Risk Stratification Classification Research Articles

Using artificial intelligence to predict post-operative outcomes in congenital heart surgeries: a systematic review

IntroductionCongenital heart disease (CHD) represents the most common group of congenital anomalies, constitutes a significant contributor to the burden of non-communicable diseases, highlighting the critical need for improved risk assessment tools. Artificial intelligence (AI) holds promise in enhancing outcome predictions for congenital cardiac surgery. This study aims to systematically review the utilization of AI in predicting post-operative outcomes in this population.MethodsFollowing PRISMA guidelines, a comprehensive search of Pubmed, Scopus, and Web of Science databases was conducted. Two independent reviewers screened articles based on predefined criteria. Included studies focused on AI models predicting various post-operative outcomes in congenital heart surgery.ResultsThe review included 35 articles, primarily published within the last four years, indicating growing interest in AI applications. Models predominantly targeted mortality and survival (n = 16), prolonged length of hospital or ICU stay (n = 7), postoperative complications (n = 6), prolonged mechanical ventilatory support time (n = 4), with additional focus on specific outcomes such as peri-ventricular leucomalacia (n = 2) and malnutrition (n = 1). Performance metrics, such as area under the curve (AUC), ranged from 0.52 to 0.997. Notably, these AI models consistently outperformed traditional risk stratification categories. For instance, in assessing the risk of morbidity and mortality, the AI models demonstrated superior performance compared to conventional methods.ConclusionAI-driven prediction models show significant promise in improving outcome predictions for congenital heart surgery. They surpass traditional risk prediction tools not only in immediate postoperative risks but also in long-term outcomes such as 1-year survival and malnutrition. Further studies with robust external validation are necessary to assess the practical applicability of these models in clinical settings.The protocol of this review was prospectively registered on PROSPERO (CRD42024550942).

Characteristics of Acute Myeloid Leukemia Patients with TP53 Alterations

Abstract Background TP53 is the most intensively studied gene in cancer. The vast majority of de novo AML patients have unaltered TP53 alleles; data from The Cancer Gene Atlas including adult patients with AML documented that ∼ 8% of cases harbor TP53 alterations. Objectives The objective of this study is to identify the features of TP53 alterations in acute myeloid leukemia (AML) patients, investigate its association with clinical findings, standard laboratory tests, morphological, cytogenetic and molecular profiles, and assess its influence on the overall outcome of patients on day 28 after induction therapy. Subjects and Methods The present study is an analytical observational study that was conducted on 60 de novo adult AML patients who presented to the Hematology/Oncology unit of Ain Shams University Hospitals. All patients’ samples were analyzed for TP53 deletion using LSI 17p13 FISH probe for detection of p53 deletion and they were further subjected to testing for TP53 Pro72Arg alterations by real time polymerase chain reaction. All patients were subjected to therapy and then followed up at day 28 to asses the outcome. Results Male gender and lower TLC were significantly associated with good outcome with (P-value ≤ 0.05). Patients with cytogenetic structural abnormalities showing 17q rearrangement showed significant association with bad outcome with (P-value ≤ 0.05) and t(8;21) was found to be highly significant in association with good outcome with (P-value< 0.01). FLT3ITD mutation was found to be highly significant in association with bad outcome with (P-value< 0.01). NPM mutation was significant in association with good outcome with (P-value ≤ 0.05). TP53 mutation was detected in 11.7 % of studied patients where 8.3% out of them were heterozygous mutation and 3.4% were homozygous. P53 mutation wassignificantly associated with advanced age, male gender and higher TLC with (P- value ≤ 0.05). Cytogenetic structural abnormalities including 11q23 rearrangement showed significant association with P53 mutation with (P-value ≤ 0.05) and t(8;21)+del 17p, del 17p& -7 and del 17p only was found to be highly significant in association with p53 mutation with (P-value< 0.01). Normal karyotyping showed negative relation with P53 mutation with (P-value ≤ 0.05) where all patients who yielded normal karyotyping (35%) were negative for P53 alterations; that doesn't necessarily denote causation, but it's an observable pattern, and del 17p that was found to be highly significant in association with p53 mutation with (P-value< 0.01). P53 mutation was found to be highly significant in association with bad outcome with (P-value < 0.01). Unfavorable and Favorable risk stratification of ELN 2022 were found to be highly significant in association with p53 mutation with (P-value< 0.01). ELN 2022 risk stratification categories were found to be highly significant in association with patients’ outcomes with (P- value < 0.01). Conclusion TP53 alterations strongly identify high-risk AML patients with poor outcome in this study yet, this needs further investigation on larger sample size with longer follow up. Investigations of TP53 mutation especially in Adult AML may help with the selection of patients in need of intensive therapeutic strategy or may help with designation of new innovative targeted therapies. Key words Tp53, Mutation, Adult Acute Myeloid Leukemia.

Nurse-Led Cardiac Rehabilitation Care Coordination Program: Improving Functional Outcomes for Patients Through Automatic Referral and Effective Care Coordination.

