Low Positive Predictive Value Research Articles

Meta-Analysis of the Performance of AI-Driven ECG Interpretation in the Diagnosis of Valvular Heart Diseases

Valvular heart diseases (VHDs) significantly impact morbidity and mortality rates worldwide. Early diagnosis improves patient outcomes. Artificial intelligence (AI) applied to electrocardiogram (ECG) interpretation presents a promising approach for early VHD detection. We conducted a meta-analysis on the efficacy of AI models in this context. We reviewed databases including PubMed, MEDLINE, Embase, Scopus, and Cochrane until August 20, 2023, focusing on AI for ECG-based VHD detection. The outcomes included pooled accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value. The pooled proportions were derived using a random-effects model with 95% confidence intervals (CIs). Study heterogeneity was evaluated with the I-squared statistic. Our analysis included 10 studies, involving ECG data from 713,537 patients. The AI algorithms mainly screened for aortic stenosis (n=6), mitral regurgitation (n=4), aortic regurgitation (n=3), mitral stenosis (n=1), mitral valve prolapse (n=2), and tricuspid regurgitation (n=1). A total of 9 studies used convolution neural network models, whereas 1 study combined the strengths of support vector machine logistic regression and multilayer perceptron for ECG interpretation. The collective AI models demonstrated a pooled accuracy of 81% (95% CI 73 to 89, I²=92%), sensitivity was 83% (95% CI 77 to 88, I²=86%), specificity was 72% (95% CI 68 to 75, I²=52%), PPV was 13% (95% CI 7 to 19, I²=90%), and negative predictive value was 99% (95% CI 97 to 99, I²=50%). The subgroup analyses for aortic stenosis and mitral regurgitation detection yielded analogous outcomes. In conclusion, AI-driven ECG offers high accuracy in VHD screening. However, its low PPV indicates the need for a combined approach with clinical judgment, especially in primary care settings.

Pendelluft in patients with acute respiratory distress syndrome during trigger and reverse triggering breaths

Pendelluft, the shift of air from non-dependent to dependent lung regions, is known to occur during active breathing in ventilated patients. However, information about pendelluft in ARDS patients under assisted mechanical ventilation is limited. In this prospectively collected and retrospectively analyzed study, we combined electrical impedance tomography and respiratory mechanics monitoring to quantitatively examine pendelluft in trigger and reverse triggering breaths in 20 mechanically ventilated patients with ARDS during the transition from controlled to active breaths under volume-cycled ventilation. Besides the 10 resting breaths in each patient, 20% of the counted active breaths were selected based on three levels of esophageal pressure swing (∆Pes): low (< 5 cm H2O, breaths = 471), moderate (≥ 5, < 10 cm H2O, breaths = 906), and high effort (≥ 10 cm H2O, breaths = 565). The pendelluft response to breathing efforts was significantly greater in trigger breaths than in reverse triggering breaths (p < 0.0001). Based on the pendelluft-∆Pes slope (ml/cmH2O), there were two distinct patterns of effort-related pendelluft (high vs. low pendelluft group). For trigger breaths, the high pendelluft group (n = 9, slope 0.7–2.4 ml/cmH2O) was significantly associated with lower peak airway/plateau pressure and lower respiratory system/lung elastance than the low pendelluft group (n = 11, slope − 0.1 to 0.3 ml/cmH2O). However, there was no difference in respiratory mechanics between high and low pendelluft groups for reverse triggering breathes. The use of ∆Pes to predict pendelluft was found to have a low positive predictive value.

Evaluation of Algorithms Using Automated Health Plan Data to Identify Breast Cancer Recurrences.

We updated algorithms to identify breast cancer recurrences from administrative data, extending previously developed methods. In this validation study, we evaluated pairs of breast cancer recurrence algorithms (vs. individual algorithms) to identify recurrences. We generated algorithm combinations that categorized discordant algorithm results as no recurrence [High Specificity and PPV (positive predictive value) Combination] or recurrence (High Sensitivity Combination). We compared individual and combined algorithm results to manually abstracted recurrence outcomes from a sample of 600 people with incident stage I-IIIA breast cancer diagnosed between 2004 and 2015. We used Cox regression to evaluate risk factors associated with age- and stage-adjusted recurrence rates using different recurrence definitions, weighted by inverse sampling probabilities. Among 600 people, we identified 117 recurrences using the High Specificity and PPV Combination, 505 using the High Sensitivity Combination, and 118 using manual abstraction. The High Specificity and PPV Combination had good specificity [98%, 95% confidence interval (CI): 97-99] and PPV (72%, 95% CI: 63-80) but modest sensitivity (64%, 95% CI: 44-80). The High Sensitivity Combination had good sensitivity (80%, 95% CI: 49-94) and specificity (83%, 95% CI: 80-86) but low PPV (29%, 95% CI: 25-34). Recurrence rates using combined algorithms were similar in magnitude for most risk factors. By combining algorithms, we identified breast cancer recurrences with greater PPV than individual algorithms, without additional review of discordant records. Researchers should consider tradeoffs between accuracy and manual chart abstraction resources when using previously developed algorithms. We provided guidance for future studies that use breast cancer recurrence algorithms with or without supplemental manual chart abstraction.

