Importance of intravenous fluid dose and composition in surgical ICU patients

External validation of a machine learning model to predict hemodynamic instability in intensive care unit

Background

Early prediction model of hemodynamic instability has the potential to improve the critical care, whereas limited external validation on the generalizability. We aimed to independently validate the Hemodynamic Stability Index (HSI), a multi-parameter machine learning model, in predicting hemodynamic instability in Asian patients.

Method

Hemodynamic instability was marked by using inotropic, vasopressor, significant fluid therapy, and/or blood transfusions. This retrospective study included among 15,967 ICU patients who aged 20 years or older (not included 20 years) and stayed in ICU for more than 6 h admitted to Taipei Veteran General Hospital (TPEVGH) between January 1, 2010, and March 31, 2020, of whom hemodynamic instability occurred in 3053 patients (prevalence = 19%). These patients in unstable group received at least one intervention during their ICU stays, and the HSI score of both stable and unstable group was calculated in every hour before intervention. The model performance was assessed using the area under the receiver operating characteristic curve (AUROC) and was compared to single indicators like systolic blood pressure (SBP) and shock index. The hemodynamic instability alarm was set by selecting optimal threshold with high sensitivity, acceptable specificity, and lead time before intervention was calculated to indicate when patients were firstly identified as high risk of hemodynamic instability.

Results

The AUROC of HSI was 0.76 (95% CI, 0.75–0.77), which performed significantly better than shock Index (0.7; 95% CI, 0.69–0.71) and SBP (0.69; 95% CI, 0.68–0.70). By selecting 0.7 as a threshold, HSI predicted 72% of all 3053 patients who received hemodynamic interventions with 67% in specificity. Time-varying results also showed that HSI score significantly outperformed single indicators even up to 24 h before intervention. And 95% unstable patients can be identified more than 5 h in advance.

Conclusions

The HSI has acceptable discrimination but underestimates the risk of stable patients in predicting the onset of hemodynamic instability in an external cohort.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13054-022-04088-9.

Ten answers to key questions for fluid management in intensive care

This review focuses on fluid management of critically ill patients. The topic is addressed based on 10 single questions with simplified answers that provide clinicians with the basic information needed at the point of care in treating patients in the Intensive Care Unit. The review has didactic purposes and may serve both as an update on fluid management and as an introduction to the subject for novices in critical care. There is an urgent need to increase awareness regarding the potential risks associated with fluid overload. Clinicians should be mindful not only of the indications for administering fluid loads and of the type of fluids administered, but also of the importance to set safety limits. Lastly, it is important to implement proactive strategies seeking to establish negative fluid balance as soon as the clinical conditions are considered to be stable and the risk of deterioration is low.

Early prediction of hemodynamic interventions in the intensive care unit using machine learning

Background

Timely recognition of hemodynamic instability in critically ill patients enables increased vigilance and early treatment opportunities. We develop the Hemodynamic Stability Index (HSI), which highlights situational awareness of possible hemodynamic instability occurring at the bedside and to prompt assessment for potential hemodynamic interventions.

Methods

We used an ensemble of decision trees to obtain a real-time risk score that predicts the initiation of hemodynamic interventions an hour into the future. We developed the model using the eICU Research Institute (eRI) database, based on adult ICU admissions from 2012 to 2016. A total of 208,375 ICU stays met the inclusion criteria, with 32,896 patients (prevalence = 18%) experiencing at least one instability event where they received one of the interventions during their stay. Predictors included vital signs, laboratory measurements, and ventilation settings.

Results

HSI showed significantly better performance compared to single parameters like systolic blood pressure and shock index (heart rate/systolic blood pressure) and showed good generalization across patient subgroups. HSI AUC was 0.82 and predicted 52% of all hemodynamic interventions with a lead time of 1-h with a specificity of 92%. In addition to predicting future hemodynamic interventions, our model provides confidence intervals and a ranked list of clinical features that contribute to each prediction. Importantly, HSI can use a sparse set of physiologic variables and abstains from making a prediction when the confidence is below an acceptable threshold.

