{ "items": [ "\n\n
BackgroundCumulative malaria parasite exposure in endemic regions often results in the acquisition of partial immunity and asymptomatic infections. There is limited information on how host-parasite interactions mediate the maintenance of chronic symptomless infections that sustain malaria transmission.MethodsHere, we determined the gene expression profiles of the parasite population and the corresponding host peripheral blood mononuclear cells (PBMCs) from 21 children (ResultsChildren with asymptomatic infections had a parasite transcriptional profile characterized by a bias toward trophozoite stage (~\u200912\u00a0h-post invasion) parasites and low parasite levels, while early ring stage parasites were characteristic of febrile malaria. The host response of asymptomatic children was characterized by downregulated transcription of genes associated with inflammatory responses, compared with children with febrile malaria,. Interestingly, the host responses during febrile infections that followed an asymptomatic infection featured stronger inflammatory responses, whereas the febrile host responses from previously uninfected children featured increased humoral immune responses.ConclusionsThe priming effect of prior asymptomatic infection may explain the blunted acquisition of antibody responses seen to malaria antigens following natural exposure or vaccination in malaria endemic areas.
\n \n\n \n \nHand, foot and mouth disease (HFMD) is highly prevalent in the Asia Pacific region, particularly in Vietnam. To develop effective interventions and efficient vaccination programs, we inferred the age-time-specific transmission patterns of HFMD serotypes enterovirus A71 (EV-A71), coxsackievirus A6 (CV-A6), coxsackievirus A10 (CV-A10), coxsackievirus A16 (CV-A16) in Ho Chi Minh City, Vietnam from a case data collected during 2013-2018 and a serological survey data collected in 2015 and 2017. We proposed a catalytic model framework with good adaptability to incorporate maternal immunity using various mathematical functions. Our results indicate the high-level transmission of CV-A6 and CV-A10 which is not obvious in the case data, due to the variation of disease severity across serotypes. Our results provide statistical evidence supporting the strong association between severe illness and CV-A6 and EV-A71 infections. The HFMD dynamic pattern presents a cyclical pattern with large outbreaks followed by a decline in subsequent years. Additionally, we identify the age group with highest risk of infection as 1-2 years and emphasise the risk of future outbreaks as over 50% of children aged 6-7 years were estimated to be susceptible to CV-A16 and EV-A71. Our study highlights the importance of multivalent vaccines and active surveillance for different serotypes, supports early vaccination prior to 1 year old, and points out the potential utility for vaccinating children older than 5 years old in Vietnam.
\n \n\n \n \nNatural killer (NK) cells are potent immune effectors that can be activated via antibody-mediated Fc receptor engagement. Using multiparameter flow cytometry, we found that NK cells degranulate and release IFN-\u03b3 upon stimulation with antibody-opsonized Plasmodium falciparum merozoites. Antibody-dependent NK (Ab-NK) activity was largely strain transcending and enhanced invasion inhibition into erythrocytes. Ab-NK was associated with the successful control of parasitemia after experimental malaria challenge in African adults. In an independent cohort study in children, Ab-NK increased with age, was boosted by concurrent P. falciparum infections, and was associated with a lower risk of clinical episodes of malaria. Nine of the 14 vaccine candidates tested induced Ab-NK, including some less well-characterized antigens: P41, P113, MSP11, RHOPH3, and Pf_11363200. These data highlight an important role of Ab-NK activity in immunity against malaria and provide a potential mechanism for evaluating vaccine candidates.
