Detection of patients with acute kidney injury by the clinical laboratory using rises in serum creatinine: comparison of proposed definitions and a laboratory delta check

Machine Learning for Predicting Risk and Prognosis of Acute Kidney Disease in Critically Ill Elderly Patients During Hospitalization: Internet-Based and Interpretable Model Study

BACKGROUND

Acute kidney disease (AKD) affects more than half of critically ill elderly patients with acute kidney injury (AKI), which leads to worse short-term outcomes.

OBJECTIVE

We aimed to establish 2 machine learning models to predict the risk and prognosis of AKD in the elderly and to deploy the models as online apps.

METHODS

Data on elderly patients with AKI (n=3542) and AKD (n=2661) from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database were used to develop 2 models for predicting the AKD risk and in-hospital mortality, respectively. Data collected from Xiangya Hospital of Central South University were for external validation. A bootstrap method was used for internal validation to obtain relatively stable results. We extracted the indicators within 24 hours of the first diagnosis of AKI and the fluctuation range of some indicators, namely delta (day 3 after AKI minus day 1), as features. Six machine learning algorithms were used for modeling; the area under the receiver operating characteristic curve (AUROC), decision curve analysis, and calibration curve for evaluating; Shapley additive explanation (SHAP) analysis for visually interpreting; and the Heroku platform for deploying the best-performing models as web-based apps.

RESULTS

For the model of predicting the risk of AKD in elderly patients with AKI during hospitalization, the Light Gradient Boosting Machine (LightGBM) showed the best overall performance in the training (AUROC=0.844, 95% CI 0.831-0.857), internal validation (AUROC=0.853, 95% CI 0.841-0.865), and external (AUROC=0.755, 95% CI 0.699-0.811) cohorts. In addition, LightGBM performed well for the AKD prognostic prediction in the training (AUROC=0.861, 95% CI 0.843-0.878), internal validation (AUROC=0.868, 95% CI 0.851-0.885), and external (AUROC=0.746, 95% CI 0.673-0.820) cohorts. The models deployed as online prediction apps allowed users to predict and provide feedback to submit new data for model iteration. In the importance ranking and correlation visualization of the model's top 10 influencing factors conducted based on the SHAP value, partial dependence plots revealed the optimal cutoff of some interventionable indicators. The top 5 factors predicting the risk of AKD were creatinine on day 3, sepsis, delta blood urea nitrogen (BUN), diastolic blood pressure (DBP), and heart rate, while the top 5 factors determining in-hospital mortality were age, BUN on day 1, vasopressor use, BUN on day 3, and partial pressure of carbon dioxide (PaCO2).

CONCLUSIONS

We developed and validated 2 online apps for predicting the risk of AKD and its prognostic mortality in elderly patients, respectively. The top 10 factors that influenced the AKD risk and mortality during hospitalization were identified and explained visually, which might provide useful applications for intelligent management and suggestions for future prospective research.

The incidence and outcome of AKI in patients with sepsis in the emergency department applying different definitions of AKI and sepsis

BACKGROUND

Sepsis is often accompanied with acute kidney injury (AKI). The incidence of AKI in patients visiting the emergency department (ED) with sepsis according to the new SOFA criteria is not exactly known, because the definition of sepsis has changed and many definitions of AKI exist. Given the important consequences of early recognition of AKI in sepsis, our aim was to assess the epidemiology of sepsis-associated AKI using different AKI definitions (RIFLE, AKIN, AKIB, delta check, and KDIGO) for the different sepsis classifications (SIRS, qSOFA, and SOFA).

METHODS

We retrospectively enrolled patients with sepsis in the ED in three hospitals and applied different AKI definitions to determine the incidence of sepsis-associated AKI. In addition, the association between the different AKI definitions and persistent kidney injury, hospital length of stay, and 30-day mortality were evaluated.

RESULTS

In total, 2065 patients were included. The incidence of AKI was 17.7-51.1%, depending on sepsis and AKI definition. The highest incidence of AKI was found in qSOFA patients when the AKIN and KDIGO definitions were applied (51.1%). Applying the AKIN and KDIGO definitions in patients with sepsis according to the SOFA criteria, AKI was present in 37.3% of patients, and using the SIRS criteria, AKI was present in 25.4% of patients. Crude 30-day mortality, prolonged length of stay, and persistent kidney injury were comparable for patients diagnosed with AKI, regardless of the definition used.

