CsFEVER and CTKFacts: acquiring Czech data for fact verification

In this paper, we examine several methods of acquiring Czech data for automated fact-checking, which is a task commonly modeled as a classification of textual claim veracity w.r.t. a corpus of trusted ground truths. We attempt to collect sets of data in form of a factual claim, evidence within the ground truth corpus, and its veracity label (supported, refuted or not enough info). As a first attempt, we generate a Czech version of the large-scale FEVER dataset built on top of Wikipedia corpus. We take a hybrid approach of machine translation and document alignment; the approach and the tools we provide can be easily applied to other languages. We discuss its weaknesses, propose a future strategy for their mitigation and publish the 127k resulting translations, as well as a version of such dataset reliably applicable for the Natural Language Inference task-the CsFEVER-NLI. Furthermore, we collect a novel dataset of 3,097 claims, which is annotated using the corpus of 2.2 M articles of Czech News Agency. We present an extended dataset annotation methodology based on the FEVER approach, and, as the underlying corpus is proprietary, we also publish a standalone version of the dataset for the task of Natural Language Inference we call CTKFactsNLI. We analyze both acquired datasets for spurious cues-annotation patterns leading to model overfitting. CTKFacts is further examined for inter-annotator agreement, thoroughly cleaned, and a typology of common annotator errors is extracted. Finally, we provide baseline models for all stages of the fact-checking pipeline and publish the NLI datasets, as well as our annotation platform and other experimental data.

Non-contact infrared thermometers compared with current approaches in primary care for children aged 5 years and under: a method comparison study

BACKGROUND

Current options for temperature measurement in children presenting to primary care include either electronic axillary or infrared tympanic thermometers. Non-contact infrared thermometers could reduce both the distress of the child and the risk of cross-infection.

OBJECTIVES

The objective of this study was to compare the use of non-contact thermometers with the use of electronic axillary and infrared tympanic thermometers in children presenting to primary care.

DESIGN

Method comparison study with a nested qualitative study.

SETTING

Primary care in Oxfordshire.

PARTICIPANTS

Children aged ≤ 5 years attending with an acute illness.

INTERVENTIONS

Two types of non-contact infrared thermometers [i.e. Thermofocus (Tecnimed, Varese, Italy) and Firhealth (Firhealth, Shenzhen, China)] were compared with an electronic axillary thermometer and an infrared tympanic thermometer.

MAIN OUTCOME MEASURES

The primary outcome was agreement between the Thermofocus non-contact infrared thermometer and the axillary thermometer. Secondary outcomes included agreement between all other sets of thermometers, diagnostic accuracy for detecting fever, parental and child ratings of acceptability and discomfort, and themes arising from our qualitative interviews with parents.

RESULTS

A total of 401 children (203 boys) were recruited, with a median age of 1.6 years (interquartile range 0.79-3.38 years). The readings of the Thermofocus non-contact infrared thermometer differed from those of the axillary thermometer by -0.14 °C (95% confidence interval -0.21 to -0.06 °C) on average with the lower limit of agreement being -1.57 °C (95% confidence interval -1.69 to -1.44 °C) and the upper limit being 1.29 °C (95% confidence interval 1.16 to 1.42 °C). The readings of the Firhealth non-contact infrared thermometer differed from those of the axillary thermometer by -0.16 °C (95% confidence interval -0.23 to -0.09 °C) on average, with the lower limit of agreement being -1.54 °C (95% confidence interval -1.66 to -1.41 °C) and the upper limit being 1.22 °C (95% confidence interval 1.10 to 1.34 °C). The difference between the first and second readings of the Thermofocus was -0.04 °C (95% confidence interval -0.07 to -0.01 °C); the lower limit was -0.56 °C (95% confidence interval -0.60 to -0.51 °C) and the upper limit was 0.47 °C (95% confidence interval 0.43 to 0.52 °C). The difference between the first and second readings of the Firhealth thermometer was 0.01 °C (95% confidence interval -0.02 to 0.04 °C); the lower limit was -0.60 °C (95% confidence interval -0.65 to -0.54 °C) and the upper limit was 0.61 °C (95% confidence interval 0.56 to 0.67 °C). Sensitivity and specificity for the Thermofocus non-contact infrared thermometer were 66.7% (95% confidence interval 38.4% to 88.2%) and 98.0% (95% confidence interval 96.0% to 99.2%), respectively. For the Firhealth non-contact infrared thermometer, sensitivity was 12.5% (95% confidence interval 1.6% to 38.3%) and specificity was 99.4% (95% confidence interval 98.0% to 99.9%). The majority of parents found all methods to be acceptable, although discomfort ratings were highest for the axillary thermometer. The non-contact thermometers required fewer readings than the comparator thermometers.

