Delta Checks Research Articles

Establishment and application of autoverification system for HbA1c testing.

This study aimed to determine autoverification rules for routine glycated hemoglobin (HbA1c) analysis based on high-performance liquid chromatography (HPLC) principle. Laboratory information system (LIS) and Bio-Rad D-100 Advisor software (Bio-Rad, Hercules, USA) with graphics recognition function were carriers for the autoverification system. A total of 105,126 HbA1c results, including 98,249 HbA1c matching fast plasma glucose (FPG) results of real-world data from May 2019 to June 2020, were collected to determine autoverification rules including flags, delta checks, reporting limits, and logical rules. The validation database was composed of 48,045 HbA1c results and 41,083 matching FPG results. Autoverification passing rate and the reduction of turnaround time (TAT) were evaluated. Four autoverification systems (A, B, C, D) were established by two types of delta check rules, 28 flags, one reporting limits, and two kinds of logical rules. The autoverification passing rates were 80.6%, 78.8%, 83.7%, and 81.3%, and the average time saved in TAT were 117.5 min, 116.7 min, 121.1 min, and 121.7 min, respectively. Autoverification system C was the optimal one. Application of distribution of FPG corresponding to HbA1c groups had better performance as logical rules. Established HbA1c autoverifcation system shortened the auditing report time and improved work efficiency.

B-108 Delta Check Performance Assessment by Using Real-WorldData at the Temple University Health System​

Abstract Background Delta check is a post-analytic tool focusing on pre-analytical, analytical and post-analytical errors. It can be sensitive in detecting isolated errors affecting a single patient, as opposed to moving patient averages, which is another QC tool for detecting errors that affect a cohort/population of patients. As a computer-based system on selected tests, delta checks are to detect excessive changes from prior value in the patient test result and to detect possible specimen errors such as misidentified or compromised samples before results are reported. It may also play a role in detecting examination(analytical) issues and clinically significant results. If delta checks exceed established limits, the result is flagged. There is a common agreement that measurands with low indices of individuality (<0.6) make good candidates for the identification of cases of sample misidentification than higher indices of individuality (>1.4). CLSI EP33 Use of Delta Checks in the Medical Laboratory guidelines recommends that laboratories assess the validity of current delta check parameters by reviewing retrospective data on flagged results to determine if errors are being identified efficiently. At Temple University Health System, delta check limits were established historically by using published data, however, its performance characteristics have not been verified previously. Verification of clinical performance of predefined delta check limits can be performed either through mathematical calculations or via clinical audits. Clinical audits are achieved by documenting the follow-up actions and outcomes of the samples flagged by a delta check rule(s). It has been suggested that this should be performed using a protocolized follow-up procedure to ensure uniformity in the investigation and management. Corey M. et al. shared an excel template, a mathematical model that can be adopted easily by replacing the example data with retrospectively extracted data and modifying the parameters to suit their laboratory’s requirements. Methods We collected up to 30,000 data points for each 12 measurand in the clinical chemistry lab retrospectively. By using the excel template, clinical sensitivity and clinical specificity has been calculated. Furthermore, we created ROC curves and evaluated them using the (YI) Youden Index and AUC (Area Under the Curve). Results In general, YI values were found to be very low, less than 0.6. Highest YI was used to establish new delta check limits for each analyte; only the current creatinine delta check limit was found as effective, and no change required. 11 out of 12 measurand were required a revision of delta check limits and were proposed to replace with lower delta checks. AUC values were found between 0.56-0.794. Na showed the highest AUC, which correlates its low individuality index (0.5) and its effects on effectiveness of the delta check system. Conclusions It is of vital importance to establish a solid follow-up evaluation for delta check alerts and distinguish between integrity issues and clinically significant changes. Conclusively, this endeavor can develop into extensive information for the application, evaluation, and follow-up of delta checks whilst improving diagnostic accuracy and quality assurance.

Establishment and discussion of autoverification rules for transfusion compatibility testing.

