Approaches to Predicting Outcomes in Patients with Acute Kidney Injury

When to start renal replacement therapy in acute kidney injury: What are we waiting for?

Acute kidney injury remains a serious condition with a high mortality risk. In the absence of any new drugs, renal replacement therapy (RRT) is the most important treatment option. Randomized controlled trials have concluded that in critically ill patients without an emergency indication for RRT, a watchful waiting strategy is safe; however, further delays in RRT did not seem to confer any benefit, rather was associated with potential harm. During this process, balancing the risks of complications due to an unnecessary intervention with the risk of not correcting a potentially life-threatening complication remains a challenge. Dynamic renal function assessment, especially dynamic assessment of renal demand-capacity matching, combined with renal biomarkers such as neutrophil gelatinase-associated lipocalin and furosemide stress test, is helpful to identify which patients and when the patients may benefit from RRT.

Using artificial intelligence to predict mortality in AKI patients: a systematic review/meta-analysis

Background

Acute kidney injury (AKI) is associated with increased morbidity/mortality. With artificial intelligence (AI), more dynamic models for mortality prediction in AKI patients have been developed using machine learning (ML) algorithms. The performance of various ML models was reviewed in terms of their ability to predict in-hospital mortality for AKI patients.

Methods

A literature search was conducted through PubMed, Embase and Web of Science databases. Included studies contained variables regarding the efficacy of the AI model [the AUC, accuracy, sensitivity, specificity, negative predictive value and positive predictive value]. Only original studies that consisted of cross-sectional studies, prospective and retrospective studies were included, while reviews and self-reported outcomes were excluded. There was no restriction on time and geographic location.

Results

Eight studies with 37 032 AKI patients were included, with a mean age of 65.3 years. The in-hospital mortality was 18.0% in the derivation and 15.8% in the validation cohorts. The pooled [95% confidence interval (CI)] AUC was observed to be highest for the broad learning system (BLS) model [0.852 (0.820-0.883)] and elastic net final (ENF) model [0.852 (0.813-0.891)], and lowest for proposed clinical model (PCM) [0.765 (0.716-0.814)]. The pooled (95% CI) AUC of BLS and ENF did not differ significantly from other models except PCM [Delong's test P = .022]. PCM exhibited the highest negative predictive value, which supports this model's use as a possible rule-out tool.

Conclusion

Our results show that BLS and ENF models are equally effective as other ML models in predicting in-hospital mortality, with variability across all models. Additional studies are needed.

Endothelium-related biomarkers enhanced prediction of kidney support therapy in critically ill patients with non-oliguric acute kidney injury

Acute kidney injury (AKI) is a common condition in hospitalized patients who often requires kidney support therapy (KST). However, predicting the need for KST in critically ill patients remains challenging. This study aimed to analyze endothelium-related biomarkers as predictors of KST need in critically ill patients with stage 2 AKI. A prospective observational study was conducted on 127 adult ICU patients with stage 2 AKI by serum creatinine only. Endothelium-related biomarkers, including vascular cell adhesion protein-1 (VCAM-1), angiopoietin (AGPT) 1 and 2, and syndecan-1, were measured. Clinical parameters and outcomes were recorded. Logistic regression models, receiver operating characteristic (ROC) curves, continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were used for analysis. Among the patients, 22 (17.2%) required KST within 72 h. AGPT2 and syndecan-1 levels were significantly greater in patients who progressed to the KST. Multivariate analysis revealed that AGPT2 and syndecan-1 were independently associated with the need for KST. The area under the ROC curve (AUC-ROC) for AGPT2 and syndecan-1 performed better than did the constructed clinical model in predicting KST. The combination of AGPT2 and syndecan-1 improved the discrimination capacity of predicting KST beyond that of the clinical model alone. Additionally, this combination improved the classification accuracy of the NRI and IDI. AGPT2 and syndecan-1 demonstrated predictive value for the need for KST in critically ill patients with stage 2 AKI. The combination of AGPT2 and syndecan-1 alone enhanced the predictive capacity of predicting KST beyond clinical variables alone. These findings may contribute to the early identification of patients who will benefit from KST and aid in the management of AKI in critically ill patients.