LIMITATIONS

A method comparison study does not compare new methods against a reference standard, which in this case would be central thermometry requiring the placement of a central line, which is not feasible or acceptable in primary care. Electronic axillary and infrared tympanic thermometers have been found to have moderate agreement themselves with central temperature measurements.

CONCLUSIONS

The 95% limits of agreement are > 1 °C for both non-contact infrared thermometers compared with electronic axillary and infrared tympanic thermometers, which could affect clinical decision-making. Sensitivity for fever was low to moderate for both non-contact thermometers.

FUTURE WORK

Better methods for peripheral temperature measurement that agree well with central thermometry are needed.

TRIAL REGISTRATION

Current Controlled Trials ISRCTN15413321.

FUNDING

This project was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 24, No. 53. See the NIHR Journals Library website for further project information.

The production of fever by influenzal viruses; effect of receptor-destroying substances

The effect of treating rabbits with materials which destroy the cell receptors for influenzal viruses upon the ability of these animals to respond with fever to injection of the PR8 and Lee strains of influenza virus and Newcastle disease virus (NDV) is described. In general, both cholera vibrio and Cl. welchii filtrates produced diminution of febrile responses. The effect of sodium periodate upon the pyrogenic reaction was not significant. Near-lethal amounts of these materials were necessary to demonstrate their protective effects against virus challenge. In order to rule out general debility as a factor in lessening the fever, it was shown that the ability of animals to respond to the pyrogenic effect of typhoid vaccine was unimpaired by injection of receptor-destroying substances. The substances tested were more effective in abolishing the febrile response to PR8 virus than to Lee virus or NDV. This finding is compatible with previous studies of the protective effect exerted by homologous and heterologous viruses. These findings give support to the hypothesis that union of virus and host receptor substance plays a part in the production of fever by these viruses.

Marrow biopsy in the diagnosis of pyrexia of undetermined origin

Marrow from 62 cases of pyrexia of undetermined origin was examined cytologically, histologically, and bacteriologically. Diagnostic findings were present in five cases (8%), but all these patients presented definite clinical or haematological indications for marrow biopsy. It is concluded that marrow examination is useful in the diagnosis of pyrexia of undetermined origin but that cases for study must be selected in the light of general principles. Fever alone, even though prolonged and undiagnosed, is not an indication for marrow biopsy.

Mechanisms of endotoxin tolerance. I. Relationship between tolerance and reticuloendothelial system phagocytic activity in the rabbit

Pyrogenic tolerance following 7 daily intravenous injections of 2.0 µg/kg E. coli endotoxin in albino rabbits was associated with significant increases in RES phagocytic activity as measured with colloidal carbon. Nevertheless, 4 hours after RES blockade with thorotrast (3 ml/kg), the tolerant rabbits exhibited significantly lower fever indices following intravenous endotoxin challenge than did non-tolerant control animals despite comparably depressed capacities to clear carbon from the blood. Moreover, plasma from rabbits tolerant to endotoxin induced significant tolerance in normal rabbits prepared by thorotrast blockade without enhancing the depressed carbon clearance. This passive protection extended to heterologous endotoxins. Analysis of the data indicates that RES blockade does not abolish tolerance; rather blockade resets the reactivity to endotoxin in the normal and tolerant animal, rendering both exquisitely reactive, but permitting retention of the major portion of tolerance. Apparently the tolerant animal possesses a dual endotoxin defense system. One system is abolished by thorotrast; the other is in part humoral, accounts for the greater portion of tolerance, and is thorotrast-resistant. The nature of the humoral component is not defined but is consistent with that of an opsonin with high endotoxin specificity.

Hyperpyrexia as a contributory factor in the toxicity of amphetamine to aggregated mice

A rise of body temperature into a range lethal to mice preceded death in groups of mice injected with amphetamine sulphate. At dose levels from 8.8 to 66.7 mg/kg, mortality was associated with the extent of rise in body temperature of the mice, irrespective of the actual dose administered. Isolated mice given comparable doses of amphetamine also showed a marked increase in body temperature. However, except in a very few cases, it did not rise into the range found to be lethal. Amphetamine was more toxic to isolated mice subjected to foot-shock than to isolated mice housed under normal conditions. This has also been shown to be related to the extent of rise in body temperature of the mice. The effect of a number of substances on grouped amphetamine toxicity was investigated. Chlorpromazine and phenoxybenzamine partially antagonized the sharp rise in temperature following the administration of amphetamine, and also significantly reduced mortality. Calcium acetylsalicylate was without effect on the rise of body temperature or on mortality. Both the hypothermic compound 4-methyl-5-(beta-chlorethyl)-thiazole (S.C.T.Z.) and L-thyroxine sodium potentiated the rise in temperature and caused a significant increase in mortality.