To develop an automated verification workflow for transfusion compatibility testing (TCT) based on the AUTO10-A guidelines and blood group serology characteristics and to conduct a simulated validation of the test and subtest results by assessing the appropriateness of the autoverification rules. The accuracy of TCT results is a fundamental prerequisite for ensuring the safety of blood transfusions. However, the verification of these results still requires manual intervention. Five autoverification rules and their standards were determined: agglutination intensity, normal results, logical relationships, delta checks and interlaboratory test comparisons. The established categories and standards for the five rules were retrospectively validated using 13 506 samples (requests) that had been manually verified in our laboratory from January 2020 to June 2023. A total of 66 638 test items were involved in the autoverification, with 3844 items violating the verification rules, resulting in a pass rate of 96.10%. Considering individual test items, four tests had a pass rate of more than 90% in both the test item result table and the test subitem result table. However, there were significant differences in the pass rates between different tests. The same conclusion can be drawn when the unit is requests. The different standards set for the agglutination intensity and the delta check in the ABO typing testing subitems showed significant differences in pass rates. The incorporation of manually verified results into the automated verification simulation indicated that the five rules established in this study have good applicability, and appropriate standards can lead to reasonable pass rates.

The preliminary study of delta checks for immunoglobulins and complements in the clinical laboratory.

To determine delta check limits for immunoglobulins and complements in outpatients and inpatients based on patient data and biological variation due to the lack of relevant studies. Patient data for IgA, IgG, IgM, IgE, C3, and C4 from 1 January 2022 to 31 December 2023 was collected from laboratory information system (LIS) in our clinical laboratory of Wuhan Union Hospital, which includes both outpatients and inpatients. The delta difference (DD), delta percent change (DPC), and reference change value (RCV) were calculated based on patient data and biological variation. For DDs, there are significant differences between outpatients and inpatients in C4, IgE, IgG, and IgM. For DPCs, the corresponding analytes which are significantly different are C3, C4, IgE, IgG, and IgM. Two sources of CVI to calculate the RCV of IgA, IgG, IgM, C3, and C4 were applied in this study, which revealed that two kinds of RCVs based on different biological variation databases are similar to each other, but both were smaller than delta check limits based on patient data, except for C4. The delta check is a useful tool to monitor potential errors which may occur in total testing process. We hope our findings could be helpful for future studies focused on delta checks in immunological analytes.

Selection of Single-Analyte Delta Check Rules with Logistic Regression for Detection of Intravenous Fluid Contamination in a Clinical Chemistry Laboratory.

The conventional single-analyte delta check, utilized for identifying intravenous fluid contamination and other preanalytical errors, is known to flag many specimens reflecting true patient status changes. This study aimed to derive delta check rules that more accurately identify contamination. Results for calcium, creatinine, glucose, sodium, and potassium were retrieved from 326 103 basic or comprehensive metabolic panels tested between February 2021 and January 2022. In total, 7934 specimens showed substantial result changes, of which 1489 were labeled as either contaminated or non-contaminated based on chart review. These labeled specimens were used to derive logistic regression models and to select the most predictive single-analyte delta checks for 4 common contaminants. Their collective performance was evaluated using a test data set from October 2023 comprising 14 717 specimens. The most predictive single-analyte delta checks included a calcium change by ≤-24% for both saline and Plasma-Lyte A contamination, a potassium increase by ≥3.0 mmol/L for potassium contamination, and a glucose increase by ≥400 mg/dL (22.2 mmol/L) for dextrose contamination. In the training data sets, multi-analyte logistic regression models performed better than single-analyte delta checks. In the test data set, logistic regression models and single-analyte delta checks demonstrated collective alert rates of 0.58% (95% CI, 0.46%-0.71%) and 0.60% (95% CI, 0.49%-0.74%), respectively, along with collective positive predictive values of 79% (95% CI, 70%-89%) and 77% (95% CI, 68%-87%). Single-analyte delta checks selected by logistic regression demonstrated a low false alert rate.

New solutions to old problems: A practical approach to identify samples with intravenous fluid contamination in clinical laboratories

ObjectivesContamination with intravenous (IV) fluids is a common cause of specimen rejection or erroneous results in hospitalized patients. Identification of contaminated samples can be difficult. Common measures such as failed delta checks may not be adequately sensitive nor specific. This study aimed to determine detection criteria using commonly ordered tests to identify IV fluid contamination and validate the use of these criteria. MethodsConfirmed contaminated and non-contaminated samples were used to identify patterns in laboratory results to develop criteria to detect IV fluid contamination. The proposed criteria were implemented at a tertiary care hospital laboratory to assess performance prospectively for 6 months, and applied to retrospective chemistry results from 3 hospitals and 1 community lab to determine feasibility and flagging rates. The algorithm was also tested at an external institution for transferability. ResultsThe proposed algorithm had a positive predictive value of 92 %, negative predictive value of 91 % and overall agreement of 92 % when two or more criteria are met (n = 214). The flagging rates were 0.03 % to 0.07 % for hospital and 0.003 % for community laboratories. ConclusionsThe proposed algorithm identified true contamination with low false flagging rates in tertiary care urban hospital laboratories. Retrospective and prospective analysis suggest the algorithm is suitable for implementation in clinical laboratories to identify samples with possible IV fluid contamination for further investigation.

