, we could not include specific doses of inotropes or other types of diuretic agents because of the relatively limited number of subject numbers that remained after propensity score matching. No records of hydration and perspiration could be acquired in the database. We had to not include the amount of hydration and perspiration, though they had effects on the urine output. And this might cause bias. Second, we were unable to investigate the effect of the clinical environment on urine output. Therefore, there could be some deviations. Finally, given the common belief among physicians that continuous infusion is likely to be more effective than bolus injection, continuous infusion of diuretics may have been more common used in more congestive patients. Therefore, patients with greater fluid overload would have a greater diuretic response, which may have introduced a degree of selection bias. Large cohort studies are needed to identify which diuretic strategy is better in the future.

## Conclusions

When there is a need to increase the dosage of a loop diuretic in an ICU patient, continuous intravenous infusion may be more efficient than bolus injection without an increased risk of AKI. The findings of this study are drawn from a relatively large database and should provide physicians with an efficient strategy that will allow them to achieve fluid management goals in their ICU patients.

## CRediT authorship contribution statement

Haoyu Weng: Visualization, Writing – review & editing. Yuxi Li: Visualization, Writing – review & editing. Xiaolu Nie: Formal analysis. Chunhui He: Investigation. Pengbin Feng: Funding acquisition, Writing – review & editing. Fengxin Zhao: Funding acquisition, Writing – review & editing. Qingjie Chen: Investigation. Wen Sun: Writing – review & editing. Jie Jiang: Writing – review & editing. Yan Zhang: Writing – review & editing. Yong Huo: Visualization. Jianping Li: Visualization.

## Declaration of Competing Interest

None.

### Funding Statement

This research was supported by a grant from the Peking University First Hospital Scientific Research Seed Fund (2018SF074).

### Acknowledgments

We thank Edanz (https://jp.edanz.com/ac) for editing a draft of this manuscript.

## References

• Mehta R.L.
Fluid overload in the ICU: evaluation and management.
BMC Nephrol. 2016; 17 (Aug 2): 109https://doi.org/10.1186/s12882-016-0323-6
• Bouchard J.
• Soroko S.B.
• Chertow G.M.
• et al.
Fluid accumulation, survival and recovery of kidney function in critically ill patients with acute kidney injury.
Kidney Int. 2009; 76 (Aug): 422-427https://doi.org/10.1038/ki.2009.159
• Boyd J.H.
• Forbes J.
• et al.
Fluid resuscitation in septic shock: a positive fluid balance and elevated central venous pressure are associated with increased mortality.
Crit Care Med. 2011; 39 (Feb): 259-265https://doi.org/10.1097/CCM.0b013e3181feeb15
• National Heart L.
• Wiedemann H.P.
• et al.
• Blood Institute Acute Respiratory Distress Syndrome Clinical Trials N
Comparison of two fluid-management strategies in acute lung injury.
N Engl J Med. 2006; 354 (Jun 15): 2564-2575https://doi.org/10.1056/NEJMoa062200
• Felker G.M.
• Lee K.L.
• Bull D.A.
• et al.
Diuretic strategies in patients with acute decompensated heart failure.
N Engl J Med. 2011; 364 (Mar 3): 797-805https://doi.org/10.1056/NEJMoa1005419
• Palazzuoli A.
• Pellegrini M.
• Ruocco G.
• et al.
Continuous versus bolus intermittent loop diuretic infusion in acutely decompensated heart failure: a prospective randomized trial.
Crit Care. 2014; 18 (Jun 28): R134https://doi.org/10.1186/cc13952
• Llorens P.
• Miro O.
• Herrero P.
• et al.
Clinical effects and safety of different strategies for administering intravenous diuretics in acutely decompensated heart failure: a randomised clinical trial.
Emerg Med J. 2014; 31 (Sep): 706-713https://doi.org/10.1136/emermed-2013-202526
• Thomson M.R.
• Nappi J.M.
• Dunn S.P.
• et al.
Continuous versus intermittent infusion of furosemide in acute decompensated heart failure.
J Card Fail. 2010; 16 (Mar): 188-193https://doi.org/10.1016/j.cardfail.2009.11.005
• Ng K.T.
• Velayit A.
• Khoo D.K.Y.
• et al.
Continuous infusion versus intermittent bolus injection of furosemide in critically ill patients: a systematic review and meta-analysis.
J Cardiothorac Vasc Anesth. 2018; (Jan 10)https://doi.org/10.1053/j.jvca.2018.01.004
• Johnson A.E.
• Pollard T.J.
• Shen L.
• et al.
MIMIC-III, a freely accessible critical care database.
Sci Data. 2016; 3 (May 24)160035https://doi.org/10.1038/sdata.2016.35
• Poggio E.D.
• Wang X.
• Greene T.
• et al.
Performance of the modification of diet in renal disease and Cockcroft-Gault equations in the estimation of GFR in health and in chronic kidney disease.
J Am Soc Nephrol. 2005; 16 (Feb): 459-466https://doi.org/10.1681/ASN.2004060447
• Metnitz P.G.
• Moreno R.P.
• Almeida E.
• et al.
SAPS 3–From evaluation of the patient to evaluation of the intensive care unit. Part 1: objectives, methods and cohort description.
Intensive Care Med. 2005; 31 (Oct): 1336-1344https://doi.org/10.1007/s00134-005-2762-6
• Moreno R.P.
• Metnitz P.G.
• Almeida E.
• et al.
SAPS 3–From evaluation of the patient to evaluation of the intensive care unit. Part 2: development of a prognostic model for hospital mortality at ICU admission.
Intensive Care Med. 2005; 31 (Oct): 1345-1355https://doi.org/10.1007/s00134-005-2763-5
1. J.A. Kellum, N. Lameire, P. Aspelin, et al. Kidney disease: improving global outcomes (KDIGO) acute kidney injury work group. KDIGO clinical practice guideline for acute kidney injury. 2012;