Development and external validation of multimodal postoperative acute kidney injury risk machine learning models

Objectives

To develop and externally validate machine learning models using structured and unstructured electronic health record data to predict postoperative acute kidney injury (AKI) across inpatient settings.

Materials and Methods

Data for adult postoperative admissions to the Loyola University Medical Center (2009-2017) were used for model development and admissions to the University of Wisconsin-Madison (2009-2020) were used for validation. Structured features included demographics, vital signs, laboratory results, and nurse-documented scores. Unstructured text from clinical notes were converted into concept unique identifiers (CUIs) using the clinical Text Analysis and Knowledge Extraction System. The primary outcome was the development of Kidney Disease Improvement Global Outcomes stage 2 AKI within 7 days after leaving the operating room. We derived unimodal extreme gradient boosting machines (XGBoost) and elastic net logistic regression (GLMNET) models using structured-only data and multimodal models combining structured data with CUI features. Model comparison was performed using the receiver operating characteristic curve (AUROC), with Delong's test for statistical differences.

Results

The study cohort included 138 389 adult patient admissions (mean [SD] age 58 [16] years; 11 506 [8%] African-American; and 70 826 [51%] female) across the 2 sites. Of those, 2959 (2.1%) developed stage 2 AKI or higher. Across all data types, XGBoost outperformed GLMNET (mean AUROC 0.81 [95% confidence interval (CI), 0.80-0.82] vs 0.78 [95% CI, 0.77-0.79]). The multimodal XGBoost model incorporating CUIs parameterized as term frequency-inverse document frequency (TF-IDF) showed the highest discrimination performance (AUROC 0.82 [95% CI, 0.81-0.83]) over unimodal models (AUROC 0.79 [95% CI, 0.78-0.80]).

Discussion

A multimodality approach with structured data and TF-IDF weighting of CUIs increased model performance over structured data-only models.

Conclusion

These findings highlight the predictive power of CUIs when merged with structured data for clinical prediction models, which may improve the detection of postoperative AKI.

The use of diagnostic tools for pediatric AKI: applying the current evidence to the bedside

Given the known deleterious consequences of acute kidney injury (AKI), exciting recent research efforts have focused on developing strategies for the earlier recognition of AKI in the pediatric population. Recognizing the limitations of serum creatinine, investigators have focused on the study of novel biomarkers and practical bedside tools for identifying patients at risk for AKI prior to a rise in serum creatinine. In PubMed, there are presently over 30 original research papers exploring the use of pediatric AKI risk prediction tools in just the last 2 years. The following review highlights the most recent advances in the literature regarding opportunities to refine our ability to detect AKI early. Importantly, this review discusses how prediction tools including novel urine and serum biomarkers, practical risk stratification tests, renal functional reserve, and electronic medical record alerts may ultimately be applied to routine clinical practice.

Timing of Kidney Support Therapy in Acute Kidney Injury: What Are We Waiting For?

The optimal timing of kidney support therapy in critically ill patients with acute kidney injury (AKI) without life-threatening complications related to AKI is controversial. Recent multicenter, randomized, controlled studies have questioned the need for earlier initiation of therapy, despite one study showing a benefit in survival and others with no differences in mortality based on the timing of kidney support therapy initiation. These findings reflect the uncertainties in decisions to initiate kidney support therapy, which should ideally be individualized according to the patient's comorbidities, severity of illness, trajectory of kidney function, and urine output as well as requirements for fluid balance and solute removal. A delayed approach could translate into a potentially reduced burden of dialysis dependence in addition to saving health resources. However, we must ascertain what constitutes the waiting period and the benefits and risks associated with this approach. This article reviews the concept of timing of dialysis in AKI, performs a critical assessment of the most important clinical trials in this topic, discusses ongoing research and knowledge gaps, and defines key research issues to address in the future.