Development and Evaluation of "The Delta Plus-Minus Even Distribution Check": A Novel Patient-Based Real-Time Quality Control Method for Laboratory Tests.

Laboratory testing of large sample numbers necessitates high-volume rapid processing, and these test results require immediate validation and a high level of quality assurance. Therefore, real-time quality control including delta checking is an important issue. Delta checking is a process of identifying errors in individual patient results by reviewing differences from previous results of the same patient (Δ value). Under stable analytical conditions, Δ values are equally positively and negatively distributed. The previous 20 Δ values from 3 tests (cholesterol, albumin, and urea nitrogen) were analyzed by calculating the R-value: "the positive Δ value ratio minus 0.5." This method of monitoring optimized R-values is referred to as the even-check method (ECM) and was compared with quality control (QC) testing in terms of error detection. Bias was observed on 4 of the 120 days for the 3 analytes measured. When QC detected errors, the ECM captured the same systematic errors and more rapidly. In contrast, the ECM did not generate an alarm for the one random error that occurred in QC. While QC did not detect any errors, the percentage of R-values exceeding the acceptable range was under 2%, the number of days generating alarms was between 16 and 21 days, with short alarm periods, and a median number of samples per alarm period between 7 and 9 samples. The ECM is a practical real-time QC method, controlled by setting R-value conditions, that quickly detects bias values.

Automating the Detection of IV Fluid Contamination Using Unsupervised Machine Learning.

Intravenous (IV) fluid contamination is a common cause of preanalytical error that can delay or misguide treatment decisions, leading to patient harm. Current approaches for detecting contamination rely on delta checks, which require a prior result, or manual technologist intervention, which is inefficient and vulnerable to human error. Supervised machine learning may provide a means to detect contamination, but its implementation is hindered by its reliance on expert-labeled training data. An automated approach that is accurate, reproducible, and practical is needed. A total of 25 747 291 basic metabolic panel (BMP) results from 312 721 patients were obtained from the laboratory information system (LIS). A Uniform Manifold Approximation and Projection (UMAP) model was trained and tested using a combination of real patient data and simulated IV fluid contamination. To provide an objective metric for classification, an "enrichment score" was derived and its performance assessed. Our current workflow was compared to UMAP predictions using expert chart review. UMAP embeddings from real patient results demonstrated outliers suspicious for IV fluid contamination when compared with the simulated contamination's embeddings. At a flag rate of 3 per 1000 results, the positive predictive value (PPV) was adjudicated to be 0.78 from 100 consecutive positive predictions. Of these, 58 were previously undetected by our current clinical workflows, with 49 BMPs displaying a total of 56 critical results. Accurate and automatable detection of IV fluid contamination in BMP results is achievable without curating expertly labeled training data.

Survey on the Implementation of Delta Checks and Panic Values in Immunoassays in Korea

Survey on the Implementation of Delta Checks and Panic Values in Immunoassays in Korea