Quantitative features of the febrile response to leucocytic pyrogen

Although the absolute febrile responses of trained individual rabbits injected intravenously with small to moderate doses of leucocytic pyrogen vary over an appreciable range, the relative responses of each rabbit to changes in dosage are satisfactorily reproducible. The quantitative dose-response relationship is characterized by a hyperthermic ceiling at which the intensity of the febrile reaction is relatively constant over a wide dosage range. Only at lower dose levels, where the dose-response curve is reasonably steep, is the magnitude of the fever produced proportional to the amount of pyrogen injected. When sufficiently large doses of LP are injected, the hyperthermic ceiling is exceeded. The fevers thus induced are biphasic in character and, in this way, resemble the usual response to bacterial endotoxin. Similar biphasic fevers result from continuous infusions of relatively low concentrations of LP at a constant rate. Repeated intermittent injections of moderate doses of LP likewise cause prolonged biphasic fevers, but, once the fever has become established, the reaction to each individual injection becomes markedly depressed. When large doses of LP are injected at daily intervals, the characteristic biphasic response occurs only following the first injection. Thereafter a state of tolerance intervenes in which the late secondary rise in temperature fails to occur. This form of tolerance lasts as long as the daily injections are continued but subsides within a few days after the injections are stopped. During the transient tolerance the rabbit's responsiveness to small doses of LP (in the sensitive range of the dose response curve) is depressed. In addition, the amount of endogenous pyrogen mobilized from the tissues by a large dose of LP is not as great as that generated in a normal rabbit. The relations of these findings to biphasic fevers, tolerance, and the accuracy of the conventional method of pyrogen assay are briefly discussed.

FEVER PRODUCED BY MICROINJECTION OF TYPHOID VACCINE INTO HYPOTHALAMUS OF CATS

In unanesthetized cats, 1 or 2.5 µliters typhoid vaccine were injected into several areas of the hypothalamus, the lateral cerebral ventricle, and other areas of the brain. The concentrations of the pyrogen used ranged from 1/8 to 1/16,000 of the dose required to produce fever by the intravenous route. Fever was obtained in all experiments in which the vaccine was applied to the anterior hypothalamus or to the lateral cerebral ventricle. No significant rise in temperature occurred following injections into other brain areas unless extremely high doses were employed. The short latency and the rapid rise in temperature following the injection of typhoid vaccine into the anterior hypothalamus suggest that pyrogens act on the cells of this diencephalic structure when fever is produced.

MODIFICATION OF HOST RESPONSES TO BACTERIAL ENDOTOXINS. I. SPECIFICITY OF PYROGENIC TOLERANCE AND THE ROLE OF HYPERSENSITIVITY IN PYROGENICITY, LETHALITY, AND SKIN REACTIVITY

Evidence is presented suggesting that the apparent non-specificity of pyrogenic tolerance observed with Gram-negative bacterial endotoxins is due to related antigenic determinants associated with the macromolecular toxins. This is based on results obtained in rabbits from pyrogenic cross-tolerance tests with selected endotoxins. In these tests, purified endotoxins from Escherichia coli (COO8) and Chromobacterium violaceum (CV) gave results one might expect with non-reciprocal cross-reacting antigens in classical immune systems. Additional evidence for an immune mechanism in tolerance is suggested by the highly significant anamnestic response observed. Lipid A, a toxic derivative of the purified COO8 endotoxin, failed to induce pyrogenic tolerance against the parent toxin. These results are explained by assuming that endotoxins have two interdependent activities associated with different portions of the macromolecule; one is assumed to be responsible for the primary toxicity, and the other is involved in secondary toxicity. The latter is dependent on the hypersensitive state of the host. Additional evidence for the role of hypersensitivity in secondary toxicity is based on the observation that adult rabbits are highly sensitive to the pyrogenic, lethal, and skin-reacting activities of endotoxin in contrast to young animals which are more resistant to all of these attributes of toxicity. In adults, the host responses to pyrogenicity, lethality, and skin reactivity could be partially inhibited by the early exposure of the animals to massive doses of endotoxin equivalent to a LD₅₀. The pyrogenic tolerance shown in these animals was specific indicating that the inhibition of the hypersusceptibility to endotoxin involved an immunological mechanism. A mechanism of endotoxin tolerance is proposed and discussed based on the induction of specific antibodies capable of assisting the RES in the clearance and destruction of endotoxin. It is suggested that the present inconsistencies relative to the chemical nature and biological activities of endotoxins might be explained on the basis of these two activities and the failure to recognize the importance of the immunological state of the host in which the toxins are tested.

