Risk Of False Positives Research Articles

Fooled by Significance Testing: An Analysis of the LOVIT Vitamin C Trial

In this article, I discuss the potential pitfalls of interpreting p values, confidence intervals, and declarations of statistical significance. To illustrate the issues, I discuss the LOVIT trial, which compared high-dose vitamin C with placebo in mechanically ventilated patients with sepsis. The primary outcome – the proportion of patients who died or had persisting organ dysfunction at day 28 – was significantly higher in patients who received vitamin C (p = .01). The authors had hypothesized that vitamin C would have a beneficial effect, although the prior evidence for benefit was weak. There was no prior evidence for a harmful effect of high-dose vitamin C. Consequently, the pretest probability for harm was low. The sample size was calculated assuming a 10% absolute risk difference, which was optimistic. Overestimating the effect size when calculating the sample size leads to low power. For these reasons, we should be skeptical that vitamin C causes harm in septic patients, despite the significant result. p-values and confidence intervals are probabilities concerning the chance of obtaining the observed data. However, we are more interested in the chance the intervention has a real effect on the outcome. That is to say, we are more interested in whether the hypothesis is true. A Bayesian approach allows us to estimate the false positive risk, which is the post-test probability there is no effect of the intervention. The false positive risk for the LOVIT trial (calculated from the published summary data using uniform priors for the parameter values) is 70%. Most likely, high-dose vitamin C does not cause harm in septic patients. Most likely it has no effect at all. If there is an effect, it is probably small and most likely beneficial.

Fooled by Significance Testing: An Analysis of the LOVIT Vitamin C Trial.

In this article, I discuss the potential pitfalls of interpreting p values, confidence intervals, and declarations of statistical significance. To illustrate the issues, I discuss the LOVIT trial, which compared high-dose vitamin C with placebo in mechanically ventilated patients with sepsis. The primary outcome - the proportion of patients who died or had persisting organ dysfunction at day 28 - was significantly higher in patients who received vitamin C (p = .01). The authors had hypothesized that vitamin C would have a beneficial effect, although the prior evidence for benefit was weak. There was no prior evidence for a harmful effect of high-dose vitamin C. Consequently, the pretest probability for harm was low. The sample size was calculated assuming a 10% absolute risk difference, which was optimistic. Overestimating the effect size when calculating the sample size leads to low power. For these reasons, we should be skeptical that vitamin C causes harm in septic patients, despite the significant result. p-values and confidence intervals are probabilities concerning the chance of obtaining the observed data. However, we are more interested in the chance the intervention has a real effect on the outcome. That is to say, we are more interested in whether the hypothesis is true. A Bayesian approach allows us to estimate the false positive risk, which is the post-test probability there is no effect of the intervention. The false positive risk for the LOVIT trial (calculated from the published summary data using uniform priors for the parameter values) is 70%. Most likely, high-dose vitamin C does not cause harm in septic patients. Most likely it has no effect at all. If there is an effect, it is probably small and most likely beneficial.

Racial disparities in Alzheimer’s Disease (AD): Correlating a novel neuropsychological screening test to established AD fluid and imaging biomarkers in African Americans and Caucasians

AbstractBackgroundThe Montreal Cognitive Assessment (MoCA) is a commonly used screening tool for the diagnosis of neurocognitive disorders such as mild cognitive impairment (MCI) and Alzheimer’s Disease (AD). The current cut‐off threshold for detection of MCI on the MoCA increases risk for false positives, particularly in ethnic minorities and those of lower educational attainment. African Americans (AA) have experienced inaccurate detection of MCI and AD, leading to a delay in diagnosis and a decreased enrollment in AD‐related clinical trials. This underscores the importance of clarified cut‐offs for AD plasma biomarkers in AA and a standardized cognitive screening test independent of language and cultural norms. The Visual‐based Cognitive Assessment Test (VCAT) is a newly developed language‐neutral cognitive screening test that has good sensitivity (85.6%) and specificity (81.1%) in diagnosing MCI and mild AD. Our study will correlate VCAT scores to MRI brain findings and established AD plasma biomarkers in AA and Caucasians with and without cardiovascular diseases (i.e. prior myocardial infarction, stroke, and chronic kidney disease).MethodsAA and Caucasians diagnosed with MCI, AD, and normal cognition have been recruited (n=50) over a period of one year (recruitment remains ongoing). Plasma samples were collected, MRI brain imaging and Neuroquant data were obtained (for volumetric analyses), and the VCAT along with the MoCA were completed. Phosphorylated‐tau 181 (p‐tau 181) and phosphorylated‐tau 217 (p‐tau 217) will be measured in the samples using the Quanterix HDX device.ResultsPreliminary results reveal that levels of p‐tau 181 are elevated (from a normative cut‐off value of 1.81 pg/mL per the Youden index) in all groups, and highest in the group of patients with both CSF findings consistent with AD and a medical history of cardiovascular diseases (MI, CKD, and/or stroke).ConclusionEstablished cut‐off values for p‐tau 181 and p‐tau 217 in the literature likely underestimate these results in AA patients. In addition to elucidating the cut‐off normative values for established plasma AD biomarkers in a diverse cohort, our goal is to validate the VCAT, a cultural and language‐neutral screening assessment. This will allow for application to multilingual populations without the need for translation of its test contents.