Dimeric immunoglobulin A as a novel diagnostic marker of measles infection.

The world is facing a measles resurgence, and improved diagnostic tests for measles infection are an urgent World Health Organization research priority. Detection of measles-specific immunoglobulin M (IgM) as a standard diagnostic test has low positive predictive value in elimination settings, and there is a need for new biomarkers of measles infection to enable enhanced surveillance and response to outbreaks. We demonstrate the detection of measles-specific dimeric immunoglobulin A (dIgA) in patients with confirmed measles infections using a new indirect enzyme-linked immunosorbent assay protocol that selects for the dIgA fraction from total IgA in the blood. The magnitude of measles-specific dIgA responses showed a low correlation with IgM responses, and our results highlight the potential of dIgA for further development as an alternative and/or complementary biomarker to IgM for serological diagnosis of measles infection.

Evaluation of Sysmex UF-5000 flow cytometer flag BACT-info for Gram discrimination in urinary tract infection

Urine culture as a gold standard for the diagnosis of urinary tract infection (UTI) involves a considerable workload in Clinical Microbiology Departments, due to the high number of samples received that will ultimately be negative. Therefore, it is necessary to use screening systems that also reduce the turnaround time for UTI diagnosis. The new flow cytometer UF-5000 (Sysmex Corporation) is able to differentiate between Gram-negative and Gram-positive bacteria using the BACT-info parameter according to manufacturer. The aim of our study was to evaluate the gram discrimination ability of the UF-5000 cytometer. A prospective study with 449 urine samples collected consecutively was conducted, in the period 7/3/2022-27/5/2022, in which the BACT-info flag was compared with urine culture as the reference method. The sensitivity obtained for both Gram-negative and Gram-positive bacteria was above 95%. However, for Gram-positive bacteria, the moderate Kappa index (0.49) and the low positive predictive value (37.1%) indicated that the correlation between BACT-info flag and urine culture was not acceptable and should not be reported to the requesting clinician. Implementation of the third generation UF-5000 cytometer represents a significant advance in the aetiological orientation of UTIs caused by Gram-negative bacteria. Reporting the Gram morphology in the urine samples reduces the response time in the microbiological diagnosis of UTI, which would have an impact on the reduction and optimisation of empirical treatment, and thus on the generation of antimicrobial resistance.

Soong Classification Using Radiographs Only Moderately Correlates With Distal Radius Plate Position on Computed Tomography

Background: The Soong classification grades the prominence of volar locking plates used to treat distal radius fractures in relation to the volar rim. The basis of the classification scheme is that increasing plate prominence over the volar rim and distal radius watershed zone is associated with increased likelihood of flexor tendon irritation and need for plate removal. However, recent studies report mixed results on the predictive value of the Soong classification for these outcomes. We hypothesized that the decreased predictive accuracy of Soong classification is due to misclassification between Soong grades secondary to a suboptimal correlation between the Soong classification on radiographs (XRs) and computed tomography (CT). Methods: Fifty volarly displaced distal radius fractures treated with a volar locking plate in the international and publicly available ICUC database were reviewed. All cases with a postoperative XR and CT were included. Soong classification of the volar locking plate in relation to the volar ulnar rim was determined on both XR and CT by 2 independent, fellowship-trained hand surgeons using CT imaging as the gold standard. The distribution of Soong grades on XR and CT was compared using Pearson’s χ2 test, and correlation was calculated using the Matthews correlation coefficient (MCC). A multi-class confusion matrix was used to calculate each grade’s positive predictive value (PPV). Results: We found an MCC of 0.65, indicating only moderate correlation between the 2 modalities. Per individual Soong grade, the PPV was the highest for grade 2 (0.96), with lower PPVs for grade 0 (0.63) and grade 1 (0.60). Conclusions: The distribution of Soong grades was significantly different when using XR versus CT ( P &lt; .001). Level of Evidence: Oxford Centre for Evidence-Based Medicine, diagnostic, level 2b

The Yield of Routine Post-Operative Doppler Ultrasound to Detect Early Post-Liver Transplantation Vascular Complications.