The aim of this investigation was to evaluate the impact of automated cardiac rehabilitation (CR) referral and nurse care coordination on patient and program outcomes. Specifically, the aim was to identify whether differences exist in physical and psychological function at CR Phase 2 enrollment and completion and CR Phase 2 participation and completion for hospitalized patients who receive in-person CR nurse visits versus phone consultation. Using a retrospective pre-/post-intervention descriptive design, a purposive sampling technique was used to select groups with matching clinical attributes. Dates were selected to mitigate the impact of COVID-19 on CR program enrollment and completion. Data were abstracted from the patient electronic medical record, telemetry documentation, and CR referral tracking tool. Patient descriptors included age, sex, cardiac diagnosis/procedure (post-coronary artery bypass graft surgery, myocardial infarction, percutaneous coronary intervention, heart failure, and aortic valve repair and replacement) and cardiac risk stratification category. Patient functional outcomes included the 6-min walk test and metabolic equivalents of task levels for functional capacity; psychological function was measured by the Patient Health Questionnaire assessment. Program outcomes included discharge to CR Phase 2 enrollment, CR sessions, and completion. Each group had 52 patients. Age was 64 ± 12yr, 68% were male. Perhaps indications for CR included coronary artery bypass graft surgery (44%), myocardial infarction (19%), percutaneous coronary intervention (20%), heart failure (10%), aortic valve repair and replacement (8%). Cardiac risk was low in 30%, intermediate in 65%, and high in 5%. The post-intervention group compared with the pre-intervention group had a shorter discharge to CR Phase 2 enrollment (35 ± 18 d vs 41 ± 28 d, P = .078) and significantly fewer sessions required for CR completion. Automated CR referral and nurse care coordination visits for hospitalized patients decreased the transition period between CR Phase 1 and 2. Patients were physically and psychologically prepared for earlier CR Phase 2 enrollment and successfully completed the program in fewer days than the pre-intervention group.

A-014 Analytical Outlier Occurrence and False Positivity Rate of the Beckman Coulter Access High-sensitivity Cardiac Troponin I Assay

Abstract Background Cardiac troponin (cTn) is the preferred biomarker for the diagnosis of acute myocardial infarction due to its high sensitivity and specificity relative to other available cardiac markers. Accurate measurement is therefore paramount to proper risk stratification of patients. Analytical outliers and false positive results have been previously reported on various platforms though the exact cause remains unknown. These false positive test results may pose a risk for the misclassification of patients and lead to adverse clinical outcomes. We evaluated the performance of the Beckman Coulter Access high-sensitivity cTn I method by measuring the occurrence of analytical outliers between duplicate results and the false positivity rate (i.e., fliers). Methods We tested 2535 samples from 1879 patients in duplicate on the Beckman Coulter Access hs-cTn I assay. These specimens had initial troponin measurements above the gender-specific 99th percentile of the upper reference limit (URL) cutoff and were immediately aliquoted, re-centrifuged, and re-tested. Non-reproducible results (outliers) were defined as having a concentration difference between the replicates exceeding 10%. Critical outliers were defined as duplicate results with concentrations that differed by greater than 10% from the initial result and the second result was classified in a lower risk stratification category than the original result. Results Approximately 12% (305) of the positive samples tested in duplicate had a percent difference that was >10% and therefore did not meet the recommended precision requirements of the assay. Of these specimens, 33 (10.8%) were classified as critical outliers because the repeat test result crossed a critical decision limit threshold. For sixteen samples the risk classification for the patient was downgraded from high (>50 ng/L) to moderate risk (13–50 ng/L (F); 20–50 ng/L (M)). Seventeen false positive results were identified in which the patients were downgraded from the high to low risk category (<13 ng/L (F); <20 ng/L (M)). Nearly 85% of the samples identified as outliers were lithium heparinized plasma. Conclusion We evaluated the reproducibility of the Beckman Coulter Access hs-cTn I method across 11 hospitals in an integrated healthcare network to characterize the occurrence of analytical outliers and false positive results. Our data indicated an outlier rate of 12% (305/2535) for positive hs-cTn I results; however, the risk category only changed for 1.3% (33/2535) of positive troponin results upon repeat analysis. Overall the false positivity rate was determined to be <0.1% (n = 55 560).

Diagnostic efficiency among Eu-/C-/ACR-TIRADS and S-Detect for thyroid nodules: a systematic review and network meta-analysis.