Conclusions

The HSI algorithm provides a single score that summarizes hemodynamic status in real time using multiple physiologic parameters in patient monitors and electronic medical records (EMR). Importantly, HSI is designed for real-world deployment, demonstrating generalizability, strong performance under different data availability conditions, and providing model explanation in the form of feature importance and prediction confidence.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13054-021-03808-x.

Ten answers to key questions for fluid management in intensive care

This review focuses on fluid management of critically ill patients. The topic is addressed based on 10 single questions with simplified answers that provide clinicians with the basic information needed at the point of care in treating patients in the Intensive Care Unit. The review has didactic purposes and may serve both as an update on fluid management and as an introduction to the subject for novices in critical care. There is an urgent need to increase awareness regarding the potential risks associated with fluid overload. Clinicians should be mindful not only of the indications for administering fluid loads and of the type of fluids administered, but also of the importance to set safety limits. Lastly, it is important to implement proactive strategies seeking to establish negative fluid balance as soon as the clinical conditions are considered to be stable and the risk of deterioration is low.

Deresuscitation in Patients with Abdominal Sepsis Carries a Lower Mortality Rate and Less Organ Dysfunction than Conservative Fluid Management

Background: The relation between deresuscitative fluid management after the resuscitation phase and clinical outcome in patients with abdominal sepsis is not completely clear. The aim of this study was to assess the contribution of deresuscitative management to death and organ dysfunction in abdominal sepsis. Methods: Consecutive patients with abdominal sepsis requiring fluid resuscitation were included in this study. According to the fluid management given in the later stage of resuscitation, a conservative group and a deresuscitative fluid management group were compared. The primary outcome was in-hospital death, whereas secondary outcomes were categorized as organ dysfunction and other adverse events. Results: A total of 138 patients were enrolled in this study. Conservative fluid management was given to 47.8% of patients, whereas deresuscitative fluid management occurred in 52.2%. The deresuscitative strategy was associated with a markedly lower prevalence of new-onset acute kidney injury and a decrease in the duration of continuous renal replacement therapy (CRRT). There was a greater risk of needing new-onset intubation and the mechanical ventilation duration in the conservative group than in the deresuscitative group. However, the deresuscitative group did not differ from the conservative group with respect to open abdomen and intra-abdominal hypertension or new-onset abdominal compartment syndrome. The conservative treatment was associated with prolonged stays as well as a higher in-hospital mortality rate. A multivariable logistic regression model showed that deresuscitative fluid management imparts a protective effect against in-hospital death (odds ratio 4.343; 95% confidence interva1 1.466-12.866; p = 0.008), whereas septic shock, source control failure, and CRRT duration were associated with a higher mortality rate. Conclusions: Fluid balance achieved using deresuscitative treatment is correlated with better outcomes in patients with abdominal sepsis, indicating that this treatment may be useful as a therapeutic strategy.

Impact of postoperative intravenous fluid administration on complications following elective hepato-pancreato-biliary surgery

BACKGROUND

The impact of perioperative intravenous fluid administration on surgical outcomes has been documented in literature, but not specifically studied in the context of hepato-pancreato-biliary (HPB) surgery. This study aimed to investigate the impact of postoperative intravenous fluid administration on intensive care unit (ICU), in this subgroup of patients.

METHODS

A single-center retrospective cohort of 241 HPB patients was assessed, focusing on intravenous fluid administration in ICU, during the first 24 h. Intravenous fluid variables were compared to hospital stay and postoperative complications. Data were assessed using Spearman's correlation test for bivariate correlations and logistic regression for multivariate analysis.

