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

BACKGROUND: 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. METHODS: 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. RESULTS: 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. CONCLUSIONS: Accurate and automatable detection of IV fluid contamination in BMP results is achievable without curating expertly labeled training data.

Limited Utility of Free Triiodothyronine Testing.

BACKGROUND: Free triiodothyronine (fT3) testing is most useful when thyroid stimulating hormone (TSH) is suppressed, and free thyroxine (fT4) is normal or decreased. These laboratory values in a symptomatic patient are referred to as T3 thyrotoxicosis. Standards for fT3 reflex testing have not been established. Herein, we examined the clinical utility of fT3 with the goal of identifying a TSH cutoff in the context of normal/decreased fT4 that maximizes the utility of measuring fT3. METHODS: TSH, fT4, and fT3 results between January 2016 and October 2021 were extracted from the laboratory information system and grouped if resulted on the same day for the same patient. Frequency of biochemical T3 thyrotoxicosis was evaluated at different TSH cutoffs and in outpatient vs inpatient settings. RESULTS: Of the 4366 TSH-fT4-fT3 results, 70 (1.6%) were consistent with biochemical T3 thyrotoxicosis. The common reasons were previously diagnosed hyperthyroidism on antithyroid medication (n = 28) or hypothyroidism on thyroid medication (n = 18) and newly diagnosed hyperthyroidism (n = 20, 0.5%). The likelihood of detecting T3 thyrotoxicosis increased with lower TSH cutoff (<0.3 µIU/mL, 10.3% vs <0.0 1µIU/mL, 27.6%). All patients with newly diagnosed hyperthyroidism had TSH <0.01 µIU/mL. Higher frequency of T3 thyrotoxicosis was observed in the outpatient setting (34%) relative to the inpatient setting (14%, P < 0.001) when TSH < 0.01 µIU/mL. CONCLUSIONS: T3 thyrotoxicosis is a relatively rare diagnosis and fT3 measurement has limited utility in the vast majority of patients. A fT3 reflex for patients with TSH <0.01 µIU/mL and normal/low fT4 may improve clinical utility and reduce unnecessary testing, especially in the outpatient setting.

Method Comparison and Workflow Differences Using the Same Free Light Chain Assay on 2 Analyzer Platforms.

BACKGROUND: The Freelite assay (The Binding Site) is utilized to quantify serum immunoglobulin free light chains (sFLC), which is crucial for diagnosing and monitoring plasma cell dyscrasias (PCDs). Using the Freelite test, we compared methods and evaluated workflow differences across two analyzer platforms. METHODS: sFLC concentrations were measured in 306 fresh serum specimens (cohort A) and 48 frozen specimens with documented sFLC >20 mg/dL (cohort B). Specimens were analyzed on the Roche cobas 8000 and Optilite analyzers using the Freelite κ and λ assays. Performance was compared using Deming regression. Workflow was compared by assessing turnaround time (TAT) and reagent usage. RESULTS: For cohort A specimens, Deming regression revealed a slope of 1.04 (95% CI, 0.88-1.02) and an intercept of -0.77 (95% CI, -0.57 to 1.85) for sFLCκ and a slope of 0.90 (95% CI, -0.04 to 1.83) and intercept of 1.59 (95% CI, -3.12 to 6.25) for sFLCλ. Regression of the κ/λ ratio revealed a slope of 2.44 (95% CI, 1.47-3.41) and intercept of -8.13 (95% CI, -16.82 to 0.58) with a concordance kappa of 0.80 (95% CI, 0.69-0.92). The proportion of specimens with TAT >60 min was 0.33% and 8% for the Optilite and cobas, respectively (P < 0.001). The Optilite required 49 (P < 0.001) and 12 (P = 0.016) fewer tests for sFLCκ and sFLCλ relative to the cobas. Cohort B specimens showed similar but more dramatic results. CONCLUSIONS: Analytical performance of the Freelite assays was comparable on the Optilite and cobas 8000 analyzers. In our study, the Optilite required less reagent, had a slightly reduced TAT, and eliminated manual dilutions for samples with sFLC concentrations >20 mg/dL.

Utility of Commercially Available Quantitative hCG Immunoassays as Tumor Markers in Trophoblastic and Non-Trophoblastic Disease.

