Estimating Degradation of Machine Learning Data Assets

Large-scale adoption of Artificial Intelligence and Machine Learning (AI-ML) models fed by heterogeneous, possibly untrustworthy data sources has spurred interest in estimating degradation of such models due to spurious, adversarial, or low-quality data assets. We propose a quantitative estimate of the severity of classifiers’ training set degradation: an index expressing the deformation of the convex hulls of the classes computed on a held-out dataset generated via an unsupervised technique. We show that our index is computationally light, can be calculated incrementally and complements well existing ML data assets’ quality measures. As an experimentation, we present the computation of our index on a benchmark convolutional image classifier.

Chasm Between Cancer Quality Measures and Electronic Health Record Data Quality

PURPOSE The Medicare Access and CHIP Reauthorization Act of 2015 (MACRA) requires eligible clinicians to report clinical quality measures (CQMs) in the Merit-Based Incentive Payment System (MIPS) to maximize reimbursement. To determine whether structured data in electronic health records (EHRs) were adequate to report MIPS CQMs, EHR data aggregated by ASCO's CancerLinQ platform were analyzed. MATERIALS AND METHODS Using the CancerLinQ health technology platform, 19 Oncology MIPS (oMIPS) CQMs were evaluated to determine the presence of data elements (DEs) necessary to satisfy each CQM and the DE percent population with patient data (fill rates). At the time of this analysis, the CancerLinQ network comprised 63 active practices, representing eight different EHR vendors and containing records for more than 1.63 million unique patients with one or more malignant neoplasms (1.73 million cancer cases). RESULTS Fill rates for the 63 oMIPS-associated DEs varied widely among the practices. The average site had at least one filled DE for 52% of the DEs. Only 35% of the DEs were populated for at least one patient record in 95% of the practices. However, the average DE fill rate of all practices was 23%. No data were found at any practice for 22% of the DEs. Since any oMIPS CQM with an unpopulated DE component resulted in an inability to compute the measure, only two (10.5%) of the 19 oMIPS CQMs were computable for more than 1% of the patients. CONCLUSION Although EHR systems had relatively high DE fill rates for some DEs, underfilling and inconsistency of DEs in EHRs render automated oncology MIPS CQM calculations impractical.

Acoustic classification of coronal stops of Eastern Punjabi

Punjabi is an Indo-Aryan language which contrasts a rich set of coronal stops at dental and retroflex places of articulation across three laryngeal configurations. Moreover, all these stops occur contrastively in various positions (word-initially, -medially, and -finally). The goal of this study is to investigate how various coronal place and laryngeal contrasts are distinguished acoustically both within and across word positions. A number of temporal and spectral correlates were examined in data from 13 speakers of Eastern Punjabi: Voice Onset Time, release and closure durations, fundamental frequency, F1-F3 formants, spectral center of gravity and standard deviation, H1*-H2*, and cepstral peak prominence. The findings indicated that higher formants and spectral measures were most important for the classification of place contrasts across word positions, whereas laryngeal contrasts were reliably distinguished by durational and voice quality measures. Word-medially and -finally, F2 and F3 of the preceding vowels played a key role in distinguishing the dental and retroflex stops, while spectral noise measures were more important word-initially. The findings of this study contribute to a better understanding of factors involved in the maintenance of typologically rare and phonetically complex sets of place and laryngeal contrasts in the coronal stops of Indo-Aryan languages.

Nursing Informatics Research Trends: Findings from an International Survey

This follow-up survey on trends in Nursing Informatics (NI) was conducted by the International Medical Informatics Association (IMIA) Student and Emerging Professionals (SEP) group as a cross-sectional study in 2019. There were 455 responses from 24 countries. Based on the findings NI research is evolving rapidly. Current ten most common trends include: clinical quality measures, clinical decision support, big data, artificial intelligence, care coordination, education and competencies, patient safety, mobile health, description of nursing practices and evaluation of patient outcomes. The findings help support the efforts to efficiently use resources in the promotion of health care activities, to support the development of informatics education and to grow NI as a profession.

Can SBAR be implemented with high fidelity and does it improve communication between healthcare workers? A systematic review

ObjectiveTo characterise the extent to which health professionals perform SBAR (situation, background, assessment, recommendation) as intended (ie, with high fidelity) and the extent to which its use improves communication clarity or other quality measures.Data sourcesMedline, Healthstar, PsycINFO, Embase and CINAHL to October 2020 and handsearching selected journals.Study selection and outcome measuresEligible studies consisted of controlled trials and time series, including simple before-after design, assessing SBAR implementation fidelity or the effects of SBAR on communication clarity or other quality measures (eg, safety climate, patient outcomes).Data extraction and synthesisTwo reviewers independently abstracted data according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses on study features, intervention details and study outcomes. We characterised the magnitude of improvement in outcomes as small (<20% relative increase), moderate (20%–40%) or large (>40%).ResultsTwenty-eight studies (3 randomised controlled trials, 6 controlled before-after studies, and 19 uncontrolled before-after studies) met inclusion criteria. Of the nine studies assessing fidelity of SBAR use, four occurred in classroom settings and three of these studies reported large improvements. The five studies assessing fidelity in clinical settings reported small to moderate effects. Among eight studies measuring communication clarity, only three reported large improvements and two of these occurred in classroom settings. Among the 17 studies reporting impacts on quality measures beyond communication, over half reported moderate to large improvements. These improvements tended to involve measures of teamwork and culture. Improvements in patient outcomes occurred only with intensive multifaceted interventions (eg, early warning scores and rapid response systems).ConclusionsHigh fidelity uptake of SBAR and improvements in communication clarity occurred predominantly in classroom studies. Studies in clinical settings achieving impacts beyond communication typically involved broader, multifaceted interventions. Future efforts to improve communication using SBAR should first confirm high fidelity uptake in clinical settings rather than assuming this has occurred.PROSPERO registration numberCRD42018111377.