RESULTS

The median volume of intravenous fluid administered in the first 24 h postoperatively was 4380 mL, of which 2200 mL was crystalloid, 1500 mL colloid and 680 mL "other" fluid. Patients with one or more complications had a higher median total intravenous fluid input (4790 vs. 4300 mL), higher colloid volume (2000 vs. 1500 mL), lower urine output (1595 vs. 1900 mL) and greater overall fluid balance (+3040 vs.+2553 mL) than those without complications. There were correlations between total intravenous fluid volume administered (r = 0.278, P < 0.001), intravenous colloid input (r = 0.278, P < 0.001), urine output (r = -0.295, P < 0.001), positive fluid balance (r = 0.344, P < 0.001) and length of hospital stay. Logistic regression model was constructed to predict the occurrence of one or more complications; total intravenous fluid volume and overall fluid balance were both independent significant predictors (OR = 2.463, P = 0.007; OR = 1.001, P = 0.011; respectively).

CONCLUSIONS

Administration of high volumes of intravenous fluids in the first 24 hours post-HPB surgery, along with higher positive fluid balance is associated with a higher rate of complications and longer hospital stay. Moreover, lower urine output is associated with longer hospital stay. Whether these are the cause of complications or the result of them remains unclear.

[Does fasting influence preload responsiveness in ASA 1 and 2 volunteers?]

INTRODUCTION

Preoperative fasting was long regarded as an important cause of fluid depletion, leading to hemodynamic instability during surgery should replenishment not be promptly instituted. Lately, this traditional point of view has been progressively challenged, and a growing number of authors now propose a more restrictive approach to fluid management, although doubt remains as to the true hemodynamic influence of preoperative fasting.

METHODS

We designed an observational, analytic, prospective, longitudinal study in which 31 ASA 1 and ASA 2 volunteers underwent an echocardiographic examination both before and after a fasting period of at least 6h. Data from both static and dynamic preload indices were obtained on both periods, and subsequently compared.

RESULTS

Static preload indices exhibited a markedly variable behaviour with fasting. Dynamic indices, however, were far more consistent with one another, all pointing in the same direction, i.e., evidencing no statistically significant change with the fasting period. We also analysed the reliability of dynamic indices to respond to known, intentional preload changes. Aortic velocity time integral (VTI) variation with the passive leg raise manoeuvre was the only variable that proved to be sensitive enough to consistently signal the presence of preload variation.

CONCLUSION

Fasting does not appear to cause a change in preload of conscious volunteers nor does it significantly alter their position in the Frank-Starling curve, even with longer fasting times than usually recommended. Transaortic VTI variation with the passive leg raise manoeuvre is the most robust dynamic index (of those studied) to evaluate preload responsiveness in spontaneously breathing patients.

Fluid Composition and Clinical Effects

The range of intravenous fluids available for therapeutic use and the differing indications are diverse. A solid understanding of the composition of different types of fluids is essential to understanding the physiologic effects following administration and the appropriate clinical application. In this review, the authors describe the different fluids commonly available and discuss the potential benefits and harms depending on the clinical circumstances.

Does fasting influence preload responsiveness in ASA 1 and 2 volunteers?

INTRODUCTION

Preoperative fasting was long regarded as an important cause of fluid depletion, leading to hemodynamic instability during surgery should replenishment is not promptly instituted. Lately, this traditional point of view has been progressively challenged, and a growing number of authors now propose a more restrictive approach to fluid management, although doubt remains as to the true hemodynamic influence of preoperative fasting.

METHODS

We designed an observational, analytic, prospective, longitudinal study in which 31 ASA 1 and ASA 2 volunteers underwent an echocardiographic examination both before and after a fasting period of at least 6hours (h). Data from both static and dynamic preload indices were obtained on both periods, and subsequently compared.

RESULTS

Static preload indices exhibited a markedly variable behaviour with fasting. Dynamic indices, however, were far more consistent with one another, all pointing in the same direction, i.e., evidencing no statistically significant change with the fasting period. We also analysed the reliability of dynamic indices to respond to known, intentional preload changes. Aortic velocity time integral (VTI) variation with the passive leg raise manoeuvre was the only variable that proved to be sensitive enough to consistently signal the presence of preload variation.

CONCLUSION

Fasting does not appear to cause a change in preload of conscious volunteers nor does it significantly alter their position in the Frank-Starling curve, even with longer fasting times than usually recommended. Transaortic VTI variation with the passive leg raise manoeuvre is the most robust dynamic index (of those studied) to evaluate preload responsiveness in spontaneously breathing patients.