\n \n\n \n \nAbstract\nBrief summary\nIn early symptomatic COVID-19 treatment, high dose oral favipiravir did not accelerate viral clearance.\n\nBackground\nFavipiravir, an anti-influenza drug, has in vitro antiviral activity against SARS-CoV-2. Clinical trial evidence to date is inconclusive. Favipiravir has been recommended for the treatment of COVID-19 in some countries.\n\nMethods\nIn a multicentre open-label, randomised, controlled, adaptive platform trial, low-risk adult patients with early symptomatic COVID-19 were randomised to one of ten treatment arms including high dose oral favipiravir (3.6g on day 0 followed by 1.6g daily to complete 7 days treatment) or no study drug. The primary outcome was the rate of viral clearance (derived under a linear mixed-effects model from the daily log10 viral densities in standardised duplicate oropharyngeal swab eluates taken daily over 8 days [18 swabs per patient]), assessed in a modified intention-to-treat population (mITT). The safety population included all patients who received at least one dose of the allocated intervention. This ongoing adaptive platform trial was registered at ClinicalTrials.gov (NCT05041907) on 13/09/2021.\n\nResults\nIn the final analysis, the mITT population contained data from 114 patients randomised to favipiravir and 126 patients randomised concurrently to no study drug. Under the linear mixed-effects model fitted to all oropharyngeal viral density estimates in the first 8 days from randomisation (4,318 swabs), there was no difference in the rate of viral clearance between patients given favipiravir and patients receiving no study drug; a -1% (95% credible interval: -14 to 14%) difference. High dose favipiravir was well-tolerated.\n\nInterpretation\nFavipiravir does not accelerate viral clearance in early symptomatic COVID-19. The viral clearance rate estimated\u00a0from quantitative measurements of oropharyngeal eluate viral densities assesses the antiviral efficacy of drugs in vivo with comparatively few studied\u00a0patients.\n
\n \n\n \n \nBackgroundImproving screening and triage practices is essential for early severity assessments at the first point of contact and ensuring timely attention by healthcare workers (HCWs). The main objective of this study was to explore the triage process among febrile patients and HCWs in the emergency department (ED) of a tertiary care hospital in a resource-constrained setting.MethodsThis qualitative study was conducted from March to May 2023 at the ED of Tribhuvan University Teaching Hospital (TUTH), Nepal. The study included in-depth interviews with febrile patients (n\u2009=\u200915) and HCWs (n\u2009=\u200915). Additionally, direct observation notes (n\u2009=\u200920) were collected to document the triage process and patients' experiences in the ED. Data underwent thematic analysis using the Interpretative Phenomenological Analysis (IPA) approach.ResultsThe ED of TUTH offered comprehensive triage services with clear delineation for the severity of febrile patients in line with the World Health Organization (WHO) guidelines. Nonetheless, challenges and constraints were identified. In the ED, evenings were generally the busiest period, and the triage process was not thorough during night shifts. Perception of triage was limited among patients and variable among HCWs. Digitalizing recordings of patient information including payment was deemed necessary for effective management of patients' waiting times at the triage station. High patient throughput added pressure on HCWs and had a potential influence on the delivery of services. Availability of medical equipment and space were also identified as challenges, with patients sometimes compelled to share beds. There were constraints related to waste disposal, hygiene, cleanliness, and the availability and maintenance of washrooms. Febrile patients experienced delays in receiving timely consultations and laboratory investigation reports, which affected their rapid diagnosis and discharge; nonetheless, patients were satisfied with the overall healthcare services received in the ED.ConclusionsImproving current triage management requires resource organization, including optimizing the waiting time of patients through a digitalized system. Urgent priorities involve upgrading visitor facilities, patient consultations, laboratory investigations, hygiene, and sanitation. HCWs' recommendations to resource the ED with more equipment, space, and beds and a dedicated triage officer to ensure 24-hour service, together with training and incentives, warrant further attention.