The performance in evaluating thyroid nodules on ultrasound varies across different risk stratification systems, leading to inconsistency and uncertainty regarding diagnostic sensitivity, specificity, and accuracy. Comparing diagnostic performance of detecting thyroid cancer among distinct ultrasound risk stratification systems proposed in the last five years. Systematic search was conducted on PubMed, EMBASE, and Web of Science databases to find relevant research up to December 8, 2022, whose study contents contained elucidation of diagnostic performance of any one of the above ultrasound risk stratification systems (European Thyroid Imaging Reporting and Data System[Eu-TIRADS]; American College of Radiology TIRADS [ACR TIRADS]; Chinese version of TIRADS [C-TIRADS]; Computer-aided diagnosis system based on deep learning [S-Detect]). Based on golden diagnostic standard in histopathology and cytology, single meta-analysis was performed to obtain the optimal cut-off value for each system, and then network meta-analysis was conducted on the best risk stratification category in each system. This network meta-analysis included 88 studies with a total of 59,304 nodules. The most accurate risk category thresholds were TR5 for Eu-TIRADS, TR5 for ACR TIRADS, TR4b and above for C-TIRADS, and possible malignancy for S-Detect. At the best thresholds, sensitivity of these systems ranged from 68% to 82% and specificity ranged from 71% to 81%. It identified the highest sensitivity for C-TIRADS TR4b and the highest specificity for ACR TIRADS TR5. However, sensitivity for ACR TIRADS TR5 was the lowest. The diagnostic odds ratio (DOR) and area under curve (AUC) were ranked first in C-TIRADS. Among four ultrasound risk stratification options, this systemic review preliminarily proved that C-TIRADS possessed favorable diagnostic performance for thyroid nodules. https://www.crd.york.ac.uk/prospero, CRD42022382818.

Cardiac MRI to Predict Sudden Cardiac Death Risk in Dilated Cardiomyopathy.

Background Sudden cardiac death (SCD) is one of the leading causes of death in individuals with nonischemic dilated cardiomyopathy (DCM). However, the risk stratification of SCD events remains challenging in clinical practice. Purpose To determine whether myocardial tissue characterization with cardiac MRI could be used to predict SCD events and to explore a SCD stratification algorithm in nonischemic DCM. Materials and Methods In this prospective single-center study, adults with nonischemic DCM who underwent cardiac MRI between June 2012 and August 2020 were enrolled. SCD-related events included SCD, appropriate implantable cardioverter-defibrillator shock, and resuscitation after cardiac arrest. Competing risk regression analysis and Kaplan-Meier analysis were performed to identify the association of myocardial tissue characterization with outcomes. Results Among the 858 participants (mean age, 48 years; age range, 18-83 years; 603 men), 70 (8%) participants experienced SCD-related events during a median follow-up of 33.0 months. In multivariable competing risk analysis, late gadolinium enhancement (LGE) (hazard ratio [HR], 1.87; 95% CI: 1.07, 3.27; P = .03), native T1 (per 10-msec increase: HR, 1.07; 95% CI: 1.04, 1.11; P < .001), and extracellular volume fraction (per 3% increase: HR, 1.26; 95% CI: 1.11, 1.44; P < .001) were independent predictors of SCD-related events after adjustment of systolic blood pressure, atrial fibrillation, and left ventricular ejection fraction. An SCD risk stratification category was developed with a combination of native T1 and LGE. Participants with a native T1 value 4 or more SDs above the mean (1382 msec) had the highest annual SCD-related events rate of 9.3%, and participants with a native T1 value 2 SDs below the mean (1292 msec) and negative LGE had the lowest rate of 0.6%. This category showed good prediction ability (C statistic = 0.74) and could be used to discriminate SCD risk and competing heart failure risk. Conclusion Myocardial tissue characteristics derived from cardiac MRI were independent predictors of sudden cardiac death (SCD)-related events in individuals with nonischemic dilated cardiomyopathy and could be used to stratify participants according to different SCD risk categories. Clinical trial registration no. ChiCTR1800017058 © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Sakuma in this issue.

Development and validation of a risk prediction score for the severity of acute hypertriglyceridemic pancreatitis in Chinese patients.

BACKGROUNDThe frequency of acute hypertriglyceridemic pancreatitis (AHTGP) is increasing worldwide. AHTGP may be associated with a more severe clinical course and greater mortality than pancreatitis caused by other causes. Early identification of patients with severe inclination is essential for clinical decision-making and improving prognosis. Therefore, we first developed and validated a risk prediction score for the severity of AHTGP in Chinese patients.AIMTo develop and validate a risk prediction score for the severity of AHTGP in Chinese patients.METHODSWe performed a retrospective study including 243 patients with AHTGP. Patients were randomly divided into a development cohort (n = 170) and a validation cohort (n = 73). Least absolute shrinkage and selection operator and logistic regression were used to screen 42 potential predictive variables to construct a risk score for the severity of AHTGP. We evaluated the performance of the nomogram and compared it with existing scoring systems. Last, we used the best cutoff value (88.16) for severe acute pancreatitis (SAP) to determine the risk stratification classification.RESULTSAge, the reduction in apolipoprotein A1 and the presence of pleural effusion were independent risk factors for SAP and were used to construct the nomogram (risk prediction score referred to as AAP). The concordance index of the nomogram in the development and validation groups was 0.930 and 0.928, respectively. Calibration plots demonstrate excellent agreement between the predicted and actual probabilities in SAP patients. The area under the curve of the nomogram (0.929) was better than those of the Bedside Index of Severity in AP (BISAP), Ranson, Acute Physiology and Chronic Health Evaluation (APACHE II), modified computed tomography severity index (MCTSI), and early achievable severity index scores (0.852, 0.825, 0.807, 0.831 and 0.807, respectively). In comparison with these scores, the integrated discrimination improvement and decision curve analysis showed improved accuracy in predicting SAP and better net benefits for clinical decisions. Receiver operating characteristic curve analysis was used to determine risk stratification classification for AHTGP by dividing patients into high-risk and low-risk groups according to the best cutoff value (88.16). The high-risk group (> 88.16) was closely related to the appearance of local and systemic complications, Ranson score ≥ 3, BISAP score ≥ 3, MCTSI score ≥ 4, APACHE II score ≥ 8, C-reactive protein level ≥ 190, and length of hospital stay.CONCLUSIONThe nomogram could help identify AHTGP patients who are likely to develop SAP at an early stage, which is of great value in guiding clinical decisions.