A Bayesian hypothesis testing-based statistical decision philosophy for structural damage detection

In this article, a new statistical decision philosophy is developed to tackle structural damage detection problems which are defined in the context of novelty detection. In line with this philosophy, structural damage detection is achieved by deriving the posterior probability of damage presence from the Bayesian testing of two competing hypotheses, with the null and alternative hypotheses representing the undamaged and damaged states of a structure of concern, respectively. To resolve the tricky problem of prior appropriateness in Bayesian hypothesis testing, a general prior specification criterion is devised based on the notion of risk management, including the risk of false positive indication (an undamaged structure is incorrectly identified as damaged) and the risk of false negative indication (a damaged structure is incorrectly identified as undamaged). To determine an optimal risk level, two principles, namely the principle of posterior probability difference minimization (PPDM) and the principle of posterior probability product maximization (PPPM), are defined. The PPDM principle is to minimize the difference between the ability of a novelty detector to avoid a false positive and its ability to avoid a false negative, and the PPPM principle is to maximize the product of the two capabilities. Both principles essentially act as a means of achieving an optimal trade-off between the false positive and false negative risks stipulated in pursuing damage detection. To demonstrate the effectiveness of the proposed statistical decision philosophy for structural damage detection, experimental data obtained from a 5-story steel frame model and a 38-story concrete building model have been investigated.

Protein A–Nanoluciferase fusion protein for generalized, sensitive detection of immunoglobulin G

Detection and quantification of antibodies, especially immunoglobulin G (IgG), is a cornerstone of ELISAs, many diagnostics, and the development of antibody-based drugs. Current state-of-the-art immunoassay techniques for antibody detection require species-specific secondary antibodies and carefully-controlled bioconjugations. Poor conjugation efficiency degrades assay performance and increases the risk of clinical false positives due to non-specific binding. We developed a generic, highly-sensitive platform for IgG quantification by fusing the IgG-Fc binding Z domain of Staphylococcal Protein A with the ultrabright bioluminescence reporter Nanoluc-luciferase (Nluc). We demonstrated the application of this fusion protein in a sandwich IgG detection immunoassay using surface-bound antigens to capture target IgG and protein A-Nanoluc fusion as the detector. We optimized the platform's sensitivity by incorporating multiple repeats of the Z domain into the fusion protein constructs. Using rabbit and mouse anti-SARS-CoV-2 Nucleoprotein IgGs as model analytes, we performed ELISAs in two different formats, either with SARS-CoV-2 Nucleoprotein as the capture antigen or with polyclonal chicken IgY as the capture antibody. Using standard laboratory equipment, the platform enabled the quantitation of antibody analytes at concentrations as low as 10 pg/mL (67 fM).

Critical item (CR) analysis expands the classification accuracy of performance validity tests based on the forced choice paradigm-Replicating previously introduced CR cutoffs within the Word Choice Test.

This study was designed to replicate previous research on critical item analysis within the Word Choice Test (WCT). Archival data were collected from a mixed clinical sample of 119 consecutively referred adults (Mage = 51.7, Meducation = 14.7). The classification accuracy of the WCT was calculated against psychometrically defined criterion groups. Critical item analysis identified an additional 2%-5% of the sample that passed traditional cutoffs as noncredible. Passing critical items after failing traditional cutoffs was associated with weaker independent evidence of invalid performance, alerting the assessor to the elevated risk for false positives. Failing critical items in addition to failing select traditional cutoffs increased overall specificity. Non-White patients were 2.5 to 3.5 times more likely to Fail traditional WCT cutoffs, but select critical item cutoffs limited the risk to 1.5-2. Results confirmed the clinical utility of critical item analysis. Although the improvement in sensitivity was modest, critical items were effective at containing false positive errors in general, and especially in racially diverse patients. Critical item analysis appears to be a cost-effective and equitable method to improve an instrument's classification accuracy. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Present status and future directions: Imaging techniques for the detection of periapical lesions.