BACKGROUND: The use of quantitative human chorionic gonadotropin (hCG) as a tumor marker is widely accepted despite lack of FDA-approval for oncology. Differences in iso- and glycoform recognition among hCG immunoassays is well established, exhibiting wide inter-method variability. Here, we assess the utility of 5 quantitative hCG immunoassays for use as tumor markers in trophoblastic and non-trophoblastic disease. METHODS: Remnant specimens were obtained from 150 patients with gestational trophoblastic disease (GTD), germ cell tumors (GCT), or other malignancies. Specimens were identified by review of results from physician-ordered hCG and tumor marker testing. Five analyzer platforms were used for split specimen analysis of hCG: Abbott Architect Total, Roche cobas STAT, Roche cobas Total, Siemens Dimension Vista Total, and Beckman Access Total. RESULTS: Frequency of elevated hCG concentrations (above reference cutoffs) was highest in GTD (100%), followed by GCT (55% to 57%), and other malignancies (8% to 23%). Overall, the Roche cobas Total detected elevated hCG in the greatest number of specimens (63/150). Detection of elevated hCG in trophoblastic disease was nearly equivalent among all immunoassays (range, 41 to 42/60). CONCLUSIONS: While no immunoassay is likely to be perfect in all clinical situations, results for the 5 hCG immunoassays evaluated suggest that all are adequate for use of hCG as a tumor marker in gestational trophoblastic disease and select germ cell tumors. Further harmonization of hCG methods is needed as serial testing for biochemical tumor monitoring must still be performed using a single method. Additional studies are needed to assess the utility of quantitative hCG as a tumor marker in other malignant disease.

Increased specimen minimum volume reduces turnaround time and hemolysis.

Quantity not sufficient (QNS) specimens with minimal blood volume for testing are common in clinical laboratories. However, there is no universal definition of minimum volume for a QNS specimen and little data is available addressing the impact of QNS / low volume specimens on turnaround time (TAT) and sample hemolysis. We compared the TAT and hemolysis index from samples ≤1.0 mL to all specimens received and quantified the number of specimens with reduced blood volume. A new QNS policy requiring ≥1.5 mL of sample in a blood tube for laboratory analysis was implemented and the results were assessed by sample hemolysis and TAT. The median laboratory TAT for samples with ≤1.0 mL of blood was 61 min (Interquartile Range, IQR: 50-82), in contrast to 28 min (26-34) for all samples. The hemolysis index for samples ≤1.0 mL was 112 (65-253) and 15 (8-29) for all samples. Requirement of a minimum volume of 1.5 mL of blood resulted in the proportion of samples with TAT ≥ 60 min to decrease from 10.4% to 4.24% in the ED, and for specimens cancelled due to hemolysis to decrease from 4.24% to 3.38%. This policy was introduced hospital wide with similar effects. Together, we correlate limited specimen volume with an increase in laboratory TAT and hemolysis. Implementation of a QNS policy of ≥1.5 mL and provider education provided a significant and durable reduction in TAT and specimen hemolysis.

Progression of SARS-CoV-2 Seroprevalence in St. Louis, Missouri, through January 2021.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) seropositivity was assessed for 3,066 individuals visiting hospitals in St. Louis, Missouri, during July 2020, November 2020, or January 2021. Seropositivity in children increased from 5.22% in July to 21.16% in January. In the same time frame, seropositivity among adults increased from 4.52% to 19.03%, prior to initiation of mass vaccination. IMPORTANCE This study determined the percentage of children and adult samples from the St. Louis metropolitan area in Missouri with SARS-CoV-2 antibodies during three collection periods spanning July 2020 to January 2021. By January 2021, 20.68% of the tested individuals had antibodies. These results show the evolution of the SARS-CoV-2 pandemic in St. Louis, Missouri, and provide a snapshot of the extent of infection just prior to the start of mass vaccination.

Evaluation of Radioactivity in Patient Specimens Received in the Core Laboratory at a National Cancer Institute (NCI) Designated Cancer Center.