• Martin S.J.
• Danziger LH.
Continuous infusion of loop diuretics in the critically ill: a review of the literature.
Crit Care Med. 1994; 22 (Aug): 1323-1329
• Makhoul N.
• Friedstrom S.
• et al.
Frusemide in pulmonary oedema: continuous versus intermittent.
Clinical Intensive Care. 2011; 8: 273-276
• Ostermann M.
• Alvarez G.
• Sharpe M.D.
• et al.
Frusemide administration in critically ill patients by continuous compared to bolus therapy.
Nephron Clin Pract. 2007; 107: c70-c76https://doi.org/10.1159/000108641
• Shah R.A.
• Subban V.
• Lakshmanan A.
• et al.
A prospective, randomized study to evaluate the efficacy of various diuretic strategies in acute decompensated heart failure.
Indian Heart J. 2014; 66 (May-Jun): 309-316https://doi.org/10.1016/j.ihj.2014.03.006
• Frea S.
• Pidello S.
• Volpe A.
• et al.
Diuretic treatment in high-risk acute decompensation of advanced chronic heart failure-bolus intermittent vs. continuous infusion of furosemide: a randomized controlled trial.
Clin Res Cardiol Off J Ger Card Soc. 2020; 109 (Apr): 417-425https://doi.org/10.1007/s00392-019-01521-y
• Ng K.T.
• Yap J.L.L.
Continuous infusion vs. intermittent bolus injection of furosemide in acute decompensated heart failure: systematic review and meta-analysis of randomised controlled trials.
Anaesthesia. 2018; 73 (Feb): 238-247https://doi.org/10.1111/anae.14038
• Mullens W.
• Damman K.
• Harjola V.P.
• et al.
The use of diuretics in heart failure with congestion - a position statement from the Heart Failure Association of the European Society of Cardiology.
Eur J Heart Fail. 2019; 21 (Feb): 137-155https://doi.org/10.1002/ejhf.1369
• Testani J.M.
• Hanberg J.S.
• Cheng S.
• et al.
Rapid and highly accurate prediction of poor loop diuretic natriuretic response in patients with heart failure.
Circ Heart Fail. 2016; 9 (Jan)e002370https://doi.org/10.1161/CIRCHEARTFAILURE.115.002370
• Galluzzo A.
• Frea S.
• Boretto P.
• et al.
Spot urinary sodium in acute decompensation of advanced heart failure and dilutional hyponatremia: insights from DRAIN trial.
Clin Res Cardiol Off J Ger Card Soc. Oct 2020; 109: 1251-1259https://doi.org/10.1007/s00392-020-01617-w