Diagnosing and treating apical periodontitis (AP) in an attempt to preserve the natural dentition, and to prevent the direct and indirect systemic effects of this condition, is the major goal in endodontics. Considering that AP is frequently asymptomatic, and is most often associated with a lesion in the periapex of the affected tooth, within the maxillary bones, imaging becomes of paramount importance for the diagnosis of the disease. The aim of this narrative review was to investigate the most relevant classic and current literature to describe which are, to date, the diagnostic imaging systems most reliable and advanced to achieve the early and predictable detection of AP, the best measures of the lesions and the disclosure of the different features of the disease. Dental panoramic tomography (DPT) is a classic exam, considered still useful to provide the basic diagnosis of AP in certain districts of the maxillary bones. Periapical radiographs (PRs) represent a valid routine examination, with few, known limitations. Cone-beam computed tomography (CBCT) is the only system that ensures the early and predictable detection of all periapical lesions in the jaws, with the minor risk of false positives. These techniques can be successfully implemented, with ultrasounds (USI) or magnetic resonance (MRI) imaging, exams that do not use ionising radiations. MRI and USI provide information on specific features of the lesions, like the presence and amount of vascular supply, their content and their relationship with the surrounding soft tissues, leading to differential diagnoses. Further, all the three-dimensional systems (CBCT, USI and MRI) allow the volumetric assessment of AP. Pioneering research on artificial intelligence is slowly progressing in the detection of periapical radiolucencies on DPTs, PRs and CBCTs, however, with promising results. Finally, it is established that all imaging techniques have to be associated with a thorough clinical examination and a good degree of calibration of the operator.

Cross-Cultural Performance Validity Testing: Managing False Positives in Examinees with Limited English Proficiency

ABSTRACT Base rates of failure (BR Fail ) on performance validity tests (PVTs) were examined in university students with limited English proficiency (LEP). BR Fail was calculated for several free-standing and embedded PVTs. All free-standing PVTs and certain embedded indicators were robust to LEP. However, LEP was associated with unacceptably high BR Fail (20–50%) on several embedded PVTs with high levels of verbal mediation (even multivariate models of PVT could not contain BR Fail ). In conclusion, failing free-standing/dedicated PVTs cannot be attributed to LEP. However, the elevated BR Fail on several embedded PVTs in university students suggest an unacceptably high overall risk of false positives associated with LEP.

Identifying signatures of natural selection in Indian populations.

In this study, we present the results of a genome-wide scan for signatures of positive selection using data from four tribal groups (Kokana, Warli, Bhil, and Pawara) and two caste groups (Deshastha Brahmin and Kunbi Maratha) from West of the Maharashtra State In India, as well as two samples of South Asian ancestry from the 1KG project (Gujarati Indian from Houston, Texas and Indian Telugu from UK). We used an outlier approach based on different statistics, including PBS, xpEHH, iHS, CLR, Tajima’s D, as well as two recently developed methods: Graph-aware Retrieval of Selective Sweeps (GRoSS) and Ascertained Sequentially Markovian Coalescent (ASMC). In order to minimize the risk of false positives, we selected regions that are outliers in all the samples included in the study using more than one method. We identified putative selection signals in 107 regions encompassing 434 genes. Many of the regions overlap with only one gene. The signals observed using microarray-based data are very consistent with our analyses using high-coverage sequencing data, as well as those identified with a novel coalescence-based method (ASMC). Importantly, at least 24 of these genomic regions have been identified in previous selection scans in South Asian populations or in other population groups. Our study highlights genomic regions that may have played a role in the adaptation of anatomically modern humans to novel environmental conditions after the out of Africa migration.

Two novel qPCR assays to enhance black rail (Laterallus jamaicensis) eDNA surveys in the United States

Black rails (Laterallus jamaicensis) are a rare and protected bird species. They are also notoriously difficult to monitor. eDNA surveys have proven beneficial for the detection of this elusive species, but additional developments are needed to enhance eDNA survey utility (including reliability of outcomes) in light of black rail status and associated high-profile conservation actions. The use of multiple assays functions to increase eDNA detection confidence by increasing the probability of amplification in low concentration, highly degraded samples (i.e., reducing the risk of false negatives) and by providing a multiplicative test for ruling out contamination as the source of qPCR amplification (i.e., reducing the risk of false positives). The two eDNA assays developed, optimized, and validated in this study will prove valuable for the continued surveillance of the secretive black rail, offering data to measure its distribution and occurrence over time and as influenced by anticipated climate-change-induced disturbances and/or habitat restoration/creation initiatives. The assays are applicable to both Eastern black rails and California black rails, and do not exhibit qPCR amplification in non-target, co-distributed taxa (including other Rallidae species) in the United States.