Determinants of Outcomes of Acute Kidney Injury: Clinical Predictors and Beyond

Acute kidney injury (AKI) is a common clinical syndrome characterized by rapid impairment of kidney function. The incidence of AKI and its severe form AKI requiring dialysis (AKI-D) has been increasing over the years. AKI etiology may be multifactorial and is substantially associated with increased morbidity and mortality. The outcome of AKI-D can vary from partial or complete recovery to transitioning to chronic kidney disease, end stage kidney disease, or even death. Predicting outcomes of patients with AKI is crucial as it may allow clinicians to guide policy regarding adequate management of this problem and offer the best long-term options to their patients in advance. In this manuscript, we will review the current evidence regarding the determinants of AKI outcomes, focusing on AKI-D.

Artificial intelligence to guide management of acute kidney injury in the ICU: a narrative review

PURPOSE OF REVIEW

Acute kidney injury (AKI) frequently complicates hospital admission, especially in the ICU or after major surgery, and is associated with high morbidity and mortality. The risk of developing AKI depends on the presence of preexisting comorbidities and the cause of the current disease. Besides, many other parameters affect the kidney function, such as the state of other vital organs, the host response, and the initiated treatment. Advancements in the field of informatics have led to the opportunity to store and utilize the patient-related data to train and validate models to detect specific patterns and, as such, predict disease states or outcomes.

RECENT FINDINGS

Machine-learning techniques have also been applied to predict AKI, as well as the patients' outcomes related to their AKI, such as mortality or the need for kidney replacement therapy. Several models have recently been developed, but only a few of them have been validated in external cohorts.

SUMMARY

In this article, we provide an overview of the machine-learning prediction models for AKI and its outcomes in critically ill patients and individuals undergoing major surgery. We also discuss the pitfalls and the opportunities related to the implementation of these models in clinical practices.

Machine Learning in the Prediction of Medical Inpatient Length of Stay

BACKGROUND

Length of stay (LOS) estimates are important for patients, doctors and hospital administrators. However, making accurate estimates of LOS can be difficult for medical patients.

AIMS

This review was conducted with the aim of identifying and assessing previous studies on the application of machine learning to the prediction of total hospital inpatient LOS for medical patients.

METHODS

A review of machine learning in the prediction of total hospital LOS for medical inpatients was conducted using the databases PubMed, EMBASE and Web of Science.

RESULTS

Of the 673 publications returned by the initial search, 21 articles met inclusion criteria. Of these articles the most commonly represented medical specialty was cardiology. Studies were also identified that had specifically evaluated machine learning LOS prediction in patients with diabetes and tuberculosis. The performance of the machine learning models in the identified studies varied significantly depending on factors including differing input datasets and different LOS thresholds and outcome metrics. Common methodological shortcomings included a lack of reporting of patient demographics and lack of reporting of clinical details of included patients.

CONCLUSIONS

The variable performance reported by the studies identified in this review supports the need for further research of the utility of machine learning in the prediction of total inpatient LOS in medical patients. Future studies should follow and report a more standardised methodology to better assess performance and to allow replication and validation. In particular, prospective validation studies and studies assessing the clinical impact of such machine learning models would be beneficial. This article is protected by copyright. All rights reserved.

Predicting acute kidney injury in critically ill patients using comorbid conditions utilizing machine learning

Background

Acute kidney injury (AKI) carries a poor prognosis. Its incidence is increasing in the intensive care unit (ICU). Our purpose in this study is to develop and externally validate a model for predicting AKI in the ICU using patient data present prior to ICU admission.

Methods

We used data of 98 472 adult ICU admissions at Mayo Clinic between 1 January 2005 and 31 December 2017 and 51 801 encounters from Medical Information Mart for Intensive Care III (MIMIC-III) cohort. A gradient-boosting model was trained on 80% of the Mayo Clinic cohort using a set of features to predict AKI acquired in the ICU.

Results

AKI was identified in 39 307 (39.9%) encounters in the Mayo Clinic cohort. Patients who developed AKI in the ICU were older and had higher ICU and in-hospital mortality compared to patients without AKI. A 30-feature model yielded an area under the receiver operating curve of 0.690 [95% confidence interval (CI) 0.682–0.697] in the Mayo Clinic cohort set and 0.656 (95% CI 0.648–0.664) in the MIMIC-III cohort.

