> top > docs > PMC:7786642 > spans > 4793-16341 > annotations

PMC:7786642 / 4793-16341 JSONTXT

Annnotations TAB JSON ListView MergeView

LitCovid-PD-HP

Id Subject Object Predicate Lexical cue hp_id
T5 6360-6367 Phenotype denotes obesity http://purl.obolibrary.org/obo/HP_0001513
T6 7664-7671 Phenotype denotes Obesity http://purl.obolibrary.org/obo/HP_0001513

LitCovid-PubTator

Id Subject Object Predicate Lexical cue tao:has_database_id
93 50-58 Disease denotes COVID-19 MESH:C000657245
94 59-65 Disease denotes deaths MESH:D003643
95 141-147 Disease denotes deaths MESH:D003643
96 190-198 Disease denotes COVID-19 MESH:C000657245
97 263-271 Disease denotes COVID-19 MESH:C000657245
98 272-278 Disease denotes deaths MESH:D003643
99 291-297 Disease denotes deaths MESH:D003643
100 342-350 Disease denotes COVID-19 MESH:C000657245
101 358-363 Disease denotes death MESH:D003643
102 518-524 Disease denotes deaths MESH:D003643
103 538-546 Disease denotes COVID-19 MESH:C000657245
104 547-553 Disease denotes deaths MESH:D003643
105 596-604 Disease denotes COVID-19 MESH:C000657245
106 605-611 Disease denotes deaths MESH:D003643
107 724-733 Disease denotes mortality MESH:D003643
110 1988-1994 Disease denotes deaths MESH:D003643
111 2215-2221 Disease denotes deaths MESH:D003643
113 2397-2400 Chemical denotes NO2
120 6472-6476 Gene denotes S1.1 Gene:6267
121 5786-5790 Gene denotes S1.1 Gene:6267
122 5564-5572 Disease denotes COVID-19 MESH:C000657245
123 5632-5640 Disease denotes COVID-19 MESH:C000657245
124 5749-5757 Disease denotes infected MESH:D007239
125 6360-6367 Disease denotes obesity MESH:D009765
130 8050-8058 Disease denotes COVID-19 MESH:C000657245
131 8059-8065 Disease denotes deaths MESH:D003643
132 8130-8139 Disease denotes mortality MESH:D003643
133 8230-8236 Disease denotes deaths MESH:D003643
136 9739-9748 Disease denotes mortality MESH:D003643
137 9941-9950 Disease denotes mortality MESH:D003643
139 10207-10211 Gene denotes S1.2 Gene:6268
141 11102-11106 Disease denotes fits MESH:D012640

LitCovid-sentences

Id Subject Object Predicate Lexical cue
T42 0-10 Sentence denotes 2 Methods
T43 12-33 Sentence denotes 2.1 Study population
T44 34-126 Sentence denotes We included all COVID-19 deaths as reported to Public Health England (PHE) by June 30, 2020.
T45 127-376 Sentence denotes These include deaths that had a laboratory confirmed report of COVID-19 (including at post-mortem) (EpiCell 2020), as well as suspected COVID-19 deaths, defined as deaths without a positive test but with mention of COVID-19 in the death certificate.
T46 377-461 Sentence denotes These definitions were consistent during the study period and over the study region.
T47 462-525 Sentence denotes The main outcome of this study was laboratory confirmed deaths.
T48 526-771 Sentence denotes We selected COVID-19 deaths up to June the 30th to ensure we captured COVID-19 deaths attributable to the first wave of epidemic that in England and Wales was over by the end of May, when all-cause mortality was no longer elevated (Kontis et al.
T49 772-778 Sentence denotes 2020).
T50 779-977 Sentence denotes Individual data on age, sex, ethnicity, lower tier local authority (LTLA) of the residential address and type of residence type (i.e. nursing homes, prisons, medical facilities etc.) were available.
T51 978-1067 Sentence denotes Population at risk in England was available through Office for National Statistics (ONS).
T52 1068-1227 Sentence denotes Information at the LSOA level about age and sex was available for 2018, whereas about ethnicity for 2011 (the most recent years available at time of analysis).
T53 1229-1245 Sentence denotes 2.2 Downscaling
T54 1246-1318 Sentence denotes There were 317 LTLAs in England in 2019 (Supplemental Material Fig. S1).
T55 1319-1433 Sentence denotes Such a coarse geographical unit is not expected to capture the strong localised spatial patterns of air-pollution.
T56 1434-1505 Sentence denotes We thus downscaled the LTLA geographical information to the LSOA level.
T57 1506-1620 Sentence denotes LSOAs are high resolution geographical units in England (32,844 units in 2011, see Supplemental Material Fig. S2).
T58 1621-1830 Sentence denotes The median population per LSOA in 2018 was 1617, varying from 591 to 14,696 (min to max) (Supplemental Material Fig. S3), and the median area per LSOA was 0.4km2, varying from 0.0002km2 to 68.4km2(min to max).
T59 1831-1907 Sentence denotes The LTLA boundaries are revised every year, whereas the LSOA ones at census.
T60 1908-2186 Sentence denotes Let l~m denote that the l-th LSOA belongs to the m-th LTLA, nijkm the number of deaths in the m-th LTLA and Pijkl the population in the i-th age group (1<, 1–4, 5–9, …, 85–90, >90), j-the sex (male or female), k-th ethnic group (White, Mixed, Asian, Black, Other) and l-th LSOA.
T61 2187-2355 Sentence denotes We sampled nijkm individual deaths at the l-th LSOA level from a Multinomial distribution with probabilities:πijkl=Pijkl/∑l~mPijkl,and repeated the procedure 100 times.
T62 2357-2370 Sentence denotes 2.3 Exposure
T63 2371-2442 Sentence denotes We considered exposure to NO2 and PM2.5 as indicators of air pollution.
T64 2443-2480 Sentence denotes We selected these pollutants because:
T65 2481-2896 Sentence denotes 1) they reflect different sources of air-pollution (NO2 reflects traffic related air-pollution, whereas PM2.5 is a combination of traffic and non-traffic sources), 2) they were considered in previous studies (Cole et al., 2020, Liang et al., 2020, Travaglio et al., 2020, Wu et al., 2020), and 3) they are responsible for the highest number of years of life lost compared to other pollutants in Europe (Ortiz 2019).
T66 2897-3020 Sentence denotes We retrieved NO2 and PM2.5 concentration in England from the Pollution Climate Mapping (PCM; https://uk-air.defra.gov.uk/).
T67 3021-3135 Sentence denotes The PCM produces annual estimates during 2001–2018 for NO2 and 2002–2018 for PM2.5 at 1x1km resolution for the UK.
T68 3136-3299 Sentence denotes The PCM model is calibrated using monitoring stations across the nation and has high predictive accuracy, R2 = 0.88 for NO2 and R2 = 0.63 for PM2.5 (Brookes 2017).
T69 3300-3450 Sentence denotes We defined long-term exposure to these compounds as the mean of the past 5 years for which data was available at the time of analysis, i.e. 2014–2018.
T70 3451-3620 Sentence denotes An alternative is calculating the median, however the distribution of the air-pollutants using any of these metrics is almost identical, (Supplemental Material Fig. S4).
T71 3621-3930 Sentence denotes We weighted the exposure using a combination of population estimates available from the fourth version of Gridded Population of the World collection at 1x1km grid as of 2020 (Center for International Earth Science Information Network - CIESIN - Columbia University 2018) and from ONS at LSOA level as of 2018.
T72 3931-4035 Sentence denotes Let Xgl be the pollutant and Pgl the population in the intersection of the g-th grid cell and l-th LSOA.
T73 4036-4199 Sentence denotes Assuming the Xg is constant (i.e. Xgl=Xg for all intersections) in the g-th grid cell, we define the population weighted version X¯lof Xgl as:X¯l=∑glP¯glXg∑glP¯gl.
T74 4200-4299 Sentence denotes To calculate P¯gl, we first compute w¯gl=wgl/∑glwgl, where wgl is the area weight per intersection.
T75 4300-4347 Sentence denotes Then calculate the population per intersection:
T76 4348-4360 Sentence denotes Pgl'=Pgw¯gl.
T77 4361-4477 Sentence denotes We then use the Pl (LSOA populations) and obtain P¯gl=vglPl, where vgl is the normalised Pgl', ie vgl= Pgl'/∑glPgl'.
T78 4479-4495 Sentence denotes 2.4 Confounders
T79 4496-4644 Sentence denotes We considered confounders related with meteorology, socio-demographics, disease spread, healthcare provision and health related variables (Table 1).
T80 4645-4858 Sentence denotes As meteorological confounders, we considered temperature and relative humidity and calculated the mean for March-June 2018 as this is the latest year with data available at 1x1km grid retrieved from the MetOffice.
T81 4859-4976 Sentence denotes We weighted temperature and relative humidity using the population weights calculated for the air-pollution exposure.
T82 4977-5106 Sentence denotes As socio-demographical confounders we considered age, sex, ethnicity, deprivation, urbanicity, population density and occupation.
T83 5107-5307 Sentence denotes Information on age (2018), sex (2018), ethnicity (2011), urbanicity (2011) and population density (2018) was available at the LSOA level from ONS (the most recent years available at time of analysis).
T84 5308-5516 Sentence denotes To adjust for deprivation, we used quintiles of the index of multiple deprivation at LSOA level in 2019 (Ministry of Housing, Communities and Local Government), excluding the dimension related to air quality.
T85 5517-5805 Sentence denotes We used estimates of occupational exposures to COVID-19, as calculated by ONS, to adjust for high risk exposure to COVID-19, defined as those with a score higher than 80/100 (corresponding to at least >1 per week exposed to someone infected, Supplemental Material Text S1.1 and Table S1).
T86 5806-5992 Sentence denotes To account for disease progression, we used the number of days since the 1st reported case and the number of positive cases in each LTLA (as of 30th of June 2020, as retrieved from PHE).
T87 5993-6154 Sentence denotes Adjustment for the latter factors is expected to attenuate geographical differences generated due to regional differences about the timing on the pandemic curve.
T88 6155-6298 Sentence denotes For healthcare provision, we used the number of intensive care unit beds per population, in February 2020 per NHS trust, as retrieved from NHS.
T89 6299-6478 Sentence denotes Last, as health-related variables, we considered smoking and obesity prevalence at the GP practice level during 2018–2019, as retrieved from PHE (Supplemental Material Text S1.1).
T90 6479-6521 Sentence denotes Table 1 Data sources used in the analysis.
T91 6522-6584 Sentence denotes Confounders Source Spatial Resolution Temporal Resolution Type
T92 6585-6668 Sentence denotes Temperature MetOfficehttps://www.metoffice.gov.uk/ 1 km2 March-June 2018 continuous
T93 6669-6758 Sentence denotes Relative humidity MetOfficehttps://www.metoffice.gov.uk/ 1 km2 March-June 2018 continuous
T94 6759-6913 Sentence denotes Index of Multiple Deprivation Ministry of Housing, Communities and Local Governmenthttps://www.gov.uk/ Lower layer super output area 2019 rank (quintiles)
T95 6914-7025 Sentence denotes Urbanicity Office for National Statisticshttps://www.ons.gov.uk/ Lower layer super output area 2011 urban/rural
T96 7026-7130 Sentence denotes Days since 1st reported case Public Health England Lower tier local authority Until 30th June continuous
T97 7131-7238 Sentence denotes Number of positive cases Public Health England Lower tier local authority Until 30th June discrete (counts)
T98 7239-7375 Sentence denotes Population density Office for National Statisticshttps://www.ons.gov.uk/ Lower layer super output area 2018 continuous (log transformed)
T99 7376-7533 Sentence denotes Number of intensive care unit beds National Health Servicehttps://www.england.nhs.uk/ National Health Service trust February 2020 continuous (per population)
T100 7534-7663 Sentence denotes Smoking Public Health Englandhttps://fingertips.phe.org.uk/ General practitioner catchment area 2018–2019 continuous (prevalence)
T101 7664-7793 Sentence denotes Obesity Public Health Englandhttps://fingertips.phe.org.uk/ General practitioner catchment area 2018–2019 continuous (prevalence)
T102 7794-7928 Sentence denotes High Risk Occupation Office for National Statisticshttps://www.ons.gov.uk/ Middle layer super output area 2011 continuous (prevalence)
T103 7930-7954 Sentence denotes 2.5 Statistical methods
T104 7955-8098 Sentence denotes We specified Bayesian hierarchical Poisson log-linear models to investigate the association of COVID-19 deaths and NO2 and PM2.5 independently.
T105 8099-8246 Sentence denotes The LSOA specific standardised mortality ratio is known to be an unstable estimator with high variance when the number of expected deaths is small.
T106 8247-8576 Sentence denotes To overcome this problem, we used a well-established hierarchical framework, specifying spatially structured and unstructured random effects, so that the model borrows strength from the other areas across the entire study region, as well as from the neighbouring ones (Best et al., 2005, Wakefield et al., 2000, Wakefield, 2006).
T107 8577-8748 Sentence denotes We model these random effects using a re-parametrisation of the Besag-York-Molliè conditional autoregressive prior distribution (Besag et al., 1991, Simpson et al., 2017).
T108 8749-8781 Sentence denotes We fitted four models including:
T109 8782-9011 Sentence denotes 1) each pollutant (model 1), 2) each pollutant and the spatial autocorrelation term (model 2), 3) each pollutant and all confounders (model 3) and 4) each pollutant, the spatial autocorrelation term and all confounders (model 4).
T110 9012-9177 Sentence denotes All models were adjusted for age, sex and ethnicity using indirect standardisation; we used the English population as the standard population to calculate the rates.
T111 9178-9320 Sentence denotes We do not report results from the joint analysis including both pollutants since they are highly correlated (Supplemental Material Figure S5).
T112 9321-9552 Sentence denotes In order to propagate the uncertainty resulted from the sampling we used for the downscaling, we fitted the models over 100 downscaled samples and then performed Bayesian model averaging to combine the estimates (Gómez-Rubio et al.
T113 9553-9559 Sentence denotes 2020).
T114 9560-9683 Sentence denotes We performed a complete case analysis since for only 1.1% of the cases information about age, sex and ethnicity is missing.
T115 9684-9897 Sentence denotes We report results as posterior median of % increase in mortality risk for every 1 μg/m3 increase in the air-pollutants, 95% credible intervals (CrI) and posterior probability that the estimated effect is positive.
T116 9898-10094 Sentence denotes We also report posterior median of spatial mortality relative risks (exponential of the spatial autocorrelation term) and posterior probabilities that the spatial relative risks are larger than 1.
T117 10095-10212 Sentence denotes The mathematical formulation of the models and prior specifications are given in the Supplemental Material Text S1.2.
T118 10213-10255 Sentence denotes All models were fitted in INLA (Rue et al.
T119 10256-10262 Sentence denotes 2009).
T120 10263-10387 Sentence denotes Covariate data and code for running the analysis are available at https://github.com/gkonstantinoudis/COVID19AirpollutionEn.
T121 10389-10414 Sentence denotes 2.6 Sensitivity analyses
T122 10415-10461 Sentence denotes We performed a series of sensitivity analyses.
T123 10462-10586 Sentence denotes First, we repeated the main analyses using data at the LTLA level with all exposures and confounding weighted by population.
T124 10587-10777 Sentence denotes Second, we examined if there is a differential effect of long-term exposure to air-pollution at the early stages of the epidemic, considering the lockdown (23rd of March 2020) as a landmark.
T125 10778-11029 Sentence denotes Third, we assessed the correlation between the latent field of the full model (model 4) with that of the model excluding or including only covariates indicating disease spread (i.e. number of tested positive cases and days since first reported cases).
T126 11030-11107 Sentence denotes Fourth, we categorised pollutants into quintiles to allow more flexible fits.
T127 11108-11185 Sentence denotes Fifth, we repeated the analysis including the suspected cases to the outcome.
T128 11186-11386 Sentence denotes Sixth, we repeated the analysis changing the definition of long-term exposure to the mean of the past 3 and 10 years for which data was available at the time of analysis, i.e. 2016–2018 and 2009–2018.
T129 11387-11548 Sentence denotes Seventh, we fitted a zero-inflated Poisson model to account for the proportion of zeros in the data (36% in the 100 samples – see Supplemental Material Fig. S6).