B-127 An Automated Approach to Obtain Delta Check Limits

Abstract Delta checking as a systematic approach to quality control specimens was first described by Nosanchuk and Gottmann in 1974. The purpose is to detect specimen errors due to short-sample aspiration, mislabeled cup or sample, incorrect patient draw, problems in analysis, and/or clerical errors. Once inaccuracy is detected, rejection and specimen recollection are the most suitable approach. The delta check procedure compares a patient’s current test results to the previous test results from the same individual using a specified time interval and delta check limit (DCL). The DCL can be expressed as absolute change, percentage change, rate of change, or rate of percentage change. This procedure is typically applied to several analytes, comparing the absolute change with a specified DCL. DCLs are commonly derived from the literature and may be optimized over time based on local patient populations. Creating and modifying DCLs poses a problem, because changes in the DCL can result in flagging too little or too many specimens, and often DCLs don’t take into consideration patient location and therefore lack specificity. Here, we describe a software tool that allows the determination of DCLs based on the individual laboratory’s or ordering unit’s patient results derived from the LIS. The software functionality was tested on a patient dataset from a large county hospital in California, USA. The software allows import of patient data for analysis and creation of population-specific DCLs. Typically, inpatient data is used, as consecutive data points within a common delta check timeframe (hours to days) are less expected for outpatients. The tool allows selection of an unlimited number of analytes of interest to be included in the DCL assessment. Furthermore, the underlying DCL timeframe that compares the two different results is customizable. Lastly, the tool provides the expected number of DCL flags based on the chosen cutoffs. This allows for proper resource allocation or adjustment of the DCL according to available resources. After the import of raw LIS data, an automatic data cleanup is performed. Then the assays of interest and, if available, the associated patient location are directly selected in the software. The dashboard provides each assay test code that is part of the delta checking process, the desired DCL, and the chosen delta check time period (either in hours or days). The resulting expected number of DCL flags is provided specifically for each test, per hour and per day. Any change to the DCL is displayed on the dashboard in real time. Delta checks are a tool to detect specimen inaccuracies resulting from pre-analytical or analytical errors. This software tool provides customized DCLs based on patient population and the individual laboratory instead of deriving them from the literature. This novel customization enables simple derivation of DCLs that are patient population-specific and optimized for each laboratory’s resources. This allows for improved detection of specimen inaccuracies and ultimately leads to improved patient care. *The product is currently under development. Its future availability cannot be guaranteed. Please contact your local Siemens Healthineers organization for further details.

B-125 Artificial Intelligence Application to Improve the Algorithm for Autoverification of Results of Biochemical Markers that Assess Thyroid Function

Abstract Background The results verification process is the most important control step of the post-analytical phase. It can be done in two ways, manually by a professional, or using autoverification systems. Appropriate results are verified and submitted to the Laboratory Information System (LIS). Corrective actions are initiated for discordant results prior to their submission to the LIS. Autoverification is a process where the laboratory tests results are published without any professional’s intervention. It is recommended that each laboratory creates the necessary cut-off values for autoverification, according to its population and its biochemical-medical criteria. In our laboratory, an autoverification algorithm is applied to evaluate thyroid function, which involves the measurement of 2 parameters: Thyrotropin (TSH) and Free Thyroxine (FT4). These parameters were chosen because they are the most used for characterizing thyroid function. The autoverification algorithm was defined for euthyroid, hypothyroid and hyperthyroid patients. Our professionals proposed a “safe range” for both parameters and we relate those parameters results for autoverification. We study the usefulness of artificial intelligence models by clustering non-supervised machine learning algorithms, to find behavioral patterns of not autoverified sample results, to support and even broaden the “safe ranges” established by our professionals. The goal was to improve the current autoverification algorithm by using artificial intelligence for thyroid function test results and to increase the autoverification percentage. Methods The autoverification process was carried out at TURNER Laboratories, Rosario, Argentina. The autoverification algorithm was designed according to the Laboratory and Clinical Standards Institute Guide AUTO10-A. The Atellica Data Manager (ADM) middleware was used to provide the connection between the Atellica Solution Immunoassay & Clinical Chemistry Analyzers and the LIS. ADM receives the results from the analyzers and processes them in consideration of the pre-analytical (age, sex, etc.), analytical (QC, flags, etc.) and post-analytical (Delta Check, etc.) conditions. It then applies the autoverification algorithm to send the results to the LIS. In the first semester of 2022, we used the autoverification algorithm with the existing “safe ranges” in 18 626 samples of outpatients between the ages of 16 and 60. On the other hand, we applied different Clustering algorithms to the samples that were not autoverified, to find behavior patterns. The ones we used are: KMeans, Fuzzy C-Means, DBSCAN, Agglomerative Clustering, Dendrogram and Affinity Propagation. We chose KMeans with k = 4. Our professionals thoroughly analyzed each of the clusters and defined “new safe ranges” for autoverification. In the second semester of 2022, 18 947 samples were analyzed and the autoverification algorithm was applied with the “safe ranges” and the “new safe ranges”. Results In the first semester of 2022, by applying the existing “safe ranges” we obtained an autoverification of 19.0%. We then simulated an autoverification with “safe ranges” and “new safe ranges” and obtained an autoverification of 71.5%. In the second semester, we applied autoverification with “safe ranges” and “new safe ranges” obtaining an autoverification of 72.7%. Conclusion The application of machine learning algorithms has significantly improved the level of autoverification of biochemical markers that evaluate thyroid function.