The Ovarian/Adnexal Reporting and Data System for Ultrasound: From Standardized Terminology to Optimal Risk Assessment and Management.

The American College of Radiology (ACR) Ovarian-Adnexal Reporting and Data System (O-RADS) lexicon and risk assessment tool for ultrasound (US) provides a framework for characterization of ovarian and adnexal pathology with the ultimate goal of harmonizing reporting and patient management strategies. Since the first O-RADS US publication in 2018, multiple validation studies have shown O-RADS US to have excellent diagnostic accuracy, with the majority of these studies using O-RADS 4 as the optimal cut-off for detecting ovarian cancer. Most of the existing validation studies include a dedicated training phase and confirm that ORADS US categories and lexicon descriptors are associated with high level inter-read agreement, regardless of radiologist training level or practice experience. O-RADS US has a similar inter-reader agreement when compared to Gynecologic Imaging Reporting and Data System (GIRADS), Assessment of Different Neoplasias in the adnexa (ADNEX), and International Tumor Analysis Group (IOTA) simple rules. System descriptors have been shown to correlate with expected malignancy rates and the O-RADS US risk stratification system has been shown to perform in the expected range of malignancy risk per category. Further directions will focus on clarifying governing concepts and lexicon terminology as well as further refining risk stratification categories based on data from published validation studies.

Risk Stratification in a Tertiary Care Spine Centre: Comparison Between STarTBack and OSPRO-YF Screening Tools.

STarT Back Screening Tool and OSPRO-YF scales have been reported to be accurate tools for estimating risk for the development of persistent pain or prolonged disability in primary care settings. We performed a comparison of construct convergent and known-group validity and ceiling floor effect (CFE) of these tools using a common sample of patients seen at a tertiary care spine centre. This was a cross-sectional study of patients with and without a work-related back injury. The Hospital Anxiety and Depression Scale (HADS) was used as the reference outcome measure for convergent validity. For known-group validity, we examined the ability of the scales to differentiate between different levels of compensation, presence of non-organic signs, and work status. The CFE values were calculated. Fifty consecutive injured workers were included along with 50 patients without an active compensation claim related to their low back pain. STarTBack and OSPRO-YF had moderate to high associations with the depression component of the HADS (0.69 to 0.77 respectively) with a statistically significant difference in favour of the OSPRO-YF. STarTBack's risk stratification categories were able to differentiate patients with a compensable injury, non-organic signs, and inability to work (p values ranging from 0.002 to < 0.001). The physical activity and work fear-avoidance beliefs constructs of the OSPRO-YF consistently outperformed other yellow flag constructs (p values ranging from 0.008 to < 0.001). The psychological sub-score of STarTBack showed a ceiling effect. There was a floor effect for the negative affect domain of OSPRO-YF. Neither total score had a floor or ceiling effect. STarTBack and OSPRO-YF are short screening tools with acceptable convergent and known-group construct validity and no floor or ceiling effect of their total score. Both tools could assist with the identification, evaluation, and management of psychological distress in patients presenting to tertiary care spine centres.

Abstract P4-06-07: Clinical risk-assessment, risk-stratification, and outcomes of ER positive, HER2 negative breast cancer patients using the Rochester modified Magee algorithm (RoMMA)