\n \n\n \n \nArtemisinin combination therapy remains effective for the treatment of falciparum malaria. However, Plasmodium falciparum can escape the effects of artemisinin by arresting their growth. The growth-arrested parasites cannot be distinguished from nonviable parasites with standard microscopy techniques due to their morphological similarities. Here, we demonstrated the efficacy of a new laboratory assay that is compatible with the artemisinin susceptibility test. As a result of the differential cell permeabilities of two DNA-binding fluorophores, growth-arrested P. falciparum can be distinguished from parasites killed by artemisinin, since the latter lose cell membrane permeability. This fluorescence-based assay increased the sensitivity and specificity of the ring survival assay in the assessment of artemisinin susceptibility. When combined with a third fluorophore-conjugated anti-human leukocyte antibody, this trio fluorophore assay became more useful in identifying growth-arrested parasites in mock human blood samples. This novel assay is a simple and rapid technique for monitoring artemisinin resistance with greater sensitivity and accuracy compared with morphology-based observations under a light microscope.
\n \n\n \n \nBackground: The Global Programme to Eliminate Lymphatic Filariasis (GPELF) aims to reduce and maintain infection levels through mass drug administration (MDA), but there is evidence of ongoing transmission post-MDA in areas where Culex mosquitoes are the main transmission vector, suggesting that a more stringent criteria is required for MDA decision-making in these settings.\nMethods: We use a transmission model to investigate how a lower prevalence threshold (<1% antigenemia (Ag) prevalence compared to <2% Ag prevalence) for MDA decision-making would impact the probability of local elimination, health outcomes, the number of MDA rounds including restarts, and programme costs associated with MDA and surveys across different scenarios. To determine the cost-effectiveness of switching to a lower threshold, we simulated 65% and 80% MDA coverage of the total population for different willingness to pay per DALY averted: India (US dollars 446.07), Tanzania (US dollars 389.83) and Haiti (US dollars 219.84).\nResults: Our results suggest that with a lower Ag threshold, there is a small proportion of simulations where extra rounds are required to reach the target, but this also reduces the need to restart MDA later in the programme. For 80% coverage, the lower threshold is cost-effective across all baseline prevalences for India, Tanzania and Haiti. For 65% MDA coverage, the lower threshold is not cost-effective due to additional MDA rounds, although it increases the probability of local elimination. Valuing the benefits of elimination to align with the GPELF goals, we find that a willingness to pay per capita government expenditure of approximately $1000 - $4000 for 1% increase in the probability of local elimination would be required to make a lower threshold cost-effective.\nConclusions: Lower Ag thresholds for stopping MDAs generally mean a higher probability of local elimination, reducing long-term costs and health impacts. However, they may also lead to an increased number of MDA rounds required to reach the lower threshold and, therefore, increased short-term costs. Collectively, our analyses highlight that lower target Ag thresholds have the potential to assist programmes in achieving lymphatic filariasis goals.
\n \n\n \n \nBackgroundPrimaquine is an 8-aminoquinoline antimalarial. It is the only widely available treatment to prevent relapses of Plasmodium vivax malaria. The 8-aminoquinolines cause dose-dependent haemolysis in glucose-6-phosphate dehydrogenase deficiency (G6PDd). G6PDd is common in malaria endemic areas but testing is often not available. As a consequence primaquine is underused.MethodsWe conducted an adaptive pharmacometric study to characterise the relationship between primaquine dose and haemolysis in G6PDd. The aim was to explore shorter and safer primaquine radical cure regimens compared to the currently recommended 8-weekly regimen (0.75 mg/kg once weekly), potentially obviating the need for G6PD testing. Hemizygous G6PDd healthy adult Thai and Burmese male volunteers were admitted to the Hospital for Tropical Diseases in Bangkok. In Part 1, volunteers were given ascending dose primaquine regimens whereby daily doses were increased from 7.5 mg up to 45 mg over 15-20 days. In Part 2 conducted at least 6 months later, a single primaquine 45 mg dose was given.Results24 volunteers were enrolled in Part 1, and 16 in Part 2 (13 participated in both studies). In three volunteers, the ascending dose regimen was stopped because of haemolysis (n=1) and asymptomatic increases in transaminases (n=2; one was hepatitis E positive). Otherwise the ascending regimens were well tolerated with no drug-related serious adverse events. In Part 1, the median haemoglobin concentration decline was 3.7 g/dL (range: 2.1-5.9; relative decline of 26% [range: 15-40%]). Primaquine doses up to 0.87 mg/kg/day were tolerated subsequently without clinically significant further falls in haemoglobin. In Part 2, the median haemoglobin concentration decline was 1.7 g/dL (range 0.9-4.1; relative fall of 12% [range: 7-30% decrease]). The ascending dose primaquine regimens gave seven times more drug but resulted in only double the haemoglobin decline.ConclusionsIn patients with Southeast Asian G6PDd variants, full radical cure treatment can be given in under 3 weeks compared with the current 8-week regimen.FundingMedical Research Council of the United Kingdom (MR/R015252/1) and Wellcome (093956/Z/10/C, 223253/Z/21/Z).Clinical trial numberThai Clinical Trial Registry: TCTR20170830002 and TCTR20220317004.