GPT-4 Underperforms Experts in Detecting IV Fluid Contamination.

Specimens contaminated with intravenous (IV) fluids are common in clinical laboratories. Current methods for detecting contamination rely on insensitive and workflow-disrupting delta checks or manual technologist review. Herein, we assessed the utility of large language models for detecting contamination by IV crystalloids and compared its performance to multiple, but variably trained healthcare personnel (HCP). Contamination of basic metabolic panels was simulated using 0.9% normal saline (NS), with (n = 30) and without (n = 30) 5% dextrose (D5NS), at mixture ratios of 0.10 and 0.25. A multimodal language model (GPT-4) and a diverse panel of 8 HCP were asked to adjudicate between real and contaminated results. Classification performance, mixture quantification, and confidence was compared by Wilcoxon rank sum. The 95% CIs for accuracy were 0.57-0.71 vs 0.73-0.80 for GPT-4 and HCP, respectively, on the NS set and 0.57-0.57 vs 0.73-0.80 on the D5NS set. HCP overestimated severity of contamination in the 0.10 mixture group (95% CI of estimate error, 0.05-0.20) for both fluids, while GPT-4 markedly overestimated the D5NS mixture at both ratios (0.16-0.33 for NS, 0.11-0.35 for D5NS). There was no correlation between reported confidence and likelihood of a correct classification. GPT-4 is less accurate than trained HCP for detecting IV fluid contamination of basic metabolic panel results. However, trained individuals were imperfect at identifying contaminated specimens implying the need for novel, automated tools for its detection.

Effective and Practical Complete Blood Count Delta Check Method and Criteria for the Quality Control of Automated Hematology Analyzers.

Delta checks increase patient safety by identifying automated hematology analyzer errors. International standards and guidelines for the complete blood count (CBC) delta check method have not been established. We established an effective, practical CBC delta check method and criteria. We assessed five delta check methods for nine CBC items (Hb, mean corpuscular volume, platelet count, white blood cell [WBC] count, and five-part WBC differential counts) using 219,804 blood samples from outpatients and inpatients collected over nine months. We adopted the best method and criteria and evaluated them using 42,652 CBC samples collected over two weeks with a new workflow algorithm for identifying test errors and corrections for Hb and platelet count. The median delta check time interval was 1 and 21 days for inpatients and outpatients (range, 1-20 and 1-222 days), respectively. We used delta values at 99.5% as delta check criteria; the criteria varied among the five methods and between outpatients and inpatients. The delta percent change (DPC)/reference range (RR) rate performed best as the delta check for CBC items. Using the new DPC/RR rate method, 1.7% of total test results exceeded the delta check criteria; the retesting and resampling rates were 0.5% and 0.001%, respectively. We developed an effective, practical delta check method, including RRs and delta check time intervals, and delta check criteria for nine CBC items. The criteria differ between outpatients and inpatients. Using the new workflow algorithm, we can identify the causes of criterion exceedance and report correct test results.

Practical delta check limits for tumour markers in different clinical settings.

Few studies have reported on delta checks for tumour markers, even though these markers are often evaluated serially. Therefore, this study aimed to establish a practical delta check limit in different clinical settings forfive tumour markers: alpha-fetoprotein, cancer antigen 19-9, cancer antigen 125, carcinoembryonic antigen, and prostate-specific antigen. Pairs of patients' results (current and previous) forfive tumour markers between 2020 and 2021 were retrospectively collected from three university hospitals. The data were classified into three subgroups, namely: health check-up recipient (subgroup H), outpatient (subgroup O), andinpatient (subgroup I) clinics. The check limits of delta percent change (DPC), absolute DPC (absDPC), and reference changevalue (RCV) for each test were determined using the development set (the first 18months, n=179,929) and then validated and simulated by applying the validation set (the last 6months, n=66,332). The check limits of DPC and absDPC for most tests varied significantly among the subgroups. Likewise, the proportions of samples requiring further evaluation, calculated by excluding samples with both current and previous results within the reference intervals, were 0.2-2.9% (lower limit of DPC), 0.2-2.7% (upper limit of DPC), 0.3-5.6% (absDPC), and 0.8-35.3% (RCV99.9%). Furthermore, high negative predictive values >0.99 were observed in all subgroups in the insilico simulation. Using real-world data, we found that DPC was the most appropriate delta-check method for tumour markers. Moreover, Delta-check limits for tumour markers should be applied based on clinical settings.

