Median Positive Predictive Value Research Articles

Noninvasive Tests for Bladder Cancer Detection and Surveillance: A Systematic Review of Commercially Available Assays

BACKGROUND: An important reason for the high health care costs associated with bladder cancer is the need for frequent cystoscopy for detection and surveillance of this disease. Cytologic analysis of voided urine specimens can assist, but is too inaccurate to replace cystoscopy. In an effort to create reliable, objective, noninvasive mechanisms for detecting bladder cancer, a number of urine-based molecular tests have been developed with the ultimate goal of reducing the frequency of cystoscopy. OBJECTIVE: To summarize the performance of urine-based biomarker tests, currently commercially available in the US, as part of the initial workup for hematuria and for bladder cancer surveillance. METHODS: In accordance with PRISMA guidelines we performed a systematic review of the literature on the performance of NMP22, BTA, UroVysion, ImmunoCyt/uCyt, CxBladder, and Bladder EpiCheck. Median sensitivity, specificity, negative (NPV) and positive predictive values (PPV) were calculated for each test based on the included studies. RESULTS: Twenty-eight studies met inclusion criteria for the performance of five urine-based biomarker tests in the setting hematuria workup. Median sensitivity ranged from 65.7% –100% and specificity ranged from 62.5% –93.8%. Median NPV ranged from 94.2% –98.3% and PPV ranged from 29% –58.7%. Fourteen studies met inclusion criteria for the performance of six tests in the setting of bladder cancer surveillance. Median sensitivity ranged from 22.6% –92.0% and specificity from 20.5% –97.9%. Median NPV ranged from 52.9% –96.5% and PPV ranged from 48.1% –75.7%. CONCLUSIONS: Our analysis finds that while these tests may provide some clinical utility, none of the assays have thus far demonstrated objective evidence to supplant the gold diagnostic standard.

Validity of Routine Health Data To Identify Safety Outcomes of Interest For Covid-19 Vaccines and Therapeutics in the Context of the Emerging Pandemic: A Comprehensive Literature Review.

Regulatory guidance encourages transparent reporting of information on the quality and validity of electronic health record data being used to generate real-world benefit-risk evidence for vaccines and therapeutics. We aimed to provide an overview of the availability of validated diagnostic algorithms for selected safety endpoints for Coronavirus disease 2019 (COVID-19) vaccines and therapeutics in the context of the emerging pandemic prior to December 2020. We reviewed the literature up to December 2020 to identify validation studies for various safety events of interest, including myocardial infarction, arrhythmia, myocarditis, acute cardiac injury, vasculitis/vasculopathy, venous thromboembolism, stroke, respiratory distress syndrome (RDS), pneumonitis, cytokine release syndrome (CRS), multiple organ dysfunction syndrome, and renal failure. We included studies published between 2015 and 2020 that were considered high quality assessed with QUADAS and that reported positive predictive values (PPVs). Out of 43 identified studies, we found that diagnostic algorithms for cardiovascular outcomes were supported by the highest number of validation studies (n=17). Accurate algorithms are available for myocardial infarction (median PPV 80%; IQR 22%), arrhythmia (PPV range >70%), venous thromboembolism (median PPV: 73%) and ischaemic stroke (PPV range ≥85%). We found a lack of validation studies for less common respiratory and cardiac safety outcomes of interest (eg, pneumonitis and myocarditis), as well as for COVID-specific complications (CRS, RDS). There is a need for better understanding of barriers to conducting validation studies, including data governance restrictions. Regulatory guidance should promote embedding validation within real-world EHR research used for decision-making.

A machine learning model for predicting congenital heart defects from administrative data.