\n \n\n \n \nBlood group O is associated with protection against severe malaria and reduced size and stability of P. falciparum-host red blood cell (RBC) rosettes compared to non-O blood groups. Whether the non-O blood groups encoded by the specific ABO genotypes AO, BO, AA, BB and AB differ in their associations with severe malaria and rosetting is unknown. The A and B antigens are host RBC receptors for rosetting, hence we hypothesized that the higher levels of A and/or B antigen on RBCs from AA, BB and AB genotypes compared to AO/BO genotypes could lead to larger rosettes, increased microvascular obstruction and higher risk of malaria pathology. We used a case-control study of Kenyan children and in vitro adhesion assays to test the hypothesis that \"double dose\" non-O genotypes (AA, BB, AB) are associated with increased risk of severe malaria and larger rosettes than \"single dose\" heterozygotes (AO, BO). In the case-control study, compared to OO, the double dose genotypes consistently had higher odds ratios (OR) for severe malaria than single dose genotypes, with AB (OR 1.93) and AO (OR 1.27) showing most marked difference (p = 0.02, Wald test). In vitro experiments with blood group A-preferring P. falciparum parasites showed that significantly larger rosettes were formed with AA and AB host RBCs compared to OO, whereas AO and BO genotypes rosettes were indistinguishable from OO. Overall, the data show that ABO genotype influences P. falciparum rosetting and support the hypothesis that double dose non-O genotypes confer a greater risk of severe malaria than AO/BO heterozygosity.
\n \n\n \n \nBackgroundWe sought to estimate SARS-CoV-2 antibody seroprevalence within representative samples of the Kenyan population during the third year of the COVID-19 pandemic and the second year of COVID-19 vaccine use.MethodsWe conducted cross-sectional serosurveys among randomly selected, age-stratified samples of Health and Demographic Surveillance System (HDSS) residents in Kilifi and Nairobi. Anti-spike (anti-S) immunoglobulin G (IgG) serostatus was measured using a validated in-house ELISA and antibody concentrations estimated with reference to the WHO International Standard for anti-SARS-CoV-2 immunoglobulin.ResultsHDSS residents were sampled in February-June 2022 (Kilifi HDSS N\u00a0=\u2009852; Nairobi Urban HDSS N\u00a0=\u2009851) and in August-December 2022 (N\u00a0=\u2009850 for both sites). Population-weighted coverage for \u22651 doses of COVID-19 vaccine were 11.1% (9.1-13.2%) among Kilifi HDSS residents by November 2022 and 34.2% (30.7-37.6%) among Nairobi Urban HDSS residents by December 2022. Population-weighted anti-S IgG seroprevalence among Kilifi HDSS residents increased from 69.1% (65.8-72.3%) by May 2022 to 77.4% (74.4-80.2%) by November 2022. Within the Nairobi Urban HDSS, seroprevalence by June 2022 was 88.5% (86.1-90.6%), comparable with seroprevalence by December 2022 (92.2%; 90.2-93.9%). For both surveys, seroprevalence was significantly lower among Kilifi HDSS residents than among Nairobi Urban HDSS residents, as were antibody concentrations (p\u00a0ConclusionMore than 70% of Kilifi residents and 90% of Nairobi residents were seropositive for anti-S IgG by the end of 2022. There is a potential immunity gap in rural Kenya; implementation of interventions to improve COVID-19 vaccine uptake among sub-groups at increased risk of severe COVID-19 in rural settings is recommended.
