Predicting worsening heart failure hospitalizations in patients with implantable cardioverter defibrillators: is it all about alerts? A pooled analysis of nine trials.

AIMS: To predict worsening heart failure hospitalizations (WHFHs) in patients with implantable defibrillators and remote monitoring, the HeartInsight algorithm (Biotronik, Berlin, Germany) calculates a heart failure (HF) score combining seven physiologic parameters: 24 h heart rate (HR), nocturnal HR, HR variability, atrial tachyarrhythmia, ventricular extrasystoles, patient activity, and thoracic impedance. We compared temporal trends of the HF score and its components 12 weeks before a WHFH with 12-week trends in patients without WHFH, to assess whether trends indicate deteriorating HF regardless of alert status. METHODS AND RESULTS: Data from nine clinical trials were pooled, including 2050 patients with a defibrillator capable of atrial sensing, ejection fraction ≤ 35%, NYHA class II/III, no long-standing atrial fibrillation, and 369 WHFH from 259 patients. The mean HF score was higher in the WHFH group than in the no WHFH group (42.3 ± 26.1 vs. 30.7 ± 20.6, P < 0.001) already at the beginning of 12 weeks. The mean HF score further increased to 51.6 ± 26.8 until WHFH (+22% vs. no WHFH group, P = 0.003). As compared to the no WHFH group, the algorithm components either were already higher 12 weeks before WHFH (24 h HR, HR variability, thoracic impedance) or significantly increased until WHFH (nocturnal HR, atrial tachyarrhythmia, ventricular extrasystoles, patient activity). CONCLUSION: The HF score was significantly higher at, and further increased during 12 weeks before WHFH, as compared to the no WHFH group, with seven components showing different behaviour and contribution. Temporal trends of HF score may serve as a quantitative estimate of HF condition and evolution prior to WHFH.

Bile Is a Selective Elevator for Mucosal Mechanics and Transport.

Mucus mechanically protects the intestinal epithelium and impacts the absorption of drugs, with a largely unknown role for bile. We explored the impacts of bile on mucosal biomechanics and drug transport within mucus. Bile diffused with square-root-of-time kinetics and interplayed with mucus, leading to transient stiffening captured in Brillouin images and a concentration-dependent change from subdiffusive to Brownian-like diffusion kinetics within the mucus demonstrated by differential dynamic microscopy. Bile-interacting drugs, Fluphenazine and Perphenazine, diffused faster through mucus in the presence of bile, while Metoprolol, a drug with no bile interaction, displayed consistent diffusion. Our findings were corroborated by rat studies, where co-dosing of a bile acid sequestrant substantially reduced the bioavailability of Perphenazine but not Metoprolol. We clustered over 50 drugs based on their interactions with bile and mucin. Drugs that interacted with bile also interacted with mucin but not vice versa. This study detailed the dynamics of mucus biomechanics under bile exposure and linked the ability of a drug to interact with bile to its abbility to interact with mucus.

Predictive algorithm for the regional spread of coronavirus infection across the Russian Federation.

BACKGROUND: Outbreaks of infectious diseases are a complex phenomenon with many interacting factors. Regional health authorities need prognostic modeling of the epidemic process. METHODS: For these purposes, various mathematical algorithms can be used, which are a useful tool for studying the infections spread dynamics. Epidemiological models act as evaluation and prognosis models. The authors outlined the experience of developing a short-term predictive algorithm for the spread of the COVID-19 in the region of the Russian Federation based on the SIR model: Susceptible (vulnerable), Infected (infected), Recovered (recovered). The article describes in detail the methodology of a short-term predictive algorithm, including an assessment of the possibility of building a predictive model and the mathematical aspects of creating such forecast algorithms. RESULTS: Findings show that the predicted results (the mean square of the relative error of the number of infected and those who had recovered) were in agreement with the real-life situation: σ(I) = 0.0129 and σ(R) = 0.0058, respectively. CONCLUSIONS: The present study shows that despite a large number of sophisticated modifications, each of which finds its scope, it is advisable to use a simple SIR model to quickly predict the spread of coronavirus infection. Its lower accuracy is fully compensated by the adaptive calibration of parameters based on monitoring the current situation with updating indicators in real-time.

The Development and Validation of Simplified Machine Learning Algorithms to Predict Prognosis of Hospitalized Patients With COVID-19: Multicenter, Retrospective Study.