BACKGROUND: Biological specimens from patients who have received radiopharmaceuticals are often collected for diagnostic testing and sent to clinical laboratories. Residual radiation has long been assumed to be minimal. However, literature is sparse and may not represent the specimen volumes or spectrum of radionuclides currently seen at National Cancer Institute (NCI)-designated cancer centers. This study examined the radiopharmaceuticals associated with patient specimens received in the hospital core laboratory and assessed the potential risk of external radiation exposure to laboratory personnel. METHODS: The types and amounts of radiopharmaceuticals administered in a large metropolitan hospital system were retrospectively examined over a 20-month study period. The associated biological specimens sent to the largest core laboratory in the system for testing were evaluated. In addition, manual survey meter assessment of random clinical specimens and weekly wipe tests were performed for 44 weeks, and wearable and environmental dosimeters were placed for 6 months. RESULTS: Over 11 000 specimens, collected within 5 physical half-lives of radiopharmaceutical administration, were processed by our laboratory. Manual survey meter assessment of random clinical specimens routinely identified radioactive specimens. If held in a closed palm for >2 min, many samples could potentially deliver a 0.02 mSv effective dose of radiation. CONCLUSIONS: The laboratory regularly receives radioactive patient specimens without radioactive labels. Although the vast majority of these are blood specimens associated with low-dose diagnostic radiopharmaceuticals, some samples may be capable of delivering a significant amount of radiation. Recommendations for laboratories associated with NCI cancer centers are given.

Seroprevalence of SARS-CoV-2 Antibodies in Children and Adults in St. Louis, Missouri, USA.

Reported coronavirus disease 2019 (COVID-19) case counts likely underestimate the true prevalence because mild or asymptomatic cases often go untested. Here, we use a sero-survey to estimate the seroprevalence of IgG antibodies against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in the St. Louis, MO, metropolitan area in a symptom-independent manner. Five hundred three adult and 555 pediatric serum/plasma samples were collected from patients presenting to Barnes-Jewish Hospital or St. Louis Children's Hospital between 14 April 2020 and 12 May 2020. We developed protocols for in-house enzyme-linked immunosorbent assays (ELISAs) using spike and nucleoprotein and used the assays to estimate a seroprevalence rate based on our samples. Overall IgG seropositivity was estimated to be 1.71% (95% credible interval [CI], 0.04% to 3.38%) in pediatric samples and 3.11% (95% CI, 0.92% to 5.32%) in adult samples. Seropositivity was significantly lower in children under 5 years of age than in adults, but rates between adults and children aged 5 or older were similar. Of the 176 samples tested from children under 4 years of age, none were positive.IMPORTANCE This study determined the percentages of both children and adult samples from the greater St. Louis metropolitan area who had antibodies to SARS-CoV-2 in late April to early May 2020. Approximately 1.7 to 3.1% of the tested individuals had antibodies, indicating that they had previously been infected by SARS-CoV-2. These results demonstrate that the extent of infection was about 10 times greater than the number of confirmed cases at that time. Furthermore, it demonstrated that by 5 years of age, children were infected to an extent similar to that of adults.

Evaluation of the Risk of Laboratory Microbial Contamination during Routine Testing in Automated Clinical Chemistry and Microbiology Laboratories.

BACKGROUND: Every clinical specimen is potentially infectious, but data regarding risk for contamination of the laboratory environment during routine testing are scarce. We assessed contamination during routine sample analysis in automated clinical chemistry and microbiology laboratories. METHODS: A fluorescent marker was applied to specimen container exteriors to assess the impact of gross contamination. Nonpathogenic MS2 virus was added to remnant blood, urine, and ESwab matrices as a biomarker of cross-contamination. Samples were processed and analyzed using Roche Cobas 8100 and ISE, c502, e602, and c702 modules (blood) and BD Kiestra total laboratory automation (blood, urine, ESwabs) over 3 experiments. Fluorescence transfer to laboratory surfaces and personnel was visualized using ultraviolet light. Surfaces were swabbed and assessed for MS2 cross-contamination by RT-PCR. Adherence to standard precautions by laboratory staff was assessed by observation. RESULTS: Fluorescence was observed on 49 of 165 (30%) laboratory surfaces and personnel and 21 of 93 (23%) total laboratory automation instruments. Fluorescence transferred most frequently to gloves (31/40), computer accessories (9/18), and specimen loading racks (12/12). None of 123 areas swabbed were positive for MS2. Improper personal protective equipment use occurred at a rate of 0.36 and 0.15 events per staff per hour in the chemistry and microbiology laboratories, respectively. Hand-washing compliance was observed for 61 of 132 (46%) staff members evaluated. CONCLUSIONS: Analysis of grossly contaminated specimens on automated chemistry and microbiology equipment elicits a low likelihood of instrument contamination. However, handling contaminated specimen containers can result in contamination of environmental laboratory surfaces, representing a source of risk that is heightened by low adherence to appropriate personal protective equipment.