Ultrasound diagnosis of hypertrophic pyloric stenosis - Time to change the criteria.

Ultrasound is the examination of choice for the diagnosis of hypertrophic pyloric stenosis (HPS). A correct diagnosis is dependent on the technique and measurement accuracy. However, in the world literature there is a wide range of values suggested for the diagnosis of this condition. The current minimum measurements used to diagnose HPS seem excessively large, and therefore, we set out to redefine these values. A retrospective study was performed on 607 patients (615 scans) being investigated for HPS. The length and transverse diameter of the pyloric canal, and thickness of the pyloric muscle were measured. All results were correlated with clinical and surgical findings. In this study, the muscle thickness in the normal group was <2.0 mm than in HPS infants having a muscle thickness of 2.0-5.0 mm. All the pyloric canal lengths in the normal group were <5.0 mm than in those with HPS having a length of 10.0-24.0 mm. The transverse diameters ranged from 6.0 to 11.0 mm in the normal group compared with those with HPS having a diameter between 8.0 and 16.0 mm. The current criteria for sonographic diagnosis of HPS should be redefined. The canal length is the single most important discriminator, with a clear separation between normal and abnormal. The commonly used 16.0-mm measurement is too long and should be reduced to 10.0 mm (without the risk of false positives). In many cases, the muscle thickness in those with HPS is as low as 2.0 mm, considerably less than the 3.0 mm that is currently used. The transverse diameter is not a useful discriminator for HPS. The use of current values will delay the diagnosis and timely treatment of this condition.

Using Immersive Virtual Reality Simulation to Ensure Competence in Contrast-Enhanced Ultrasound.

Contrast-enhanced ultrasound (CEUS) is used in various medical specialties as a diagnostic imaging tool and for procedural guidance. Experience in the procedure is currently attained via supervised clinical practice that is challenged by patient availability and risks. Prior simulation-based training and subsequent assessment could improve and ensure competence before performance on patients, but no simulator currently exists. Immersive virtual reality (IVR) is a new promising simulation tool that can replicate complex interactions and environments that are unfeasible to achieve by traditional simulators. This study was aimed at developing an IVR simulation-based test for core CEUS competencies and gathering validity evidence for the test in accordance with Messick's framework. The test was developed by IVR software specialists and clinical experts in CEUS and medical education and imitated a CEUS examination of a patient with a focal liver lesion with emphasis on the pre-contrast preparations. Twenty-five medical doctors with varying CEUS experience were recruited as test participants, and their results were used to analyze test quality and to establish a pass/fail standard. The final test of 23 test items had good internal reliability (Cronbach's α=0.85) and discriminatory abilities. The risks of false positives and negatives (9.1% and 23.6%, respectively) were acceptable for the test to be used as a certification tool prior to supervised clinical training in CEUS.

Enrichment analysis on regulatory subspaces: A novel direction for the superior description of cellular responses to SARS-CoV-2

StatementEnrichment analysis of cell transcriptional responses to SARS-CoV-2 infection from biclustering solutions yields broader coverage and superior enrichment of GO terms and KEGG pathways against alternative state-of-the-art machine learning solutions, thus aiding knowledge extraction. Motivation and methodsThe comprehensive understanding of the impacts of SARS-CoV-2 virus on infected cells is still incomplete. This work aims at comparing the role of state-of-the-art machine learning approaches in the study of cell regulatory processes affected and induced by the SARS-CoV-2 virus using transcriptomic data from both infectable cell lines available in public databases and in vivo samples. In particular, we assess the relevance of clustering, biclustering and predictive modeling methods for functional enrichment. Statistical principles to handle scarcity of observations, high data dimensionality, and complex gene interactions are further discussed. In particular, and without loos of generalization ability, the proposed methods are applied to study the differential regulatory response of lung cell lines to SARS-CoV-2 (α-variant) against RSV, IAV (H1N1), and HPIV3 viruses. ResultsGathered results show that, although clustering and predictive algorithms aid classic stances to functional enrichment analysis, more recent pattern-based biclustering algorithms significantly improve the number and quality of enriched GO terms and KEGG pathways with controlled false positive risks. Additionally, a comparative analysis of these results is performed to identify potential pathophysiological characteristics of COVID-19. These are further compared to those identified by other authors for the same virus as well as related ones such as SARS-CoV-1. The findings are particularly relevant given the lack of other works utilizing more complex machine learning algorithms within this context.