BACKGROUND: The current COVID-19 pandemic is unprecedented; under resource-constrained settings, predictive algorithms can help to stratify disease severity, alerting physicians of high-risk patients; however, there are only few risk scores derived from a substantially large electronic health record (EHR) data set, using simplified predictors as input. OBJECTIVE: The objectives of this study were to develop and validate simplified machine learning algorithms that predict COVID-19 adverse outcomes; to evaluate the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and calibration of the algorithms; and to derive clinically meaningful thresholds. METHODS: We performed machine learning model development and validation via a cohort study using multicenter, patient-level, longitudinal EHRs from the Optum COVID-19 database that provides anonymized, longitudinal EHR from across the United States. The models were developed based on clinical characteristics to predict 28-day in-hospital mortality, intensive care unit (ICU) admission, respiratory failure, and mechanical ventilator usages at inpatient setting. Data from patients who were admitted from February 1, 2020, to September 7, 2020, were randomly sampled into development, validation, and test data sets; data collected from September 7, 2020, to November 15, 2020, were reserved as the postdevelopment prospective test data set. RESULTS: Of the 3.7 million patients in the analysis, 585,867 patients were diagnosed or tested positive for SARS-CoV-2, and 50,703 adult patients were hospitalized with COVID-19 between February 1 and November 15, 2020. Among the study cohort (n=50,703), there were 6204 deaths, 9564 ICU admissions, 6478 mechanically ventilated or EMCO patients, and 25,169 patients developed acute respiratory distress syndrome or respiratory failure within 28 days since hospital admission. The algorithms demonstrated high accuracy (AUC 0.89, 95% CI 0.89-0.89 on the test data set [n=10,752]), consistent prediction through the second wave of the pandemic from September to November (AUC 0.85, 95% CI 0.85-0.86) on the postdevelopment prospective test data set [n=14,863], great clinical relevance, and utility. Besides, a comprehensive set of 386 input covariates from baseline or at admission were included in the analysis; the end-to-end pipeline automates feature selection and model development. The parsimonious model with only 10 input predictors produced comparably accurate predictions; these 10 predictors (age, blood urea nitrogen, SpO2, systolic and diastolic blood pressures, respiration rate, pulse, temperature, albumin, and major cognitive disorder excluding stroke) are commonly measured and concordant with recognized risk factors for COVID-19. CONCLUSIONS: The systematic approach and rigorous validation demonstrate consistent model performance to predict even beyond the period of data collection, with satisfactory discriminatory power and great clinical utility. Overall, the study offers an accurate, validated, and reliable prediction model based on only 10 clinical features as a prognostic tool to stratifying patients with COVID-19 into intermediate-, high-, and very high-risk groups. This simple predictive tool is shared with a wider health care community, to enable service as an early warning system to alert physicians of possible high-risk patients, or as a resource triaging tool to optimize health care resources.

Baseline Values of the Compensatory Reserve Index in a Healthy Pediatric Population.

The objective of this study is to describe the compensatory reserve index (CRI) baseline values in a healthy cohort of healthy pediatric patients and evaluate the existing correlation with other physiological parameters that influence compensatory hemodynamic mechanisms. CRI is a computational algorithm that has been broadly applied to non-invasively estimate hemodynamic vascular adaptations during acute blood loss. So far, there is a lack of baseline values from healthy individuals, which complicates accurately estimating the severity of the hemodynamic compromise. Additionally, the application of this technology in pediatric populations is limited to a few reports, highlighting a marked variability by age, weight, and other physiological parameters. The CRI of 205 healthy subjects from 0 to 60 years of age were prospectively evaluated from January to February 2020 at several public outpatient clinics in El Salvador; vital signs and sociodemographic data were also collected during this period. After data collection, baseline values were classified for each age group. Multiple correlation models were tested between the CRI and the other physiological parameters. CRI value varies significantly for each age group, finding for patients under 18 years old a mean value lower than 0.6, which is currently considered the lower normal limit for adults. CRI presents strong correlations with other physiological variables such as age, weight, estimated blood volume, and heart rate (R > 0.8, R2 > 0.6, p < 0.0001). There is significant variability in the CRI normal values observed in healthy patients based on age, weight, estimated blood volume, and heart rate.

An efficient strategy for predicting river dissolved oxygen concentration: application of deep recurrent neural network model.