Early detection of liver transplantation (LT) vascular complications enables timely management. Our aim was to assess if routine Doppler ultrasound (rDUS) improves the detection of hepatic artery thrombosis (HAT), portal vein thrombosis (PVT) and hepatic venous outflow obstruction (HVOO). We retrospectively analysed timing and outcomes, number needed to diagnose one complication (NND) and positive predictive value (PPV) of rDUS on post-operative day (POD) 0,1 and 7 in 708 adult patients who underwent primary LT between 2010-2022. We showed that HAT developed in 7.1%, PVT in 8.2% and HVOO in 3.1% of patients. Most early complications were diagnosed on POD 0 (26.9%), 1 (17.3%) and 5 (17.3%). rDUS correctly detected 21 out of 26 vascular events during the protocol days. PPV of rDUS was 53.8%, detection rate 1.1% and NND was 90.5. Median time to diagnosis was 4days for HAT and 47days for PVT and 21days for HVOO. After intervention, liver grafts were preserved in 57.1%. In conclusion, rDUS protocol helps to detect first week's vascular events, but with low PPV and a high number of ultrasounds needed.

Proposed New Cutoff Value for Troponin in the Diagnosis of Acute Myocardial Infarction (AMI) in Patients with Congestive Heart Failure (CHF) and/or End Stage Renal Disease (ESRD)

Abstract Introduction/Objective In a recent publication, we assessed the efficacy of two serum biomarkers for AMI, cardiac Troponin I (cTnI) and creatine kinase-MB (CK-MB), for patients presenting to our emergency department (ED) with acute chest pain. We found that two conditions, CHF and ESRD in 85 patients, caused a significant number of false positive results that resulted in lowering the specificity and positive predictive value of both tests, which were already reduced by the low prevalence of AMI in our patient population. Removal of these patients resulted in significant increases in the two latter parameters. In this study, we focus on cTnI to determine if changing the cutoff value can increase the sensitivity and specificity of diagnosing acute MI in patients with CHF and ESRD. Methods/Case Report A cohort of 380 patients who presented with acute chest pain in the ED and worked up for acute MI were included, 85 of whom had CHF and/or ESRD. A total of 47 patients (15.4% prevalence) were definitively diagnosed with AMI. An initial serum cTnI sample was collected for each patient. Specificity, sensitivity, positive predictive value (PPV) and negative predictive value (NPV) were then compared using the routine cutoff cTnI value (0.05 ng/mL) and the higher cutoff of 0.2 ng/mL. Results (if a Case Study enter NA) Using the original cutoff of 0.05 ng/mL for cTnI in the patients with CHF or ESRD, the sensitivity and NPV were both 100%, while the specificity and PPV were low at 40.3% and 23%, respectively. Interestingly, when the cutoff was raised from 0.05 to 0.2 ng/mL, the sensitivity decreased by about 8% to 92.3 (12 out of 13 patients with AMI), and the specificity increased almost two-fold to 76.4%. The PPV increased to 40.4% and NPV was 98.2%. These values are comparable with the sensitivity, specificity, PPV and NPV of the total population at 85.3%, 82.4%, 38.7% and 97.7%, respectively. Conclusion The original cutoff value for cTnI in patients with ESRD and CHF produces a very low specificity and PPV. The results of this study suggest that an acceptable cutoff for cTnI to rule-out AMI patients who have CHF and or ESRD is 0.2 ng/mL which is four times higher than the original, but has comparable specificity, sensitivity, PPV and PPV to the total patient population. We note that this study must be regarded as a pilot study since the numbers of patients are small. We recommend a larger study to determine the validity of these conclusions.

Residual Component Culture: Application of Standardized Transfusion Reaction Culture Criteria to Reduce Culture Rates

Abstract Blood products are often cultured as part of a transfusion reaction (TRXN) workup in order to “not miss” a septic TRXN. These cultures are vulnerable to secondary contamination and are often challenging to interpret due to the lack of a corresponding patient blood culture. Although bacterial contamination of blood products remains a leading cause of transfusion related fatalities, the actual prevalence of primary contamination of blood components is quite low, leading to a low positive predictive value (PPV) for component culture. Given recently mandated mitigation strategies to reduce bacterial contamination of platelets, the PPV of component culture is likely to further decline, making appropriate selection criteria for culture even more critical to conserve laboratory resources and maximize clinical utility. At our institution the decision to culture is at the discretion of the pathology resident and/or transfusion medicine physician. We sought to review our current culture rates and determine if applying standardized indications for component culture could reduce culture rates without impacting diagnostic utility. Clinical symptoms and vital signs were reviewed for the 24-hour period after the reaction for all TRXNs reported from July 2021 to June 2022. Of the 105 reactions reported during the study timeframe, 39 (37%) were cultured, with one positive culture involving a platelet unit. This was deemed a true septic reaction given concordant cultures in the patient. Modified forms of the AABB and BEST culture criteria were then applied to the reported reactions. Modifications were made to provide specific definitions for tachycardia (heart rate &amp;gt; 100bpm and at least 15% increase) and hypotension (SBP &amp;lt; 80mmhg and at least a 30mmHg decrease). Application of the AABB criteria reduces the number of cultures to 22 (21% of reactions) whereas the BEST criteria increases the number of cultures to 51 (49% of reactions). Both criteria captured the true positive septic reaction. Neither criteria were completely concordant with our current practice. Of the 39 reactions selected for culture by our institution, 26 (67%) did not meet AABB criteria and 11 (28%) did not met BEST criteria; 9 and 22 different reactions would have been cultured based on AABB and BEST criteria respectively. Although we are not able to confirm if culturing of the discrepant units would have yielded additional positive cultures, retrospective chart review did not reveal any additional cases of clinically significant sepsis. The modified AABB criteria decreased the culture rate by 56% while still capturing the true positive septic TRXN suggesting that it would be the best option to reduce the cost/burden of culturing products without reducing diagnostic accuracy. Both published criteria do not take into consideration duration of symptoms and clinical interventions, thus clinical judgement remains critical to the decision to culture.

