Clinical Prediction Score Research Articles

Diagnostic Value of Clinical Prediction Scores for Acute Appendicitis in Children Younger than 4 Years.

Acute appendicitis can be difficult to diagnose, especially in children < 4 years old. The aim of the present study was to assess the diagnostic value of Alvarado score (AS), appendicitis inflammatory response (AIR) score, and pediatric appendicitis score (PAS) in children younger than 4 years. All children younger than 4 years who underwent appendicectomy between 2005 and 2019 were included retrospectively. The diagnostic performance of the scores was analyzed using the area under the receiver-operating characteristic (ROC) curve and by calculating the diagnostic performances at optimal criterion value cutoff points. In this study, 100 children were included (58 boys and 42 girls) with a median age of 39.5 (12-47) months. Ninety children were diagnosed with pathologically proven acute appendicitis. The area under ROC curve of AS was 0.73, AIR score was 0.79, and PAS was 0.69 (p > 0.05, respectively). In children with low risk of acute appendicitis, negative predictive values were 75.0% for AS, 50.0% for AIR score, and 66.7% for PAS (p < 0.05, respectively). The positive predictive values in children with high risk of acute appendicitis were of 92.7% for AS, 92.6% for AIR score, and 93.6% for PAS (p > 0.05, respectively). AS, AIR score, and PAS plus positive ultrasonography have 0.58, 0.49, and 0.88 area under ROC curve. The three scores can be of assistance in the suspicion of acute appendicitis. PAS markedly improved combined with positive ultrasonography, but none can be used in setting the diagnosis of acute appendicitis in young children.

A scoring system derived from electronic health records to identify patients at high risk for noninvasive ventilation failure

ObjectiveTo develop and validate a clinical risk prediction score for noninvasive ventilation (NIV) failure defined as intubation after a trial of NIV in non-surgical patients.DesignRetrospective cohort study of a multihospital electronic health record database.PatientsNon-surgical adult patients receiving NIV as the first method of ventilation within two days of hospitalization.MeasurementPrimary outcome was intubation after a trial of NIV. We used a non-random split of the cohort based on year of admission for model development and validation. We included subjects admitted in years 2010–2014 to develop a risk prediction model and built a parsimonious risk scoring model using multivariable logistic regression. We validated the model in the cohort of subjects hospitalized in 2015 and 2016.Main resultsOf all the 47,749 patients started on NIV, 11.7% were intubated. Compared with NIV success, those who were intubated had worse mortality (25.2% vs. 8.9%). Strongest independent predictors for intubation were organ failure, principal diagnosis group (substance abuse/psychosis, neurological conditions, pneumonia, and sepsis), use of invasive ventilation in the prior year, low body mass index, and tachypnea. The c-statistic was 0.81, 0.80 and 0.81 respectively, in the derivation, validation and full cohorts. We constructed three risk categories of the scoring system built on the full cohort; the median and interquartile range of risk of intubation was: 2.3% [1.9%–2.8%] for low risk group; 9.3% [6.3%–13.5%] for intermediate risk category; and 35.7% [31.0%–45.8%] for high risk category.ConclusionsIn patients started on NIV, we found that in addition to factors known to be associated with intubation, neurological, substance abuse, or psychiatric diagnoses were highly predictive for intubation. The prognostic score that we have developed may provide quantitative guidance for decision-making in patients who are started on NIV.

Validation of a Community-Acquired Pneumonia Score To Improve Empiric Antibiotic Selection at an Academic Medical Center.

The 2019 American Thoracic Society and the Infectious Diseases Society of America community-acquired pneumonia (CAP) guidelines recommend that drug-resistant pathogens (DRP) be empirically covered if locally validated risk factors are present. This retrospective case-control validation study evaluated the performance of the drug resistance in pneumonia (DRIP) clinical prediction score. Two hundred seventeen adult patients with ICD-10 (https://www.who.int/classifications/classification-of-diseases) pneumonia diagnosis, positive confirmed microbiologic data, and clinical signs and symptoms were included. A DRIP score of ≥4 was used to assess model performance. Logistic regression was used to select for significant predictors and create a modified DRIP score, which was evaluated to define clinical application. The DRIP score predicted pneumonia due to a DRP with a sensitivity of 67% and specificity of 73%. The area under the receiver operating characteristic (AUROC) curve was 0.76 (95% confidence interval [CI], 0.69 to 0.82). From regression analysis, prior infection with a DRP and antibiotics in the last 60 days, yielding scores of 2 points and 1 point, respectively, remained local risk factors in predicting drug-resistant pneumonia. Sensitivity (47%) and specificity (94%) were maximized at a threshold of ≥2 in the modified DRIP model. Therefore, prior infection with a DRP remained the only clinically relevant predictor for drug-resistant pneumonia. The original DRIP score demonstrated a decreased performance in our patient population and behaved similarly to other clinical prediction models. Empiric CAP therapy without anti-methicillin-resistant Staphylococcus aureus and antipseudomonal coverage should be considered for noncritically ill patients without a drug resistant pathogen infection in the past year. Our data support the necessity of local validation to authenticate clinical risk predictors for drug-resistant pneumonia.