Conclusions

Using machine learning, AKI among ICU patients can be predicted using information available prior to admission. This model is independent of ICU information, making it valuable for stratifying patients at admission.

Comparing Machine Learning Algorithms for Predicting Acute Kidney Injury

Prior studies have used vital signs and laboratory measurements with conventional modeling techniques to predict acute kidney injury (AKI). The purpose of this study was to use the trend in vital signs and laboratory measurements with machine learning algorithms for predicting AKI in ICU patients. The eICU Collaborative Research Database was queried for five consecutive days of laboratory measurements per patient. Patients with AKI were identified and trends in vital signs and laboratory values were determined by calculating the slope of the least-squares-fit linear equation using three days for each value. Different machine learning classifiers (gradient boosted trees [GBT], logistic regression, and deep learning) were trained to predict AKI using the laboratory values, vital signs, and slopes. There were 151,098 ICU stays identified and the rate of AKI was 5.6 per cent. The best performing algorithm was GBT with an AUC of 0.834 ± 0.006 and an F-measure of 42.96 per cent ± 1.26 per cent. Logistic regression performed with an AUC of 0.827 ± 0.004 and an F-measure of 28.29 per cent ± 1.01 per cent. Deep learning performed with an AUC of 0.817 ± 0.005 and an F-measure of 42.89 per cent ± 0.91 per cent. The most important variable for GBT was the slope of the minimum creatinine (30.32%). This study identifies the best performing machine learning algorithms for predicting AKI using trends in laboratory values in ICU patients. Early identification of these patients using readily available data indicates that incorporating machine learning predictive models into electronic medical record systems is an inevitable requisite for improving patient outcomes.

KNOWLEDGE AND CARE PRACTICE OF NURSES OF INTENSIVE CARE UNITS REGARDING ACUTE KIDNEY INJURY

ABSTRACT Objective: to evaluate the knowledge and care practice of nurses in the care of patients with acute kidney injury in an intensive care unit. Method: cross-sectional study with 136 nurses from seven large public hospitals. Knowledge was measured by a questionnaire with 25 objective questions; and care practice, by a checklist with 15 questions. The instrument was created for this research and evaluated by judges regarding reliability, criterion and construct. Correlation tests, bivariate and multivariate analyses were used for data analysis. Results: the percentage of nurses' knowledge about acute kidney injury was 44.96%. The questions with the highest rates of correct answers dealt with nursing care. The percentage of execution of the practice was 47.54%. The most complete care was: applies protocol if the patient becomes hypotensive (89.7%); and checks skin condition, respiratory pattern and peripheral perfusion in complications (88.2%). Regarding professional data, it was observed that having a specialization in intensive care (p=0.034) and attending nephrology in specialization (p=0.030) were determining factors for greater knowledge, while specialization in intensive care (p=0.019) was a determining factor for practice. Conclusion: nurses obtained inadequate knowledge and care practice. It was observed that professionals with specialization in intensive care who attended a discipline or training in the area of nephrology showed better knowledge and care practices, when compared to those who did not. These data contribute to the construction of institutional policies that prioritize permanent education strategies in intensive care units.

Plasma protein C levels are directly associated with better outcomes in patients with severe burns