CONCLUSION

The incidence of AKI in patients with sepsis is highly dependent on how patients with sepsis are categorised and how AKI is defined. When AKI (any definition) was already present at the ED, 30-day mortality was high (22.2%). The diagnosis of AKI in sepsis can be considered as a sign of severe disease and helps to identify patients at high risk of adverse outcome at an early stage.

[Laboratory diagnostics in medicine]

The development of clinical laboratory diagnostics is in line with the evidence-based medicine, which requires that clinical decisions have to be based on diagnostic methods with proven informativity. This creates a request for the scientific validity of the use of laboratory researches and application of probabilistic interpretation tools corresponding to the tasks. The concept of indefiniteness (analytical, biological and clinical) is at the heart of interpretation of laboratory results. The inclusion of laboratory research in clinical guidelines, the choice and appointment of this research to the patient should not be made from the position of ideas about increasing or decreasing the laboratory index in the disease, but on the basis of its scientifically proven characteristics as a laboratory biomarker sensitivity, specificity, predictive value, as well as the relationship with certain clinical events, outcomes, risks. These characteristics are probabilistic and can be defined.

National survey on delta checks in clinical laboratories in China

Background

This study aimed to understand the status quo of delta checks in Chinese clinical laboratories through a nationwide online survey.

Methods

The survey was divided into two parts. The first part was a general situation survey in which clinical laboratories had to provide information about the laboratories, including delta checks used. In the second part, clinical laboratories were asked to record the delta check alerts generated in their laboratories from June 1st, 2019 to June 30th, 2019.

Results

The most frequently used analytes in delta checks were potassium (K), glucose (Glu), creatinine (Cre) for clinical chemistry and hemoglobin (Hgb), platelet (PLT) count and white blood cell (WBC) count for clinical hematology. The median maximum time interval between specimens for all analytes was 5 days. The most commonly used delta check calculation modes in Chinese clinical laboratories were percentage change and absolute change. K and Hgb were the analytes most involved in clinical chemistry and clinical hematology delta check alerts. The most common causes of delta check alerts were that the patients had received treatment, which was followed by the change in the patient’s physiological state and interference from hemolysis, lipemia and icterus. The two most common outcomes of delta check alerts were ‘no problems found, standard report issued’ and ‘no problems found, report issued with comment’.

Conclusions

This study was the first nationwide survey of delta checks in China, the results of which help us to understand the current situation of delta checks in Chinese clinical laboratories.

Molecular nephrology: types of acute tubular injury

The acute loss of kidney function has been diagnosed for many decades using the serum concentration of creatinine - a muscle metabolite that is an insensitive and non-specific marker of kidney function, but is now used for the very definition of acute kidney injury (AKI). Fortunately, myriad new tools have now been developed to better understand the relationship between acute tubular injury and elevation in serum creatinine (SCr). These tools include unbiased gene and protein expression analyses in kidney, urine and blood, the localization of specific gene transcripts in pathological biopsy samples by rapid in-situ RNA technology and single-cell RNA-sequencing analyses. However, this molecular approach to AKI has produced a series of unexpected problems, because the expression of specific kidney-derived molecules that are indicative of injury often do not correlate with SCr levels. This discrepancy between kidney injury markers and SCr level can be reconciled by the recognition that many separate subtypes of AKI exist, each with distinct patterning of molecular markers of tubular injury and SCr data. In this Review, we describe the weaknesses of isolated SCr-based diagnoses, the clinical and molecular subtyping of acute tubular injury, and the role of non-invasive biomarkers in clinical phenotyping. We propose a conceptual model that synthesizes molecular and physiological data along a time course spanning from acute cellular injury to organ failure.

Design and evaluation of a LIS-based autoverification system for coagulation assays in a core clinical laboratory

Background

The autoverification system for coagulation consists of a series of rules that allow normal data to be released without manual verification. With new advances in medical informatics, the laboratory information system (LIS) has growing potential for the autoverification, allowing rapid and accurate verification of clinical laboratory tests. The purpose of the study is to develop and evaluate a LIS-based autoverification system for validation and efficiency.

