Dataset Information

A study of the transferability of influenza case detection systems between two large healthcare systems.

ABSTRACT:

Objectives

This study evaluates the accuracy and transferability of Bayesian case detection systems (BCD) that use clinical notes from emergency department (ED) to detect influenza cases.

Methods

A BCD uses natural language processing (NLP) to infer the presence or absence of clinical findings from ED notes, which are fed into a Bayesain network classifier (BN) to infer patients' diagnoses. We developed BCDs at the University of Pittsburgh Medical Center (BCDUPMC) and Intermountain Healthcare in Utah (BCDIH). At each site, we manually built a rule-based NLP and trained a Bayesain network classifier from over 40,000 ED encounters between Jan. 2008 and May. 2010 using feature selection, machine learning, and expert debiasing approach. Transferability of a BCD in this study may be impacted by seven factors: development (source) institution, development parser, application (target) institution, application parser, NLP transfer, BN transfer, and classification task. We employed an ANOVA analysis to study their impacts on BCD performance.

Results

Both BCDs discriminated well between influenza and non-influenza on local test cases (AUCs > 0.92). When tested for transferability using the other institution's cases, BCDUPMC discriminations declined minimally (AUC decreased from 0.95 to 0.94, p<0.01), and BCDIH discriminations declined more (from 0.93 to 0.87, p<0.0001). We attributed the BCDIH decline to the lower recall of the IH parser on UPMC notes. The ANOVA analysis showed five significant factors: development parser, application institution, application parser, BN transfer, and classification task.

Conclusion

We demonstrated high influenza case detection performance in two large healthcare systems in two geographically separated regions, providing evidentiary support for the use of automated case detection from routinely collected electronic clinical notes in national influenza surveillance. The transferability could be improved by training Bayesian network classifier locally and increasing the accuracy of the NLP parser.

SUBMITTER: Ye Y

PROVIDER: S-EPMC5381795 | biostudies-literature | 2017

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A study of the transferability of influenza case detection systems between two large healthcare systems.

Ye Ye Y Wagner Michael M MM Cooper Gregory F GF Ferraro Jeffrey P JP Su Howard H Gesteland Per H PH Haug Peter J PJ Millett Nicholas E NE Aronis John M JM Nowalk Andrew J AJ Ruiz Victor M VM López Pineda Arturo A Shi Lingyun L Van Bree Rudy R Ginter Thomas T Tsui Fuchiang F

PloS one 20170405 4

<h4>Objectives</h4>This study evaluates the accuracy and transferability of Bayesian case detection systems (BCD) that use clinical notes from emergency department (ED) to detect influenza cases.<h4>Methods</h4>A BCD uses natural language processing (NLP) to infer the presence or absence of clinical findings from ED notes, which are fed into a Bayesain network classifier (BN) to infer patients' diagnoses. We developed BCDs at the University of Pittsburgh Medical Center (BCDUPMC) and Intermountai ...[more]

PMID: 28380048

Similar Datasets

Project description:BackgroundThe World Health Organisation recommends outpatient influenza-like illness (ILI) and inpatient severe acute respiratory illness (SARI) surveillance. We evaluated two influenza surveillance systems in South Africa: one for ILI and another for SARI.MethodologyThe Viral Watch (VW) programme has collected virological influenza surveillance data voluntarily from patients with ILI since 1984 in private and public clinics in all 9 South African provinces. The SARI surveillance programme has collected epidemiological and virological influenza surveillance data since 2009 in public hospitals in 4 provinces by dedicated personnel. We compared nine surveillance system attributes from 2009-2012.ResultsWe analysed data from 18,293 SARI patients and 9,104 ILI patients. The annual proportion of samples testing positive for influenza was higher for VW (mean 41%) than SARI (mean 8%) and generally exceeded the seasonal threshold from May to September (VW: weeks 21-40; SARI: weeks 23-39). Data quality was a major strength of SARI (most data completion measures >90%; adherence to definitions: 88-89%) and a relative weakness of the VW programme (62% of forms complete, with limited epidemiologic data collected; adherence to definitions: 65-82%). Timeliness was a relative strength of both systems (e.g. both collected >93% of all respiratory specimens within 7 days of symptom onset). ILI surveillance was more nationally representative, financially sustainable and expandable than the SARI system. Though the SARI programme is not nationally representative, the high quality and detail of SARI data collection sheds light on the local burden and epidemiology of severe influenza-associated disease.ConclusionsTo best monitor influenza in South Africa, we propose that both ILI and SARI should be under surveillance. Improving ILI surveillance will require better quality and more systematic data collection, and SARI surveillance should be expanded to be more nationally representative, even if this requires scaling back on information gathered.

