Search

Abstract Inadequate at-home management and self-awareness of heart failure (HF) exacerbations are known to be leading causes of the greater than 1 million estimated HF-related hospitalizations in the USA alone. Most current at-home HF management protocols include paper guidelines or exploratory health applications that lack rigor and validation at the level of the individual patient. We report on a novel triage methodology that uses machine learning predictions for real-time detection and assessment of exacerbations. Medical specialist opinions on statistically and clinically comprehensive, simulated patient cases were used to train and validate prediction algorithms. Model performance was assessed by comparison to physician panel consensus in a representative, out-of-sample validation set of 100 vignettes. Algorithm prediction accuracy and safety indicators surpassed all individual specialists in identifying consensus opinion on existence/severity of exacerbations and appropriate treatment response. The algorithms also scored the highest sensitivity, specificity, and PPV when assessing the need for emergency care. Lay summary Here we develop a machine-learning approach for providing real-time decision support to adults diagnosed with congestive heart failure. The algorithm achieves higher exacerbation and triage classification performance than any individual physician when compared to physician consensus opinion. Graphical abstract

TOPIC: Chest Infections TYPE: Original Investigations PURPOSE: Viral Respiratory illnesses such as Covid-19 and Influenza pose significant health challenges worldwide. There are more than 150M confirmed cases of Covid-19 with a reported 3.15M deaths (as of April, 2021). The WHO reports there to be ~ 1 billion influenza cases and 290-650K influenza-associated deaths annually. A signature feature of these illnesses is an early infection period that, if insufficiently recognized and controlled early, can lead to viral spread and avoidable morbidity/mortality. The need for personalized, remote care tools that facilitate early detection and triage of viral illness has never been greater. To address this gap, we developed an institutional software, Vironix, that uses machine-learned (ML) prediction models to enable real-time risk stratification and decision support for global organizations. METHODS: ML models were trained on clinical characteristic data from East and South Asia, Western Europe, and USA. Algorithms take an input of symptom, profile, biometric, and exposure data and return an assessment of disease severity. Covid-19 algorithms were validated on computer generated patient vigenttes and deployed in the Vironix web app among 22 participants in a small business commercial pilot for member self-screening. Members conducted daily health assessments and received personalized decision support while organization managers received work-from-home recommendations and compliant symptom monitoring without seeing member health data. For influenza, Vironix ML algorithms were tested on a dataset (with a 90/10 train test split) collected from one academic and two community emergency rooms from March 2014 to July 2017 (Hong et al.). RESULTS: ML-predictions showed 87.6% accuracy, 85.5% sensitivity, and 87.8% specificity in identifying severe Covid-19 presentations in an out-of-sample validation set of 5,000 patient cases. After 4-months pilot use, Vironix issued 14 stay-at-home and 10 healthcare escalation recommendations while maintaining 30-day and 7-day user retention of 66% and 72%, greatly exceeding common app adoption rates. ML predictions for the Influenza data set showed 67.8% accuracy, 71.7% sensitivity, and 65.4% specificity in identifying admissible or dischargeable presentations of influenza in an out-of-sample validation set of 56,000 patient cases. CONCLUSIONS: Covid-19 ML-severity assessments showed strong accuracy, sensitivity, and specificity in identifying severe clinical presentations. The deployed web-app showed high adoption with members receiving relevant decision support. Flu algorithm performance could be bolstered by inclusion of biometric features. Additional controlled trials could be conducted to establish validated markers of health improvement and early illness detection resulting from Vironix use. The overall methodology for mapping clinical characteristic data into patient scenarios for training ML classifiers of health deterioration is generalizable for a variety of potential software and hardware deployments across disease spaces. CLINICAL IMPLICATIONS: The technology detailed in this study represents a potential low cost, scalable, hardware/software agnostic, global solution for early detection and intervention on infectious respiratory illness. These solutions can be integrated into remote care and institutional wellness workflows to support public health initiatives. DISCLOSURES: No relevant relationships by Anna Berryman, source=Web Response No relevant relationships by Shreyas Iyer, source=Web Response No relevant relationships by Vinay Konda, source=Web Response Advisory Committee Member relationship with ABMRCC Please note: $1-$1000 by Chris Landon, source=Web Response, value=Consulting fee Removed 04/28/2021 by Chris Landon, source=Web Response Consultant relationship with ABM Respiratory Please note: 11/20 - date Added 04/30/2021 by Chris Landon, source=Web Response, value=Consulting fee no disclosure on file for Nicholas Mark;No relevant relationships by James Morrill, source=Web R sponse No relevant relationships by Sriram Ramanathan, source=Web Response Owner/Founder relationship with Vironix Health, Inc Please note: 05/2020 - Present Added 04/28/2021 by Sumanth Swaminathan, source=Web Response, value=Ownership interest Owner/Founder relationship with Vironix Health Please note: 04/2020-Now Added 05/10/2021 by Botros Toro, source=Web Response, value=Ownership interest Consultant relationship with Vironix Please note: 2019-present Added 04/28/2021 by Nicholas Wysham, source=Web Response, value=Ownership interest