\n \n\n \n \nBackgroundSingle low-dose primaquine (SLDPQ) effectively blocks the transmission of Plasmodium falciparum malaria, but anxiety remains regarding its haemolytic potential in patients with glucose-6-phopshate dehydrogenase (G6PD) deficiency. We, therefore, examined the independent effects of several factors on haemoglobin (Hb) dynamics in falciparum-infected children with a particular interest in SLDPQ and G6PD status.MethodsThis randomised, double-blind, placebo-controlled, safety trial was conducted in Congolese and Ugandan children aged 6\u00a0months-11\u00a0years with acute uncomplicated P. falciparum and day (D) 0 Hbs\u2009\u2265\u20096\u00a0g/dL who were treated with age-dosed SLDPQ/placebo and weight-dosed artemether lumefantrine (AL) or dihydroartemisinin piperaquine (DHAPP). Genotyping defined G6PD (G6PD c.202T allele), haemoglobin S (HbS), and \u03b1-thalassaemia status. Multivariable linear and logistic regression assessed factor independence for continuous Hb parameters and Hb recovery (D42 Hb\u2009>\u2009D0 Hb), respectively.ResultsOne thousand one hundred thirty-seven children, whose median age was 5\u00a0years, were randomised to receive: AL\u2009+\u2009SLDPQ (n\u2009=\u2009286), AL\u2009+\u2009placebo (286), DHAPP\u2009+\u2009SLDPQ (283), and DHAPP\u2009+\u2009placebo (282). By G6PD status, 284 were G6PD deficient (239 hemizygous males, 45 homozygous females), 119 were heterozygous females, 418 and 299 were normal males and females, respectively, and 17 were of unknown status. The mean D0 Hb was 10.6 (SD 1.6) g/dL and was lower in younger children with longer illnesses, lower mid-upper arm circumferences, splenomegaly, and \u03b1-thalassaemia trait, who were either G6PDd or heterozygous females. The initial fractional fall in Hb was greater in younger children with higher D0 Hbs and D0 parasitaemias and longer illnesses but less in sickle cell trait. Older G6PDd children with lower starting Hbs and greater factional falls were more likely to achieve Hb recovery, whilst lower D42 Hb concentrations were associated with younger G6PD normal children with lower fractional falls,\u00a0sickle cell disease, \u03b1-thalassaemia silent carrier and trait, and late treatment failures. Ten blood transfusions were given in the first week (5 SLDPQ, 5 placebo).ConclusionsIn these falciparum-infected African children, posttreatment Hb changes were unaffected by SLDPQ, and G6PDd patients had favourable posttreatment Hb changes and a higher probability of Hb recovery. These reassuring findings support SLDPQ deployment without G6PD screening in Africa.Trial registrationThe trial is registered at ISRCTN 11594437.