Application of artificial intelligence in chronic liver diseases: a systematic review and meta-analysis

BackgroundThe gold standard for the diagnosis of liver fibrosis and nonalcoholic fatty liver disease (NAFLD) is liver biopsy. Various noninvasive modalities, e.g., ultrasonography, elastography and clinical predictive scores, have been used as alternatives to liver biopsy, with limited performance. Recently, artificial intelligence (AI) models have been developed and integrated into noninvasive diagnostic tools to improve their performance.MethodsWe systematically searched for studies on AI-assisted diagnosis of liver fibrosis and NAFLD on MEDLINE, Scopus, Web of Science and Google Scholar. The pooled sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and diagnostic odds ratio (DOR) with their 95% confidence intervals (95% CIs) were calculated using a random effects model. A summary receiver operating characteristic curve and the area under the curve was generated to determine the diagnostic accuracy of the AI-assisted system. Subgroup analyses by diagnostic modalities, population and AI classifiers were performed.ResultsWe included 19 studies reporting the performances of AI-assisted ultrasonography, elastrography, computed tomography, magnetic resonance imaging and clinical parameters for the diagnosis of liver fibrosis and steatosis. For the diagnosis of liver fibrosis, the pooled sensitivity, specificity, PPV, NPV and DOR were 0.78 (0.71–0.85), 0.89 (0.81–0.94), 0.72 (0.58–0.83), 0.92 (0.88–0.94) and 31.58 (11.84–84.25), respectively, for cirrhosis; 0.86 (0.80–0.90), 0.87 (0.80–0.92), 0.85 (0.75–0.91), 0.88 (0.82–0.92) and 37.79 (16.01–89.19), respectively; for advanced fibrosis; and 0.86 (0.78–0.92), 0.81 (0.77–0.84), 0.88 (0.80–0.93), 0.77 (0.58–0.89) and 26.79 (14.47–49.62), respectively, for significant fibrosis. Subgroup analyses showed significant differences in performance for the diagnosis of fibrosis among different modalities. The pooled sensitivity, specificity, PPV, NPV and DOR were 0.97 (0.76–1.00), 0.91 (0.78–0.97), 0.95 (0.87–0.98), 0.93 (0.80–0.98) and 191.52 (38.82–944.81), respectively, for the diagnosis of liver steatosis.ConclusionsAI-assisted systems have promising potential for the diagnosis of liver fibrosis and NAFLD. Validations of their performances are warranted before implementing these AI-assisted systems in clinical practice.Trial registration: The protocol was registered with PROSPERO (CRD42020183295).

Wavelet-based neurovascular coupling can predict brain abnormalities in neonatal encephalopathy

BackgroundHypoxic-ischemic encephalopathy (HIE) is a leading cause of morbidity and mortality in neonates, but quantitative methods to predict outcomes early in their course of illness remain elusive. Real-time physiologic biomarkers of neurologic injury are needed in order to predict which neonates will benefit from therapies. Neurovascular coupling (NVC) describes the correlation of neural activity with cerebral blood flow, and the degree of impairment could predict those at risk for poor outcomes. ObjectiveTo determine if neurovascular coupling (NVC) calculated in the first 24-hours of life based on wavelet transform coherence analysis (WTC) of near-infrared spectroscopy (NIRS) and amplitude-integrated electroencephalography (aEEG) can predict abnormal brain MRI in neonatal HIE. MethodsWTC analysis was performed between dynamic oscillations of simultaneously recorded aEEG and cerebral tissue oxygen saturation (SctO2) signals for the first 24 h after birth. The squared cross-wavelet coherence, R2, of the time–frequency domain described by the WTC, is a localized correlation coefficient (ranging between 0 and 1) between these two signals in the time–frequency domain. Statistical analysis was based on Monte Carlo simulation with a 95% confidence interval to identify the time–frequency areas from the WTC scalograms. Brain MRI was performed on all neonates and classified as normal or abnormal based on an accepted classification system for HIE. Wavelet metrics of % significant SctO2-aEEG coherence was compared between the normal and abnormal MRI groups. ResultThis prospective study recruited a total of 36 neonates with HIE. A total of 10 had an abnormal brain MRI while 26 had normal MRI. The analysis showed that the SctO2-aEEG coherence between the group with normal and abnormal MRI were significantly different (p = 0.0007) in a very low-frequency (VLF) range of 0.06–0.2 mHz. Using receiver operating characteristic (ROC) curves, the use of WTC-analysis of NVC had an area under the curve (AUC) of 0.808, and with a cutoff of 10% NVC. Sensitivity was 69%, specificity was 90%, positive predictive value (PPV) was 94%, and negative predictive value (NPV) was 52% for predicting brain injury on MRI. This was superior to the clinical Total Sarnat score (TSS) where AUC was 0.442 with sensitivity 61.5%, specificity 30%, PPV 75%, and NPV 31%. ConclusionNVC is a promising neurophysiological biomarker in neonates with HIE, and in our prospective cohort was superior to the clinical Total Sarnat score for prediction of abnormal brain MRI.