Dissolved oxygen (DO) concentration in water is one of the key parameters for assessing river water quality. Artificial intelligence (AI) methods have previously proved to be accurate tools for DO concentration prediction. This study presents the implementation of a deep learning approach applied to a recurrent neural network (RNN) algorithm. The proposed deep recurrent neural network (DRNN) model is compared with support vector machine (SVM) and artificial neural network (ANN) models, formerly shown to be robust AI algorithms. The Fanno Creek in Oregon (USA) is selected as a case study and daily values of water temperature, specific conductance, streamflow discharge, pH, and DO concentration are used as input variables to predict DO concentration for three different lead times ("t + 1," "t + 3," and "t + 7"). Based on Pearson's correlation coefficient, several input variable combinations are formed and used for prediction. The model prediction performance is evaluated using various indices such as correlation coefficient, Nash-Sutcliffe efficiency, root mean square error, and mean absolute error. The results identify the DRNN model ([Formula: see text]) as the most accurate among the three models considered, highlighting the potential of deep learning approaches for water quality parameter prediction.

The application of a neural network to predict hypotension and vasopressor requirements non-invasively in obstetric patients having spinal anesthesia for elective cesarean section (C/S).

BACKGROUND: Neural networks are increasingly used to assess physiological processes or pathologies, as well as to predict the increased likelihood of an impending medical crisis, such as hypotension. METHOD: We compared the capabilities of a single hidden layer neural network of 12 nodes to those of a discrete-feature discrimination approach with the goal being to predict the likelihood of a given patient developing significant hypotension under spinal anesthesia when undergoing a Cesarean section (C/S). Physiological input information was derived from a non-invasive blood pressure device (Caretaker [CT]) that utilizes a finger cuff to measure blood pressure and other hemodynamic parameters via pulse contour analysis. Receiver-operator-curve/area-under-curve analyses were used to compare performance. RESULTS: The results presented here suggest that a neural network approach (Area Under Curve [AUC] = 0.89 [p < 0.001]), at least at the implementation level of a clinically relevant prediction algorithm, may be superior to a discrete feature quantification approach (AUC = 0.87 [p < 0.001]), providing implicit access to a plurality of features and combinations thereof. In addition, the expansion of the approach to include the submission of other physiological data signals, such as heart rate variability, to the network can be readily envisioned. CONCLUSION: This pilot study has demonstrated that increased coherence in Arterial Stiffness (AS) variability obtained from the pulse wave analysis of a continuous non-invasive blood pressure device appears to be an effective predictor of hypotension after spinal anesthesia in the obstetrics population undergoing C/S. This allowed us to predict specific dosing thresholds of phenylephrine required to maintain systolic blood pressure above 90 mmHg.

The Compensatory Reserve Index Responds to Acute Hemodynamic Changes in Patients with Congenital Heart Disease: A Proof of Concept Study.

Patients with congenital heart disease (CHD) who undergo cardiac procedures may become hemodynamically unstable. Predictive algorithms that utilize dense physiologic data may be useful. The compensatory reserve index (CRI) trends beat-to-beat progression from normovolemia (CRI = 1) to decompensation (CRI = 0) in hemorrhagic shock by continuously analyzing unique sets of features in the changing pulse photoplethysmogram (PPG) waveform. We sought to understand if the CRI accurately reflects changing hemodynamics during and after a cardiac procedure for patients with CHD. A transcatheter pulmonary valve replacement (TcPVR) model was used because left ventricular stroke volume decreases upon sizing balloon occlusion of the right ventricular outflow tract (RVOT) and increases after successful valve placement. A single-center, prospective cohort study was performed. The CRI was continuously measured to determine the change in CRI before and after RVOT occlusion and successful TcPVR. Twenty-six subjects were enrolled with a median age of 19 (interquartile range (IQR) 13-29) years. The mean (± standard deviation) CRI decreased from 0.66 ± 0.15 1-min before balloon inflation to 0.53 ± 0.16 (p = 0.03) 1-min after balloon deflation. The mean CRI increased from a pre-valve mean CRI of 0.63 [95% confidence interval (CI) 0.56-0.70] to 0.77 (95% CI 0.71-0.83) after successful TcPVR. In this study, the CRI accurately reflected acute hemodynamic changes associated with TcPVR. Further research is justified to determine if the CRI can be useful as an early warning tool in patients with CHD at risk for decompensation during and after cardiac procedures.