The Economic and Humanistic Burden of Severe Sepsis

Sepsis and severe sepsis in particular remain a major health problem worldwide. Their cost to society extends well beyond lives lost, as the impact of survivorship is increasingly felt. A review of the medical literature was completed in MEDLINE using the search phrases "severe sepsis" and "septic shock" and the MeSH terms "epidemiology", "statistics", "mortality", "economics", and "quality of life". Results were limited to human trials that were published in English from 2002 to 2014. Articles were classified by dominant themes to address epidemiology and outcomes, including quality of life of both patient and family caregivers, as well as societal costs. The severity of sepsis is determined by the number of organ failures and the presence of shock. In most developed countries, severe sepsis and septic shock account for disproportionate mortality and resource utilization. Although mortality rates have decreased, overall mortality continues to increase and is projected to accelerate as people live longer with more chronic illness. Among those who do survive, impaired quality of life, increased dependence, and rehospitalization increase healthcare consumption and, along with increased mortality, all contribute to the humanistic burden of severe sepsis. A large part of the economic burden of severe sepsis occurs after discharge. Initial inpatient costs represent only 30 % of the total cost and are related to severity and length of stay, whereas lost productivity and other indirect medical costs following hospitalization account for the majority of the economic burden of sepsis. Timeliness of treatment as well as avoidance of intensive care unit (ICU)-acquired illness/morbidity lead to important differences in both cost and outcome of treatment for severe sepsis and represent areas where improvement in care is possible. The degree of sophistication of a health system from a national perspective results in significant differences in resource use and outcomes for patients with serious infections. Comprehensive understanding of the cost and humanistic burden of severe sepsis provides an initial practical framework for health policy development and resource use.

Volume expansion in the first 4 days of shock: a prospective multicentre study in 19 French intensive care units

PURPOSE

To describe the current practices of volume expansion in French intensive care units (ICU).

METHODS

In 19 ICUs, we prospectively observed the prescription and monitoring practices of volume expansion in consecutive adult patients with shock [sustained hypotension and/or need of vasopressor therapy, associated with at least tachycardia and/or sign (s) of hypoperfusion]. Patients were included at the time of prescription of the first fluid bolus (FB). Thereafter, all the FBs administered during the 96 h following shock onset were surveyed. An FB was defined as an intravenous bolus of at least 100 ml of a blood volume expander intended to rapidly improve the patient's circulatory condition.

RESULTS

We included 777 patients [age: 63 ± 15 years; female gender: 274 (35 %); simplified acute physiology score II: 55.9 ± 20.6; ICU length of stay: 6 days (interquartile range (IQR) 3-13); ICU mortality: 32.8 %] and surveyed 2,694 FBs. At enrolment mean arterial pressure was 63 mmHg (IQR 55-71). The most frequent triggers of FB were hypotension, low urine output, tachycardia, skin mottling and hyperlactataemia. Amount of fluid given at each FB was highly variable between centres. Crystalloids were used in 91 % (2,394/2,635) and synthetic colloids in 3.3 % (87/2,635) of FBs. Overall, clinicians used any kind of haemodynamic assessment (central venous pressure measurement, predictive indices of fluid responsiveness, echocardiography, cardiac output monitoring or a combination of these) in 23.6 % (635/2,694) of all FBs surveyed, with an important between-centre heterogeneity.

CONCLUSIONS

High between-centre variability characterised all the aspects of FB prescription and monitoring, but overall haemodynamic exploration to help guide and monitor FB was infrequent.