Intracerebral Hemorrhage Hematoma Expansion Prediction with Deep Learning

Background: Hematoma expansion (HE) is associated with poor clinical outcomes following intracerebral hemorrhage (IcH). Existing predictors of HE from non-contrast computed tomography (NCCT) have low sensitivity and high inter-observer variability. Availability of an objective, standardized method to predict HE remains an unmet need. Methods: This multicenter observational cohort study analyzed a retrospective dataset of 2350 patients (December 2011 to June 2018, NCCT only, 86 centers), and a prospective dataset of 460 patients (March 2018 to February 2019, NCCT plus clinical data, 28 centers) with acute IcH . Prediction of HE, defined as a hematoma growth of > 6ml within 48 hours was the primary outcome of the study. A deep learning system (DLS) was developed and validated to autonomously predict HE based on the NCCT images using retrospectively collected data. A multivariate logistic regression model and a five-point score based on clinical variables and DLS prediction score were then developed using prospectively collected data. Findings: The DLS, using only NCCT, achieved HE prediction AUC of 0·783 (95% CI, 0·693-0·871) on the prospective validation dataset. Multivariate analysis of clinical variables indicated a higher risk of HE in patients with higher baseline NIHSS (OR: 4·37(1·56-12·25)), shorter onset-to-NCCT time (OR: 2·92(1·13-7·56)), and lack of antihypertensive therapy history (OR: 2·61 (1·05-6·53)) . Inclusion of these clinical variables along with DLS prediction, further improved the HE prediction accuracy, achieving an AUC of 0·812 (0·732-0·891) which was higher than all the previous methods. Finally, an easy-to-use five-point score was developed, which achieved an AUC of 0·789 (0·707-0·871). Interpretation: Deep learning systems could provide an automated, objective, multicenter generalizable, and observer bias-free solution for HE prediction from NCCT images. Inclusion of clinical factors may further improve the prediction accuracy and the system can be used to identify patients at risk of HE leading to a personalized treatment planning. Funding: Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences, Beijing Municipal Committee of Science and Technology, Ministry of Science and Technology of the People’s Republic of China, National Natural Science Foundation of China, and Beijing Nova Program. Declaration of Interest: None to declare. Ethical Approval: This study was approved by the Institutional Review Board of Beijing TianTan hospital which covers all the hospitals contributing data to this study as part of the Chinese Stroke Center Alliance

Consensus Document on Anticoagulant Management of Deep Vein Thrombosis

Deep-vein thrombosis (DVT) is encountered commonly in clinical practice. Recently, the European Society of Cardiology released guidelines on the diagnosis and management of venous thromboembolism. In this article, the authors reviewed the important aspects of DVT and advised on the clinical diagnosis and management of the same. Further, assessment of bleeding risk, risk of recurrences of DVT, and management of DVT in special situations is reviewed briefly. This will help physicians effectively diagnose and manage DVT in routine practice. Some of the essential clinical pearls include the following. Clinical risk prediction scores such as modified Wells score and Padua prediction score are effective tools used for screening DVT. Along with D-dimer assessment, venous ultrasound is helpful in confirming the DVT in “likely DVT” patients. Anticoagulation should be offered to all the patients diagnosed with DVT. Newer direct oral anticoagulants (DOACs) should be preferred. Among four commonly used DOACs, dabigatran and edoxaban use should be preceded by parenteral anticoagulants for at least 10 days without any drug overlap, whereas rivaroxaban and apixaban should be used as single-drug approach. Treatment duration should be 3–6 months and extended treatment should be based on the assessment of risk of bleeding and recurrences. Special situations such as DVT in upper extremity and in the elderly should be managed as recommended otherwise. For DVT pregnancy, low-molecular-weight heparin should be the anticoagulant of choice.

Hepatitis C Prevalence and Validation of a Clinical Prediction Score for Targeted Screening among People Living with HIV in Ghana.

WHO recommends hepatitis C (HCV) screening for all people living with HIV (PLHIV). Yet, HCV coinfection was shown to be rare in some Sub-Saharan HIV cohorts, and targeted testing was suggested more efficient for such settings. We studied HCV prevalence among Ghanaian PLHIV, and assessed the external validity of a score to guide targeted testing. This score was initially derived from a Cambodian HIV cohort, and uses as predictors: age, household member/partner with liver disease, diabetes, generalized pruritus, AST, platelets, and AST-to-platelet ratio index. We enrolled 4,023 PLHIV, most from Greater Accra and Central regions, 28.4% were male, median age was 47 years, and high-risk behavior was reported to be rare. HCV seroprevalence was 0.57%, and HCV-RNA was detectable in 0.5%. Sequencing revealed genotype 1(b) and 2(q/r) infections. The discriminatory performance of the score was suboptimal in the Ghanaian setting. The area under the curve was 0.69 (95% CI 0.59-0.79). HCV coinfection prevalence was very low in this Ghanaian PLHIV cohort with reported low-risk of onward transmission. To avoid the cost of screening all PLHIV in similar cohorts in resource-constrained settings, further research to develop better tools/scores to guide targeted HCV testing is needed.

A clinical prediction score for transient versus persistent childhood immune thrombocytopenia

Unlabelled Box AbstractBackgroundChildhood immune thrombocytopenia (ITP) is an autoimmune bleeding disorder. The prognosis (transient, persistent, or chronic ITP) remains difficult to predict. The morbidity is most pronounced in children with persistent and chronic ITP. Clinical characteristics are associated with ITP outcomes, but there are no validated multivariate prediction models. ObjectiveDevelopment and external validatation of the Childhood ITP Recovery Score to predict transient versus persistent ITP in children with newly diagnosed ITP. MethodsPatients with a diagnosis platelet count ≤ 20 × 109/L and age below 16 years were included from two prospective multicenter studies (NOPHO ITP study, N = 377 [development cohort]; TIKI trial, N = 194 [external validation]). The primary outcome was transient ITP (complete recovery with platelets ≥100 × 109/L 3 months after diagnosis) versus persistent ITP. Age, sex, mucosal bleeding, preceding infection/vaccination, insidious onset, and diagnosis platelet count were used as predictors. ResultsIn external validation, the score predicted transient versus persistent ITP at 3 months follow‐up with an area under the receiver operating characteristic curve of 0.71. In patients predicted to have a high chance of recovery, we observed 85%, 90%, and 95% recovered 3, 6, and 12 months after the diagnosis. For patients predicted to have a low chance of recovery, this was 32%, 46%, and 71%. The score also predicted cessation of bleeding symptoms and the response to intravenous immunoglobulins (IVIg). ConclusionThe Childhood ITP Recovery Score predicts prognosis and may be useful to individualize clinical management. In future research, the additional predictive value of biomarkers can be compared to this score. A risk calculator is available (http://www.itprecoveryscore.org).