Predictors of perioperative complications in paediatric cranial vault reconstruction surgery: a multicentre observational study from the Pediatric Craniofacial Collaborative Group.

BACKGROUND: The current incidence of major complications in paediatric craniofacial surgery in North America has not been accurately defined. In this report, the Pediatric Craniofacial Collaborative Group evaluates the incidence and determines the independent predictors of major perioperative complications using a multicentre database. METHODS: The Pediatric Craniofacial Surgery Perioperative Registry was queried for subjects undergoing complex cranial vault reconstruction surgery over a 5-year period. Major perioperative complications were identified through a structured a priori consensus process. Logistic regression was applied to identify predictors of a major perioperative complication with bootstrapping to evaluate discrimination accuracy and provide internal validity of the multivariable model. RESULTS: A total of 1814 patients from 33 institutions in the US and Canada were analysed; 15% were reported to have a major perioperative complication. Multivariable predictors included ASA physical status 3 or 4 (P=0.005), craniofacial syndrome (P=0.008), antifibrinolytic administered (P=0.003), blood product transfusion >50 ml kg-1 (P<0.001), and surgery duration over 5 h (P<0.001). Bootstrapping indicated that the predictive algorithm had good internal validity and excellent discrimination and model performance. A perioperative complication was estimated to increase the hospital length of stay by an average of 3 days (P<0.001). CONCLUSIONS: The predictive algorithm can be used as a prognostic tool to risk stratify patients and thereby potentially reduce morbidity and mortality. Craniofacial teams can utilise these predictors of complications to identify high-risk patients. Based on this information, further prospective quality improvement initiatives may decrease complications, and reduce morbidity and mortality.

A Driving-Force Predictor for Estimating Pathway Activation in Patient-Specific Models of Deep Brain Stimulation.

OBJECTIVE: Detailed biophysical modeling of deep brain stimulation (DBS) provides a theoretical approach to quantify the cellular response to the applied electric field. However, the most accurate models for performing such analyses, patient-specific field-cable (FC) pathway-activation models (PAMs), are so technically demanding to implement that their use in clinical research is greatly limited. Predictive algorithms can simplify PAM calculations, but they generally fail to reproduce the output of FC models when evaluated over a wide range of clinically relevant stimulation parameters. Therefore, we set out to develop a novel driving-force (DF) predictive algorithm (DF-Howell), customized to the study of DBS, which can better match FC results. METHODS: We developed the DF-Howell algorithm and compared its predictions to FC PAM results, as well as to the DF-Peterson algorithm, which is currently the most accurate and generalizable DF-based method. Comparison of the various methods was quantified within the context of subthalamic DBS using activation thresholds of axons representing the internal capsule, hyperdirect pathway, and cerebellothalamic tract for various combinations of fiber diameters, stimulus pulse widths, and electrode configurations. RESULTS: The DF-Howell predictor estimated activation of the three axonal pathways with less than a 6.2% mean error with respect to the FC PAM for all 21 cases tested. In 15 of the 21 cases, DF-Howell outperformed DF-Peterson in estimating pathway activation, reducing mean-errors up to 22.5%. CONCLUSIONS: DF-Howell represents an accurate predictor for estimating axonal pathway activation in patient-specific DBS models, but errors still exist relative to FC PAM calculations. Nonetheless, the tractability of DF algorithms helps to reduce the technical barriers for performing accurate biophysical modeling in clinical DBS research studies.

HapLogic: A Predictive Human Leukocyte Antigen-Matching Algorithm to Enhance Rapid Identification of the Optimal Unrelated Hematopoietic Stem Cell Sources for Transplantation.

The search for a suitable human leukocyte antigen (HLA)-matched unrelated adult stem cell donor (URD) or umbilical cord blood unit (UCB) is a complex process. The National Marrow Donor Program (NMDP) developed a search algorithm known as HapLogic, which is currently provided within the NMDP Traxis application. The HapLogic algorithm has been in use since 2006 and has advanced URD/UCB HLA-matching technology. The algorithm has been shown to have high predictive accuracy, which can streamline URD/UCB selection and drive efficiencies in the search process to the benefit of the stem cell transplantation community. Here, we describe the fundamental components of the NMDP matching algorithm, output, validation, and future directions.

iPREDICT: Characterization of Asthma Triggers and Selection of Digital Technology to Predict Changes in Disease Control.