Methods

Autoverification decision rules, including quality control, analytical error flag, critical value, limited range check, delta check and logical check, as well as patient’s historical information, were integrated into the LIS. Autoverification limited range was constructed based on 5 and 95% percentiles. The four most commonly used coagulation assays, prothrombin time (PT), activated partial thromboplastin time (APTT), thrombin time (TT), and fibrinogen (FBG), were followed by the autoverification protocols. The validation was assessed by the autoverification passing rate, the true-positive cases, the true-negative cases, the false-positive cases, the false-negative cases, the sensitivity and the specificity; the efficiency was evaluated in the turnaround time (TAT).

Results

A total of 157,079 historical test results of coagulation profiles from January 2016 to December 2016 were collected to determine the distribution intervals. The autoverification passing rate was 77.11% (29,165/37,821) based on historical patient data. In the initial test of the autoverification version in June 2017, the overall autoverification passing rate for the whole sample was 78.75% (11,257/14,295), with 892 true-positive cases, 11,257 true-negative cases, 2146 false-positive cases, no false-negative cases, sensitivity of 100% and specificity of 83.99%. After formal implementation of the autoverification system for 6 months, 83,699 samples were assessed. The average overall autoverification passing rate for the whole sample was 78.86% and the 95% confidence interval (CI) of the passing rate was [78.25, 79.59%]. TAT was reduced from 126 min to 101 min, which was statistically significant (P < 0.001, Mann-Whitney U test).

Conclusions

The autoverification system for coagulation assays based on LIS can halt the samples with abnormal values for manual verification, guarantee medical safety, minimize the requirements for manual work, shorten TAT and raise working efficiency.

Decision Support Tools within the Electronic Health Record

Laboratory tests are an integral part of the electronic health record (EHR). Providing clinical decision support (CDS) for the ordering, collection, reporting, viewing, and interpretation of laboratory testing is a fundamental function of the EHR. The implementation of a sustainable, effective laboratory CDS program requires a commitment to standardization and harmonization of the laboratory dictionaries that are the foundation of laboratory-based CDS. In this review, the authors provide an overview of the tools available within the EHR to improve decision making throughout the entire laboratory testing process, from test order to clinical action.

Impact of Daily Electronic Laboratory Alerting on Early Detection and Clinical Documentation of Acute Kidney Injury in Hospital Settings

Acute kidney injury, especially early-stage disease, is a common hospital comorbidity requiring timely recognition and treatment. We investigated the effect of daily laboratory alerting of patients at risk for acute kidney injury as measured by documented International Classification of Diseases diagnoses. A quasi-experimental study was conducted at 8 New York hospitals between January 1, 2014, and June 30, 2017. Education of clinical documentation improvement specialists, physicians, and nurses was conducted from July 1, 2014, to December 31, 2014, prior to initiating daily hospital-wide laboratory acute kidney injury alerting on January 1, 2015. Incidence based on documented International Classification of Diseases diagnosis of acute kidney injury and acute tubular necrosis during the intervention periods (3 periods of 6 months each: January 1 to June 30 of 2015, 2016, and 2017) were compared to one preintervention period (January 1, 2014, to June 30, 2014). The sample consisted of 269 607 adult hospital discharges, among which there were 39 071 episodes based on laboratory estimates and 27 660 episodes of documented International Classification of Diseases diagnoses of acute kidney injury or acute tubular necrosis. Documented incidence improved significantly from the 2014 preintervention period (5.70%; 95% confidence interval: 5.52%-5.88%) to intervention periods in 2015 (9.89%; 95% confidence interval, 9.66%-10.12%; risk ratio = 1.73, P < .001), 2016 (12.76%; 95% confidence interval, 12.51%-13.01%; risk ratio = 2.24, P < .001), and 2017 (12.49%; 95% confidence interval, 12.24%-12.74%; risk ratio = 2.19, P < .001). A multifactorial intervention comprising daily laboratory alerting and education of physicians, nurses, and clinical documentation improvement specialists led to increased recognition and clinical documentation of acute kidney injury.