1482. Local validation of the drug resistance in pneumonia clinical prediction score at a large academic medical center and a community hospital

BackgroundCommunity-acquired pneumonia (CAP) is responsible for ~1 million emergency department (ED) visits yearly and the leading cause of infection-related deaths. Given that increasing antibiotic resistance rates complicate appropriate empiric antibiotic selection, clinicians may benefit from tools to help identify patients at risk for drug-resistant pathogens (DRPs). Limitations of traditional tools, such as healthcare-associated pneumonia criteria (HCAP), have led to development of novel scoring tools such as the drug resistance in pneumonia (DRIP) score. Webb et al. showed the DRIP score was more predictive of CAP caused by DRPs than HCAP criteria. The objective of this study was to validate the DRIP score in a local population of hospitalized patients at an academic and a community medical center.MethodsPatients who presented to the ED between May 2017 and May 2019 were included in this retrospective review. Patients were included if they were ≥ 18 years diagnosed with CAP by radiographic evidence with respiratory culture positivity and susceptibility results. Exclusion criteria were: presence of non-bacterial non-respiratory pathogens, patients with cystic fibrosis, lung transplant or systemic co-infections. The primary outcome was validation of the DRIP score by comparing the sensitivity, specificity, negative and positive predictive values (NPV/PPV) to the derivation and validation study by Webb et al. Secondary outcomes were the percentage of CAP cases with DRPs and the predictability of DRP using the DRIP score versus HCAP criteria.ResultsA total of 164 patients were included; 60.4% were male with a median age of 70 years. The primary outcome shown in Table 1 demonstrated similar sensitivity, specificity, NPV, and PPV of the DRIP score to those in the study by Webb et al. Staphylococcus aureus (32.9%) and Streptococcus pneumoniae (27.4%) were the most commonly isolated pathogens and CAP due to DRPs occurred in 30.5% of patients. The DRIP score also demonstrated improved performance in predicting DRPs in CAP compared to the HCAP Criteria as shown in Table 2.Table 1. DRIP Score Validation Table 2. Predictability of the DRIP score vs. HCAP criteria ConclusionOur results further validate the DRIP score derived by Webb et al. in predicting DRPs in CAP. These results encourage a local prospective evaluation of the DRIP score as an antimicrobial stewardship tool.Disclosures All Authors: No reported disclosures

Clinical diagnostic predictive score for Meckel diverticulum

Background/Purpose: Meckel diverticulum (MD) is present in 2% of the population. Many practitioner feel the diagnosis relies upon technetium-99 m pertechnetate scintigraphy. When negative, patients undergo additional invasive procedures delaying definitive therapy. This study aims to identify features of bleeding MD and generate a risk score, which could preclude unnecessary testing and facilitate earlier operation.Methods: All patients <18-years-old who presented with hematochezia from 2005 to 2015 were identified. MD diagnosis was based on histopathology of operative tissue. Controls were patients with hematochezia undergoing colonoscopy. A points system was used generate a risk score.Results: A total of 215 patients presented with hematochezia out of which 42 patients with MD were identified. Predictive variables included infant (OR 7, 95%CI 2–29) and toddler (OR 20, 95%CI 8–50) age groups, duration <6 days (OR 18, 95%CI 8–43), presence of large blood volume (OR 16, 95% CI 7–36), hemoglobin <7 g/dL (OR 6, 95% CI 3–15) and transfusion requirement (OR 16, 95% CI 7–38). A score of 6 or higher is highly suggestive of MD.Conclusions: This scoring system identifies children with bleeding MD who may benefit from exploratory surgery without undergoing endoscopy. This novel scoring system can be applied to provide accurate clinical diagnosis, reduce unnecessary tests and allow prompt surgical management.

A validated prediction score for having two or more embryos for cryopreservation following freeze-all IVF cycles: an analysis utilizing SART CORS database.

To develop and validate a prediction score for having 2 or more embryos cryopreserved following an IVF cycle without a fresh transfer such that an embryo selection method may be applicable. We also developed a counseling tool on the probability of not having any embryos following an IVF cycle without a fresh transfer. We split the data into a development set and a validation set by region within the USA using a coin flip approach and subsequently performed a logistic regression model to identify factors most predictive of cryopreservation of 2 or more embryos in the development set. This model was validated in the validation set. Subsequently, a clinical prediction score was derived using the model coefficients and the predictive accuracy measured with the concordance (c) statistic. A total of 31,537 potential freeze-all cycles were reported to the Society for Assisted Reproductive Technology in 2014. Of these, 57.87% produced and cryopreserved two or more embryos. We identified that age, AMH, and the number of eggs retrieved were the most significant predictors of having 2 or more embryos cryopreserved with a validated c-statistic of 0.84 (95% CI: 0.83 to 0.85). A clinical prediction score was derived from the model. 28.9% of freeze-all cycles had no embryos created from the IVF cycle despite a cycle start and an egg retrieval. The number of eggs retrieved was the most significant predictor of having no embryos available for a transfer, with a c-statistic of 0.80 when modeled as the only predictor variable. We derived counseling tools with acceptable discrimination for use in clinical practice (c-statistics > 0.7). Our study further suggests that the number of eggs retrieved from an IVF cycle is most predictive of having 2 or more embryos cryopreserved and not having any embryos after an IVF cycle, suggesting that clinicians should strive to optimize oocyte yield especially in poor prognosis patients. The probability of having two more embryos cryopreserved in a freeze-all IVF cycle such that an embryo selection method is applicable can be predicted with acceptable precision prior to the IVF cycle and excellent precision following egg retrieval using the predictionscore.