Shock score for prediction of clinical outcomes among stable patients with acute symptomatic pulmonary embolism

BackgroundThe Composite Pulmonary Embolism Shock (CPES) score has been developed to identify normotensive patients with acute pulmonary embolism (PE) and a low cardiac index (referred to as normotensive shock). We aimed to externally assess the validity of this model for predicting a complicated course among hemodynamically stable patients with acute PE. MethodsUsing prospectively collected data from the PROgnosTic valuE of Computed Tomography scan (PROTECT) study, we calculated the CPES score for each patient and the proportion of patients with a score > 3. We calculated the test performance characteristics to predict a complicated course (i.e., death from any cause, hemodynamic collapse, or recurrent PE) and the discriminatory power using the area under the receiver operating characteristic curve. ResultsSixty-three of the 848 (7.4 %) patients had a complicated course during the 30-day follow-up period. Of the 848 enrolled patients, the CPES score was positive (i.e., score > 3) in 78 (9.2 %). The specificity was 92.1 % (723/785), the positive predictive value was 20.5 % (16/78), and the positive likelihood ratio was 3.22 for a complicated course. The areas under the receiver operating characteristic curve for a complicated course were 0.71 (95 % confidence interval [CI], 0.65–0.78). With the higher score risk classification threshold (cutoff score > 4), the proportion of patients designated as positive was 2.1 %, and the specificity was 98.1 %. When echocardiographic right ventricle (RV) dysfunction was replaced by computed tomographic RV enlargement, the specificity was 85.4 %, the positive predictive value was 14.2 %, and the positive likelihood ratio was 2.06 for a complicated course. When analyses were restricted to the subgroup of patients with intermediate-risk PE, the specificity and the positive predictive value for a complicated course were identical to the overall cohort. ConclusionsThe CPES score has acceptable C-statistic, excellent specificity, and low positive predictive value for identification of hemodynamic deterioration in normotensive patients with PE.ClinicalTrials.gov number: NCT02238639.

Inappropriate requests for cranial CT scans in emergency departments increase overuse and reduce test performance.

The number of non-traumatic urgent cranial computed tomography (NT-UCCT) is exponentially increasing but limited research has been conducted on the quality of clinical justification. Accordingly, we aimed (1) to assess how clinical information in the electronic NT-UCCT request agreed with that provided in the patient's emergency department discharge summary and (2) to analyze the potential effect of those discrepancies on the NT-UCCT overload. Patients undergoing NT-UCCT in 2017-2021 were randomly selected for this retrospective research-board-approved study. Signs and symptoms (S/S) in electronic request and emergency department discharge summary, acute and relevant computed tomography (CT) findings (acute ischemia or hemorrhage, masses, brain edema, or previously undetected hydrocephalus), and final diagnosis at emergency department discharge summary were collected. Concordance between digital request and emergency department discharge summary and their association with both acute and relevant CT findings and final diagnosis were analyzed. We recruited 156 patients: 80 men; mean age, 55. Acute, relevant CT findings were detected in 28 cases (17.9%). The final diagnosis was neurological disease, non-neurological disease, and no definitive diagnosis in 46 (29.5%), 58 (37.2%), and 51 (32.7%) cases, respectively. Full agreement between the electronic request and emergency department discharge summary occurred in only 36 patients (23.1%). Motor deficit was the most frequent false positive electronic request S/S (18; 11.54%), having low positive predictive value (30.30%; 95%CI 15.59-48.71%) andworst association with acute relevant CT findings than when true positive (OR 2.54; 95%CI 0.04-6.21 vs. OR 6.26, 95%CI 2.21-17.78). Nausea/vomiting was the third most common false negative electronic request S/S (13; 10.26%) and reduced the likelihood of acute and relevant CT findings (OR 0.126; 95%CI 0.016-0.971; p = 0.020). False S/S in electronic request predominated in non-neurological diseases (50-60.2% vs. 33-39.8%; p = 0.068). Discrepancies between electronic request and emergency department discharge summary were observed in >75% of patients, leading to unnecessary NT-UCCT tests.