Abstract INTRODUCTION: In 2013, Klein and Dabbs et al. published three linear regression equations (the new Magee equations) using different combinations of standard histopathologic variables. In 2015, our group published a modification of the new Magee equations. Based on this modification, we published an algorithmic approach using an average Modified Magee equation. This algorithmic approach supported reflex Oncotype DX (ODX) testing based on several cutoff points using an average Modified Magee score. In 2019, we validated this algorithmic approach as the Rochester Modified Magee Algorithm (RoMMA) in a multi-institutional study, with outcome data in 247 patients, suggesting that ER positive breast cancer patients with an average Modified Magee score of ≤ 12 had a low risk of breast cancer recurrence. There has been limited published outcome data on the Magee equation since that 2019 study. We have further refined our risk-stratification approach, with additional outcome data in 416 ER positive breast cancer patients. METHODS: 416 patients with an ODX recurrence score who had at least five years of follow-up data or a breast cancer recurrence were included in the final outcome analysis (2008-2017). All patients received either Tamoxifen or an aromatase inhibitor. None of the patients received adjuvant systemic chemotherapy. The average Modified Magee score was calculated and patients were stratified into four risk-stratification categories: 1) very low, 2) low ≤ 50 years of age, 3) low &amp;gt; 50 years of age, and 4) high. We compared these four risk-stratification categories, with outcomes, between the average Modified Magee score and the ODX recurrence score. A p-value of &amp;lt; 0.05 was considered statistically significant. RESULTS: 27/416 (6.5%) patients had a recurrence of breast cancer. When comparing the same risk category groups, there was no significant difference between the average Modified Magee score and the ODX recurrence score (Table 1). CONCLUSION: Our study further reinforces that breast cancer patients can be confidently stratified into low and high risk recurrence groups using the average Modified Magee score. The average Modified Magee score may be an alternative to ODX for clinical risk-assessment and risk-stratification, particularly in lower risk patients, offering breast cancer patients increased access to clinical risk-assessment and risk-stratification, both domestically and internationally, with a potential significant cost savings for health care systems. A large prospective evaluation, similar to the studies done by ODX, using multi-institutional data or data from studies like the NSABP trial B-14 and the NSABP trial B-20, is necessary. Table 1.Risk-stratification categories and outcomes using the average Modified Magee score (aMMs) and the Oncotype DX recurrence score (ODXRS)RECURRENCENO RECURRENCEP-VALUEVERY LOW (N)aMMs ≤ 12 (76)1 (1.3)75 (98.7)0.65ODXRS &amp;lt; 11 (108)4 (3.7)104 (96.3)LOW ≤ 50 years of age (N)aMMs &amp;gt; 12, ≤ 18 (50)3 (6.0)47 (94.0)1.0ODXRS 11 - 15 (32)1 (3.1)31 (96.9)LOW &amp;gt;50 years of age (N)aMMs &amp;gt; 12, ≤ 18 (153)8 (5.2)145 (94.8)0.52ODXRS 11 - 25 (214)16 (7.5)198 (92.5)HIGH (N)aMMs &amp;gt; 18 (137)15 (10.9)122 (89.1)1.0ODXRS ≥ 16 - 25 (33)* and ODXRS &amp;gt; 25 (29)**6 (9.7)56 (90.3)* Patients ≤ 50 years of age with an ODXRS of ≥16 - 25. ** All patients with an ODXRS of &amp;gt; 25 Citation Format: Numbereye Numbere, Ioana Moisini, Ajay Dhakal, Kristin Skinner, Michelle Shayne, Mary Ann Gimenez Sanders, Huina Zhang, David Hicks, Bradley M Turner. Clinical risk-assessment, risk-stratification, and outcomes of ER positive, HER2 negative breast cancer patients using the Rochester modified Magee algorithm (RoMMA) [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P4-06-07.

PD-L1 Expression in Papillary Thyroid Cancer

Abstract There is limited data on prevalence of PD-L1 expression in papillary thyroid cancer (PTC). Anti- PD-L1 therapy is now being trialled for head and neck cancers including thyroid cancer. Whether presence of PD-L1 staining in PTC might result in a better response to anti-PD-L1 therapy is uncertain. We evaluated the frequency of tumoral PD-L1 expression using standard antibody (SP263) and clinicopathologic factors including mutation status - BRAF, NRAS, KRAS using OncoFocus Mass Array System in an Australian PTC cohort with low or high risk of recurrence /treatment response as per 2015 ATA guidelines. The cohort had 128 patients; 66% were females; 40% had both multifocal and bilateral PTC; 47% had low recurrence risk, of whom 92% had excellent treatment response; 34% had high recurrence risk, of whom 37% had structural incomplete treatment response. There was nil (0%) PD-L1 staining in 59%, 1% staining in 20%, 5-10% staining in 11%, &amp;gt;10% staining in rest of the cohort; 30%, 40% and 70% staining was seen in 2% each. PD-L1 in immune cells was positive in 62%. There was no significant association between PD-L1 tumour status and age, gender, size of tumour, multifocality, bilaterality, both multifocality and bilaterality, lymphovascular invasion, perineural invasion, extrathyroidal extension (ETE), histological variants, risk stratification, treatment response, metastases or mutation status. Among PD-L1 negative vs positive patients, BRAFV600E was present in 64.9% vs 67.5%, NRAS in 8.8% vs 0% and nil mutation in 26 % vs 33 % p=0.15. The staining thresholds were chosen according to existing literature. When analysed according to the percentage of PD-L1 staining (0%, 1-10% and &amp;gt;=10% thresholds), there was no significant association between PD-L1 staining category and other features, except for ETE (35% vs 65% vs 27%; p=0.007), a relationship that remained significant in a multivariable-adjusted model (OR 9.9, p=0.04). PD-L1 staining thresholds of &amp;lt;5% and &amp;gt;=5% were significantly associated with the size of tumour (24.7±15.4 vs 33.3±21.8; p=0.02) and histological variants (57 % vs 61% for classic variants and 27% vs 11% for follicular variants, p=0.03). This is the first study to our knowledge that analysed PD-L1 staining based on the risk stratification categories. The majority of PTC showed negative or low staining intensity for PD-L1 expression. There was no significant association between PD-L1 staining, clinicopathologic characteristics and mutation status of the PTC. Relatively less ETE was seen in the group that showed &amp;gt;=10% and also &amp;gt;=5% PD-L1 staining as compared to the group that showed 1% &amp;lt;10% and 1% to &amp;lt;5% staining which is an interesting observation and needs further investigation. The significant variation in reported PD-L1 staining in current literature reiterates that the method is yet to be optimised. Moreover the relationship between PD-L1 staining and PTC remains to be further investigated.