Designing and evaluating autoverification rules for thyroid function profiles and sex hormone tests

Purpose Following the analytical phase, the current practice of many hospital laboratories involves the manual verification of all test results followed by the production of the report. However, manual verification is a time-consuming and tedious process. In this paper, we provide a detailed description of how to design autoverification rules for thyroid function test profiles and sex hormones. Materials and methods We used DM2 (Data manager 2) to construct the algorithm and build the database for autoverification of thyroid function test profiles and sex hormones, with reference to Boolean logic, Auto 10-A and CLSI'88. The rules consist of checking quality control, instrument error flags, critical values, the analytical measurement range (AMR), the limit range, consistency check and delta check. Firstly, we established the rules in the DM2, collected clinical specimens for validation, then tested the rules in a 'live' environment. Results Agreement was achieved between manual verification by two senior laboratory personnel and verification using the autoverification rules in 99.78% of the cases. The total autoverification rate for all tests was 77.06%. Following implementation of the rules, the laboratory turnaround time (TAT) was reduced by 54.55% and staffing numbers fell from three to two whole time equivalents (WTE). Statistical analysis resulted in a kappa statistic of 0.99 ( P < 0.001). Moreover, after implementing the autoverification rules, the error rate fell to 0.04%, indicating that errors were almost completely eliminated. Conclusion Implementing autoverification rules can reduce TAT, minimize the number of samples that require manual verification and allow for a reduction in staffing numbers. It also allows laboratory staff to devote more time and effort to the handling of problematic test results and contributing to improved patient care.

Determining the utility of creatinine delta checks: A large retrospective analysis

INTRODUCTION

Delta checks are a long-standing practice for identifying errors in the laboratory. However, with the decrease in errors due to laboratory automation, their utility is unclear. The objective of this retrospective analysis was to determine whether establishment of a creatinine delta check would be an effective means for capturing true laboratory error.

METHODS

All patients with a minimum of two creatinine results during March of 2015 were selected for review (n = 23,410 creatinine results). The lowest % change for a previously confirmed creatinine error in our laboratory was approximately 60%; therefore only results that changed by at least ±60% (n = 254) were reviewed. The etiology of creatinine value change was categorized as laboratory error, pathologic change, or non-pathologic change, based upon chart review.

RESULTS

1.2% (3/254) of reviewed delta checks were determined to reflect 2 instances of true laboratory error that went unrecognized by laboratory staff. 91.3% (232/254) of the delta checks were determined to reflect a pathologic or dialysis-related change in creatinine levels. The remaining 7.5% of delta checks (19/234) were deemed to be non-pathologic changes in creatinine.

DISCUSSION

This study identified two instances of laboratory error reflected by 3 delta checks (1.2%); the vast majority (91.3%) of creatinine results that changed by ±60% were pathologic or dialysis-related. Thus, establishment of a ±60% delta check for creatinine would overwhelmingly flag true biological change and would not be an efficient means for identifying rare laboratory errors. Clinical laboratories should perform similar retrospective analyses prior to enacting delta checks to determine whether they will effectively capture laboratory error.

Precision Medicine for Acute Kidney Injury (AKI): Redefining AKI by Agnostic Kidney Tissue Interrogation and Genetics

Acute kidney injury (AKI) currently is diagnosed by a temporal trend of a single blood analyte: serum creatinine. This measurement is neither sensitive nor specific to kidney injury or its protean forms. Newer biomarkers, neutrophil gelatinase-associated lipocalin (NGAL, Lipocalin 2, Siderocalin), or kidney injury molecule-1 (KIM-1, Hepatitis A Virus Cellular Receptor 1), accelerate the diagnosis of AKI as well as prospectively distinguish rapidly reversible from prolonged causes of serum creatinine increase. Nonetheless, these biomarkers lack the capacity to subfractionate AKI further (eg, sepsis versus ischemia versus nephrotoxicity from medications, enzymes, or metals) or inform us about the primary and secondary sites of injury. It also is unknown whether all nephrons are injured in AKI, whether all cells in a nephron are affected, and whether injury responses can be stimulus-specific or cell type-specific or both. In this review, we summarize fully agnostic tissue interrogation approaches that may help to redefine AKI in cellular and molecular terms, including single-cell and single-nuclei RNA sequencing technology. These approaches will empower a shift in the current paradigm of AKI diagnosis, classification, and staging, and provide the renal community with a significant advance toward precision medicine in the analysis AKI.