International Classification of Diseases (ICD) codes recorded in administrative data are often used to identify congenital heart defects (CHD). However, these codes may inaccurately identify true positive (TP) CHD individuals. CHD surveillance could be strengthened by accurate CHD identification in administrative records using machine learning (ML) algorithms. To identify features relevant to accurate CHD identification, traditional ML models were applied to a validated dataset of 779 patients; encounter level data, including ICD-9-CM and CPT codes, from 2011 to 2013 at four US sites were utilized. Five-fold cross-validation determined overlapping important features that best predicted TP CHD individuals. Median values and 95% confidence intervals (CIs) of area under the receiver operating curve, positive predictive value (PPV), negative predictive value, sensitivity, specificity, and F1-score were compared across four ML models: Logistic Regression, Gaussian Naive Bayes, Random Forest, and eXtreme Gradient Boosting (XGBoost). Baseline PPV was 76.5% from expert clinician validation of ICD-9-CM CHD-related codes. Feature selection for ML decreased 7138 features to 10 that best predicted TP CHD cases. During training and testing, XGBoost performed the best in median accuracy (F1-score) and PPV, 0.84 (95% CI: 0.76, 0.91) and 0.94 (95% CI: 0.91, 0.96), respectively. When applied to the entire dataset, XGBoost revealed a median PPV of 0.94 (95% CI: 0.94, 0.95). Applying ML algorithms improved the accuracy of identifying TP CHD cases in comparison to ICD codes alone. Use of this technique to identify CHD cases would improve generalizability of results obtained from large datasets to the CHD patient population, enhancing public health surveillance efforts.

External Validation and Retraining of DeepBleed: The First Open-Source 3D Deep Learning Network for the Segmentation of Spontaneous Intracerebral and Intraventricular Hemorrhage.

The objective of this study was to assess the performance of the first publicly available automated 3D segmentation for spontaneous intracerebral hemorrhage (ICH) based on a 3D neural network before and after retraining. We performed an independent validation of this model using a multicenter retrospective cohort. Performance metrics were evaluated using the dice score (DSC), sensitivity, and positive predictive values (PPV). We retrained the original model (OM) and assessed the performance via an external validation design. A multivariate linear regression model was used to identify independent variables associated with the model's performance. Agreements in volumetric measurements and segmentation were evaluated using Pearson's correlation coefficients (r) and intraclass correlation coefficients (ICC), respectively. With 1040 patients, the OM had a median DSC, sensitivity, and PPV of 0.84, 0.79, and 0.93, compared to thoseo f 0.83, 0.80, and 0.91 in the retrained model (RM). However, the median DSC for infratentorial ICH was relatively low and improved significantly after retraining, at p < 0.001. ICH volume and location were significantly associated with the DSC, at p < 0.05. The agreement between volumetric measurements (r > 0.90, p > 0.05) and segmentations (ICC ≥ 0.9, p < 0.001) was excellent. The model demonstrated good generalization in an external validation cohort. Location-specific variances improved significantly after retraining. External validation and retraining are important steps to consider before applying deep learning models in new clinical settings.

PD39-08 INTER-CENTER VARIABILITY IN POSITIVE PREDICTIVE VALUES OF PIRADS SCORE IN PATIENTS UNDERGOING PROSTATE BIOPSY: RESULTS FROM THE PROMOD STUDY GROUP

PD39-08 INTER-CENTER VARIABILITY IN POSITIVE PREDICTIVE VALUES OF PIRADS SCORE IN PATIENTS UNDERGOING PROSTATE BIOPSY: RESULTS FROM THE PROMOD STUDY GROUP

Evaluation of Endoscopic Response Using Deep Neural Network in Esophageal Cancer Patients Who Received Neoadjuvant Chemotherapy.