Use of airway pressure-based indices to detect high and low inspiratory effort during pressure support ventilation: a diagnostic accuracy study

BackgroundAssessment of the patient’s respiratory effort is essential during assisted ventilation. We aimed to evaluate the accuracy of airway pressure (Paw)-based indices to detect potential injurious inspiratory effort during pressure support (PS) ventilation.MethodsIn this prospective diagnostic accuracy study conducted in four ICUs in two academic hospitals, 28 adult acute respiratory failure patients undergoing PS ventilation were enrolled. A downward PS titration was conducted from 20 cmH2O to 2 cmH2O at a 2 cmH2O interval. By performing an end-expiratory airway occlusion maneuver, the negative Paw generated during the first 100 ms (P0.1) and the maximal negative swing of Paw (∆Pocc) were measured. After an end-inspiratory airway occlusion, Paw reached a plateau, and the magnitude of change in plateau from peak Paw was measured as pressure muscle index (PMI). Esophageal pressure was monitored and inspiratory muscle pressure (Pmus) and Pmus–time product per minute (PTPmus/min) were used as the reference standard for the patient’s effort. High and low effort was defined as Pmus > 10 and < 5 cmH2O, or PTPmus/min > 200 and < 50 cmH2O s min−1, respectively.ResultsA total of 246 levels of PS were tested. The low inspiratory effort was diagnosed in 145 (59.0%) and 136 (55.3%) PS levels using respective Pmus and PTPmus/min criterion. The receiver operating characteristic area of the three Paw-based indices by the respective two criteria ranged from 0.87 to 0.95, and balanced sensitivity (0.83–0.96), specificity (0.74–0.88), and positive (0.80–0.91) and negative predictive values (0.78–0.94) were obtained. The high effort was diagnosed in 34 (13.8%) and 17 (6.9%) support levels using Pmus and PTPmus/min criterion, respectively. High receiver operating characteristic areas of the three Paw-based indices by the two criteria were found (0.93–0.95). A high sensitivity (0.80–1.00) and negative predictive value (0.97–1.00) were found with a low positive predictive value (0.23–0.64).ConclusionsBy performing simple airway occlusion maneuvers, the Paw-based indices could be reliably used to detect low inspiratory efforts. Non-invasive and easily accessible characteristics support their potential bedside use for avoiding over-assistance. More evaluation of their performance is required in cohorts with high effort.

Diagnosis of Endometriosis Based on Comorbidities: A Machine Learning Approach.

Endometriosis is defined as the presence of estrogen-dependent endometrial-like tissue outside the uterine cavity. Despite extensive research, endometriosis is still an enigmatic disease and is challenging to diagnose and treat. A common clinical finding is the association of endometriosis with multiple diseases. We use a total of 627,566 clinically collected data from cases of endometriosis (0.82%) and controls (99.18%) to construct and evaluate predictive models. We develop a machine learning platform to construct diagnostic tools for endometriosis. The platform consists of logistic regression, decision tree, random forest, AdaBoost, and XGBoost for prediction, and uses Shapley Additive Explanation (SHAP) values to quantify the importance of features. In the model selection phase, the constructed XGBoost model performs better than other algorithms while achieving an area under the curve (AUC) of 0.725 on the test set during the evaluation phase, resulting in a specificity of 62.9% and a sensitivity of 68.6%. The model leads to a quite low positive predictive value of 1.5%, but a quite satisfactory negative predictive value of 99.58%. Moreover, the feature importance analysis points to age, infertility, uterine fibroids, anxiety, and allergic rhinitis as the top five most important features for predicting endometriosis. Although these results show the feasibility of using machine learning to improve the diagnosis of endometriosis, more research is required to improve the performance of predictive models for the diagnosis of endometriosis. This state of affairs is in part attributed to the complex nature of the condition and, at the same time, the administrative nature of our features. Should more informative features be used, we could possibly achieve a higher AUC for predicting endometriosis. As a result, we merely perceive the constructed predictive model as a tool to provide auxiliary information in clinical practice.

Validity of16 AHRQ Patient Safety Indicators to identify in-hospital complications: a medical record review across nine Swiss hospitals.

The validity of the Agency for Healthcare Research and Quality's Patient Safety Indicators (PSIs) has been established in the USA and Canada. However, these indicators are also used for hospital benchmarking and cross-country comparisons in other nations with different health-care settings and coding systems as well as missing present on admission (POA) flags in the administrative data. This study sought to comprehensively assess and compare the validity of 16 PSIs in Switzerland, where they have not been previously applied. We performed a medical record review using administrative and electronic medical record data from nine Swiss hospitals. Seven independent reviewers evaluated 1245 cases at various hospitals using retrospective data from the years 2014-18. True positives, false positives, positive predictive values (PPVs), and reasons for misclassification were compared across all investigated PSIs, and the documentation quality of the PSIs was examined. PSIs 6 (iatrogenic pneumothorax), 10 (postoperative acute kidney injury), 11 (postoperative respiratory failure), 13 (postoperative sepsis), 14 (wound dehiscence), 17 (birth trauma), and 18 and 19 (obstetric trauma with or without instrument) showed high PPVs (range: 90-99%) and were not strongly influenced by missing POA information. In contrast, PSIs 3 (pressure ulcer), 5 (retained surgical item), 7 (central venous catheter-related bloodstream infection), 8 (fall with hip fracture), and 15 (accidental puncture/laceration) showed low PPVs (range: 18-49%). In the case of PSIs 3, 8, and 12 (perioperative embolism/thrombosis), the low PPVs were largely due to the lack of POA information. Additionally, it was found that the documentation of PSI 3 in discharge letters could be improved. We found large differences in validity across the 16 PSIs in Switzerland. These results can guide policymakers in Switzerland and comparable health-care systems in selecting and prioritizing suitablePSIs for quality initiatives. Furthermore, the national introduction of a POA flag would allow for the inclusion of additional PSIs in quality monitoring.

Population-wide evaluation of artificial intelligence and radiologist assessment of screening mammograms

ObjectivesTo validate an AI system for standalone breast cancer detection on an entire screening population in comparison to first-reading breast radiologists.Materials and methodsAll mammography screenings performed between August 4, 2014, and August 15, 2018, in the Region of Southern Denmark with follow-up within 24 months were eligible. Screenings were assessed as normal or abnormal by breast radiologists through double reading with arbitration. For an AI decision of normal or abnormal, two AI-score cut-off points were applied by matching at mean sensitivity (AIsens) and specificity (AIspec) of first readers. Accuracy measures were sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and recall rate (RR).ResultsThe sample included 249,402 screenings (149,495 women) and 2033 breast cancers (72.6% screen-detected cancers, 27.4% interval cancers). AIsens had lower specificity (97.5% vs 97.7%; p < 0.0001) and PPV (17.5% vs 18.7%; p = 0.01) and a higher RR (3.0% vs 2.8%; p < 0.0001) than first readers. AIspec was comparable to first readers in terms of all accuracy measures. Both AIsens and AIspec detected significantly fewer screen-detected cancers (1166 (AIsens), 1156 (AIspec) vs 1252; p < 0.0001) but found more interval cancers compared to first readers (126 (AIsens), 117 (AIspec) vs 39; p < 0.0001) with varying types of cancers detected across multiple subgroups.ConclusionStandalone AI can detect breast cancer at an accuracy level equivalent to the standard of first readers when the AI threshold point was matched at first reader specificity. However, AI and first readers detected a different composition of cancers.Clinical relevance statementReplacing first readers with AI with an appropriate cut-off score could be feasible. AI-detected cancers not detected by radiologists suggest a potential increase in the number of cancers detected if AI is implemented to support double reading within screening, although the clinicopathological characteristics of detected cancers would not change significantly.Key Points• Standalone AI cancer detection was compared to first readers in a double-read mammography screening population.• Standalone AI matched at first reader specificity showed no statistically significant difference in overall accuracy but detected different cancers.• With an appropriate threshold, AI-integrated screening can increase the number of detected cancers with similar clinicopathological characteristics.

Abstract 18812: Premortem Risk Assessment Among Countywide Sudden Deaths

Introduction: Heart failure (HF) and left ventricular ejection fraction (EF) assessment have been hallmarks of sudden cardiac death (SCD) risk stratification for over 20 years. However, their sensitivity, specificity, and positive predictive value (PPV) for true arrhythmic death in the general population are unknown because nearly all SCD studies presume arhythmic causes. Aims: To define sensitivity, specificity, and PPV of current SCD risk stratification criteria for the autopsy-confirmed arrhythmic cause among countywide sudden deaths. Methods: POST SCD (POstmortem SysTematic Investigation of Sudden Cardiac Death) is a prospective cohort study using autopsy, clinical records, and toxicology to adjudicate arrhythmic (SAD; potentially rescuable with implantable defibrillator) or non-arrhythmic (non-SAD; e.g., tamponade, overdose) causes among presumed SCDs (pSCDs) in San Francisco County. We analyzed sensitivity, specificity, and PPV for syncope, EF≤35%, diagnosed HF, or the presence of any of 3 factors to discriminate SADs among pSCDs. Results: Between February 2011 and September 2022, 845 of 904 (93%) pSCDs 18-90 years had complete medical records available, of which 491 (58%) were SADs. Seventy-three pSCDs had history of syncope, with similar prevalence in SADs and non-SADs (p=0.4), while prevalence of HF was higher in SADs vs. non-SADs (88/491[18%] vs. 39/354 [11%], p=0.003). EF assessment was performed in 206 pSCDs (24%, 133 SADs, 73 non-SADs); prevalence of EF ≤ 35% was similar between SADs and non-SADs (p=0.112). A greater proportion of SADs had ≥1 factor than non-SADs (122/491[25%] vs. 65/354 [18%], p=0.015). Sensitivity of syncope, EF≤35%, HF, ≥1 factor for SAD was 9, 26, 17, 25%, specificity was 92, 82, 90, 82%, and PPV was 60, 73, 69, 65%, respectively. Conversely, 369 of 491 (75%) SADs had no risk factors before sudden death. Defibrillators were implanted in 4/491 (0.8%) SADs. Conclusions: In this 11-year countywide postmortem study, conventional risk factors had low sensitivity and PPV but high specificity for autopsy-confirmed arrhythmic cause of sudden death. Three-fourths of SADs had no risk factors before sudden death, highlighting the need for improved risk prediction in the general population.

Abstract 16342: Age- and Sex-Differences in Accuracy of ICD-10 Codes for Identifying Adults With Pulmonary Embolism: PE-EHR+ Study

Background Age and sex-differences exist in presentation of adults with pulmonary embolism (PE). However, it remains unclear if these differences influence the accuracy of International Classification of Diseases, Tenth Revision (ICD-10) codes used for identification of PE. Methods: Using multicenter data from Mass General-Brigham Health System (2016-2021), patients were randomly selected from 3 groups: 578 patients with Principal Discharge codes for PE, 578 with Secondary Discharge codes for PE, and 578 without codes for PE. The reference standard for identification of PE was manual chart review by 2 independent physicians. Sensitivity, specificity, and positive and negative predictive values were determined by age (&lt;65 vs ≥65 years) and sex (female vs male). Weighted estimates were obtained by using the total number of hospitalizations. Results: Among 1712 patients included in final analyses, 896 (52.3%) were female and 775 (45.3%) were ≥65 years. Use of Principal-or-Secondary ICD-10 codes had a similar sensitivity by sex (99.5% vs 100% among females vs males) and age (99.8% among both groups) in the study sample. The weighted sensitivity among females was 72.8% (vs 100% for males) and 80.4% among patients ≥65 years (vs 85.2% for &lt;65 years). In the weighted cohort, the positive predictive value was 78.2% among females (vs 80.3% for males) and 78.3% among those ≥65 years (vs 80.0% for &lt;65 years) (Table). Age and sex-differences in the diagnostic accuracy were similarly observed in analyses restricted to Principal, or to Secondary codes. Conclusions: Principal-or-Secondary ICD-10 codes had lower sensitivity and positive predictive value among females and older adults, though the sex-differences in sensitivity may be influenced by limited number of false negatives in this study. Additionally, Secondary codes had higher false positives among older adults. These findings may have important implications for identification of patients with PE using ICD-10 codes.

ATA risk stratification in papillary thyroid microcarcinoma has low positive predictive value when identifying recurrence

BackgroundRising incidence of papillary thyroid microcarcinomas (PTMC) has raised concerns for overdiagnosis. Utility of the American Thyroid Association Risk Stratification System (ATA-RSS) 2015 in predicting risk of disease recurrence in patients with PTMC was assessed. MethodsElectronic health records of patients who underwent total thyroidectomy were queried. ATA-RSS 2015 risk stratification was performed on those with PTMC, and validity for predicting disease recurrence was calculated. ResultsWith 10-year median follow up, recurrence was higher in PTMC patients with high/intermediate vs low ATA risk (33 % vs 4 %, p = 0.002). Sensitivity of ATA-RSS for detecting recurrence was 60 %, specificity 90 %, PPV 33.3 %, NPV 96.6 %, and accuracy 88 %. When microscopic extrathyroidal extension (ETE) was excluded as an intermediate risk criterion, PPV improved to 50 % and accuracy improved to 92.5 % ConclusionsATA-RSS 2015 predicts recurrence in PTMC with high NPV but low PPV. Exclusion of microscopic ETE improved PPV, which may help prevent overtreatment.