Project description:BACKGROUND:Patient safety research is conducted predominantly in hospital settings, with a dearth of insight from primary care, despite suggestions that 2.2% of primary care consultations result in a patient safety incident. This study aimed to assess the feasibility of an intervention intended to improve patient safety in general practice. METHODS:A randomised controlled feasibility study was conducted with general practices in the Republic of Ireland (N = 9) and Northern Ireland (N = 2), randomly assigned to the intervention (N = 5) or control (N = 6) group. The nine-month intervention consisted of: 1) repeated safety climate (SC) measurement (using GP-SafeQuest questionnaire) and feedback (comparative anonymised practice-level SC data), and 2) patient record reviews using a specialised trigger tool to identify instances of undetected patient harm. For control practices, SC was measured at baseline and study end only. The intervention's perceived usefulness and feasibility were explored via an end-of-study questionnaire and semi-structured interviews. RESULTS:Thirteen practices were invited; 11 participated; 10 completed the study. At baseline, 84.8% of intervention practice staff (39/46) and 77.8% (42/54) of control practice staff completed the SC questionnaire; at the study terminus, 78.3% (36/46) of intervention practice staff and 68.5% (37/54) of control practice staff did so. Changes in SC scores, indicating improvement, were observed among the intervention practices but not in the control group. The trigger tool was applied to 188 patient records; patient safety incidents of varying severity were detected in 19.1% (36/188). Overall, 59% of intervention practice team members completed the end-of-study questionnaire, with the majority in both healthcare systems responding positively about the intervention. Interviews (N = 9) identified the intervention's usefulness in informing practice management and patient safety issues, time as a barrier to its use, and the value of group discussion of feedback. CONCLUSION:This feasibility study suggests that a definitive randomised controlled trial of the intervention is warranted. Our findings suggest that the intervention is feasible, useful, and sustainable. Practices were willing to be recruited into the study, response and retention rates were acceptable, and there is possible evidence of a positive effect of the intervention. TRIAL REGISTRATION:The trial registration number is: ISRCTN11426121 (retrospectively registered 12th June 2018).

Project description:BackgroundElectronic health records (EHR) are increasingly used for studying multimorbidities. However, concerns about accuracy, completeness, and EHRs being primarily designed for billing and administration raise questions about the consistency and reproducibility of EHR-based multimorbidity research.MethodsUtilizing phecodes to represent the disease phenome, we analyzed pairwise comorbidity strengths using a dual logistic regression approach and constructed multimorbidity as an undirected weighted graph. We assessed the consistency of the multimorbidity networks within and between two major EHR systems at local (nodes and edges), meso (neighboring patterns), and global (network statistics) scales. We present case studies to identify disease clusters and uncover clinically interpretable disease relationships. We provide an interactive web tool and a knowledge base combing data from multiple sources for online multimorbidity analysis.FindingsAnalyzing data from 500,000 patients across Vanderbilt University Medical Center and Mass General Brigham health systems, we observed a strong correlation in disease frequencies ( Kendall's τ = 0.643) and comorbidity strengths (Pearson ρ = 0.79). Consistent network statistics across EHRs suggest a similar structure of multimorbidity networks at various scales. Comorbidity strengths and similarities of multimorbidity connection patterns align with the disease genetic correlations. Graph-theoretic analyses revealed a consistent core-periphery structure, implying efficient network clustering through threshold graph construction. Using hydronephrosis as a case study, we demonstrated the network's ability to uncover clinically relevant disease relationships and provide novel insights.InterpretationOur findings demonstrate the robustness of large-scale EHR data for studying complex disease interactions. The alignment of multimorbidity patterns with genetic data suggests the potential utility for uncovering shared etiology of diseases. The consistent core-periphery network structure offers a strategic approach to analyze disease clusters. This work also sets the stage for advanced disease modeling, with implications for precision medicine.FundingVUMC Biostatistics Development Award, UL1 TR002243, R21DK127075, R01HL140074, P50GM115305, R01CA227481.