We previously reported that endoscopic response evaluation can preoperatively predict the prognosis and distribution of residual tumors after neoadjuvant chemotherapy (NAC). In this study, we developed artificial intelligence (AI)-guided endoscopic response evaluation using a deep neural network to discriminate endoscopic responders (ERs) in patients with esophageal squamous cell carcinoma (ESCC) after NAC. Surgically resectable ESCC patients who underwent esophagectomy following NAC were retrospectively analyzed in this study. Endoscopic images of the tumors were analyzed using a deep neural network. The model was validated with a test data set using 10 newly collected ERs and 10 newly collected non-ER images. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the endoscopic response evaluation by AI and endoscopists were calculated and compared. Of 193 patients, 40 (21%) were diagnosed as ERs. The median sensitivity, specificity, PPV, and NPV values for ER detection in 10 models were 60%, 100%, 100%, and 71%, respectively. Similarly, the median values by the endoscopist were 80%, 80%, 81%, and 81%, respectively. This proof-of-concept study using a deep learning algorithm demonstrated that the constructed AI-guided endoscopic response evaluation after NAC could identify ER with high specificity and PPV. It would appropriately guide an individualized treatment strategy that includes an organ preservation approach in ESCC patients.

Development and retrospective evaluation of a clinical decision support system for the efficient detection of drug-related problems by clinical pharmacists

BackgroundClinical decision support systems (CDSS) can help identify drug-related problems (DRPs). However, the alert specificity remains variable. Defining more relevant alerts for detecting DRPs would improve CDSS.AimDevelop electronic queries that assist pharmacists in conducting medication reviews and an assessment of the performance of this model to detect DRPs.MethodElectronic queries were set up in CDSS using “triggers” from electronic health records: drug prescriptions, laboratory values, medical problems, vital signs, demographics. They were based on a previous study where 315 patients admitted in internal medicine benefited from a multidisciplinary medication review (gold-standard) to highlight potential DRPs. Electronic queries were retrospectively tested to assess performance in detecting DRPs revealed with gold-standard. For each electronic query, sensitivity, specificity, positive and negative predictive value were computed.ResultsOf 909 DRPs, 700 (77.8%) were used to create 366 electronic queries. Electronic queries correctly detected 77.1% of DRPs, median sensitivity and specificity reached 100.0% (IQRs, 100.0%–100.0%) and 99.7% (IQRs, 97.0%–100.0%); median positive predictive value and negative predictive value reached 50.0% (IQRs, 12.5%–100.0%) and 100.0% (IQRs, 100.0%–100.0%). Performances varied according to “triggers” (p < 0.001, best performance in terms of predictive positive value when exclusively involving drug prescriptions).ConclusionElectronic queries based on electronic heath records had high sensitivity and negative predictive value and acceptable specificity and positive predictive value and may contribute to facilitate medication review. Implementing some of these electronic queries (the most effective and clinically relevant) in current practice will allow a better assessment of their impact on the efficiency of the clinical pharmacist.

Agreement between estimated computed tomography perfusion ischemic core and follow-up infarct on diffusion-weighted imaging

BackgroundComputed tomography perfusion (CTP) is frequently performed during the diagnostic workup of acute ischemic stroke patients. Yet, ischemic core estimates vary widely between different commercially available software packages. We assessed the volumetric and spatial agreement of the ischemic core on CTP with the follow-up infarct on diffusion-weighted imaging (DWI) using an automated software.MethodsWe included successfully reperfused patients who underwent endovascular treatment (EVT) with CTP and follow-up DWI between November 2017 and September 2020. CTP data were processed with a fully automated software using relative cerebral blood flow (rCBF) < 30% to estimate the ischemic core. The follow-up infarct was segmented on DWI imaging data, which were acquired at approximately 24 h. Ischemic core on CTP was compared with the follow-up infarct lesion on DWI using intraclass correlation coefficient (ICC) and Dice similarity coefficient (Dice).ResultsIn 59 patients, the median estimated core volume on CTP was 16 (IQR 8–47) mL. The follow-up infarct volume on DWI was 11 (IQR 6–42) mL. ICC was 0.60 (95% CI 0.33–0.76), indicating moderate volumetric agreement. Median Dice was 0.20 (IQR 0.01–0.35). The median positive predictive value was 0.24 (IQR 0.05–0.57), and the median sensitivity was 0.3 (IQR 0.13–0.47). Severe core overestimation on computed tomography perfusion > 50 mL occurred in 4/59 (7%) of the cases.ConclusionsIn patients with successful reperfusion after EVT, CTP-estimated ischemic core showed moderate volumetric and spatial agreement with the follow-up infarct lesion on DWI, similar to the most used commercially available CTP software packages. Severe ischemic core overestimation was relatively uncommon.