Abstract 14368: Artificial Intelligence-Electrocardiography to Predict Time to Atrial Fibrillation: An Analysis of Mayo Clinic Study of Aging

Introduction: An artificial intelligence (AI) algorithm applied to electrocardiography (ECG) during normal sinus rhythm (NSR) has been shown to predict concomitant atrial fibrillation (AF). We sought to characterize the value of AI-ECG as a predictor of future AF and assess its performance compared to other clinical prediction scores in a population-based sample. Methods: We calculated the AI-ECG probability during NSR in patients who enrolled in the population-based Mayo Clinic Study of Aging with at least one ECG in NSR within two years prior and no history of AF at the time of the baseline study visit. The cumulative incidence of AF was estimated for strata defined by AI-ECG probability and CHARGE-AF score. Cox proportional hazards were fit to assess the independent prognostic value and interaction of AI-ECG probability and CHARGE-AF score. Concordance (c) statistics were calculated for AI-ECG probability, CHARGE-AF score and combined AI-ECG and CHARGE-AF score. Results: A total of 1,936 patients with a median age 75.8 (quartile range [QR] 70.4, 81.8) years, median CHARGE-AF score 14.0 (QR 13.2, 14.7) and median CHADS2VASC score 3 (QR 2, 4) were included in the analysis. The cumulative incidence of AF increased in a stepwise fashion across quartiles of AI-ECG probability and CHARGE-AF score (Figure 1). When compared in the same model, both AI-ECG probability (hazard ratio [HR] 1.76, 95% confidence interval [CI] 1.51-2.04) and CHARGE-AF score (HR 1.90, 95% CI 1.58-2.28) independently predicted AF without significant interaction (p=0.54). C statistics were 0.69 (95% CI 0.66-0.72) for AI-ECG probability, 0.69 (95% CI 0.66-0.71) for CHARGE-AF and 0.72 (95% CI 0.69-0.75) for combined AI-ECG and CHARGE-AF score. Conclusions: In the Mayo Clinic Study of Aging, both the AI-ECG probability and CHARGE-AF score independently predicted time to AF. The AI-ECG may offer a means to assess risk with a single test without requiring manual or automated clinical data abstraction.

Comparative Analysis of Clinical Prediction Scores in Acquired Thrombotic Thrombocytopenic Purpura: The Superiority of Plasmic Score

Background: Acquired thrombotic thrombocytopenic purpura (aTTP) is a rare and life-threatening hematologic disorder. The decision to start plasma exchange is challenging in clinical practice, due to aTTP rarity, and the extensive differential diagnosis with other thrombotic microangiopathies (TMA). Although the ADAMTS13 activity measurement is essential to aTTP diagnosis, a rapid quantification is unavailable in most centers, being necessary to develop clinical prediction models to evaluate severe enzymatic deficiency. The first model developed was Bentley score, followed by French score and more recently PLASMIC score. Studies comparing these 3 methods are scarce. The criteria of published scoring models are summarized in Table1 Aim: To compare and evaluate the discriminative power of the Bentley, French and PLASMIC scores in predicting severe ADAMTS13 deficiency (≤10%) in our group of patients (pts). Methods: We performed a retrospective analysis of pts with ADAMTS13 activity tested between 2008 and 2019, in a single center. Pts were divided in two groups: ADAMTS13 activity ≤10% (ADAMTS13≤10%) and &amp;gt;10% (ADAMTS13&amp;gt;10%) and then stratified according to the 3 scores. We evaluate sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) to predict low ADAMTS13 activity. Diagnostic accuracy was tested using the area under the ROC curve. A P-value of &amp;lt;0.05 was considered statistically significant. Results: ADAMTS13 activity was evaluated in 119 pts with suspected aTTP. There was a female predominance (60.5%; n=72) and the mean age was 54±20 years. Thirty-one pts had ADAMTS13≤10% and 27 (87.1%) of them had been diagnosed with aTTP, two were inherited thrombotic thrombocytopenic purpura and the other two were sepsis associated TMA. Comparing the groups ADAMTS13≤10% and ADAMTS13&amp;gt;10%, pts in the first were younger (46±18 vs 57±21 years; p=0.008), had lower platelet count (23.04±27.1 vs 47.79±37.6x109/L; p&amp;lt;0.001), and lower creatinine value (1.25±0.8 vs 2.92±4.2 mg/dL; p&amp;lt;0.017). ADAMTS13≤10% group had a higher reticulocyte count (5.99±5.0 vs 3.46±3.4%; p=0.005) and higher indirect bilirubin levels (2.59±2.1 vs 1.59±2.2 mg/dL; p=0.013). There were no statistical differences in hemoglobin, MCV, INR, D-Dimers and antinuclear antibodies (ANA's) positivity among the two groups. PLASMIC score was calculated in all pts. Due to missing data, Bentley score was only applied in 44.5% (n=53) and French score in 54.6% (n= 65) pts. Using a high-risk PLASMIC score as the predictor, the sensitivity was 80.6%, specificity 78.4%, PPV 56.8% and NPV 92.0%. The C-statistics for high PLASMIC score cut-offs (&amp;gt;5 points) was 0.80 (95% confidence interval [CI] 0.70-0.89) (fig.1). Applying high-risk Bentley score vs low-to-intermediate-risk score, the sensitivity was 44.4%, specificity 94.3%, PPV 80.0% and NPV 76.7%. The C-statistics was 0.69 (95% CI 0.53-0.86) using the high Bentley score cut-offs (fig.2). The French score had higher sensitivity (83.3%), but lower specificity (36.2%), with 33.3% of PPV and 85.0% of NPV. Using the high French score cut-offs, the C-statistics was 0.60 (95% CI 0.45-0.75) (fig.3). There were not statistically significant differences between PLASMIC and Bentley scores in the ability to predict low ADAMTS13 activity, and both performed better than French score (p&amp;lt;0.001; p=0.02, respectively). Conclusion: In our cohort, the PLASMIC score showed to be the most balanced and efficient clinical prediction score, with high sensitivity and specificity. Bentley score was also a good predictor of ADAMTS13 activity, but had lower sensitivity and less applicability; French score was comparatively inferior. Pts with low PLASMIC score are highly unlikely to have ADAMTS13 deficiency. PLASMIC score is easy to calculate and readily usable in most emergency settings while the French and Bentley scores require additional laboratory data, which is not always accessible. These results support the previously published data, showing PLASMIC score as a helpful complementary tool in clinical practice recognizing high-risk pts when ADAMTS13 activity measurement is not available. Disclosures No relevant conflicts of interest to declare.