Background: Chronic hypersensitivity pneumonitis (cHP) is a disease caused by exposure to inhaled environmental antigens. Diagnosis of cHP is influenced by the awareness of the disease prevalence, which varies significantly in different regions, and how clinicians utilize relevant clinical information. We conducted a retrospective study to evaluate how clinicians in the Southeast United States, where the climate is humid favoring mold growth, diagnosed cHP using items identified in the international modified Delphi survey of experts, i.e., environmental exposure, CT imaging and lung pathology, Methods: We searched Duke University Medical Center database for patients over the age of 18 with a diagnosis of cHP (ICD-9 code: 495) between Jan. 1, 2008 to Dec. 31, 2013 using a query tool, Duke Enterprise Data Unified Content Explorer (DEDUCE). Results: Five hundred patients were identified and 261 patients had cHP confirmed in clinic notes by a pulmonologist or an allergist. About half of the patients lived in the Research Triangle area where our medical center is located, giving an estimated prevalence rate of 6.5 per 100,000 persons. An exposure source was mentioned in 69.3% of the patient. The most common exposure sources were environmental molds (43.1%) and birds (26.0%). We used Venn diagram to evaluate how the patients met the three most common cHP diagnostic criteria: evidence of environmental exposures (history or precipitin) (E), chest CT imaging (C) and pathology from lung biopsies (P). Eighteen patients (6.9%) met none of three criteria. Of the remaining 243 patients, 135 patients (55.6%) had one (E 35.0%, C 3.3%, P 17.3%), 81 patients (33.3%) had two (E+C 12.3%, E+P 17.3%, C+P 4.9%), and 27 patients (11.1%) had all three criteria (E+C+P). Overall, 50.6% of patients had pathology from lung biopsy compared to 31.6% with CT scan. Conclusions: Environmental mold was the most common exposure for cHP in the Southeast United States. Lung pathology was available in more than half of cHP cases in our tertiary care center, perhaps reflecting the complexity of referrals. Differences in exposure sources and referral patterns should be considered in devising future diagnostic pathways or guidelines for cHP.

Background: Over 30% of patients with asthma have suboptimal control and frequently experience acute/sub-acute exacerbations. Improved patient-centered tools are needed to facilitate early identification and management of these events. To address this gap, we developed an on-demand asthma triage application which uses machine learning algorithms to enable real-time triage advice delivered directly to the patient. Objectives: Study aims were to test algorithm accuracy in simulated patient cases and assess therapeutic benefit to real-life asthmatics in an observational trial. Methods: The application was trained on opinions of six pulmonologists triaging over 1900 simulated cases covering the clinically relevant health variable space. The algorithm outputs 1) presence of exacerbation and 2) a triage recommendation from 4 choices (no action, continue usual treatment, call MD, and go to ER). Initial validation of the algorithm’s accuracy was through comparison to consensus (mode) of 8 pulmonologists. The algorithm was subsequently embedded in a mobile phone, and an observational trial on asthma control, anxiety, quality of life, and user experience was conducted. Results and Conclusion: Using physician consensus as a standard, the algorithm correctly assigned the exacerbation and triage classes with accuracy of 96% and 84% respectively, better than all 8 MDs. The algorithm also demonstrated superior accuracy and sensitivity in asthma scenarios requiring emergency care. Clinical trial data indicates statistically significant improvements in asthma control, quality of life and anxiety. Further clinical testing is needed to confirm these promising initial findings.