Protein C circulates in human plasma to regulate inflammation and coagulation. It has shown a crucial role in wound healing in animals, and low plasma levels predict the presence of a wound in diabetic patients. However, no detailed study has measured protein C levels in patients with severe burns over the course of a hospital admission. A severe burn is associated with dysfunction of inflammation and coagulation as well as a significant risk of morbidity and mortality. The current methods of burn assessment have shortcomings in reliability and have limited prognostic value. The discovery of a biomarker that estimates burn severity and predicts clinical events with greater accuracy than current methods may improve management, resource allocation and patient counseling. This is the first study to assess the potential role of protein C as a biomarker of burn severity. We measured the plasma protein C levels of 86 patients immediately following a severe burn, then every three days over the first three weeks of a hospital admission. We also analysed the relationships between burn characteristics, blood test results including plasma protein C levels and clinical events. We used a primary composite outcome of increased support utilisation defined as: a mean intravenous fluid administration volume of five litres or more per day over the first 72 h of admission, a length of stay in the intensive care unit of more than four days, or greater than four surgical procedures during admission. The hypothesis was that low protein C levels would be negatively associated with increased support utilisation. At presentation to hospital after a severe burn, the mean plasma protein C level was 76 ± 20% with a range of 34-130% compared to the normal range of 70-180%. The initial low can be plausibly explained by impaired synthesis, increased degradation and excessive consumption of protein C following a burn. Levels increased gradually over six days then remained at a steady-state until the end of the inpatient study period, day 21. A multivariable regression model (Nagelkerke's R² = 0.83) showed that the plasma protein C level on admission contributed the most to the ability of the model to predict increased support utilisation (OR = 0.825 (95% CI = 0.698-0.977), P = 0.025), followed by burn size (OR = 1.252 (95% CI = 1.025-1.530), P = 0.027), burn depth (partial thickness was used as the reference, full thickness OR = 80.499 (1.569-4129.248), P = 0.029), and neutrophil count on admission (OR = 1.532 (95% CI = 0.950-2.473), P = 0.08). Together, these four variables predicted increased support utilisation with 93.2% accuracy, 83.3% sensitivity and 97.6% specificity. However if protein C values were disregarded, only 49.5% of the variance was explained, with 82% accuracy, 63% sensitivity and 91.5% specificity. Thus, protein C may be a useful biomarker of burn severity and study replication will enable validation of these novel findings.

Electronic health records accurately predict renal replacement therapy in acute kidney injury

Background

Electronic health records (EHR) detect the onset of acute kidney injury (AKI) in hospitalized patients, and may identify those at highest risk of mortality and renal replacement therapy (RRT), for earlier targeted intervention.

Methods

Prospective observational study to derive prediction models for hospital mortality and RRT, in inpatients aged ≥18 years with AKI detected by EHR over 1 year in a tertiary institution, fulfilling modified KDIGO criterion based on serial serum creatinine (sCr) measures.

Results

We studied 3333 patients with AKI, of 77,873 unique patient admissions, giving an AKI incidence of 4%. KDIGO AKI stages at detection were 1(74%), 2(15%), 3(10%); corresponding peak AKI staging in hospital were 61, 20, 19%. 392 patients (12%) died, and 174 (5%) received RRT. Multivariate logistic regression identified AKI onset in ICU, haematological malignancy, higher delta sCr (sCr rise from AKI detection till peak), higher serum potassium and baseline eGFR, as independent predictors of both mortality and RRT. Additionally, older age, higher serum urea, pneumonia and intraabdominal infections, acute cardiac diseases, solid organ malignancy, cerebrovascular disease, current need for RRT and admission under a medical specialty predicted mortality. The AUROC for RRT prediction was 0.94, averaging 0.93 after 10-fold cross-validation. Corresponding AUROC for mortality prediction was 0.9 and 0.9 after validation. Decision tree analysis for RRT prediction achieved a balanced accuracy of 70.4%, and identified delta-sCr ≥ 148 μmol/L as the key factor that predicted RRT.

Conclusion

Case fatality was high with significant renal deterioration following hospital-wide AKI. EHR clinical model was highly accurate for both RRT prediction and for mortality; allowing excellent risk-stratification with potential for real-time deployment.

Electronic supplementary material

The online version of this article (10.1186/s12882-019-1206-4) contains supplementary material, which is available to authorized users.

The Development of a Machine Learning Inpatient Acute Kidney Injury Prediction Model

OBJECTIVES

To develop an acute kidney injury risk prediction model using electronic health record data for longitudinal use in hospitalized patients.

DESIGN

Observational cohort study.

SETTING

Tertiary, urban, academic medical center from November 2008 to January 2016.