Pre-Stage Acute Kidney Injury Can Predict Mortality and Medical Costs in Hospitalized Patients

The significance of minimal increases in serum creatinine below the levels indicative of the acute kidney injury (AKI) stage is not well established. We aimed to investigate the influence of pre-stage AKI (pre-AKI) on clinical outcomes. We enrolled a total of 21,261 patients who were admitted to the Seoul National University Bundang Hospital from January 1, 2013 to December 31, 2013. Pre-AKI was defined as a 25–50% increase in peak serum creatinine levels from baseline levels during the hospital stay. In total, 5.4% of the patients had pre-AKI during admission. The patients with pre-AKI were predominantly female (55.0%) and had a lower body weight and lower baseline levels of serum creatinine (0.63 ± 0.18 mg/dl) than the patients with AKI and the patients without AKI (P < 0.001). The patients with pre-AKI had a higher prevalence of diabetes mellitus (25.1%) and malignancy (32.6%). The adjusted hazard ratio of in-hospital mortality for pre-AKI was 2.112 [95% confidence interval (CI), 1.143 to 3.903]. In addition, patients with pre-AKI had an increased length of stay (7.7 ± 9.7 days in patients without AKI, 11.4 ± 11.4 days in patients with pre-AKI, P < 0.001) and increased medical costs (4,061 ± 4,318 USD in patients without AKI, 4,966 ± 5,099 USD in patients with pre-AKI, P < 0.001) during admission. The adjusted hazard ratio of all-cause mortality for pre-AKI during the follow-up period of 2.0 ± 0.6 years was 1.473 (95% CI, 1.228 to 1.684). Although the adjusted hazard ratio of pre-AKI for overall mortality was not significant among the patients admitted to the surgery department or who underwent surgery, pre-AKI was significantly associated with mortality among the non-surgical patients (adjusted HR 1.542 [95% CI, 1.330 to 1.787]) and the patients admitted to the medical department (adjusted HR 1.384 [95% CI, 1.153 to 1.662]). Pre-AKI is associated with increased mortality, longer hospital stay, and increased medical costs during admission. More attention should be paid to the clinical significance of pre-AKI.

Development and Implementation of Autoverification Rules for ELISA Results of HBV Serological Markers

Autoverification is a process of using computer-based rules to verify clinical laboratory test results without manual review. But to date, there are few published articles on the use of autoverification over the course of years in a clinical laboratory. In our study, we firstly described the development and implementation of autoverification rules for enzyme-linked immunosorbent assay (ELISA) results of hepatitis B virus (HBV) serological markers in a clinical immunology laboratory. We designed the autoverification rules for HBV by using Boolean logic on five clinically used serological markers in accordance with the framework of AUTO-10A, issued by the American Clinical Laboratory Standards Institute in 2006. The rules were written into the laboratory information system (LIS) and installed in the computer, so we could use the LIS to screen the test results. If the results passed the autoverification rules, they could be sent to doctors immediately. To evaluate the autoverification rules, we applied the real-time data of 11,585 patients with the autoverification rules. The autoverification rate of the five HBV serological markers was 79.5%. Furthermore, the turnaround time (TAT) was reduced by 38% (78 minutes vs. 126 minutes). The error rate was nearly eliminated. These results show that using LIS with autoverification rules can shorten TAT, enhance efficiency, and reduce manual review errors.

Enhanced creatinine and estimated glomerular filtration rate reporting to facilitate detection of acute kidney injury

OBJECTIVES

While acute kidney injury (AKI) can be diagnosed based on specified increases in a patient's plasma creatinine level, standard creatinine reporting methods typically only flag creatinine results as abnormal when outside the reference range and often fail to identify rising creatinine values indicative of AKI. Here, we evaluate the impact of this limitation in standard creatinine reporting and develop and implement an enhanced creatinine reporting algorithm.