Deep learning diagnostic and risk-stratification pattern detection for COVID-19 in digital lung auscultations: clinical protocol for a case\u2013control and prospective cohort study

BackgroundLung auscultation is fundamental to the clinical diagnosis of respiratory disease. However, auscultation is a subjective practice and interpretations vary widely between users. The digitization of auscultation acquisition and interpretation is a particularly promising strategy for diagnosing and monitoring infectious diseases such as Coronavirus-19 disease (COVID-19) where automated analyses could help decentralise care and better inform decision-making in telemedicine. This protocol describes the standardised collection of lung auscultations in COVID-19 triage sites and a deep learning approach to diagnostic and prognostic modelling for future incorporation into an intelligent autonomous stethoscope benchmarked against human expert interpretation.MethodsA total of 1000 consecutive, patients aged ≥ 16 years and meeting COVID-19 testing criteria will be recruited at screening sites and amongst inpatients of the internal medicine department at the Geneva University Hospitals, starting from October 2020. COVID-19 is diagnosed by RT-PCR on a nasopharyngeal swab and COVID-positive patients are followed up until outcome (i.e., discharge, hospitalisation, intubation and/or death). At inclusion, demographic and clinical data are collected, such as age, sex, medical history, and signs and symptoms of the current episode. Additionally, lung auscultation will be recorded with a digital stethoscope at 6 thoracic sites in each patient. A deep learning algorithm (DeepBreath) using a Convolutional Neural Network (CNN) and Support Vector Machine classifier will be trained on these audio recordings to derive an automated prediction of diagnostic (COVID positive vs negative) and risk stratification categories (mild to severe). The performance of this model will be compared to a human prediction baseline on a random subset of lung sounds, where blinded physicians are asked to classify the audios into the same categories.DiscussionThis approach has broad potential to standardise the evaluation of lung auscultation in COVID-19 at various levels of healthcare, especially in the context of decentralised triage and monitoring.Trial registration: PB_2016-00500, SwissEthics. Registered on 6 April 2020.

Evaluation of the American Association of Cardiovascular and Pulmonary Rehabilitation Exercise Risk Stratification Classification Tool Without Exercise Testing.

The American Association of Cardiovascular and Pulmonary Rehabilitation (AACVPR) recommends that patients starting cardiac rehabilitation (CR) undergo stratification to identify risk for exercise-related adverse events (AE), but this tool has not been recently evaluated. Among patients who enrolled in CR in 2016, we used the AACVPR risk stratification tool to evaluate the risk for AE and clinical events (CE). We defined AE as signs or symptoms that precluded or interrupted exercise during CR, and CE as events requiring an urgent evaluation outside of CR exercise sessions. During the study period, 657 patients with cardiovascular diagnoses were included and classified as high (58%), medium (31%), or low risk (11%). Over the course of CR (76 d, 17 sessions), there were 63 AE and 33 CE. Adverse events were mostly minor (no cardiac arrests or deaths) and managed by CR staff members. When compared with the low- or medium-risk groups, the high-risk group was more likely to have AE (HR 3.0 [95% CI, 1.7-5.9], P = .002) and CE (HR 3.7 [95% CI, 1.5-10.8], P = .002) with fair model discrimination (area under the curve: 0.637, P < .001). The AACVPR risk stratification tool was predictive of both AE and CE with fair discrimination, although event rates were low and mostly minor. Thus, the AACVPR model may require reevaluation to better identify truly at-risk patients for major AE.

Fully Automatic Coronary Calcium Score Software Empowered by Artificial Intelligence Technology: Validation Study Using Three CT Cohorts.

ObjectiveThis study aimed to validate a deep learning-based fully automatic calcium scoring (coronary artery calcium [CAC]_auto) system using previously published cardiac computed tomography (CT) cohort data with the manually segmented coronary calcium scoring (CAC_hand) system as the reference standard.Materials and MethodsWe developed the CAC_auto system using 100 co-registered, non-enhanced and contrast-enhanced CT scans. For the validation of the CAC_auto system, three previously published CT cohorts (n = 2985) were chosen to represent different clinical scenarios (i.e., 2647 asymptomatic, 220 symptomatic, 118 valve disease) and four CT models. The performance of the CAC_auto system in detecting coronary calcium was determined. The reliability of the system in measuring the Agatston score as compared with CAC_hand was also evaluated per vessel and per patient using intraclass correlation coefficients (ICCs) and Bland-Altman analysis. The agreement between CAC_auto and CAC_hand based on the cardiovascular risk stratification categories (Agatston score: 0, 1–10, 11–100, 101–400, > 400) was evaluated.ResultsIn 2985 patients, 6218 coronary calcium lesions were identified using CAC_hand. The per-lesion sensitivity and false-positive rate of the CAC_auto system in detecting coronary calcium were 93.3% (5800 of 6218) and 0.11 false-positive lesions per patient, respectively. The CAC_auto system, in measuring the Agatston score, yielded ICCs of 0.99 for all the vessels (left main 0.91, left anterior descending 0.99, left circumflex 0.96, right coronary 0.99). The limits of agreement between CAC_auto and CAC_hand were 1.6 ± 52.2. The linearly weighted kappa value for the Agatston score categorization was 0.94. The main causes of false-positive results were image noise (29.1%, 97/333 lesions), aortic wall calcification (25.5%, 85/333 lesions), and pericardial calcification (24.3%, 81/333 lesions).ConclusionThe atlas-based CAC_auto empowered by deep learning provided accurate calcium score measurement as compared with manual method and risk category classification, which could potentially streamline CAC imaging workflows.