TOLERANCE OF DOG LUNG TO HYPERTHERMIC PERFUSION

Hyperthermia of the left lung of the dog was produced by pulmonary artery perfusion in 29 animals. It was found that the lung tolerates tissue temperatures up to 43 C for periods as long as 60 min without significant permanent disturbance of structure or function as measured by blood gas analysis, angiography, and contralateral pulmonary artery ligation. Elevation of the lung temperature above 43 C for periods longer than 10–12 min resulted in partial or total destruction of pulmonary tissue. Data are presented regarding perfusion techniques, factors affecting lung tolerance to hyperthermia, and possible areas for future study.

THE EFFECT OF PHAGOCYTOSIS ON THE RELEASE OF ENDOGENOUS PYROGEN BY POLYMORPHONUCLEAR LEUKOCYTES

1. Phagocytosis promotes the release of endogenous pyrogen from polymorphonuclear leucocytes. 2. The release of pyrogen, though initiated by the phagocytic event, is not synchronous with it. 3. The postphagocytic release mechanism is not inhibited by sodium fluoride and, therefore, appears not to require continued production of energy by the cell. 4. The release process, on the other hand, is inhibited by arsenite, suggesting the participation of one or more sulfhydryl-dependent enzymes in the over-all reaction. 5. Particle for particle, the ingestion of heat-killed rough pneumococci causes the release of approximately 100 times as much pyrogen as the ingestion of polystyrene beads of the same size. 6. The pyrogen release mechanism of polymorphonuclear leucocytes separated directly from blood, unlike that of granulocytes in acute inflammatory exudates, is not readily activated by incubation of the cells in K-free saline. Despite this difference, both blood and exudate leucocytes following phagocytosis release large amounts of pyrogen, even in the presence of K(+). The fact that the postphagocytic reaction is uninhibited by the concentrations of K(+) which are present in plasma and extracellular fluids, suggests that this mechanism of pyrogen release may well operate in vivo. 7. As might be expected from the foregoing observations, the intravenous injection of a sufficiently large number of heat-killed pneumococci causes fever in the intact host. Intravenously injected polystyrene beads, on the other hand, are significantly less pyrogenic. Evidence is presented to support the conclusion that the fever in both instances is caused by pyrogen released from the circulating leucocytes which have phagocyted the injected particles. 8. The possible relationships of these findings to the pathogenesis of fevers caused by acute bacterial infections are discussed.

ELECTROLYTIC FACTORYS INFLUENCING THE RELEASE OF ENDOGENOUS PYROGEN FROM POLYMORPHONUCLEAR LEUKOCYTES

The metabolic reactions responsible for the release of endogenous pyrogen from rabbit granulocytes incubated in 0.15 M NaCl are specifically inhibited by the presence of K⁺ (and by related alkali metal ions, Rb⁺ and Cs⁺) in the medium. The inhibitory action of K⁺ apparently involves penetration of the cell membrane and is directly antagonized by the cardiac glycoside, ouabain. It is concluded, therefore, that the inhibition of pyrogen release by extracellular K⁺ is due to transport of K⁺ into the cell. Although the precise molecular mechanisms which are responsible for the release of pyrogen from granulocytes incubated in K-free saline have not been elucidated, further study of the process has revealed: (a) that it is preceded by the accumulation of pyrogen within the cell, (b) that it depends upon the catalytic action of one or more sulfhydryl-containing enzymes, (c) that it does not require energy, either from glycolysis or from reactions depending on molecular oxygen, and (d) that its inhibition by K⁺ and by arsenite is qualitatively similar to the depression caused by these same reagents on the release of other leucocytic proteins; i.e., lysozyme and aldolase.

DIFFERENTIAL EFFECTS OF HEAT ON SPLENIC AND RENAL VASCULAR BEDS

The effects on vascular resistance of increasing blood temperature, from hypothermic to hyperthermic levels, were evaluated on denervated perfused spleens and kidneys in the dog. Temperature of the perfusate (blood) was increased for variable intervals of time from the hypothermic level of 32 C to the hyperthermic level of 42 C. The calculated resistance in the splenic vascular bed increased, whereas the resistance decreased in the renal vascular bed. It is concluded that heat has a differential effect in the splenic and renal vascular bed, causing vasoconstriction in the spleen and vasodilation in the kidney.