Purpose: The iPREDICT program aimed to develop an integrated digital health solution capable of continuous data streaming, predicting changes in asthma control, and enabling early intervention. Patients and Methods: As part of the iPREDICT program, asthma triggers were characterized by surveying 221 patients (aged ≥18 years) with self-reported asthma for a risk-benefit analysis of parameters predictive of changes in disease control. Seventeen healthy volunteers (age 25-65 years) tested 13 devices to measure these parameters and assessed their usability attributes. Results: Patients identified irritants such as chemicals, allergens, weather changes, and physical activity as triggers that were the most relevant to deteriorating asthma control. Device testing in healthy volunteers revealed variable data formats/units and quality issues, such as missing data and low signal-to-noise ratio. Based on user preference and data capture validity, a spirometer, vital sign monitor, and sleep monitor formed the iPREDICT integrated system for continuous data streaming to develop a personalized/predictive algorithm for asthma control. Conclusion: These findings emphasize the need to systematically compare devices based on several parameters, including usability and data quality, to develop integrated digital technology programs for asthma care.

A Study on Optimization and Evaluation of the Visualization of Complex Algorithm Results in Remote Monitoring of COPD.

BACKGROUND: Artificial intelligence (AI) can potentially increase the quality of telemonitoring in chronic obstructive pulmonary disease (COPD). However, the output from AI is often difficult for clinicians to understand due to the complexity. This challenge may be accommodated by visualizing the AI results, however it hasn't been studied how this could be done specifically, i.e., considering which visual elements to include. AIM: To investigate how complex results from a predictive algorithm for patients with COPD can be translated into easily understandable data for the clinicians. METHODS: Semi-structured interviews were conducted to explore clinicians' needs when visualizing the results of a predictive algorithm. This formed a basis for creating a prototype of an updated user interface. The user interface was evaluated using usability tests through the "Think aloud" method. RESULTS: The clinicians pointed out the need for visualization of exacerbation alerts and the development in patients' data. Furthermore, they wanted the system to provide more information about what caused exacerbation alerts. Elements such as color and icons were described as particularly useful. The usability of the prototype was primarily assessed as easily understandable and advantageous in connection to the functions of the predictive algorithm. CONCLUSION: Predictive algorithm use in telemonitoring of COPD can be optimized by clearly visualizing the algorithm's alerts, clarifying the reasons for algorithm output, and by providing a clear overview of the development in the patient's data. This can contribute to clarity when the clinicians should act and why they should act on alerts from predictive algorithms.

An Exploratory Review on the Potential of Artificial Intelligence for Early Detection of Acute Kidney Injury in Preterm Neonates.

A preterm birth is a live birth that occurs before 37 completed weeks of pregnancy. Approximately 15 million babies are born preterm annually worldwide, indicating a global preterm birth rate of about 11%. Up to 50% of premature neonates in the gestational age (GA) group of <29 weeks' gestation will develop acute kidney injury (AKI) in the neonatal period; this is associated with high mortality and morbidity. There are currently no proven treatments for established AKI, and no effective predictive tool exists. We propose that the development of advanced artificial intelligence algorithms with neural networks can assist clinicians in accurately predicting AKI. Clinicians can use pathology investigations in combination with the non-invasive monitoring of renal tissue oxygenation (rSO2) and renal fractional tissue oxygenation extraction (rFTOE) using near-infrared spectroscopy (NIRS) and the renal resistive index (RRI) to develop an effective prediction algorithm. This algorithm would potentially create a therapeutic window during which the treating clinicians can identify modifiable risk factors and implement the necessary steps to prevent the onset and reduce the duration of AKI.

Systematic review identifies the design and methodological conduct of studies on machine learning-based prediction models.