Delta checks.

Delta check is an electronic error detection tool. It compares the difference in sequential results within a patient against a predefined limit, and when exceeded, the delta check rule is considered triggered. The patient results should be withheld for review and troubleshooting before releasing to the clinical team for patient management. Delta check was initially developed as a tool to detect wrong-blood-in-tube (sample misidentification) errors. It is now applied to detect errors more broadly within the total testing process. Recent advancements in the theoretical understanding of delta check has allowed for more precise application of this tool to achieve the desired clinical performance and operational set up. In this Chapter, we review the different pre-implementation considerations, the foundation concepts of delta check, the process of setting up key delta check parameters, performance verification and troubleshooting of a delta check flag.

Detecting Exogenous Contamination of Clinical Blood Samples with Crystalloid Solutions: A Data Driven Approach to Derive and Validate Delta Checks

Abstract Pre-analytical error may occur when collecting blood samples from intravenous (IV) lines. Blood samples can be inadvertently diluted by exogenous solutions if an insufficient amount of blood is discarded prior to collection. Rules built in the laboratory information system or middleware can be used to alert laboratory staff to specimens that require co-investigation with the clinical team to determine if sample dilution is possible or likely. Currently, however, there is a lack of literature to guide laboratorians on how to define quality rules of this nature, particularly for detecting subtle, rather than gross contamination by commonly administered IV-fluids. Accordingly, the primary objective of this study is to derive sensitive and specific, multivariate delta checks to detect more subtle blood sample contamination by commonly used crystalloid solutions. To derive the rules, we began with in vitro experiments by spiking increasing volumes of major IV-fluids (normal saline (NS), lactated ringers (LR), and 5% dextrose (D5W)) into clinical blood samples that were collected from healthy donors (n=3). Crystalloid solutions were serially spiked into blood samples at 10% (solution:sample) increments for NS and LR from 10% up to 90%, and in 5% increments for D5W, up to 50%. Basic metabolic panel (BMP) analytes were measured and compared between neat and contrived samples. All testing was performed on Roche Cobas 8000 chemistry auto-analyzers. Based on in vitro data, we derived multivariate delta checks using analytes in a BMP that would reflect 10-40% contamination by a given fluid. We then performed a retrospective data analysis on more than 28 thousand, serially collected BMP results to identify samples that were flagged by these rules. Chart review was then performed to identify EHR-based evidence of temporally associated administration of crystalloid solutions through a peripheral IV access line. Increasing sample dilution from the in vitro study showed significant changes in several BMP analytes across all fluid types. With respect to NS, the most significant changes relative to baseline were observed with potassium, calcium, and chloride, while for LR, the most significant changes relative to baseline were observed with glucose, bicarbonate, and calcium. For D5W, significant changes were observed with glucose, calcium, chloride and sodium. Accordingly, delta check rules were implemented using relative changes of the aforementioned analytes. On chart review, we found 80-100% of flagged samples came from patients with administration of that given fluid in the last 12-24 hours. The results from this study support delta checks that consider multiple analytes, and fluid-specific bias profiles, as this may provide a more sensitive and specific approach to detecting contamination of clinical blood samples by crystalloid solutions. Further evaluation using retrospective data analysis will be used to validate these rules sensitivity and specificity followed by prospective testing before implementation.

Machine Learning Models Can Accurately Detect Crystalloid Contamination of Basic Metabolic Panels

