Performance evaluation of three semantic expansions to query PubMed.

BACKGROUND: PubMed is one of the most important basic tools to access medical literature. Semantic query expansion using synonyms can improve retrieval efficacy. OBJECTIVE: The objective was to evaluate the performance of three semantic query expansion strategies. METHODS: Queries were built for forty MeSH descriptors using three semantic expansion strategies (MeSH synonyms, UMLS mappings, and mappings created by the CISMeF team), then sent to PubMed. To evaluate expansion performances for each query, the first twenty citations were selected, and their relevance were judged by three independent evaluators based on the title and abstract. RESULTS: Queries built with the UMLS expansion provided new citations with a slightly higher mean precision (74.19%) than with the CISMeF expansion (70.28%), although the difference was not significant. Inter-rater agreement was 0.28. Results varied greatly depending on the descriptor selected. DISCUSSION: The number of citations retrieved by the three strategies and their precision varied greatly according to the descriptor. This heterogeneity could be explained by the quality of the synonyms. Optimal use of these different expansions would be through various combinations of UMLS and CISMeF intersections or unions. CONCLUSION: Information retrieval tools should propose different semantic expansions depending on the descriptor and the search objectives.

Community-acquired Klebsiella pneumoniae pneumonia in ICU: a multicenter retrospective study.

BACKGROUND: Alongside the recent worldwide expansion of hypervirulent Klebsiella pneumoniae (KP) infections, the available literature regarding cases of community acquired pneumonias (KP-CAP) remains scarce but reports a strikingly high and early mortality. We performed a retrospective multicenter study (7 ICU in France) between 2015 and 2019, comparing prognosis and severity of KP-CAP versus Streptococcus pneumoniae - CAP (SP-CAP). METHODS: For each KP-CAP, three SP-CAP admitted in ICUs within the same center and within the same 6-month window were selected. When available, KP strains were studied, and bacterial virulence was genetically assessed for virulence factors. The primary outcome was in-hospital mortality. Associations between clinical outcomes and type of infection were tested using univariate and multivariate logistic regressions, adjusted for pairing variables. RESULTS: Twenty-seven KP-CAP and 81 SP-CAP were included. Respective in-hospital mortality rates were 59% (n = 16) and 17% (n = 14, p < 0.001), despite adequate antibiotic therapy. KP-CAP median time from admission to death was 26.9 h [IQR 5.75-44 h] and were significantly associated with higher rates of multiple organ failures (93% vs. 42%, p < 0.001), disseminated intravascular coagulation (12% vs. 1.3%, p = 0.046), septic shock (median lactate on ICU admission 4.60 vs. 2.90 mmol/L, p = 0.030) and kidney failure (KDIGO-3: 87% vs. 44%, p < 0.001). Interestingly, alcoholism was the only identified predisposing factor of KP-CAP. Severity on ICU admission (2-fold higher for KP-CAP) was the only factor associated with mortality in a multivariate analysis. CONCLUSION: We described a strong association between KP-CAP infection and higher and earlier mortality when compared to SP-CAP. Moreover, alcoholism was the sole predisposing factor associated with KP-CAP infection. These findings should raise awareness of clinicians involved in the management of severe CAP about this microbiological etiology. Future prospective studies are needed to confirm these results and to design strategies to improve the prognosis of such infections.

Modeling the impact of national and regional lockdowns on the 2020 spring wave of COVID-19 in France.

Several countries have implemented lockdowns to control their COVID-19 epidemic. However, questions like "where" and "when" still require answers. We assessed the impact of national and regional lockdowns considering the French first epidemic wave of COVID-19 as a case study. In a regional lockdown scenario aimed at preventing intensive care units (ICU) saturation, almost all French regions would have had to implement a lockdown within 10 days and 96% of ICU capacities would have been used. For slowly growing epidemics, with a lower reproduction number, the expected delays between regional lockdowns increase. However, the public health costs associated with these delays tend to grow with time. In a quickly growing pandemic wave, defining the timing of lockdowns at a regional rather than national level delays by a few days the implementation of a nationwide lockdown but leads to substantially higher morbidity, mortality, and stress on the healthcare system.

[Covid-19 : un pari épidémiologique perdu ?] / Covid-19: a lost epidemiological bet ?