Diagnostic Performance of Red Blood Cell Indices, Serum Erythropoietin, and JAK2 Mutation Testing for the Evaluation of Polycythemia Vera at High Altitude

Introduction: Polycythemia vera (PV) is a clonal myeloproliferative neoplasm associated with mutation of Janus kinase 2 ( JAK2) in virtually all patients. Patients with PV typically present with constitutional symptoms and thrombotic complications related to myeloproliferation but can present with only mild or no symptoms. In 2016, the World Health Organization (WHO) diagnostic criteria for PV were revised to use lower hemoglobin and hematocrit cut-off levels in order to avoid missing masked polycythemia. Accordingly, a hemoglobin concentration of more than 16 g/dL (HCT&amp;gt;48%) for females and 16.5 g/dL (HCT&amp;gt;49%) for males can be considered for PV evaluation (previously &amp;gt;18.5 g/dL for men and &amp;gt;16.5 g/dL for women). In conjunction, concerns arise about the unnecessary evaluation of individuals with otherwise high-normal hemoglobin levels, more so for those living at high altitude. Altitude exposure, amongst others, decreases plasma volume to increase [Hb] which may accordingly muddle the diagnosis of PV at altitude. Nonetheless, Red Cell Mass (RCM) is seldom assessed in clinical practice. In addition, red blood cell count, related indices, and serum erythropoietin may also be affected by high altitude, and their diagnostic performance in the evaluation of PV at altitude has not been prospectively assessed. Method: We collected lab records for all JAK2 requests over three years (2020-2022) in two referral hospitals serving a population residing at an altitude of 2,270-3,000 meters above sea level. Indications for requests, as well as laboratory data and imaging, were gathered. Complete blood count (CBC) and other laboratory values were collected when ordered within one week of JAK2 testing. For CBC, EDTA anticoagulated samples were processed using Sysmex XN-3100. For JAK2 testing, RNA isolated from peripheral blood samples (using the MagNA Pure System) was reverse-transcribed, amplified, and sequenced for any mutations involving exons 12, 13, 14, or 15 using an ABI 3730 XL genetic analyzer. For serum EPO, frozen serum samples were outsourced to Bio Scientia Labs, Germany. Reference values for serum erythropoietin were lab generated at 4.3-29 mIU/ml and were not validated for our altitude. . Results:Of the 208 JAK2 tests performed during the reviewed period, 132 (63.4%) were for polycythemia (Table 1). The JAK2 V617F mutation was found in 22 (16.7%) patients, with 10 (45.5%) of them being men. No JAK2 mutation other than the exon 14 V617F substitution was found. JAK2-positive patients had a median age of 61 years, while JAK2-unmutated patients had a median age of 47 years. Overt polycythemia, according to the WHO 2016 classification, was identified in 15 (68.2%) , while the remainder had masked PV ( defined as increased RBCs with normal hemoglobin/ HCT). Patients with PV had significantly higher RBCs and lower MCV/MCH/MCHC compared to those with JAK2-unmutated polycythemia. They also exhibited significant elevation of RDW-CW and significant neutrophilia, basophilia, thrombocytosis, and splenomegaly. All PV patients with available serum EPO tests had low values (median 2 mIU/ml). Conclusion: At altitude, patients with PV show the same age and sex characteristics as those living at low elevations. Patients with PV have higher RBCs and lower MCV/MCH/MCHC compared to those with JAK-unmutated (secondary) polycythemia. RDW is significantly increased in patients with PV, reflecting an associated iron deficiency. Serum EPO performed well at altitude to suggest the presence of PV. The low rate of positive JAK2 mutation in our high-altitude cohort may reflect poor selection of subjects and over-workup of otherwise altitude-associated erythrocytosis. Isolated mild polycythemia at altitude remains common. Using hemoglobin/hematocrit alone to select patients for PV evaluation has a low positive predictive value and may be associated with unnecessary workup. RBC indices and RDW, along with an evaluation for iron status, might help select those more likely to have PV and could decrease unnecessary workups.

Right Atrial Enlargement on Electrocardiogram in Previously Healthy Young Patients

ObjectivesTo assess whether right atrial enlargement (RAE) on electrocardiogram (ECG) correlates with true RAE on echocardiogram in previously healthy young patients and to understand which patients with RAE on ECG may warrant additional testing. Study DesignA single-center, retrospective review of previously healthy young patients with 1) ECGs that were read as RAE by a pediatric cardiologist and 2) echocardiograms obtained within 90 days of the ECG. ECGs were reviewed to confirm RAE and determine which leads met criteria. The echocardiograms were then reviewed and RA measurements with Z-scores obtained. A Z-score >2 was considered positive for RAE on echocardiogram. Results162 patients with median age 10.8 years were included in the study. A total of 23 patients had true RAE on echocardiogram, giving a positive predictive value (PPV) of 14%. In patients <1 year of age, the PPV increased to 35%. In patients older than 1 year, the PPV was low at 7%. Patients with true RAE were more likely to meet criteria for RAE in the anterior precordial leads (V1-V3) (48% vs 5%, p< 0.001) and meet criteria for right ventricular hypertrophy (RVH) (22% vs 6%, p=0.023). ConclusionOur findings show that RAE on ECG has a low PPV for RAE on echocardiogram in previously healthy young patients. The highest yield for RAE on echocardiogram was observed in patients who were <1 year of age, had RAE in the anterior precordial leads, or displayed RVH on ECG.