Association between SARS-CoV-2 Neutralizing Antibodies and Commercial Serological Assays.

BACKGROUND: Commercially available SARS-CoV-2 serological assays based on different viral antigens have been approved for the qualitative determination of anti-SARS-CoV-2 antibodies. However, there are limited published data associating the results from commercial assays with neutralizing antibodies. METHODS: Sixty-six specimens from 48 patients with PCR-confirmed COVID-19 and a positive result by the Roche Elecsys Anti-SARS-CoV-2, Abbott SARS-CoV-2 IgG, or EUROIMMUN SARS-CoV-2 IgG assays and 5 control specimens were analyzed for the presence of neutralizing antibodies to SARS-CoV-2. Correlation, concordance, positive percent agreement (PPA), and negative percent agreement (NPA) were calculated at several cutoffs. Results were compared in patients categorized by clinical outcomes. RESULTS: The correlation between SARS-CoV-2 neutralizing titer (EC50) and the Roche, Abbott, and EUROIMMUN assays was 0.29, 0.47, and 0.46, respectively. At an EC50 of 1:32, the concordance kappa with Roche was 0.49 (95% CI; 0.23-0.75), with Abbott was 0.52 (0.28-0.77), and with EUROIMMUN was 0.61 (0.4-0.82). At the same neutralizing titer, the PPA and NPA for the Roche was 100% (94-100) and 56% (30-80); Abbott was 96% (88-99) and 69% (44-86); and EUROIMMUN was 91% (80-96) and 81% (57-93) for distinguishing neutralizing antibodies. Patients who were intubated, had cardiac injury, or acute kidney injury from COVID-19 infection had higher neutralizing titers relative to those with mild symptoms. CONCLUSIONS: COVID-19 patients generate an antibody response to multiple viral proteins such that the calibrator ratios on the Roche, Abbott, and EUROIMMUN assays are all associated with SARS-CoV-2 neutralization. Nevertheless, commercial serological assays have poor NPA for SARS-CoV-2 neutralization, making them imperfect proxies for neutralization.

Urinary Tract Infection With Gram-Positive Bacteria Does Not Cause False-Positive Results with the Urine-Based Human Chorionic Gonadotropic Point-of-Care Assay.

BACKGROUND: False-positive results for human chorionic gonadotropic (hCG) on point-of-care (POC) devices can occur for a variety of technical and biological reasons. It has been postulated that urinary tract infection can result in false-positive POC hCG assays, but the cause of this phenomenon remains elusive. Gram-positive bacteria have previously been reported to express an hCG-like molecule. We investigated whether urinary tract infection with Gram-positive bacteria can result in false-positive POC hCG. METHODS: We utilized remnant clinical urine specimens that had been submitted for culture as part of evaluation for urinary tract infection. Urine specimens with >100,000 colony-forming units per milliliter of Gram-positive bacteria (n = 95) were tested on ICON 20 POC hCG tests (Beckman Coulter). Specimens from adult patients that had been collected for clinical testing in the prior 48 hours were included in the study, and only 1 specimen per patient was included. RESULTS: Of 95 patients with Gram-positive urine specimens, 42 (44%) were female, and the median age was 62 years. The most common bacteria identified during clinical urine culture of these patients' specimens were coagulase-negative Staphylococcus species (36/95, 38%), Enterococcus species (34/95, 36%), and Streptococcus agalactiae (9/95, 9%). Five of 95 (5.3%) urine specimens were positive for POC hCG. Chart review revealed that 3 specimens were from pregnant women and 2 were from patients with cancer diagnoses. CONCLUSIONS: Urine specimens from patients suspected to have urinary tract infection with Gram-positive organisms did not cause positive results on POC hCG test devices.