PATIENTS

All adult inpatients without pre-existing renal failure at admission, defined as first serum creatinine greater than or equal to 3.0 mg/dL, International Classification of Diseases, 9th Edition, code for chronic kidney disease stage 4 or higher or having received renal replacement therapy within 48 hours of first serum creatinine measurement.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Demographics, vital signs, diagnostics, and interventions were used in a Gradient Boosting Machine algorithm to predict serum creatinine-based Kidney Disease Improving Global Outcomes stage 2 acute kidney injury, with 60% of the data used for derivation and 40% for validation. Area under the receiver operator characteristic curve (AUC) was calculated in the validation cohort, and subgroup analyses were conducted across admission serum creatinine, acute kidney injury severity, and hospital location. Among the 121,158 included patients, 17,482 (14.4%) developed any Kidney Disease Improving Global Outcomes acute kidney injury, with 4,251 (3.5%) developing stage 2. The AUC (95% CI) was 0.90 (0.90-0.90) for predicting stage 2 acute kidney injury within 24 hours and 0.87 (0.87-0.87) within 48 hours. The AUC was 0.96 (0.96-0.96) for receipt of renal replacement therapy (n = 821) in the next 48 hours. Accuracy was similar across hospital settings (ICU, wards, and emergency department) and admitting serum creatinine groupings. At a probability threshold of greater than or equal to 0.022, the algorithm had a sensitivity of 84% and a specificity of 85% for stage 2 acute kidney injury and predicted the development of stage 2 a median of 41 hours (interquartile range, 12-141 hr) prior to the development of stage 2 acute kidney injury.

CONCLUSIONS

Readily available electronic health record data can be used to predict impending acute kidney injury prior to changes in serum creatinine with excellent accuracy across different patient locations and admission serum creatinine. Real-time use of this model would allow early interventions for those at high risk of acute kidney injury.

Phenotyping of Acute Kidney Injury: Beyond Serum Creatinine

Acute kidney injury (AKI) is a common complication in hospitalized patients and is associated with adverse short- and long-term outcomes. AKI is diagnosed by serum creatinine (SCr)-based consensus definitions that capture an abrupt decrease in glomerular filtration rate associated with AKI. However, SCr-based AKI definitions lack sensitivity and specificity for diagnosing structural kidney injury. Moreover, AKI is a heterogeneous condition consisting of distinct phenotypes based on its etiology, prognosis, and molecular pathways, and that may potentially require different therapies. SCr-based AKI definitions provide no information on these AKI phenotypes. This review highlights traditional and novel tools that overcome the limitations of SCr-based AKI definitions to improve AKI phenotyping.

Infrequent Provision of Palliative Care to Patients with Dialysis-Requiring AKI

BACKGROUND AND OBJECTIVES

The use of palliative care in AKI is not well described. We sought to better understand palliative care practice patterns for hospitalized patients with AKI requiring dialysis in the United States.

DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS

Using the 2012 National Inpatient Sample, we identified patients with AKI and palliative care encounters using validated International Classification of Diseases, Ninth Revision, Clinical Modification codes. We compared palliative care encounters in patients with AKI requiring dialysis, patients with AKI not requiring dialysis, and patients without AKI. We described the provision of palliative care in patients with AKI requiring dialysis and compared the frequency of palliative care encounters for patients with AKI requiring dialysis with that for patients with other illnesses with similarly poor prognoses. We used logistic regression to determine factors associated with the provision of palliative care, adjusting for demographics, hospital-level variables, and patient comorbidities.

RESULTS

We identified 3,031,036 patients with AKI, of whom 91,850 (3%) received dialysis. We observed significant patient- and hospital-level differences in the provision of palliative care for patients with AKI requiring dialysis; adjusted odds were 26% (95% confidence interval, 12% to 38%) lower in blacks and 23% (95% confidence interval, 3% to 39%) lower in Hispanics relative to whites. Lower provision of palliative care was observed for rural and urban nonteaching hospitals relative to urban teaching hospitals, small and medium hospitals relative to large hospitals, and hospitals in the Northeast compared with the South. After adjusting for age and sex, there was low utilization of palliative care services for patients with AKI requiring dialysis (8%)-comparable with rates of utilization by patients with other illnesses with poor prognosis, including cardiogenic shock (9%), intracranial hemorrhage (10%), and acute respiratory distress syndrome (10%).

CONCLUSIONS

The provision of palliative care varied widely by patient and facility characteristics. Palliative care was infrequently used in hospitalized patients with AKI requiring dialysis, despite its poor prognosis and the regular application of life-sustaining therapy.