Evolution of fluid therapy

The human organism consists of evolutionary conserved mechanisms to prevent death from hypovolaemia. Intravenous fluid therapy to support these mechanisms had first been published about 180 years ago. The present review depicts the evolution of fluid therapy from early, not well-defined solutions up to modern balanced fluids. Notably, evidence accumulates that the most commonly used fluid (i.e. 0.9% saline) has no advantage over balanced solutions, increases the risk of acute kidney injury and should therefore be abandoned. Notably, in published trials, the prognostically important 'golden hours' of shock, where fluid therapy may be essential, have not been adequately addressed. It is therefore unclear whether negative data on colloids in some trials reflect real harm or rather inadequate use. Future studies should focus on optimal protocols for initiation, dosing and discontinuation of fluid therapy in specific disease entities. Moreover, the practice of de-resuscitation after fluid-based haemodynamic stabilization should be further investigated.

Fluids are drugs: type, dose and toxicity

PURPOSE OF REVIEW

We discuss the formulation of a prescription for intravenous (i.v.) fluid therapy (a 'volume prescription') for critically ill patients: pros/cons of different fluid types; accurate dosing; and qualitative and quantitative toxicities. Updated physiologic concepts are invoked and results of recent major clinical trials on i.v. fluid therapy in the acutely ill are interpreted.

RECENT FINDINGS

Context is vital and any fluid can be harmful if dosed incorrectly. When contrasting 'crystalloid versus colloid', differences in efficacy are modest, but differences in safety are significant. Differences in chloride load and strong ion difference appear to be clinically important. Quantitative toxicity is mitigated when dosing is based on dynamic parameters that predict volume responsiveness. Qualitative toxicity for colloids (even with newer hydroxyethyl starch 130/0.4 solutions) and isotonic saline remain a concern.

SUMMARY

Similar to any drug used in acutely ill patients, clinicians ordering a volume prescription must recognize that context is crucial. Physiologically balanced crystalloids may be the 'default' fluid for acutely ill patients, and the role for colloids is unclear. Optimal dosing involves assessment of volume responsiveness.

Incidence and prognosis of dysnatraemia in critically ill patients: analysis of a large prevalence study

BACKGROUND

The objective of this study is to assess the impact of dysnatraemia on mortality among intensive care unit (ICU) patients in a large, international cohort.

MATERIAL AND METHODS

Analysis of the Extended Prevalence of Infection in Intensive Care (EPIC II) study, a 1-day (8 May 2007) worldwide multicenter, prospective point prevalence study. Hyponatraemia was categorized as mild (130-134 mM/L), moderate (125-129 mM/L) or severe (< 125 mM/L). Hypernatraemia was also categorized as mild (146-150 mM/L), moderate (151-155 mM/L) or severe (> 155 mM/L). Patients with normal serum sodium (135-145 mM/L) constituted the reference group. The main outcome was hospital mortality. Analysis was conducted separately for patients admitted on the study day (25.8%) and those already present on the ICU (74.2%).

RESULTS

Serum sodium was measured in 13 276 of the 13 796 patients (96.2%). A total of 3815 patients (28.7%) had dysnatraemia: 12.9% with hyponatraemia and 15.8% with hypernatraemia. The prevalence of dysnatraemia was significantly greater in patients already present on the ICU prior to the study day than for those just admitted (13.1% vs. 12.3% for hyponatraemia and 17.1% vs. 12.1% for hypernatraemia, both P < 0.001). Hospital mortality rates were higher in patients with dysnatraemia than in those with normal sodium levels and were directly related to the severity of hypo- and hypernatraemia. This association between dysnatraemia and mortality was similar in infected and noninfected patients (P = 0.061).

CONCLUSIONS

Dysnatraemia is more frequent during the ICU stay than on the day of admission. Dysnatraemia in the ICU - even mild - is an independent predictor of increased hospital mortality.

Fluid management in cardiac surgery: colloid or crystalloid?

The crystalloid-colloid debate has raged for decades, with the publication of many meta-analyses, yet no consensus. There are important differences between colloids and crystalloids, and these differences have direct relevance for cardiac surgical patients. Rather than asking crystalloid or colloid, we believe better questions to ask are (1) High or low chloride content? and (2) Synthetic or natural colloid? In this paper we review the published literature regarding fluid therapy in cardiac surgery and explain the background to these two important and unanswered questions.