Remote monitoring using donor-derived, cell-free DNA after kidney transplantation during the coronavirus disease 2019 pandemic.

BackgroundDonor-derived, cell-free DNA (dd-cfDNA) level correlates with allograft injury with clinical validity and utility for quiescence and active acute rejection (AR) in kidney transplant recipients. We analyzed trends in dd-cfDNA level immediately preceding and during the coronavirus disease 2019 (COVID-19) pandemic with implemented “shelter in place” and a tele-health strategy with remote home phlebotomy to limit COVID-19 exposure.MethodsDuring COVID-19 in the United States (US), we surveyed weekly (January 6, 2020-May 25, 2020) metrics for dd-cfDNA corresponding to both a low risk for active rejection (dd-cfDNA < 0.5%) and cohorts with indeterminate levels of 0.5% to 1.0% and > 1.0%. During the study timeframe, over 11,000 patient samples (67%) from 150 kidney transplantation centers were transitioned from standard facility-based to remote phlebotomy.ResultsThe proportion of dd-cfDNA samples, analyzed in 21 weekly aggregated cohorts by risk-stratification category, was unchanged during the COVID-19 escalation in the US. Linearized slopes for numbers of samples corresponding to indeterminate risk for AR cohorts of > 1.0% and 0.5% to 1.0% were -0.31 and -0.12, respectively; indicating that prevalence of these “at risk for AR cohorts” decreased during remote surveillance. Approximately 73% of samples corresponded to low risk of AR (dd-cfDNA < 0.5%), while an additional 15% of samples had dd-cfDNA level ≤ 1.0%.ConclusionThe combination of remote home phlebotomy including dd-cfDNA analysis and a tele-health program offer a new paradigm that may substantially improve patient compliance and assuage anxiety regarding the state of kidney allograft health during the COVID-19 pandemic. Further prospective multi-center studies with robust outcomes data are warranted.

Impact of expert review of histological diagnosis of papillary and follicular thyroid cancer.

Histologic and pTNM classification of differentiated thyroid cancer (DTC) is mandatory to assess risk of relapse, risk of death, and radioactive iodine administration. The impact of an expert central review of external pathology reports has not yet been reported. Monocentric retrospective study to evaluate the difference between initial and second-opinion histopathologic diagnosis for DTC patients referred for post-operative radioactive iodine administration between January 2014 and December 2016. We evaluated major discordance (change of diagnosis from malignant to benign or in main histological subtype or a description of aggressive pathological subtypes), minor discordance (change in histological subtype or description of an aggressive component, multifocality or extrathyroidal extension), and change in ATA classification. A second-opinion histological diagnosis was available for 199 patients. A major discordance was observed in 42 (21%) cases (changes in malignancy in 4 cases, changes in main histological subtype in 22, changes in aggressive pathology variants of PTC in 16). One hundred and four minor discordances were observed regarding 92 patients. These histopathological changes led to changes in the ATA 2015 risk stratification classification in 61 (31%) of cases. There were no predictive factors of major/minor histologic changes or ATA risk stratification changes. Expert central review of pathology has an impact on the 2015 ATA risk stratification classification that can lead to changes in the management of patients with differentiated thyroid cancer.

Making a Case for Adjusting NHSN SSI Risk Stratification Classification for Use of Enhanced Electronic Infection Surveillance

Background: A large healthcare system in Georgia went live with an enhanced electronic infection surveillance system in August of 2018. The system was employed at its facilities using a staggered approach. Prior to the implementation of this infection surveillance platform, the healthcare system performed healthcare-associated infection (HAI) surveillance using an in-house culture-based system. The NHSN estimates that culture-based surveillance misses 50%–60% of true surgical site infections (SSIs). Due to the lack of clinical-based detection methods (eg, radiologic imaging), we were unable to appropriately detect all patient harm using the old surveillance system. Method: A retrospective analysis was performed to assess the change in HAI for colon (COLO), abdominal hysterectomy (HYST), hip prosthesis (HPRO), and knee prosthesis (KPRO). SSI cases that met NHSN surveillance criteria were reviewed to determine whether they would have been identified prior to launching the new enhanced electronic surveillance system. Results: Systemwide, 8 of 26 COLO SSIs (31%) and 9 of 18 HYST SSIs (50%) would have not been detected using our old surveillance system. HPRO SSIs and KPRO SSIs identified by our new surveillance system were detected using our old surveillance system, and no change was observed. Conclusion: This analysis showed an increase in COLO SSIs and HYST SSIs from enhanced surveillance. Electronic surveillance systems are not considered as a risk factor in the NHSN annual facility survey that aids in calculating a facility’s standardized infection ratio (SIR). These data help support NHSN consideration of modifying the logistic regression calculation used for the complex SSI models. This revision would allow facilities to compare themselves equitably to those using electronic infection surveillance.Funding: NoneDisclosures: None