Project description:ObjectivesTo examine the factors associated with COVID-19 vaccine receipt among healthcare workers and the role of vaccine confidence in decisions to vaccinate, and to better understand concerns related to COVID-19 vaccination.DesignCross-sectional anonymous survey among front-line, support service and administrative healthcare workers.SettingTwo large integrated healthcare systems (one private and one public) in New York City during the initial roll-out of the COVID-19 vaccine.Participants1933 healthcare workers, including nurses, physicians, allied health professionals, environmental services staff, researchers and administrative staff.Primary outcome measuresThe primary outcome was COVID-19 vaccine receipt during the initial roll-out of the vaccine among healthcare workers.ResultsAmong 1933 healthcare workers who had been offered the vaccine, 81% had received the vaccine at the time of the survey. Receipt was lower among black (58%; OR: 0.14, 95% CI 0.1 to 0.2) compared with white (91%) healthcare workers, and higher among non-Hispanic (84%) compared with Hispanic (69%; OR: 2.37, 95% CI 1.8 to 3.1) healthcare workers. Among healthcare workers with concerns about COVID-19 vaccine safety, 65% received the vaccine. Among healthcare workers who agreed with the statement that the vaccine is important to protect family members, 86% were vaccinated. Of those who disagreed, 25% received the vaccine (p<0.001). In a multivariable analysis, concern about being experimented on (OR: 0.44, 95% CI 0.31 to 0.6), concern about COVID-19 vaccine safety (OR: 0.39, 95% CI 0.28 to 0.55), lack of influenza vaccine receipt (OR: 0.28, 95% CI 0.18 to 0.44), disagreeing that COVID-19 vaccination is important to protect others (OR: 0.37, 95% CI 0.27 to 0.52) and black race (OR: 0.38, 95% CI 0.24 to 0.59) were independently associated with COVID-19 vaccine non-receipt. Over 70% of all healthcare workers responded that they had been approached for vaccine advice multiple times by family, community members and patients.ConclusionsOur data demonstrated high overall receipt among healthcare workers. Even among healthcare workers with concerns about COVID-19 vaccine safety, side effects or being experimented on, over 50% received the vaccine. Attitudes around the importance of COVID-19 vaccination to protect others played a large role in healthcare workers' decisions to vaccinate. We observed striking inequities in COVID-19 vaccine receipt, particularly affecting black and Hispanic workers. Further research is urgently needed to address issues related to vaccine equity and uptake in the context of systemic racism and barriers to care. This is particularly important given the influence healthcare workers have in vaccine decision-making conversations in their communities.

Project description:Background and aimsLarge clades of angiosperms are often characterized by diverse interactions with pollinators, but how these pollination systems are structured phylogenetically and biogeographically is still uncertain for most families. Apocynaceae is a clade of >5300 species with a worldwide distribution. A database representing >10 % of species in the family was used to explore the diversity of pollinators and evolutionary shifts in pollination systems across major clades and regions.MethodsThe database was compiled from published and unpublished reports. Plants were categorized into broad pollination systems and then subdivided to include bimodal systems. These were mapped against the five major divisions of the family, and against the smaller clades. Finally, pollination systems were mapped onto a phylogenetic reconstruction that included those species for which sequence data are available, and transition rates between pollination systems were calculated.Key resultsMost Apocynaceae are insect pollinated with few records of bird pollination. Almost three-quarters of species are pollinated by a single higher taxon (e.g. flies or moths); 7 % have bimodal pollination systems, whilst the remaining approx. 20 % are insect generalists. The less phenotypically specialized flowers of the Rauvolfioids are pollinated by a more restricted set of pollinators than are more complex flowers within the Apocynoids + Periplocoideae + Secamonoideae + Asclepiadoideae (APSA) clade. Certain combinations of bimodal pollination systems are more common than others. Some pollination systems are missing from particular regions, whilst others are over-represented.ConclusionsWithin Apocynaceae, interactions with pollinators are highly structured both phylogenetically and biogeographically. Variation in transition rates between pollination systems suggest constraints on their evolution, whereas regional differences point to environmental effects such as filtering of certain pollinators from habitats. This is the most extensive analysis of its type so far attempted and gives important insights into the diversity and evolution of pollination systems in large clades.

Dataset Information

A study of the transferability of influenza case detection systems between two large healthcare systems.

Objectives

Methods

Results

Conclusion

Publications

A study of the transferability of influenza case detection systems between two large healthcare systems.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets