A Review on Fact Extraction and Verification

We study the fact-checking problem, which aims to identify the veracity of a given claim. Specifically, we focus on the task of Fact Extraction and VERification (FEVER) and its accompanied dataset. The task consists of the subtasks of retrieving the relevant documents (and sentences) from Wikipedia and validating whether the information in the documents supports or refutes a given claim. This task is essential and can be the building block of applications such as fake news detection and medical claim verification. In this article, we aim at a better understanding of the challenges of the task by presenting the literature in a structured and comprehensive way. We describe the proposed methods by analyzing the technical perspectives of the different approaches and discussing the performance results on the FEVER dataset, which is the most well-studied and formally structured dataset on the fact extraction and verification task. We also conduct the largest experimental study to date on identifying beneficial loss functions for the sentence retrieval component. Our analysis indicates that sampling negative sentences is important for improving the performance and decreasing the computational complexity. Finally, we describe open issues and future challenges, and we motivate future research in the task.

Maternal and Neonatal Risk Factors Affecting the Occurrence of Neurodevelopmental Disorders: A Population-Based Nationwide Study

To investigate the maternal and neonatal risk factors related to pregnancy and birth affecting the occurrence of neurodevelopmental disorders to their children using the medical claim data for the whole population. The study was conducted on all the babies born in Korea from 2005 to 2009 based on data from the National Health Information Database. All birth records were tracked from birth to December 31, 2015. To analyze factors related to the mother, data related to the mother of the newborn were collected. Increased maternal age was found to increase the risk of cerebral palsy (adjusted odds ratio [aOR] = 1.46, 95% confidence interval [CI] [1.22, 1.75]) and autism spectrum disorder (aOR = 1.48, 95% CI [1.25, 1.76]), while lowering the risk of intellectual disability (aOR = 1.83, 95% CI [1.33, 2.53]) and speech and language impairment (aOR = 1.41, 95% CI [1.08, 1.83]) compared with the reference group aged 25 to 29 years old. The incidence affected by socioeconomic factors varied according to the types of disorders. Among various risk factors, prematurity or low birth weight, problems associated with amniotic fluid or amniotic membrane, preeclampsia or eclampsia, and cesarean section affect the incidence of neurodevelopmental disorders. To reduce the incidence or severity of neurodevelopmental disorders, a better understanding of the risk factors of neurodevelopmental disorders is important. The results of this study can be used as basic data to help such understanding.

The risk of laryngitis with herpes zoster infection: A nested case-control study using data from the Korean National Sample Cohort

Background Whether herpes zoster infection (HZI) affects laryngitis incidence remains unknown. Objective The purpose of this population-based retrospective study was to analyze the relationship between laryngitis and HZI using data from the Korean Health Insurance Review and Assessment Service—National Sample Cohort. Methods This study analyzed 1,197,093 medical claim codes from 2018. Patients with HZI (ICD-10: B02) were retrospectively identified. Laryngeal diseases were defined by ICD-10 codes for five subgroups: 1) malignant disease, 2) benign disease, 3) vocal cord palsy, 4) inflammatory disease, and 5) reflux disease. Results Among the Korean population older than 20 years, 12,809 experienced HZI. Subjects with HZI were more likely to be older (mean age: 51.54 years vs. 48.06 years, p <0.0001). The proportion of subjects with laryngeal disease was higher in those with HZI than in those without HZI (55.55% vs. 41.37%, p <0.0001). Laryngeal disease was significantly associated with HZI in multiple regression analysis (odds ratio (OR) = 1.77, 95% confidence interval: 1.71–1.84) after adjusting for age, sex, hypertension, diabetes, dyslipidemia, ischemic heart disease, cerebral stroke, and depression. Among laryngeal disease subgroups, inflammatory disease (OR = 1.05; 95% CI: 1.01–1.09) and reflux (OR = 1.20; 95% CI: 1.15–1.25) were associated with HZI. Conclusions HZI is independently associated with laryngitis. Results of this study have implications for etiological investigations and prevention strategies for laryngitis.

Drug Adherence and Persistence of Patients with Moderate to Severe Psoriasis Treated with Biologic Medications in a US Commercially Insured Population

<b><i>Background:</i></b> Adalimumab (ADA), certolizumab pegol (CER), etanercept (ETA), guselkumab (GUS), ixekizumab (IXE), secukinumab (SEC), and ustekinumab (UST) are biologic medications approved in the USA for the treatment of moderate to severe psoriasis. We examined drug adherence and persistence of patients with moderate to severe psoriasis who initiated these seven biologic medications. <b><i>Methods:</i></b> Adult patients with ≥1 pharmacy/medical claim for any of the seven psoriasis medications and ≥1 diagnosis of psoriasis in the previous 6 months between July 1, 2014 and June 30, 2019 were selected from the IBM MarketScan® Commercial Claims and Encounters Database. The index date was defined as the date of the first prescription fill. Patients were required to have continuous health plan enrollment during the 6 months prior to their index date and ≥9 months after. Patients were grouped into seven study cohorts based upon their index biologic medication. Adherence was measured using the proportion of days covered (PDC) and defined by a PDC ≥80%. Adherence and persistence with index biologic medications were examined during fixed follow-up periods of 3, 6, and 9 months, with a subpopulation analysis carried out among patients with 12 months of follow-up. <b><i>Results:</i></b> Among psoriasis patients with ≥9 months of continuous enrollment included in the study population, the number of those who initiated each biologic medication was 10,324 for ADA, 431 for CER, 3,092 for ETA, 821 for GUS, 1,766 for IXE, 4,132 for SEC, and 5,441 for UST. The mean age at the time of initiating biologic treatment was 46.9 years. During the 9-month follow-up period, the proportions of adherent patients (i.e., PDC ≥80%) were numerically higher among those treated with UST (59.9%) and GUS (56.9%), followed by those treated with SEC (46.1%), IXE (45.5%), ADA (44.7%), ETA (33.9%), and CER (22.0%). The proportions of patients who were persistent with their index biologic medication during the 9-month follow-up period were numerically higher among those treated with UST (70.1%) and GUS (67.8%), followed by those treated with IXE (47.3%), SEC (46.9%), ADA (28.7%), CER (14.8%), and ETA (10.7%). <b><i>Conclusions:</i></b> In this large healthcare claims database analysis of psoriasis patients treated with seven different biologic medications, adherence was numerically higher among those treated with UST or GUS. UST and GUS were also associated with numerically greater persistence.

MIHARI project, a preceding study of MID-NET, adverse event detection database of Ministry Health of Japan—Validation study of the signal detection of adverse events of drugs using export data from EMR and medical claim data

We studied the effectiveness of the direct data collection from electronic medical records (EMR) when it is used for monitoring adverse drug events and also detection of already known adverse events. In this study, medical claim data and SS-MIX2 standardized storage data were used to identify four diseases (diabetes, dyslipidemia, hyperthyroidism, and acute renal failure) and the validity of the outcome definitions was evaluated by calculating positive predictive values (PPV). The maximum positive predictive value (PPV) for diabetes based on medical claim data was 40.7% and that based on prescription data from SS-MIX2 Standardized Storage was 44.7%. The PPV for dyslipidemia was 50% or higher under either of the conditions. The PPV for hyperthyroidism based on disease name data alone was 20–30%, but exceeded 60% when prescription data was included in the evaluation. Acute renal failure was evaluated using information from medical records in addition to the data. The PPV for acute renal failure based on the data of disease names and laboratory examination results was slightly higher at 53.7% and increased to 80–90% when patients who previously had a high serum creatinine (Cre) level were excluded. When defining a disease, it is important to include the condition specific to the disease; furthermore, it is very useful if laboratory examination results are also included. Therefore, the inclusion of laboratory examination results in the definitions, as in the present study, was considered very useful for the analysis of multi-center SS-MIX2 standardized storage data.

The Effects of Previous Thyroid Disease on the Susceptibility to, Morbidity of, and Mortality Due to COVID-19: A Nationwide Cohort Study in South Korea

This study aimed to investigate the associations of the susceptibility to, morbidity of, and mortality due to coronavirus disease 2019 (COVID-19) with thyroid diseases. Korea National Health Insurance Database Coronavirus disease 2019 (NHID-COVID-19) medical claim code data from 2015 to 2020 were analyzed. A total of 8070 COVID-19 patients and 32,280 matched control participants were evaluated for histories of hypothyroidism, hyperthyroidism, Graves’ disease, thyroiditis, and autoimmune thyroiditis. The relationships of susceptibility to, morbidity of, and mortality due to COVID-19 with hypothyroidism, hyperthyroidism, Graves’ disease, thyroiditis, and autoimmune thyroiditis were analyzed using a conditional logistic regression. Hypothyroidism, hyperthyroidism, Graves’ disease, thyroiditis, and autoimmune thyroiditis were not associated with susceptibility to, morbidity of, or mortality due to COVID-19. Graves’ disease was related to higher odds of mortality due to COVID-19 in the adjusted model but the confidence interval (CI) was wide, probably due to the small number of deaths among patients with Graves’ disease (aOR = 11.43, 95% CI = 1.29–101.22, p = 0.029). Previous histories of hypothyroidism, hyperthyroidism, Graves’ disease, thyroiditis, and autoimmune thyroiditis were not related to susceptibility to COVID-19. In addition, prior histories of thyroid diseases were not related to increased risks of COVID-19-related morbidity and mortality.

The effects of income level on susceptibility to COVID-19 and COVID-19 morbidity/mortality: A nationwide cohort study in South Korea (Preprint)

BACKGROUND Low economic status has been highlighted as a factor affecting vulnerability to COVID-19 infection. OBJECTIVE This study aimed to investigate the association of income level with susceptibility to coronavirus disease 2019 (COVID-19) and COVID-19 morbidity and mortality. METHODS Using the Korea National Health Insurance COVID-19 Database cohort, medical claim data from 2015 through 2020 were collected. A total of 7,943 patients who were diagnosed with COVID-19 from 1 January 2020 to 4 June 2020 were included. A total of 118,914 participants had negative COVID-19 PCR tests. Income levels were classified by 20th percentiles based on 2019 Korean National Health Insurance premiums. The 20th percentile income levels were categorized into three groups (low, middle, and high). The relation of income level with susceptibility to COVID-19 and COVID-19 morbidity and mortality was analyzed using logistic regression analysis adjusted for age, sex, Charlson Comorbidity Index (CCI), and hypertension. Subgroup analyses were conducted according to age, sex, CCI score, and history of hypertension. RESULTS A high income level was related to lower odds of COVID-19 infection (adjusted odds ratio [aOR] = 0.79, 95% confidence interval [CI] = 0.75–0.83, P <.001). The negative association between income level and COVID-19 infection was maintained in all subgroups. Income level did not show an association with COVID-19 morbidity and mortality in the overall study population or in the subgroup analyses. CONCLUSIONS Patients with low income levels were susceptible to COVID-19 infection. However, there was no relation of COVID-19 morbidity and mortality with income level in the Korean population.

Healthcare Costs Associated with Complications in Patients with Type 2 Diabetes among 1.85 Million Adults in Beijing, China

We aimed to provide reliable regression estimates of expenditures associated with various complications in type 2 diabetics in China. In total, 1,859,039 type 2 diabetes patients with complications were obtained from the Beijing Medical Claim Data for Employees database from 2008 to 2016. We estimated costs for complications using a generalized estimating equation model adjusted for age, sex, and the incidence of various complications. The average total cost for diabetic patients with complications was 17.12 thousand RMB. Prescribed drugs accounted for 63.4% of costs. We observed a significant increase in costs in the first year after the onset of complications. Compared with costs before the incidence of complications, the additional costs per person in the first year and >1 year after the event would be 10,631.16 RMB and 1150.71 RMB for cardiovascular disease, 1017.62 RMB and 653.82 RMB for cerebrovascular disease, and 301.14 RMB and 624.00 RMB for kidney disease, respectively. The estimated coefficients for outpatient visits were relatively lower than those of inpatient visits. Complications in diabetics exert a significant impact on total healthcare costs in the first year of their onset and in subsequent years. Our estimates may assist policymakers in quantifying the economic burden of diabetes complications.

Profiles of 2-Drug Regimen in People Living with HIV-1 in A Real World Setting: A Large-Scale Medical Claim Database Analysis in Japan

Background Regimen simplification to 2-drug antiretroviral therapy (2-ART) may address potential tolerability issues, increase adherence, and reduce toxicity and potential drug-drug-interactions among people living with HIV-1 (PLWH). However, real-world treatment patterns and patient profiles associated with 2-ART are unclear. Methods This retrospective observational cohort study employed a large-scale medical claim database of Japanese hospitals to extract data on 4,293 PLWH aged ≥18 years with diagnosis of HIV and treated with any ART regimens between April 2008 and April 2019. A 2-ART cohort was compared with a 3-drug antiretroviral therapy (3-ART) cohort in terms of patient characteristics, comorbid conditions, and treatment patterns. Treatment switching rates were calculated for each cohort followed by sensitivity analysis to confirm the robustness of the findings.Results There were 94 patients identified in the 2-ART cohort. Compared to the standard 3-ART cohort (n=3,993), the 2-ART cohort was older (mean age 54.4 vs 43.4 years), with a lower proportion of males (87.2% vs 93.8%), higher Charlson Comorbidity Index (CCI) (mean score 6.9 vs 5.3), more co-medications (mean 8.3 vs 5.0), and a higher percentage of AIDS-defining conditions (66.0% vs 42.8%). The most common 2-ART were protease inhibitor (PI) + integrase strand transfer inhibitor (INSTI) and non-nucleoside reverse transcriptase inhibitor (NNRTI) + INSTI (33.0% and 31.9%, respectively). Overall, most of the regimens were nucleoside reverse transcriptase inhibitor (NRTI)-sparing (71.3%), with a decreasing trend over time (76.2% to 70.2%). ART regimen switch occurred more often in the 2-ART cohort than in the 3-ART cohort (33.0% vs 21.2%). Conclusion The profiles of patients on 2-ART in Japan were demonstrated to be complex. Most patients were treated with NRTI-sparing regimens which may reflect an effort to reduce treatment-related toxicities.