\n \n\n \n \nAbstractHospital-based transmission had a dominant role in Middle East respiratory syndrome coronavirus (MERS-CoV) and severe acute respiratory syndrome coronavirus (SARS-CoV) epidemics1,2, but large-scale studies of its role in the SARS-CoV-2 pandemic are lacking. Such transmission risks spreading the virus to the most vulnerable individuals and can have wider-scale impacts through hospital\u2013community interactions. Using data from acute hospitals in England, we quantify within-hospital transmission, evaluate likely pathways of spread and factors associated with heightened transmission risk, and explore the wider dynamical consequences. We estimate that between June 2020 and March 2021 between 95,000 and 167,000 inpatients acquired SARS-CoV-2 in hospitals (1% to 2% of all hospital admissions in this period). Analysis of time series data provided evidence that patients who themselves acquired SARS-CoV-2 infection in hospital were the main sources of transmission to other patients. Increased transmission to inpatients was associated with hospitals having fewer single rooms and lower heated volume per bed. Moreover, we show that reducing hospital transmission could substantially enhance the efficiency of punctuated lockdown measures in suppressing community transmission. These findings reveal the previously unrecognized scale of hospital transmission, have direct implications for targeting of hospital control measures and highlight the need to design hospitals better equipped to limit the transmission of future high-consequence pathogens.
\n \n\n \n \nWe investigate the subsurface composition of turbid materials at the micro scale by means of a portable non-invasive technique, micro-spatially offset Raman spectroscopy (micro-SORS), combined with shifted excitation Raman difference spectroscopy (SERDS). This combination enables the microscale layer analysis and allows to deal effectively with highly fluorescing samples as well as ambient light, all in a form of an in-house portable prototype device optimised for applications in heritage science. The instrument comprises ability to simultaneously collect multiple spectra by means of an optical fibre bundle, thus reducing the dead time and simplifying the ease of deployment of the technique. The performance of the synergy between micro-SORS and 785 nm SERDS dual-wavelength diode laser is demonstrated on a stratified mock-up painting samples including highly fluorescing painted layers. This instrumental approach could be ground-breaking in heritage science, due to the largely unmet need of analysing the molecular composition of subsurface of artworks non-invasively and in situ, and in the presence of fluorescent background and ambient light. Moreover, many other fields are expected to benefit from this technological advancement such as solar energy, forensic and food analytical areas.
\n \n\n \n \nThis study explores the possibility of using microspatially offset Raman spectroscopy (micro-SORS) imaging to reconstruct noninvasively letters and figures hidden by opaque layers. Micro-SORS experiments were conducted on mockup samples that mimic real situations encountered in the cultural heritage field, such as sealed letters with inaccessible text and original documents. Subsurface images were obtained using both the characteristic Raman bands of the hidden compounds and their different optical properties from the remaining matrix. In the latter case, contrast obtained through observing a difference in the overall spectral intensity and fluorescence profile rather than any specific Raman bands were used to track the images within the hidden layer. This approach opens new prospects for the use of micro-SORS in heritage science, with applications in the field that include the study of objects covered by opaque overlayers not only through their Raman signatures but also through differences in their optical properties (e.g., fluorescence emission, absorption).
\n \n\n \n \nBackground: Growth faltering is well-recognized during acute childhood illness and growth acceleration during convalescence, with or without nutritional therapy, may occur. However, there are limited recent data on growth after hospitalization in low- and middle-income countries. Methods: We evaluated growth following hospitalization among children aged 2\u201323 months in sub-Saharan Africa and South Asia. Between November 2016 and January 2019, children were recruited at hospital admission and classified as: not-wasted (NW), moderately-wasted (MW), severely-wasted (SW), or having nutritional oedema (NO). We describe earlier (discharge to 45-days) and later (45- to 180-days) changes in length-for-age [LAZ], weight-for-age [WAZ], mid-upper arm circumference [MUACZ], weight-for-length [WLZ] z-scores, and clinical, nutritional, and socioeconomic correlates. Findings: We included 2472 children who survived to 180-days post-discharge: NW, 960 (39%); MW, 572 (23%); SW, 682 (28%); and NO, 258 (10%). During 180-days, LAZ decreased in NW (\u22120.27 [\u22120.36, \u22120.19]) and MW (\u22120.23 [\u22120.34, \u22120.11]). However, all groups increased WAZ (NW, 0.21 [95% CI: 0.11, 0.32]; MW, 0.57 [0.44, 0.71]; SW, 1.0 [0.88, 1.1] and NO, 1.3 [1.1, 1.5]) with greatest gains in the first 45-days. Of children underweight (