BACKGROUND AND OBJECTIVES: We sought to summarize the study design, modelling strategies, and performance measures reported in studies on clinical prediction models developed using machine learning techniques. METHODS: We search PubMed for articles published between 01/01/2018 and 31/12/2019, describing the development or the development with external validation of a multivariable prediction model using any supervised machine learning technique. No restrictions were made based on study design, data source, or predicted patient-related health outcomes. RESULTS: We included 152 studies, 58 (38.2% [95% CI 30.8-46.1]) were diagnostic and 94 (61.8% [95% CI 53.9-69.2]) prognostic studies. Most studies reported only the development of prediction models (n = 133, 87.5% [95% CI 81.3-91.8]), focused on binary outcomes (n = 131, 86.2% [95% CI 79.8-90.8), and did not report a sample size calculation (n = 125, 82.2% [95% CI 75.4-87.5]). The most common algorithms used were support vector machine (n = 86/522, 16.5% [95% CI 13.5-19.9]) and random forest (n = 73/522, 14% [95% CI 11.3-17.2]). Values for area under the Receiver Operating Characteristic curve ranged from 0.45 to 1.00. Calibration metrics were often missed (n = 494/522, 94.6% [95% CI 92.4-96.3]). CONCLUSION: Our review revealed that focus is required on handling of missing values, methods for internal validation, and reporting of calibration to improve the methodological conduct of studies on machine learning-based prediction models. SYSTEMATIC REVIEW REGISTRATION: PROSPERO, CRD42019161764.

The Onco-Nephrology Field: The Role of Personalized Chemotherapy to Prevent Kidney Damage.

In recent years, the onco-nephrology field has acquired a relevant role in internal medicine due to the growing number of cases of renal dysfunction that have been observed in cancer patients. This clinical complication can be induced by the tumor itself (for example, due to obstructive phenomena affecting the excretory tract or by neoplastic dissemination) or by chemotherapy, as it is potentially nephrotoxic. Kidney damage can manifest as acute kidney injury or represent a worsening of pre-existing chronic kidney disease. In cancer patients, physicians should try to set preventive strategies to safeguard the renal function, avoiding the concomitant use of nephrotoxic drugs, personalizing the dose of chemotherapy according to the glomerular filtration rate (GFR) and using an appropriate hydration therapy in combination with nephroprotective compounds. To prevent renal dysfunction, a new possible tool useful in the field of onco-nephrology would be the development of a personalized algorithm for the patient based on body composition parameters, gender, nutritional status, GFR and genetic polymorphisms.

An integrated precision medicine approach in major depressive disorder: a study protocol to create a new algorithm for the prediction of treatment response.

Major depressive disorder (MDD) is the most common psychiatric disease worldwide with a huge socio-economic impact. Pharmacotherapy represents the most common option among the first-line treatment choice; however, only about one third of patients respond to the first trial and about 30% are classified as treatment-resistant depression (TRD). TRD is associated with specific clinical features and genetic/gene expression signatures. To date, single sets of markers have shown limited power in response prediction. Here we describe the methodology of the PROMPT project that aims at the development of a precision medicine algorithm that would help early detection of non-responder patients, who might be more prone to later develop TRD. To address this, the project will be organized in 2 phases. Phase 1 will involve 300 patients with MDD already recruited, comprising 150 TRD and 150 responders, considered as extremes phenotypes of response. A deep clinical stratification will be performed for all patients; moreover, a genomic, transcriptomic and miRNomic profiling will be conducted. The data generated will be exploited to develop an innovative algorithm integrating clinical, omics and sex-related data, in order to predict treatment response and TRD development. In phase 2, a new naturalistic cohort of 300 MDD patients will be recruited to assess, under real-world conditions, the capability of the algorithm to correctly predict the treatment outcomes. Moreover, in this phase we will investigate shared decision making (SDM) in the context of pharmacogenetic testing and evaluate various needs and perspectives of different stakeholders toward the use of predictive tools for MDD treatment to foster active participation and patients' empowerment. This project represents a proof-of-concept study. The obtained results will provide information about the feasibility and usefulness of the proposed approach, with the perspective of designing future clinical trials in which algorithms could be tested as a predictive tool to drive decision making by clinicians, enabling a better prevention and management of MDD resistance.

Predicting misdiagnosed adult-onset type 1 diabetes using machine learning.

AIMS: It is now understood that almost half of newly diagnosed cases of type 1 diabetes are adult-onset. However, type 1 and type 2 diabetes are difficult to initially distinguish clinically in adults, potentially leading to ineffective care. In this study a machine learning model was developed to identify type 1 diabetes patients misdiagnosed as type 2 diabetes. METHODS: In this retrospective study, a machine learning model was developed to identify misdiagnosed type 1 diabetes patients from a population of patients with a prior type 2 diabetes diagnosis. Using Ambulatory Electronic Medical Records (AEMR), features capturing relevant information on age, demographics, risk factors, symptoms, treatments, procedures, vitals, or lab results were extracted from patients' medical history. RESULTS: The model identified age, BMI/weight, therapy history, and HbA1c/blood glucose values among top predictors of misdiagnosis. Model precision at low levels of recall (10 %) was 17 %, compared to <1 % incidence rate of misdiagnosis at the time of the first type 2 diabetes encounter in AEMR. CONCLUSIONS: This algorithm shows potential for being translated into screening guidelines or a clinical decision support tool embedded directly in an EMR system to reduce misdiagnosis of adult-onset type 1 diabetes and implement effective care at the outset.