COVID-19: A LOST EPIDEMIOLOGICAL BET? For the last two years, the world has been sailing from one epidemic wave to another. From lockdowns to curfews, strategies have changed over time, whether on travel restrictions, mask requirements, or vaccination. The health crisis has never ceased to toss us from one extreme to the other; each step further testing the resilience of our health system and the population's trust in its leaders. By reviewing some key stages of this pandemic crisis, we highlight why the fight against an emerging virus forces us to trials and errors, which must nevertheless be corrected in the light of the knowledge acquired. Therefore, changes in strategy become necessary even if they undermine the confidence of the population; confidence that cannot be maintained without a perfect understanding of this specific context.

COVID-19 : UN PARI ÉPIDÉMIOLOGIQUE PERDU ? Depuis deux ans, le monde navigue d'une vague épidémique à l'autre. De confinements en couvre-feux, les stratégies ont changé avec le temps, que ce soit sur la restriction des voyages, le port du masque, ou la vaccination. La crise sanitaire n'aura eu de cesse de nous ballotter d'un extrême à l'autre ; chaque étape mettant un peu plus à l'épreuve la résilience de notre système de santé et le crédit de ses dirigeants. Passer en revue quelques étapes clés de cette pandémie, permet d'éclaircir pourquoi la lutte contre un virus émergent nous contraint au tâtonnement et aux erreurs, qui doivent néanmoins être corrigés à l'aune des connaissances acquises. Dès lors, les changements de stratégies deviennent nécessaires même s'ils mettent à mal la confiance de la population ; une confiance qui ne peut être gardée sans une parfaite compréhension de ce contexte si particulier.

Evaluating COVID-19 Booster Vaccination Strategies in a Partially Vaccinated Population: A Modeling Study.

BACKGROUND: Several countries are implementing COVID-19 booster vaccination campaigns. The objective of this study was to model the impact of different primary and booster vaccination strategies. METHODS: We used a compartmental model fitted to hospital admission data in France to analyze the impact of primary and booster vaccination strategies on morbidity and mortality, assuming waning of immunity and various levels of virus transmissibility during winter. RESULTS: Strategies prioritizing primary vaccinations were systematically more effective than strategies prioritizing boosters. Regarding booster strategies targeting different age groups, their effectiveness varied with immunity and virus transmissibility levels. If the waning of immunity affects all adults, people aged 30 to 49 years should be boosted in priority, even for low transmissibility levels. CONCLUSIONS: Increasing the primary vaccination coverage should remain a priority. If a plateau has been reached, boosting the immunity of younger adults could be the most effective strategy, especially if SARS-CoV-2 transmissibility is high.

A modelling study investigating short and medium-term challenges for COVID-19 vaccination: From prioritisation to the relaxation of measures.

BACKGROUND: The roll-out of COVID-19 vaccines is a multi-faceted challenge whose performance depends on pace of vaccination, vaccine characteristics and heterogeneities in individual risks. METHODS: We developed a mathematical model accounting for the risk of severe disease by age and comorbidity, and transmission dynamics. We compared vaccine prioritisation strategies in the early roll-out stage and quantified the extent to which measures could be relaxed as a function of the vaccine coverage achieved in France. FINDINGS: Prioritizing at-risk individuals reduces morbi-mortality the most if vaccines only reduce severity, but is of less importance if vaccines also substantially reduce infectivity or susceptibility. Age is the most important factor to consider for prioritization; additionally accounting for comorbidities increases the performance of the campaign in a context of scarce resources. Vaccinating 90% of ≥65 y.o. and 70% of 18-64 y.o. before autumn 2021 with a vaccine that reduces severity by 90% and susceptibility by 80%, we find that control measures reducing transmission rates by 15-27% should be maintained to remain below 1000 daily hospital admissions in France with a highly transmissible variant (basic reproduction number R0  = 4). Assuming 90% of ≥65 y.o. are vaccinated, full relaxation of control measures might be achieved with a vaccine coverage of 89-100% in 18-64 y.o or 60-69% of 0-64 y.o. INTERPRETATION: Age and comorbidity-based vaccine prioritization strategies could reduce the burden of the disease. Very high vaccination coverage may be required to completely relax control measures. Vaccination of children, if possible, could lower coverage targets necessary to achieve this objective.

Identification of the Best Semantic Expansion to Query PubMed Through Automatic Performance Assessment of Four Search Strategies on All Medical Subject Heading Descriptors: Comparative Study.