\n \n\n \n \nThe COVID CIRCLE initiative Research Project Tracker by UKCDR and GloPID-R and associated living mapping review (LMR) showed the importance of sharing and analysing data on research at the point of funding to improve coordination during a pandemic. This approach can also help with research preparedness for outbreaks and hence our new programme the Pandemic Preparedness: Analytical Capacity and Funding Tracking Programme (Pandemic PACT) has been established. The LMR described in this protocol will provide an open, accessible, near-real-time overview of the funding landscape for a wide range of infectious disease and pandemic preparedness research using a rich database. The underpinning database will feed into an online funding tracking dashboard, with visualisations and advanced exploration features. The database is the expansion of the previous UKCDR and GloPID-R COVID-19 Research Project database with addition of the priority diseases from the WHO Blueprint list plus initial additions of pandemic influenza, mpox and plague. We have captured as much information as possible about grants and their outputs and ensured that our metadata and database is aligned to the FAIR (findable, accessible, interoperable and reproducible) principles by design. Prior to the public release, the data are processed by researchers and supported with artificial intelligence to create a curated data product. We anticipate that the database, this associated LMR and online funding tracking dashboard, will be a useful resource for funders, policy makers and researchers. In the future, our work will inform a more coordinated approach to research funding by providing evidence and data, including identification of gaps in funding allocation with a particular focus on low- and middle-income countries.
\n \n\n \n \nAbstract\nThe emergence and spread of artemisinin resistant Plasmodium falciparum, first in the Greater Mekong Subregion (GMS), and now in East Africa, is a major threat to global malaria eliminations ambitions. To investigate the artemisinin resistance mechanism, transcriptome analysis was conducted of 577 P. falciparum isolates collected in the GMS between 2016\u20132018. A specific artemisinin resistance-associated transcriptional profile was identified that involves a broad but discrete set of biological functions related to proteotoxic stress, host cytoplasm remodeling and REDOX metabolism. The artemisinin resistance-associated transcriptional profile evolved from initial transcriptional responses of susceptible parasites to artemisinin. The genetic basis for this adapted response is likely to be complex.
\n \n\n \n \nAbstract\nSpecialised pre-trained language models are becoming more frequent in Natural language Processing (NLP) since they can potentially outperform models trained on generic texts. BioBERT (Sanh et al., Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter. arXiv preprint arXiv: 1910.01108, 2019) and BioClinicalBERT (Alsentzer et al., Publicly available clinical bert embeddings. In Proceedings of the 2nd Clinical Natural Language Processing Workshop, pp. 72\u201378, 2019) are two examples of such models that have shown promise in medical NLP tasks. Many of these models are overparametrised and resource-intensive, but thanks to techniques like knowledge distillation, it is possible to create smaller versions that perform almost as well as their larger counterparts. In this work, we specifically focus on development of compact language models for processing clinical texts (i.e. progress notes, discharge summaries, etc). We developed a number of efficient lightweight clinical transformers using knowledge distillation and continual learning, with the number of parameters ranging from \n\n\n\n$15$\n\n\n million to \n\n\n\n$65$\n\n\n million. These models performed comparably to larger models such as BioBERT and ClinicalBioBERT and significantly outperformed other compact models trained on general or biomedical data. Our extensive evaluation was done across several standard datasets and covered a wide range of clinical text-mining tasks, including natural language inference, relation extraction, named entity recognition and sequence classification. To our knowledge, this is the first comprehensive study specifically focused on creating efficient and compact transformers for clinical NLP tasks. The models and code used in this study can be found on our Huggingface profile at https://huggingface.co/nlpie and Github page at https://github.com/nlpie-research/Lightweight-Clinical-Transformers, respectively, promoting reproducibility of our results.
\n \n\n \n \n