PM 7/20 (3) Erwinia amylovora

PM 7/20 (3) <i>Erwinia amylovora</i>

The Importance of Random Slopes in Mixed Models for Bayesian Hypothesis Testing

Mixed models are gaining popularity in psychology. For frequentist mixed models, previous research showed that excluding random slopes—differences between individuals in the direction and size of an effect—from a model when they are in the data can lead to a substantial increase in false-positive conclusions in null-hypothesis tests. Here, I demonstrated through five simulations that the same is true for Bayesian hypothesis testing with mixed models, which often yield Bayes factors reflecting very strong evidence for a mean effect on the population level even if there was no such effect. Including random slopes in the model largely eliminates the risk of strong false positives but reduces the chance of obtaining strong evidence for true effects. I recommend starting analysis by testing the support for random slopes in the data and removing them from the models only if there is clear evidence against them.

Cascaded learning in intravascular ultrasound: coronary stent delineation in manual pullbacks.

Purpose: Implanting stents to re-open stenotic lesions during percutaneous coronary interventions is considered a standard treatment for acute or chronic coronary syndrome. Intravascular ultrasound (IVUS) can be used to guide and assess the technical success of these interventions. Automatically segmenting stent struts in IVUS sequences improves workflow efficiency but is non-trivial due to a challenging image appearance entailing manifold ambiguities with other structures. Manual, ungated IVUS pullbacks constitute a challenge in this context. We propose a fully data-driven strategy to first longitudinally detect and subsequently segment stent struts in IVUS frames. Approach: A cascaded deep learning approach is presented. It first trains an encoder model to classify frames as "stent," "no stent," or "no use." A segmentation model then delineates stent struts on a pixel level only in frames with a stent label. The first stage of the cascade acts as a gateway to reduce the risk for false positives in the second stage, the segmentation, which is trained on a smaller and difficult-to-annotate dataset. Training of the classification and segmentation model was based on 49,888 and 1826 frames of 74 sequences from 35 patients, respectively. Results: The longitudinal classification yielded Dice scores of 92.96%, 82.35%, and 94.03% for the classes stent, no stent, and no use, respectively. The segmentation achieved a Dice score of 65.1% on the stent ground truth (intra-observer performance: 75.5%) and 43.5% on all frames (including frames without stent, with guidewires, calcium, or without clinical use). The latter improved to 49.5% when gating the frames by the classification decision and further increased to 57.4% with a heuristic on the plausible stent strut area. Conclusions: A data-driven strategy for segmenting stents in ungated, manual pullbacks was presented-the most common and practical scenario in the time-critical clinical workflow. We demonstrated a mitigated risk for ambiguities and false positive predictions.

False-positive and false-negative risks for individual multicentre trials in critical care.

In medical research, null hypothesis significance testing (NHST) is the dominant framework for statistical inference. NHST involves calculating P-values and confidence intervals to quantify the evidence against the null hypothesis of no effect. However, P-values and confidence intervals cannot tell us the probability that the hypothesis is true. In contrast, false-positive risk (FPR) and false-negative risk (FNR) are post-test probabilities concerning the truth of the hypothesis, that is to say, the probability a real effect exists. We calculated the FPR or FNR for 53 individual multicentre trials in critical care based on a pretest probability of 0.5 that the hypothesis was true. For trials reporting statistical significance, the FPR varied between 0.1% and 57.6%. For trials reporting non-significance, the FNR varied between 1.7% and 36.9%. Twenty-six of 47 trials (55.3%) reporting non-significance provided strong or very strong evidence in favour of the null hypothesis; the remaining trials provided limited evidence. There was no obvious relationship between the P-value and the FNR. The FPR and FNR showed marked variability, indicating that the probability of a real or absent treatment effect differed substantially between trials. Only one trial reporting statistical significance provided convincing evidence of a real treatment effect, and nearly half of all trials reporting non-significance provided limited evidence for the absence of a treatment effect. Our findings suggest that the quality of evidence from multicentre trials in critical care is highly variable.

Bayesian identification of structural coefficients in causal models and the causal false-positive risk of confounders and colliders in linear Markovian models

BackgroundCausal inference has seen an increasing popularity in medical research. Estimation of causal effects from observational data allows to draw conclusions from data when randomized controlled trials cannot be conducted. Although the identification of structural causal models (SCM) and the calculation of structural coefficients has received much attention, a key requirement for valid causal inference is that conclusions are drawn based on the true data-generating model.MethodsIt remains widely unknown how large the probability is to reject the true structural causal model when observational data from it is sampled. The latter probability – the causal false-positive risk – is crucial, as rejection of the true causal model can induce bias in the estimation of causal effects. In this paper, the widely used causal models of confounders and colliders are studied regarding their causal false-positive risk in linear Markovian models. A simulation study is carried out which investigates the causal false-positive risk in Gaussian linear Markovian models. Therefore, the testable implications of the DAG corresponding to confounders and colliders are analyzed from a Bayesian perspective. Furthermore, the induced bias in estimating the structural coefficients and causal effects is studied.ResultsResults show that the false-positive risk of rejecting a true SCM of even simple building blocks like confounders and colliders is substantial. Importantly, estimation of average, direct and indirect causal effects can become strongly biased if a true model is rejected. The causal false-positive risk may thus serve as an indicator or proxy for the induced bias.ConclusionWhile the identification of structural coefficients and testable implications of causal models have been studied rigorously in the literature, this paper shows that causal inference also must develop new concepts for controlling the causal false-positive risk. Although a high risk cannot be equated with a substantial bias, it is indicative of the induced bias. The latter fact calls for the development of more advanced risk measures for committing a causal type I error in causal inference.

Overall performance of a drug\u2013drug interaction clinical decision support system: quantitative evaluation and end-user survey

BackgroundClinical decision support systems are implemented in many hospitals to prevent medication errors and associated harm. They are however associated with a high burden of false positive alerts and alert fatigue. The aim of this study was to evaluate a drug–drug interaction (DDI) clinical decision support system in terms of its performance, uptake and user satisfaction and to identify barriers and opportunities for improvement.MethodsA quantitative evaluation and end-user survey were performed in a large teaching hospital. First, very severe DDI alerts generated between 2019 and 2021 were evaluated retrospectively. Data collection comprised alert burden, override rates, the number of alert overrides reviewed by pharmacists and the resulting pharmacist recommendations as well as their acceptance rate. Second, an e-survey was carried out among prescribers to assess satisfaction, usefulness and relevance of DDI alerts as well as reasons for overriding.ResultsA total of 38,409 very severe DDI alerts were generated, of which 88.2% were overridden by the prescriber. In 3.2% of reviewed overrides, a recommendation by the pharmacist was provided, of which 79.2% was accepted. False positive alerts were caused by a too broad screening interval and lack of incorporation of patient-specific characteristics, such as QTc values. Co-prescribing of a non-vitamin K oral anticoagulant and a low molecular weight heparin accounted for 49.8% of alerts, of which 92.2% were overridden. In 88 (1.1%) of these overridden alerts, concurrent therapy was still present. Despite the high override rate, the e-survey revealed that the DDI clinical decision support system was found useful by prescribers.ConclusionsIdentified barriers were the lack of DDI-specific screening intervals and inclusion of patient-specific characteristics, both leading to a high number of false positive alerts and risk for alert fatigue. Despite these barriers, the added value of the DDI clinical decision support system was recognized by prescribers. Hence, integration of DDI-specific screening intervals and patient-specific characteristics is warranted to improve the performance of the DDI software.

The impact of the ultrasound classification on the rate of thyroid surgery indications: a 577 cases series.

Nodules of the thyroid gland are common but rarely malignant. Their management can range from simple monitoring to surgery. The use of ultrasound and fine needle aspiration can reduce the rate of unnecessary surgeries. However, there is a risk of false positives and false negatives of malignancy that only pathology can avoid. The objective of this study is to assess the impact of ultrasound classification on the rate of surgical indications. Between 2013 and 2017, the ultrasound classification was gradually adopted in our daily practice to become now routine. During this period, we conducted a retrospective study of all the patients who presented to our department for one or more thyroid nodules. A total of 577 patients were included in the study. We compared two groups, a first where the ultrasound classification was used and a second where this classification was not used. In the end, we found that this classification significantly reduced the surgical indication by 19% while increasing the malignancy detection rate in operated patients by 21%. The use of ultrasound classification reduces the indications for surgery while increasing the rate of malignancy in operated patients. The generalization of the use of the ultrasound classification score is strongly recommended in daily practice.