METHODS

We evaluated 59,712 plasma creatinine results collected over approximately 3 months, using computational simulations and statistical analyses.

RESULTS

Our analyses demonstrated that 29% of creatinine results substantially increased over the patient's baseline and concerning for AKI remained within the normal reference range. These concerning results would not be flagged as abnormal using standard reporting. Likewise, we found that simple delta checks are also insensitive at AKI detection. To improve creatinine reporting, we developed and implemented an algorithm within our laboratory information system to alert clinicians to rising creatinine results, which we describe in this report.

CONCLUSION

While both creatinine reference limits and simple delta checks are insensitive for AKI identification, a simple algorithm can be implemented within a common laboratory information system to enhance AKI identification.

The definition of acute kidney injury and its use in practice

Acute kidney injury (AKI) is a common syndrome that is independently associated with increased mortality. A standardized definition is important to facilitate clinical care and research. The definition of AKI has evolved rapidly since 2004, with the introduction of the Risk, Injury, Failure, Loss, and End-stage renal disease (RIFLE), AKI Network (AKIN), and Kidney Disease Improving Global Outcomes (KDIGO) classifications. RIFLE was modified for pediatric use (pRIFLE). They were developed using both evidence and consensus. Small rises in serum creatinine are independently associated with increased mortality, and hence are incorporated into the current definition of AKI. The recent definition from the international KDIGO guideline merged RIFLE and AKIN. Systematic review has found that these definitions do not differ significantly in their performance. Health-care staff caring for children or adults should use standard criteria for AKI, such as the pRIFLE or KDIGO definitions, respectively. These efforts to standardize AKI definition are a substantial advance, although areas of uncertainty remain. The new definitions have enabled the use of electronic alerts to warn clinicians of possible AKI. Novel biomarkers may further refine the definition of AKI, but their use will need to produce tangible improvements in outcomes and cost effectiveness. Further developments in AKI definitions should be informed by research into their practical application across health-care providers. This review will discuss the definition of AKI and its use in practice for clinicians and laboratory scientists.

Volumetric contrast-enhanced ultrasound imaging of renal perfusion

OBJECTIVES

To determine whether volumetric contrast-enhanced ultrasound (US) imaging has the potential to monitor changes in renal perfusion after vascular injury.

METHODS

Volumetric contrast-enhanced US uses a series of planar image acquisitions, capturing the nonlinear second harmonic signal from microbubble contrast agents flowing in the vasculature. Tissue perfusion parameters (peak intensity [IPK], time to peak intensity [TPK], wash-in rate [WIR], and area under the curve [AUC]) were derived from time-intensity curve data collected during in vitro flow phantom studies and in vivo animal studies of healthy and injured kidneys. For the flow phantom studies, either the contrast agent concentration was held constant (10 μL/L) with varying volumetric flow rates (10, 20, and 30 mL/min), or the flow rate was held constant (30 mL/min) with varying contrast agent concentrations (5, 10, and 20 μL/L). Animal studies used healthy rats or those that underwent renal ischemia-reperfusion injury. Renal studies were performed with healthy rats while the transducer angle was varied for each volumetric contrast-enhanced US image acquisition (reference or 0°, 45°, and 90°) to determine whether repeated renal perfusion measures were isotropic and independent of transducer position. Blood serum biomarkers and immunohistology were used to confirm acute kidney injury.

RESULTS

Flow phantom results revealed a linear relationship between microbubble concentrations injected into the flow system and the IPK, WIR, and AUC (R(2) > 0.56; P < .005). Furthermore, there was a linear relationship between volume flow rate changes and the TPK, WIR, and AUC (R(2) > 0.77; P < .005). No significant difference was found between the transducer angle during data acquisition and any of the perfusion measures (P > .60). After induction of renal ischemia-reperfusion injury in the rat animal model (n = 4), volumetric contrast-enhanced US imaging of the injured kidney revealed an initial reduction in renal perfusion compared to control animals, followed by progressive recovery of vascular function.

CONCLUSIONS

Volumetric contrast-enhanced US-based renal perfusion imaging may prove clinically feasible for detecting and monitoring acute kidney injury.