Machine Learning for Prediction of Cancer-Associated Venous Thromboembolism

Background: Several clinical prediction scores have been designed to assess the risk of cancer-associated thrombosis (CAT). The most commonly used in current clinical practice is the Khorana score, however it is applicable only to patients prior to initiation of chemotherapy. We now apply machine learning with clinical, demographic, and genomics parameters to predict CAT events. Methods: The random survival forest (RSF) ensemble learning method was selected to illustrate a machine approach to CAT prediction. The cohort consisted of 14,223 individuals with a solid tumor malignancy and MSK IMPACT somatic genomic data collected during the years 2014 to 2016. CAT was defined as the diagnosis of lower extremity deep vein thrombosis (proximal or distal) or pulmonary embolism, incidental or symptomatic. Covariates considered for inclusion in the model consisted of tumor type, metastatic status, age, exposure to cytotoxic chemotherapy in the month before cohort entry, time elapsed since cancer diagnosis, time elapsed since tumor sampling, normalized mean blood cell counts (white cell count, hemoglobin, platelet count) in the prior 3 months, normalized mean prothrombin time (PT) and activated partial thromboplastin time (aPTT) in the prior 3 months, body mass index (BMI), and presence or absence of a somatic genetic alteration for oncogenes/tumor suppressor genes with an alteration frequency ≥ 1.5% (n = 56). The primary endpoint consisted of time to CAT episode. The C-index for models including different covariates was derived from the test holdout sample using repeated 10-fold cross-validation. The C-index, measuring the relative agreement between the RSF predicted risk and the CAT times of patients, has values between 0.5 and 1.0 with the latter indicating perfect agreement. Results: 12,040 patients were included in the final analysis. There were 855 CAT events during the observation period. The most common tumor types were lung (17%), breast (15%) and colorectal cancer (9%). Blood cell count data and coagulation parameters were missing for 8% and 51% of patients respectively. Using cross-validation, the baseline model with cancer type and metastatic status had a C-index of 0.62 (95% CI = 0.61-0.64), which increased to 0.65 (95% CI = 0.63-0.66) with the addition of chemotherapy, age, time from tissue sampling, time from cancer diagnosis and BMI. Further adding genetic data increased the C-index to 0.68 (95% CI = 0.66-0.69). Replacing genetic data in this model with cell counts and coagulation parameters resulted in a C-index of 0.69 (95% CI = 0.68-0.70). The model with all available covariates had a C-index of 0.70 (95% CI = 0.69-0.71). The cumulative incidence of CAT at 6 months for 5 categories of predicted risk using the model with all available covariates is plotted in Figure A. Scaled Ishwaran-Kogalur Variable Importance (VIMP) values, presented in Figure B, indicate that cancer type and prior chemotherapy are the two top factors for model performance. Conclusions: Machine learning is a promising approach in the search of more accurate and generalizable models for prediction of CAT. In the application described here, the use of random survival forests performed well without information about future chemotherapy administration. Additional work is needed to identify the optimal algorithm and covariates, including better delineation of which cancer genomic information should be retained. Future models will have to be validated independently before being used for patient care. Disclosures Mantha: Physicians Education Resource: Honoraria; MJH Associates: Honoraria. Bolton:GRAIL: Research Funding. Soff:Bristol-Myers Squibb, Pfizer: Honoraria; Dova Pharmaceuticals: Honoraria; Janssen Scientific Affairs: Honoraria; Amgen: Research Funding; Janssen Scientific Affairs: Research Funding; Amgen: Honoraria; Dova Pharmaceuticals: Research Funding.