Locally Advanced Gastrointestinal Stromal Tumor in a 33-Year-Old Woman Seeking to Conceive

Gastrointestinal stromal tumors (GISTs) are rare neoplasms of the gastrointestinal tract. They commonly present with nonspecific symptoms and thus are often discovered incidentally. They are best identified by CT scan, and most stain positive for CD117 (C-Kit), CD34, and/or DOG-1. Several risk stratification classification systems have been developed based on tumor size, mitotic rate, location, and perforation. Traditional chemotherapy and radiation therapy have been very ineffective, making surgery the mainstay of treatment. The discovery of mutations associated with these tumors has revolutionized the treatment approach. Imatinib mesylate, a selective tyrosine kinase receptor inhibitor, used as adjuvant or neoadjuvant therapy, has greatly improved the morbidity and mortality associated with GISTs. As the survival of patients has increased with the long-term use of targeted therapies, quality-of-life issues now have become much more relevant and have come to the forefront of care. We present a young woman who was successfully treated for GIST but now faces associated long-term adverse effects of imatinib, including the challenge of preserving fertility and the potential for childbearing.

Is it necessary to obtain a second review on outside pathology for patients referred with non-muscle invasive (NMI) bladder cancer?

e17049 Background: Correct pathological staging is paramount in treatment recommendations for patients with non-muscle invasive disease. As a referral center, we sought to evaluate the quality of outside pathology determined by College of American Pathology (CAP) recommendation as well as discordance in AUA risk group stratification. Methods: We retrospectively reviewed our database of patients who originally underwent resection of bladder tumors at external facilities that were subsequently referred to our cancer institute between 2015 and 2018. 75 initial pathologies were overread by one of our GU pathologists all with non-muscle invasive bladder cancer. We evaluated the discordance rate between outside pathology reports and our overread using the CAP criteria for reporting. Additionally, we included the association in risk stratification category as well as the change in risk stratification group following overread. Cohen’s kappa (κ) statistics were used to evaluate concordance in pathology report between LCI and external facilities. Comparisons of risk stratification between LCI and external facilities were analyzed using Fisher’s exact test. Results: 5 criteria for quality were evaluated to assess reporting. A relatively high agreement in reporting tumor grade between LCI and external facilities (κ = 0.65, p &lt; 0.001) and moderate agreement in microscopic extent (κ = 0.41, p &lt; 0.001). LVI was not commented on in 58.7% of outside reports. 6/12 (50%) of patients were upstaged from Low Risk (LR) to Intermediate Risk (IR), 2/11 (18%) from IR to High Risk (HR), and 6/46 (13%) from HR to MIBC (Table). Conclusions: Initial pathology reports from outside facilities were often lacking minimum criteria as recommended by the CAP. Furthermore, a significant number of patients were upstaged after review, including 13% of HR patients being overread as muscle invasive disease. Second review of outside pathology should strongly be considered as re-review may have implications on treatment recommendations. [Table: see text]

Morphophenotypic Classification of Hepatocellular Carcinoma: the Biliary/Stem Cell Subgroup and Worst Outcome—Implications on Patient Selection

BackgroundHepatocellular carcinoma (HCC) is one of the most common cancers worldwide and the third cause of cancer-related death. Current clinical/pathological criteria contribute to risk stratification, but are far from the desired on individualized medicine. Recently, HCC classifications have been published based on immunohistochemical and morphological features. MethodsA retrospective review of patients submitted to surgical treatment—partial hepatectomy (PH) or liver transplantation (LT), with pathological diagnosis of HCC, in a 9-year period (2007–2015) was performed. ResultsApplying the classification of Srivastava et al. (#1), based on the expression of CD31, p53, AFP and CD44, tumour size and presence of vascular invasion, HCC were categorized as low- and high-risk HCC. With the classification of Tsujikawa et al. (#2), HCC were classified into biliary/stem cell marker positive, Wnt signalling positive and the “all negative” HCC, according to the expression of CK19, SALL4, β-catenin glutamine synthetase, EpCAM and p53. There were sixty-six patients (53 males; 13 females), with median age of 64.5±9.46 years (range 38–86), with solitary HCC, comprehending 37 PH (56.1%) and 29 LT (43.9%). The mean overall survival (OS) was 75.4±6.9 months. Biliary/stem cell type of HCC was a predictive factor of worse OS on the overall population (24.4 versus 78.3 months, p = 0.032) and in PH cohort (11.5 versus 64.01 months, p = 0.016), on uni- and multivariate analyses. ConclusionThese results support the relevance of a risk stratification classification of HCC. Classification #2 seems adequate to our reality demonstrating OS impact, allowing its application in future biopsies, prompting individualized medicine.