Abstract A common cause of erroneous laboratory results is preanalytical contamination of samples with intravenous crystalloids. Rules-based approaches, such as delta checks or feasibility cut-offs, can detect these errors, but rely on prior laboratory results and lack sensitivity. In this study, we aimed to improve the detection of crystalloid contamination using two machine learning workflows for the detection of contamination by the four most common crystalloids; normal saline (NS), lactated Ringer’s (LR), and their dextrose-containing counterparts (D5NS and D5LR). First, we employed a semi-supervised approach using manifold approximation and contrastive learning. To do this, we aggregated five years' worth of basic metabolic panel results (n = 67 million), then simulated contaminated samples by mixing random subsets of samples in silico with the crystalloid solutions at different ratios. The dataset was partitioned into a 70:30 split between training and testing sets with five-fold cross-validation. An autoencoder was trained to create an embedding of the data by minimizing the triplet loss. This embedding was then projected onto a two-dimensional manifold. When the four crystalloid solutions and the independent test set were mapped to the manifold, decision boundaries were drawn based on the threshold that maximized the F1 statistic, and labels were assigned. This method detected the 10% in silico mixtures with an area under the receiver operating characteristic curve (AUC) of 0.97, 0.94, 0.99, and 0.99 for NS, LR, D5NS, and D5LR respectively. Next, we evaluated a fully-supervised approach by training and validating Random Forest and XGBoost models using the same training and testing sets as above. Both models performed well, with XGBoost slightly outperforming Random Forest. The XGBoost model demonstrated AUCs of 0.992, 0.990, 0.998, and 0.996 for the NS, LR, D5NS, and D5LR mixtures. This high-level performance was maintained even as the proportion of mixed samples in the dataset was decreased from 50% to 1% to more accurately reflect real-world contamination rates. For NS and LR, the variables with the highest relative importance were calcium and chloride. For their D5 counterparts, glucose and calcium were most important. Creatinine and blood urea nitrogen were the least important variables in both approaches. Overall, the fully-supervised learning approach outperformed that of the contrastive embeddings for the specific task of identifying contamination from pre-defined crystalloids. One advantage of the semi-supervised approach is that it could be adapted to identify outliers that are not pre-defined. However, considering the relative ease of translating tree-based models into practice, we conclude that our XGBoost model offers the most promising solution and plan to prospectively validate this model for real-time detection of sample contamination.

Contamination of clinical blood samples with crystalloid solutions: An experimental approach to derive multianalyte delta checks

BackgroundLaboratorians are left unguided by a paucity of literature on how to configure rules for the detection of intravenous (IV) fluid contamination in blood samples. We designed a study to determine the in vitro effect of increasing blood sample contamination from commonly used crystalloid solutions and how these observations can guide the derivation of multianalyte delta checks to detect such pre-analytical error. MethodsIn this study, we spiked increasing volumes of commonly used IV fluids (normal saline (NS), lactated ringers (LR), and 5% dextrose) into blood samples that were collected from healthy donors. Routine chemistry analytes were measured and compared between neat and contrived samples. From these observations, we derived several permutations of multianalyte delta checks using the basic metabolic panel framework and evaluated rule performance using retrospective data. ResultsThe wet chemistry experiments showed that increasing the volume of crystalloid solution contamination significantly changed several analytes. Subsequently derived multianalyte delta check procedures were applied to retrospective data. For all IV fluids tested, smaller magnitudes of analyte change resulted in more samples flagged. ConclusionMultianalyte delta checks may be an effective method for the detection of IV fluid contamination.

M104 Application and modification of reference change values for delta checks in clinical laboratory

M104 Application and modification of reference change values for delta checks in clinical laboratory

Comparison of two (data mining) indirect approaches for between-subject biological variation determination

BackgroundBetween-subject biological variation (CVg) is an important parameter in several aspects of laboratory practice, including setting of analytical performance specification, delta checks and calculation of index of individuality. Using simulations, we compare the performance of two indirect (data mining) approaches for deriving CVg. MethodsThe expected mean squares (EMS) method was compared against that proposed by Harris and Fraser. Using numerical simulations, d the percentage difference in the mean between the non-pathological and pathological populations, CVi the within-subject coefficient of variation of the non-pathological distribution, f the fraction of pathological values, and e the relative increase in CVi of the pathological distribution were varied for a total of 320 conditions to examine the impact on the relative fractional of error of the recovered CVg compared to the true value. ResultsComparing the two methods, the EMS and Harris and Fraser’s approaches yielded similar performance of 158 conditions and 157 conditions within ± 0.20 fractional error of the true underlying CVg, for the normal and lognormal distributions, respectively. It is observed that both EMS and Harris and Fraser’s method performed better using the calculated CVi rather than the actual (‘presumptive’) CVi. The number of conditions within 0.20 fractional error of the true underlying CVg did not differ significantly between the normal and lognormal distributions. The estimation of CVg improved with decreasing values of f, d and CViCVg. DiscussionsThe two statistical approaches included in this study showed reliable performance under the simulation conditions examined.