Derivation and validation of a scoring system to predict after discharge risk of cardiac events in patients with acute myocardial infarction undergoing percutaneous coronary revascularization

Abstract Background In the acute coronary syndromes (ACS) setting, despite the extensive use of coronary revascularization and P2Y12 inhibitors such as prasugrel and ticagrelor, with a more pronounced inhibitory effect on platelets than clopidogrel, the rate of death and recurrent myocardial infarction (MI) at 1 year is still high. In this clinical setting the development of a risk score which takes into account patient's and procedural characteristics could represent a useful tool to identify patients at high risk for ischaemic events at 1 year who could take advantage from more aggressive secondary prevention strategies. Purpose The aim of our study was to develop a risk score to predict 1-year probability of after discharge cardiac events (recurrent MI and cardiac death) in patients with acute MI treated with percutaneous coronary intervention (PCI). Methods We prospectively enrolled all consecutive patients hospitalized for acute MI between 2003 and 2017 treated with PCI with/without stent placement at our center. We excluded patients who died in-hospital or who experienced in-hospital recurrent MI and patients undergoing surgical revascularization by coronary artery bypass graft (CABG). The patients of the final study cohort were therefore randomly assigned to either a derivation sample (60%) or a validation sample (40%). Based on the multivariate analysis we developed a point system according to the “Framingham Risk Score” method. Results The final study cohort, represented by 4922 patients, was split in a derivation sample of 2972 patients and in a validation sample of 1950 patients: in both groups the median age was around 70 years; the male prevalence was 73%; 65% of patients were dagnosed with ST-segment elevation MI. The clinical prediction score underlined as risk factors for recurrent cardiac events older age, diabetes mellitus, peripheral arterial disease, prior MI, Killip class &amp;gt;2 at presentation, higher platelet count and creatinine values, lower left ventricular ejection fraction; radial access and the use of second generation drug eluting stents resulted to be protective. This model showed a good discrimination power in both the derivation and the validation samples with an area under the curve (AUC) of 0.75 and 0.71, respectively. The calibration showed a good concordance between predicted and observed events in both the derivation and the validation samples. Same results were observed in patients with/without ST-segment elevation MI and in gender subgroups. Conclusions The present study, conducted retrospectively on a large population of patients with acute MI treated with PCI enrolled prospectively, enabled us to derivate and validate a risk score of cardiac death and recurrent MI at 1 year which took into account both clinical and procedural characteristics and which demonstrated a good predictive performance. After-discharge events by risk subgroups Funding Acknowledgement Type of funding source: None

Monogenic variants in dystonia: an exome-wide sequencing study

Dystonia is a clinically and genetically heterogeneous condition that occurs in isolation (isolated dystonia), in combination with other movement disorders (combined dystonia), or in the context of multisymptomatic phenotypes (isolated or combined dystonia with other neurological involvement). However, our understanding of its aetiology is still incomplete. We aimed to elucidate the monogenic causes for the major clinical categories of dystonia. For this exome-wide sequencing study, study participants were identified at 33 movement-disorder and neuropaediatric specialty centres in Austria, Czech Republic, France, Germany, Poland, Slovakia, and Switzerland. Each individual with dystonia was diagnosed in accordance with the dystonia consensus definition. Index cases were eligible for this study if they had no previous genetic diagnosis and no indication of an acquired cause of their illness. The second criterion was not applied to a subset of participants with a working clinical diagnosis of dystonic cerebral palsy. Genomic DNA was extracted from blood of participants and whole-exome sequenced. To find causative variants in known disorder-associated genes, all variants were filtered, and unreported variants were classified according to American College of Medical Genetics and Genomics guidelines. All considered variants were reviewed in expert round-table sessions to validate their clinical significance. Variants that survived filtering and interpretation procedures were defined as diagnostic variants. In the cases that went undiagnosed, candidate dystonia-causing genes were prioritised in a stepwise workflow. We sequenced the exomes of 764 individuals with dystonia and 346 healthy parents who were recruited between June 1, 2015, and July 31, 2019. We identified causative or probable causative variants in 135 (19%) of 728 families, involving 78 distinct monogenic disorders. We observed a larger proportion of individuals with diagnostic variants in those with dystonia (either isolated or combined) with coexisting non-movement disorder-related neurological symptoms (100 [45%] of 222; excepting cases with evidence of perinatal brain injury) than in those with combined (19 [19%] of 98) or isolated (16 [4%] of 388) dystonia. Across all categories of dystonia, 104 (65%) of the 160 detected variants affected genes which are associated with neurodevelopmental disorders. We found diagnostic variants in 11 genes not previously linked to dystonia, and propose a predictive clinical score that could guide the implementation of exome sequencing in routine diagnostics. In cases without perinatal sentinel events, genomic alterations contributed substantively to the diagnosis of dystonic cerebral palsy. In 15 families, we delineated 12 candidate genes. These include IMPDH2, encoding a key purine biosynthetic enzyme, for which robust evidence existed for its involvement in a neurodevelopmental disorder with dystonia. We identified six variants in IMPDH2, collected from four independent cohorts, that were predicted to be deleterious de-novo variants and expected to result in deregulation of purine metabolism. In this study, we have determined the role of monogenic variants across the range of dystonic disorders, providing guidance for the introduction of personalised care strategies and fostering follow-up pathophysiological explorations. Else Kröner-Fresenius-Stiftung, Technische Universität München, Helmholtz Zentrum München, Medizinische Universität Innsbruck, Charles University in Prague, Czech Ministry of Education, the Slovak Grant and Development Agency, the Slovak Research and Grant Agency.

Clinical Risk Prediction Scores in Coronavirus Disease 2019: Beware of Low Validity and Clinical Utility.

Several risk stratification tools were developed to predict disease progression in coronavirus disease 2019, with no external validation to date. We attempted to validate three previously published risk-stratification tools in a multicenter study. Primary outcome was a composite outcome of development of severe coronavirus disease 2019 disease leading to ICU admission or death censored at hospital discharge or 30 days. We collected data from 169 patients. Patients were 73 years old (59–82 yr old), 66 of 169 (39.1%) were female, 57 (33.7%) had one comorbidity, and 80 (47.3%) had two or more comorbidities. Area under the receiver operating characteristic curve (95% CI) for the COVID-GRAM score was 0.636 (0.550–0.722), for the CALL score 0.500 (0.411–0.589), and for the nomogram 0.628 (0.543–0.714).

S0131 Addition of Serum Bicarbonate and Anion Gap to Bedside Index of Severity in Acute Pancreatitis Score Do Not Improve Its Predictive Performance

INTRODUCTION: Reliable and easy to use clinical prediction score is important in optimal management of acute pancreatitis. Currently, Bedside Index of Severity in Acute Pancreatitis (BISAP) score is the most accurate and easy to use clinical prediction score for acute pancreatitis (AP). Patients with AP are often hypovolemic at the time of presentation, which is a contributing factor to AP associated multi-organ dysfunction. We hypothesize that an elevated serum bicarbonate level and anion gap, which are readily available in most clinical settings, could improve performance of the BISAP score. Here, we assessed the performance of BISAP including serum bicarbonate in prognosticating moderate and severe acute pancreatitis. METHODS: Consecutive adults with acute pancreatitis admitted to our academic and two community hospitals between Jan 2014 and Dec 2019 were included in this study. We excluded patients with pregnancy or with diseases independently affecting their bicarbonate level (i.e., diabetic ketoacidosis, chronic renal failure stage 4 and 5). In cases of recurrent pancreatitis, only the initial episode was included. The severity of pancreatitis was defined based on the revised Atlanta classification: mild, moderate and severe. Performance of BISAP with and without bicarbonate levels and the anion gap was assessed using the ROC analysis. RESULTS: During the study period, 3 hospitals together had 653 admission for acute pancreatitis; a total of 444 unique patient admission were included. Of these, 316 patients had mild, 94 moderate and 34 patients had severe AP. Clinical characteristics of the cohort by severity of pancreatitis are summarized in Table 1. Bicarbonate levels were slightly lower in patients with moderate to severe acute pancreatitis, P = 0.038). The area under the curve (AUC) for BISAP score was 0.87 (95% CI 0.81–0.93) for severe and 0.72 (0.67–0.77) for moderate and severe pancreatitis combined. Inclusion of bicarbonate levels and the anion gap into BISAP improved AUC to 0.88 (0.81–0.95) and 0.89 (0.84–0.95), respectively (Table 2). This difference was not statistically significant. CONCLUSION: Inclusion of serum bicarbonate levels and the anion gap to the BISAP score did not improve its performance significantly.Table 1.: Clinical and laboratory characteristicsTable 2.: Discrimination by predictors—Two Revised Atlanta Classification

Risk Score For Rehospitalization Among Home Health Care Patients with Heart Failure

Introduction Heart failure (HF) is one of the leading causes of rehospitalization in the United States. Rehospitalizations are a costly outcome and are challenging to reduce due to multiple factors. Some interventions for reducing readmissions utilize predictive models to forecast adverse events such as mortality and rehospitalizations. Many rehospitalization risk scores were developed using hospital-based data from electronic health records (EHR). However, no rehospitalization risk score uses home healthcare EHR data. This study evaluated the production of a rehospitalization risk score using data from the Outcome and Assessment Information Set (OASIS) among home healthcare patients with HF after hospital discharge. Aim To produce a risk score for rehospitalization using OASIS among home healthcare patients with HF after hospital discharge. Methods This study was a secondary data analysis of the OASIS C version that was collected from January 1, 2011 to August 31, 2013. The dataset included N=836 patients. A multivariable logistic regression model was fitted to the rehospitalization outcome, using interactive backwards elimination variable selection. The final model included 9 predictor variables in a simple-to-use risk score that has a possible range (0 to 41). Regression coefficients were divided by the smallest regression coefficient and rounded to the nearest integer. For each patient risk factor, the integer weightings were summed to obtain the risk score. The predictive accuracy of the risk score was measured using the area under the receiver operating characteristic curve (i.e., c-statistic). The model was then internally validated using 10-fold cross validation. Results The median age was 79.5 years old, 43% were male, and 82% were white. We identified 9 significant risk factors: multiple hospitalization in the past 12 months, frequency of pain interfering, urinary incontinence, ability to grooming, ability to dress lower body, ability to feed, prior ability with everyday activities, the level of assistance, and Charlson comorbidity index score. For example, patients with grooming utensils must be placed within reach was a significantly greater risk compared to being able to groom self-unaided. The clinical prediction score which combined the patient's risk factors achieved a c-statistic of 0.64 (95%CI: 0.61 to 0.68). In the 10-fold cross-validation, the risk score had a c-statistic of 0.62 (95%CI: 0.58 to 0.65), which is the validated prediction accuracy expected in future home healthcare patients with HF. Conclusions This study developed a risk score for rehospitalization using OASIS among home healthcare patients with HF after hospital discharge. The risk scores of specific items such as the identified risk factors in this study could be an indicator of decision-making for personalized care.