Abstract 164: Efficacy Of Thoracic Impedance For Ventilation Detection During Continuous Chest Compressions

Introduction: Current technologies to identify ventilations during out-of-hospital cardiac arrest (OHCA) are limited. Thoracic impedance (TI) recorded by defibrillator pads can be used to identify ventilations, but these are often obscured by artifacts due to chest compressions (CC) and electrode motion. We tested a novel automatic algorithm to detect ventilations in TI during continuous CC. Methods: We analyzed Philips MRx CPR processed data from Dallas - Fort Worth patients in the Pragmatic Airway Resuscitation Trial. We extracted one-minute TI segments during continuous CC with concurrent capnogram recordings. We applied a novel algorithm to identify ventilation in the TI signal, which combined a) adaptive signal processing to remove CC artifacts, b) a recurrent neural network to discriminate TI fluctuations as ventilations, and c) a quality control stage to anticipate TI segments in which ventilation detection could be defective. We evaluated the performance of the algorithm in detecting ground truth capnogram ventilations in terms of sensitivity (SE), proportion of correctly identified ventilations, and Positive Predictive Value (PPV), proportion of detections corresponding to actual ventilations. Results: We extracted 2,551 one-minute TI segments from 367 OHCA patients; median of 6 (IQR 3-10) minutes per patient. Median patient-wise SE and PPV were 86.5 (IQR 71.6 - 95.1) % and 85.4 (68.3 - 94.7) %, respectively. The quality control was successful at discriminating TI segments by ventilation detection performance (Figure). For the 75% and 50% of data with highest quality estimates, SE improved to 90.6 (76.8 - 96.7) % and 95.5 (87.1 - 100.0) %, respectively, and PPV improved to 89.5 (75.0 - 96.4) % and 94.9 (87.3 - 99.8) %, respectively. Conclusions: We demonstrated the accuracy of an algorithm to detect ventilations using TI during continuous CC. Ventilation detection using TI during continuous CC is feasible.

Evaluation of three commercial ELISA tests for serological detection of maedi-visna virus using Bayesian latent class analysis

Early and accurate diagnosis is fundamental for successful surveillance and control of maedi-visna virus (MVV). MVV was detected in Norway in 2019, almost 14 years after the previous outbreak. Genetic analysis indicates persistence of the virus in the sheep population since 2005. The virus was not detected despite continuous serological surveillance. This emphasises the need for improved surveillance, which relies on an understanding of both diagnostic test performance, sampling strategy and the prevalence of the disease. This study therefore aims to evaluate three commercial ELISA tests for MVV antibodies. We conducted a retrospective study using 615 samples from six flocks diagnosed with MVV in 2019. We ran all samples with the following three tests: ID Screen® MVV/CAEV Indirect (IDvet, Grabels, France), IDEXX MVV/CAEV p28 Ab Verification Test (IDEXX Laboratories, Maine, USA) and Elitest MVV/CAEV (Hyphen Biomed, Neuville-sur-Oise, France), hereinafter referred to as ID Screen, IDEXXp28 and Elitest respectively. Without a perfect reference test, we used Bayesian latent class analysis, including conditional dependence between tests, to estimate diagnostic accuracy and true prevalence in the flocks. Using recommended cut-off values, we found that ID Screen and Elitest had significantly higher sensitivity (Se) estimates (99.3 % [97.4–100.0, 95 % Posterior Credible Interval] and 97.4 % [94.1–99.7 %], respectively) than IDEXXp28 (79.5 % [72.3–86.0 %]), while IDEXXp28 and ID Screen had significantly higher specificity (Sp) estimates than Elitest (99.7 % [99.1–100.0], 99.1 % [98.0–99.8 %] and 93.7 % [91.4–95.7 %], respectively). The estimated true prevalence in the six flocks ranged from a median of 0.8–93.5 %. Combining ID Screen and Elitest in serial interpretation showed the highest median Se and Sp (96.7 % [92.0–99.1] and 100.0 % [99.9–100.0], respectively), as well as the highest median positive predictive value (PPV) for the population with the lowest prevalence. Our study supports the use of ID Screen for screening. Further verification with Elitest in serial interpretation will enhance the PPV.