Background: Current at-home asthma management protocols are crowded with paper guidelines and exploratory health apps that lack rigor and validation at the level of the individual patient. No clear medical consensus has emerged regarding the efficacy of such approaches. We developed a novel digital therapeutic application that uses machine learning predictions for real-time detection of exacerbations and on-demand decision support while obviating the need for burdensome daily symptom entry. Methods: Physician opinion on a statistically and clinically comprehensive set of simulated patient cases was used to train a set of prediction algorithms. The accuracy of the models was assessed by comparison to physician panel consensus in a representative, out-of-sample validation set of 100 vignettes. Algorithms were subsequently deployed in a mobile application and evaluated in a 6 month, pre-post, observational trial of 25 patients with persistent asthma. Outcome data was collected using scored assessments of asthma control (ACT), quality of life (EQ5D5L) and anxiety (AIR). Findings: Algorithm accuracy and safety indicators surpassed all individual pulmonologists in both identifying exacerbations and identifying the consensus triage. The algorithm was also the top performer in sensitivity, specificity, and PPV when predicting a patient's need for emergency care. The observational trial yielded statistically and clinically significant improvement in mean difference scores in asthma control, 4.8 [95\% CI 2.1 - 7.5] (p = 0.004); quality of life, 15.7 [8.0 - 23.3] (p = 0.001); and anxiety, -3.1 [(-5.2) - (-1)] (p = 0.001). Interpretation: A mobile application equipped with a highly accurate machine-learning triage algorithm presents a promising and viable support tool with strong early indications of therapeutic value to patients. A randomized control trial is required to prove causality. Funding: Revon Systems Inc, eThera Technologies, NSF Award No. (FAIN):1820049. Declaration of Interest: We would like to disclose that Anthony N. Gerber is a consultant for eThera Inc and holds stock options. This does not alter our adherence to the Lancet Respiratory Medicine policies on sharing data and materials. Ethical Approval: The study was approved by Quorum Review, an independent Ethics Review Board.

Spirometric Volumes and Breathlessness Across Levels of Airflow Limitation : The COPDGene Study.

COPD patients are burdened with a daily risk of acute exacerbation and loss of control, which could be mitigated by effective, on-demand decision support tools. In this study, we present a machine learning-based strategy for early detection of exacerbations and subsequent triage. Our application uses physician opinion in a statistically and clinically comprehensive set of patient cases to train a supervised prediction algorithm. The accuracy of the model is assessed against a panel of physicians each triaging identical cases in a representative patient validation set. Our results show that algorithm accuracy and safety indicators surpass all individual pulmonologists in both identifying exacerbations and predicting the consensus triage in a 101 case validation set. The algorithm is also the top performer in sensitivity, specificity, and ppv when predicting a patient’s need for emergency care.

Purpose Communication in the intensive care unit (ICU) is an important component of quality ICU care. In this report, we evaluate the long-term effects of a quality improvement (QI) initiative, based on the VALUE communication strategy, designed to improve communication with family members of critically ill patients. Materials and Methods We implemented a multifaceted intervention to improve communication in the ICU and measured processes of care. Quality improvement components included posted VALUE placards, templated progress note inclusive of communication documentation, and a daily rounding checklist prompt. We evaluated care for all patients cared for by the intensivists during three separate 3 week periods, pre, post, and 3 years following the initial intervention. Results Care delivery was assessed in 38 patients and their families in the pre-intervention sample, 27 in the post-intervention period, and 41 in follow-up. Process measures of communication showed improvement across the evaluation periods, for example, daily updates increased from pre 62% to post 76% to current 84% of opportunities. Conclusions Our evaluation of this quality improvement project suggests persistence and continued improvements in the delivery of measured aspects of ICU family communication. Maintenance with point-of-care-tools may account for some of the persistence and continued improvements.

We report a case of Cryptococcus neoformans pneumonia in a patient taking ruxolitinib, a janus kinase 1,2 inhibitor approved for the treatment of myelofibrosis. We hypothesize that ruxolitinib contributed to this infection through its effects on cell-mediated immunity. Clinicians should be aware of the potential for intracellular or opportunistic infections associated with this novel drug class.