BACKGROUND: With the continuous expansion of available biomedical data, efficient and effective information retrieval has become of utmost importance. Semantic expansion of queries using synonyms may improve information retrieval. OBJECTIVE: The aim of this study was to automatically construct and evaluate expanded PubMed queries of the form "preferred term"[MH] OR "preferred term"[TIAB] OR "synonym 1"[TIAB] OR "synonym 2"[TIAB] OR …, for each of the 28,313 Medical Subject Heading (MeSH) descriptors, by using different semantic expansion strategies. We sought to propose an innovative method that could automatically evaluate these strategies, based on the three main metrics used in information science (precision, recall, and F-measure). METHODS: Three semantic expansion strategies were assessed. They differed by the synonyms used to build the queries as follows: MeSH synonyms, Unified Medical Language System (UMLS) mappings, and custom mappings (Catalogue et Index des Sites Médicaux de langue Française [CISMeF]). The precision, recall, and F-measure metrics were automatically computed for the three strategies and for the standard automatic term mapping (ATM) of PubMed. The method to automatically compute the metrics involved computing the number of all relevant citations (A), using National Library of Medicine indexing as the gold standard ("preferred term"[MH]), the number of citations retrieved by the added terms ("synonym 1"[TIAB] OR "synonym 2"[TIAB] OR …) (B), and the number of relevant citations retrieved by the added terms (combining the previous two queries with an "AND" operator) (C). It was possible to programmatically compute the metrics for each strategy using each of the 28,313 MeSH descriptors as a "preferred term," corresponding to 239,724 different queries built and sent to the PubMed application program interface. The four search strategies were ranked and compared for each metric. RESULTS: ATM had the worst performance for all three metrics among the four strategies. The MeSH strategy had the best mean precision (51%, SD 23%). The UMLS strategy had the best recall and F-measure (41%, SD 31% and 36%, SD 24%, respectively). CISMeF had the second best recall and F-measure (40%, SD 31% and 35%, SD 24%, respectively). However, considering a cutoff of 5%, CISMeF had better precision than UMLS for 1180 descriptors, better recall for 793 descriptors, and better F-measure for 678 descriptors. CONCLUSIONS: This study highlights the importance of using semantic expansion strategies to improve information retrieval. However, the performances of a given strategy, relatively to another, varied greatly depending on the MeSH descriptor. These results confirm there is no ideal search strategy for all descriptors. Different semantic expansions should be used depending on the descriptor and the user's objectives. Thus, we developed an interface that allows users to input a descriptor and then proposes the best semantic expansion to maximize the three main metrics (precision, recall, and F-measure).

Word Embedding for the French Natural Language in Health Care: Comparative Study.

BACKGROUND: Word embedding technologies, a set of language modeling and feature learning techniques in natural language processing (NLP), are now used in a wide range of applications. However, no formal evaluation and comparison have been made on the ability of each of the 3 current most famous unsupervised implementations (Word2Vec, GloVe, and FastText) to keep track of the semantic similarities existing between words, when trained on the same dataset. OBJECTIVE: The aim of this study was to compare embedding methods trained on a corpus of French health-related documents produced in a professional context. The best method will then help us develop a new semantic annotator. METHODS: Unsupervised embedding models have been trained on 641,279 documents originating from the Rouen University Hospital. These data are not structured and cover a wide range of documents produced in a clinical setting (discharge summary, procedure reports, and prescriptions). In total, 4 rated evaluation tasks were defined (cosine similarity, odd one, analogy-based operations, and human formal evaluation) and applied on each model, as well as embedding visualization. RESULTS: Word2Vec had the highest score on 3 out of 4 rated tasks (analogy-based operations, odd one similarity, and human validation), particularly regarding the skip-gram architecture. CONCLUSIONS: Although this implementation had the best rate for semantic properties conservation, each model has its own qualities and defects, such as the training time, which is very short for GloVe, or morphological similarity conservation observed with FastText. Models and test sets produced by this study will be the first to be publicly available through a graphical interface to help advance the French biomedical research.

Word Embedding for French Natural Language in Healthcare: A Comparative Study.

Structuring raw medical documents with ontology mapping is now the next step for medical intelligence. Deep learning models take as input mathematically embedded information, such as encoded texts. To do so, word embedding methods can represent every word from a text as a fixed-length vector. A formal evaluation of three word embedding methods has been performed on raw medical documents. The data corresponds to more than 12M diverse documents produced in the Rouen hospital (drug prescriptions, discharge and surgery summaries, inter-services letters, etc.). Automatic and manual validation demonstrates that Word2Vec based on the skip-gram architecture had the best rate on three out of four accuracy tests. This model will now be used as the first layer of an AI-based semantic annotator.