Screening tools to identify patients with unmet palliative care needs in the emergency department: A systematic review.

This systematic review identified and assessed psychometric properties of the available screening tools to identify patients with unmet palliative care (PC) needs in the emergency department (ED). A comprehensive search of electronic databases and the gray literature was conducted. Two independent reviewers completed study screening and inclusion, data extraction, and quality assessment. A descriptive summary of the results was reported using median of medians and interquartile ranges (IQRs). A total of 35 studies were included, involving the assessment of 14 unique screening tools. The most commonly used screening tool was the surprise question (SQ; n=12 studies), followed by the Palliative Care and Rapid Emergency Screening (P-CaRES) tool (n=8), and the screening for palliative and end-of-life care needs in the emergency department (SPEED) instrument (n=4). Twelve of the included studies reported on the psychometric properties of the screening tools, of which eight of these studies assessed the performance of the SQ to predict patient mortality. Overall, the median sensitivity (63%, IQR 38%-78%) and specificity (75%, IQR 57%-84%) of the SQ to predict mortality at 1 or 12 months was moderate. While the median positive predictive value of the SQ was low (32%, IQR 16%-40%), the median negative predictive value was high (91%, IQR 88%-95%). Across the studies, the proportion of patients identified as having unmet PC based on the criteria of the screening tools ranged from 5% to 83%. This review identified 14 unique screening tools used to identify adult patients with unmet PC needs in the ED. One screening tool, the SQ, was found to have moderate sensitivity and specificity to accurately predict future patient mortality. Additional research is needed to better understand the clinical value of this and the other available tools prior to their widespread implementation.

A pilot study to test the validity of a piezoelectric intra-splint force detector for monitoring of sleep bruxism in comparison to portable polysomnography

To test the validity of a force-based detection system (ISFD: intra-splint force detector) to record sleep bruxism (SB) in comparison to portable polysomnography (PSG). Simultaneous portable PSG recordings with a masseter electromyography (EMG) channel and ISFD with a deformation-sensitive piezoelectric film were performed on six participants with definite SB. First, simulated bruxism behaviors (static clenching, grinding, tapping, and rhythmic clenching) were recorded using both EMG and ISFD. Using these data, interval and duration criteria for ISFD data conditioning were established. Then, portable PSG recordings were conducted with the ISFD during sleep. Using the above criteria, ISFD events were compared with EMG-based SB episodes (the gold standard), and the sensitivity and positive predictive value of ISFD events were calculated. Spearman's correlation coefficients between true-positive ISFD events and SB episodes were then calculated. Among the tested conditioning criteria, a 3-s interval combined with a 1-s duration was selected. The median sensitivity and positive predictive value for the ISFD were 0.861 and 0.585, respectively. The duration of true-positive ISFD events was correlated with that of EMG-based SB episodes (rho = 0.658, P < 0.01). ISFD has validity for SB detection and could be an alternative to single-channel EMG-based recordings.

"Pink Pattern" Visualized in Magnifying Endoscopy With Narrow-Band Imaging Is a Novel Feature of Early Differentiated Gastric Cancer: A Bridge Between Endoscopic Images and Histopathological Changes.

Background: A pink color change occasionally found by us under magnifying endoscopy with narrow-band imaging (ME-NBI) may be a special feature of early gastric cancer (EGC), and was designated the “pink pattern”. The purposes of this study were to determine the relationship between the pink pattern and the cytopathological changes in gastric cancer cells and whether the pink pattern is useful for the diagnosis of EGC.Methods: The color features of ME-NBI images and pathological images of cancerous gastric mucosal surfaces were extracted and quantified. The cosine similarity was calculated to evaluate the correlation between the pink pattern and the nucleus-to-cytoplasm ratio of cancerous epithelial cells. Two diagnostic tests were performed by 12 endoscopists using stored ME-NBI images of 185 gastric lesions to investigate the diagnostic efficacy of the pink pattern for EGC. The diagnostic values, such as the area under the curve (AUC), the accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), of test 1 and test 2 were compared.Results: The cosine similarity between the color values of ME-NBI images and pathological images of 20 lesions was at least 0.744. The median AUC, accuracy, sensitivity, specificity, PPV, and NPV of test 2 were significantly better than those of test 1 for all endoscopists and for the junior and experienced groups.Conclusions: The pink pattern observed in ME-NBI images correlated strongly with the change in the nucleus-to-cytoplasm ratio of gastric epithelial cells, and could be considered a useful marker for the diagnosis of differentiated EGC.

Quantifying throwing load in handball: a method for measuring the number of throws

ABSTRACT Shoulder injuries are a common problem in handball. One likely cause of such injuries is excessive throwing. However, it is difficult to measure the number of player throws in large cohort studies using existing methods accurately. Therefore, the purpose of this study is to develop and validate a method for identifying overhead throws using a low-cost inertial measurement unit (IMU) worn on the wrist. In a two-stage approach, we developed a threshold-based automatic identification method for overhead throws in a laboratory study using the IMU. Subsequently, we validated the suggested thresholds in a field setting by comparing throws identified by the threshold-method to throws identified by video recordings of handball practices. The best set of threshold values resulted in a per-player median sensitivity of 100% (range: 84–100%) and a median positive predictive value (PPV) of 96% (range: 86–100%) in the development study. In the validation study, the per-player median sensitivity dropped to 78% sensitivity (range: 52–91%), while the per-player median PPV dropped to 79% (range: 47–90%). The proposed method is a promising method for automatically identifying handball throws in a cheap and feasible way.

A Detector for Premature Atrial and Ventricular Complexes.

The relationship between premature atrial complexes (PACs) and atrial fibrillation (AF), stroke and myocardium degradation is unclear. Current PAC detectors are beat classifiers that attain low sensitivity on PAC detection. The lack of a proper PAC detector hinders the study of the implications of this event and its monitoring. In this work a PAC and ventricular detector is presented. Two PhysioNet open-source databases were used: the long-term ST database (LTSTDB) and the supraventricular arrhythmia database (SVDB). A combination of heart rate variability (HRV) and morphological features were used to classify beats. Morphological features were extracted from the ECG as well as on the 4th scale of the discrete wavelet transform (DWT). After feature selection, a random forest algorithm was trained for a binary classification of PAC (S) vs. others and for a multi-labels classification to discriminate between normal (N), S and ventricular (V) beats. The algorithm was tested in a 10-fold cross-validation following a patient-wise train-test division (i.e., no beats belonging to the same patient were included both in the test and train set). The resultant median sensitivity, specificity and positive predictive value (PPV) were 99.29, 99.54, and 100% for (N), 95.83, 99.39, and 35.68% for (S), 100, 99.90, and 79.63% for (V). The proposed method attains a greater PAC and ventricular beat sensitivity and PPV than the state-of-the-art classifiers.

Online control of an assistive active glove by slow cortical signals in patients with amyotrophic lateral sclerosis

Objective. A brain–computer interface (BCI) allows users to control external devices using brain signals that can be recorded non-invasively via electroencephalography (EEG). Movement related cortical potentials (MRCPs) are an attractive option for BCI control since they arise naturally during movement execution and imagination, and therefore, do not require an extensive training. This study tested the feasibility of online detection of reaching and grasping using MRCPs for the application in patients suffering from amyotrophic lateral sclerosis (ALS). Approach. A BCI system was developed to trigger closing of a soft assistive glove by detecting a reaching movement. The custom-made software application included data collection, a novel method for collecting the input data for classifier training from the offline recordings based on a sliding window approach, and online control of the glove. Eight healthy subjects and two ALS patients were recruited to test the developed BCI system. They performed assessment blocks without the glove active (NG), in which the movement detection was indicated by a sound feedback, and blocks (G) in which the glove was controlled by the BCI system. The true positive rate (TPR) and the positive predictive value (PPV) were adopted as the outcome measures. Correlation analysis between forehead EEG detecting ocular artifacts and sensorimotor area EEG was conducted to confirm the validity of the results. Main results. The overall median TPR and PPV were >0.75 for online BCI control, in both healthy individuals and patients, with no significant difference across the blocks (NG versus G). Significance. The results demonstrate that cortical activity during reaching can be detected and used to control an external system with a limited amount of training data (30 trials). The developed BCI system can be used to provide grasping assistance to ALS patients.

Intuitive real-time control strategy for high-density myoelectric hand prosthesis using deep and transfer learning

Myoelectric hand prostheses offer a way for upper-limb amputees to recover gesture and prehensile abilities to ease rehabilitation and daily life activities. However, studies with prosthesis users found that a lack of intuitiveness and ease-of-use in the human-machine control interface are among the main driving factors in the low user acceptance of these devices. This paper proposes a highly intuitive, responsive and reliable real-time myoelectric hand prosthesis control strategy with an emphasis on the demonstration and report of real-time evaluation metrics. The presented solution leverages surface high-density electromyography (HD-EMG) and a convolutional neural network (CNN) to adapt itself to each unique user and his/her specific voluntary muscle contraction patterns. Furthermore, a transfer learning approach is presented to drastically reduce the training time and allow for easy installation and calibration processes. The CNN-based gesture recognition system was evaluated in real-time with a group of 12 able-bodied users. A real-time test for 6 classes/grip modes resulted in mean and median positive predictive values (PPV) of 93.43% and 100%, respectively. Each gesture state is instantly accessible from any other state, with no mode switching required for increased responsiveness and natural seamless control. The system is able to output a correct prediction within less than 116 ms latency. 100% PPV has been attained in many trials and is realistically achievable consistently with user practice and/or employing a thresholded majority vote inference. Using transfer learning, these results are achievable after a sensor installation, data recording and network training/fine-tuning routine taking less than 10 min to complete, a reduction of 89.4% in the setup time of the traditional, non-transfer learning approach.

Avoidable Serum Potassium Testing in the Cardiac ICU: Development and Testing of a Machine-Learning Model.

To create a machine-learning model identifying potentially avoidable blood draws for serum potassium among pediatric patients following cardiac surgery. Retrospective cohort study. Tertiary-care center. All patients admitted to the cardiac ICU at Boston Children's Hospital between January 2010 and December 2018 with a length of stay greater than or equal to 4 days and greater than or equal to two recorded serum potassium measurements. None. We collected variables related to potassium homeostasis, including serum chemistry, hourly potassium intake, diuretics, and urine output. Using established machine-learning techniques, including random forest classifiers, and hyperparameter tuning, we created models predicting whether a patient's potassium would be normal or abnormal based on the most recent potassium level, medications administered, urine output, and markers of renal function. We developed multiple models based on different age-categories and temporal proximity of the most recent potassium measurement. We assessed the predictive performance of the models using an independent test set. Of the 7,269 admissions (6,196 patients) included, serum potassium was measured on average of 1 (interquartile range, 0-1) time per day. Approximately 96% of patients received at least one dose of IV diuretic and 83% received a form of potassium supplementation. Our models predicted a normal potassium value with a median positive predictive value of 0.900. A median percentage of 2.1% measurements (mean 2.5%; interquartile range, 1.3-3.7%) was incorrectly predicted as normal when they were abnormal. A median percentage of 0.0% (interquartile range, 0.0-0.4%) critically low or high measurements was incorrectly predicted as normal. A median of 27.2% (interquartile range, 7.8-32.4%) of samples was correctly predicted to be normal and could have been potentially avoided. Machine-learning methods can be used to predict avoidable blood tests accurately for serum potassium in critically ill pediatric patients. A median of 27.2% of samples could have been saved, with decreased costs and risk of infection or anemia.

Abstract 16730: Computerized Prediction of Avoidable Serum Potassium Testing in Critically Ill Cardiac Patients

Introduction: Electrolytes are frequently monitored among critically ill patients, especially among cardiac patients due to higher risk of arrhythmias. However, a percentage of these blood draws may potentially be saved. To date, no studies have investigated strategies to safely reduce the density of electrolyte monitoring in critically ill patients. Hypothesis: We hypothesized that machine learning models can identify potentially avoidable blood draws for serum potassium (K) among pediatric patients following cardiac surgery. Methods: We retrospectively reviewed data of all patients admitted to the CICU at Boston Children’s Hospital during 2010-2018, having a length of stay ≥4 days and ≥2 recorded serum K measurements. We collected variables related to K homeostasis, including serum chemistry, hourly K intake, diuretics, and urine output. Using established machine learning techniques (Random Forest classifiers and hyperparameters) we created models predicting whether a patient’s K would be normal or abnormal based on the most recent K level, medications administered, urine output, and markers of renal function. We developed multiple models based on different age categories and temporal proximity of the most recent K measurement. We assessed the predictive performance of the models using an independent test set. Results: Of the 7,269 admissions (6,196 patients) included, 95,674 serum K was measured on average of 1 (IQR 0-1) time per day. 96% of patients received at least one dose of IV diuretic and 83% received a form of K supplementation. Our models predicted a normal K value with a median positive predictive value of 0.90. A median percentage of 2.1% measurements (mean 2.5%, IQR 1.3%-3.7%) were incorrectly predicted as normal when they were abnormal. A median percentage of 0.0% (IQR 0.0%-0.4%) were incorrectly predicted as normal while being critically low or high. A median of 27.2% (IQR 7.8%-32.4%) of samples were correctly predicted to be normal and could have been potentially avoided. Conclusions: Machine-learning methods can be used to accurately predict avoidable blood tests for serum K in critically ill pediatric patients. A median of 27.2% of samples could have been saved, with decreased costs and risk of infection or anemia.

Pre-clinical animal models are poor predictors of human toxicities in phase 1 oncology clinical trials

BackgroundOur objective was to determine the correlation between preclinical toxicity found in animal models (mouse, rat, dog and monkey) and clinical toxicity reported in patients participating in Phase 1 oncology clinical trials.MethodsWe obtained from two major early-Phase clinical trial centres, preclinical toxicities from investigational brochures and clinical toxicities from published Phase 1 trials for 108 drugs, including small molecules, biologics and conjugates. Toxicities were categorised according to Common Terminology Criteria for Adverse Events version 4.0. Human toxicities were also categorised based on their reported clinical grade (severity). Positive predictive values (PPV) and negative predictive values (NPV) were calculated to determine the probability that clinical studies would/would not show a particular toxicity category given that it was seen in preclinical toxicology analysis. Statistical analyses also included kappa statistics, and Matthews (MCC) and Spearman correlation coefficients.ResultsOverall, animal toxicity did not show strong correlation with human toxicity, with a median PPV of 0.65 and NPV of 0.50. Similar results were obtained based on kappa statistics and MCC.ConclusionsThere is an urgent need to assess more novel approaches to the type and conduct of preclinical toxicity studies in an effort to provide better predictive value for human investigation.