Development and evaluation of a predictive algorithm for unsatisfactory response among patients with pulmonary arterial hypertension using health insurance claims data.

OBJECTIVE: This study aimed to develop and validate a predictive algorithm for unsatisfactory response to initial pulmonary arterial hypertension (PAH) therapy using health insurance claims. METHODS: Adult patients with PAH initiated on a first PAH therapy (index date) were identified from Optum's de-identified Clinformatics Data Mart Database (1/1/2010-12/31/2019). A random survival forest algorithm was developed using patient-month data and predicted the "survival function" (i.e. risk of not having unsatisfactory response) over time. For each patient-month observation, risk factors were assessed in the 12 months prior. Unsatisfactory response was defined as the first instance of (1) new PAH therapy, (2) PAH-related hospitalization or emergency room visit, (3) lung transplant or atrial septostomy, (4) PAH-related death or (5) chronic oxygen therapy initiation. To facilitate use in clinical practice, a simplified risk score was also developed based on a linear combination of the most important risk factors identified in the algorithm. RESULTS: In total, 4781 patients were included (median age = 69.0 years; 58.6% female). Over a median follow-up of 14.0 months, 3169 (66.3%) had an unsatisfactory response. The most important risk factors included in the algorithm were healthcare resource use (i.e. PAH-related outpatient visits, pulmonologist visits, cardiologist visits, all-cause hospitalizations), time since first PAH diagnosis, time since index date, Charlson Comorbidity Index, dyspnea, and age. Predictive accuracy was good for the full algorithm (C-statistic: 0.732) but was slightly lower for the simplified risk score (C-statistic: 0.668). CONCLUSION: The present claims-based algorithm performed well in predicting time to unsatisfactory response following initial PAH therapy.

Using real-world data to dynamically predict flares during tapering of biological DMARDs in rheumatoid arthritis: development, validation, and potential impact of prediction-aided decisions.

BACKGROUND: Biological disease-modifying antirheumatic drugs (bDMARDs) are effective in the treatment of rheumatoid arthritis. However, as bDMARDs may also lead to adverse events and are expensive, tapering them is of great clinical interest. Tapering according to disease activity-guided dose optimization (DGDO) does not seem to affect long term remission rates, but flares are frequent during this process. Our objective was to develop a model for the prediction of flares during bDMARD tapering using data from routine care and to evaluate its potential clinical impact. METHODS: We used a joint latent class model to repeatedly predict the probability of a flare occurring within the next 3 months. The model was developed using longitudinal data on disease activity (DAS28) and other routine care data from two clinics. Predictive accuracy was assessed in cross-validation and external validation was performed with data from the DRESS (Dose REduction Strategy of Subcutaneous tumor necrosis factor inhibitors) trial. Additionally, we simulated the reduction in number of flares and bDMARD dose when implementing the model as a decision aid during bDMARD tapering in the DRESS trial. RESULTS: Data from 279 bDMARD courses were used for model development. The final model included two latent DAS28-trajectories, bDMARD type and dose, disease duration, and seropositivity. The area under the curve of the final model was 0.76 (0.69-0.83) in cross-validation and 0.68 (0.62-0.73) in external validation. In simulation of prediction-aided decisions, the mean number of flares over 18 months decreased from 1.21 (0.99-1.43) to 0.75 (0.54-0.96). The reduction in he bDMARD dose was mostly maintained, increasing from 54 to 64% of full dose. CONCLUSIONS: We developed a dynamic flare prediction model, exclusively based on data typically available in routine care. Our results show that using this model to aid decisions during bDMARD tapering may significantly reduce the number of flares while maintaining most of the bDMARD dose reduction. TRIAL REGISTRATION: The clinical impact of the prediction model is currently under investigation in the PATIO randomized controlled trial (Dutch Trial Register number NL9798).