3 ms·

I don’t think it’s cherry picking so much as maybe working with the limitations of merging 50 datasets vs 2? They at least acknowledge the dataset bias in the

by rsfern 5y ago

I don’t think it’s cherry picking so much as maybe working with the limitations of merging 50 datasets vs 2? They at least acknowledge the dataset bias in the discussion section, and they cite a couple similar findings from other countries, and also they start with this: > By November 30, 2021, approximately 130,781 COVID-19– associated deaths, one in six of all U.S. deaths from COVID-19, had occurred in California and New York

tomohawk 5y ago
CDC is a federal institution. They have one dataset.
- rsfern 5y ago
  I guess, but it’s a heterogeneous dataset. They even discuss in this manuscript differences between how CA reports stuff compared with NY. (This is kind of a separate issue where CDC could have lead on data standardization, but not everyone in the US agrees about the Feds playing that role in this case; kind of a separate issue) The CDC also has covidnet, which doesn’t have the data heterogeneity problem so much, but I’m not sure it has the vaccination status breakdown needed for this study. CDC has famously not done a good job breaking that down… https://www.cdc.gov/coronavirus/2019-ncov/covid-data/covid-net/purpose-methods.html https://www.cdc.gov/coronavirus/2019-ncov/covid-data/covid-n... Anyway, none of this is to say that they’re definitely not doing a cherry picking exercise, but at least one should read the paper and decide if the choices they make about data sources and analysis are reasonable. Not super productive to just immediately assume bad faith, IMO. Especially since the conclusions are consistent with research from other institutions/countries
- dredmorbius 5y ago
  False, as is specifically noted in TFA discussing data sources. Those are: Statewide immunization databases in California are the California Immunization Registry, Regional Immunization Data Exchange, and San Diego Immunization Registry; the laboratory system is the California COVID Reporting System (CCRS). In New York, immunization information systems include Citywide Immunization Registry and the New York State Immunization Information System; the laboratory system is the Electronic Clinical Laboratory Reporting System (ECLRS). California data were matched between the immunization and case registries using a probabilistic algorithm with exact match for zip code and date of birth and fuzzy match on first name and last name. New York data were matched to the ECLRS with the use of a deterministic algorithm based on first name, last name, and date of birth. In California, person-level hospitalization data from CCRS and supplementary hospitalization reports were used to identify COVID-19–associated hospitalizations. Specifically acknowledged are: Dana Jaffe, California Department of Public Health; Rebecca Hoen, Meng Wu, New York State Department of Health; Citywide Immunization Registry Program, New York City Department of Health and Mental Hygiene. From whom those data were likely in substantial part obtained.

by rsfern 5y ago

tomohawk 5y ago

CDC is a federal institution. They have one dataset.

rsfern 5y ago
I guess, but it’s a heterogeneous dataset. They even discuss in this manuscript differences between how CA reports stuff compared with NY. (This is kind of a separate issue where CDC could have lead on data standardization, but not everyone in the US agrees about the Feds playing that role in this case; kind of a separate issue) The CDC also has covidnet, which doesn’t have the data heterogeneity problem so much, but I’m not sure it has the vaccination status breakdown needed for this study. CDC has famously not done a good job breaking that down… https://www.cdc.gov/coronavirus/2019-ncov/covid-data/covid-net/purpose-methods.html https://www.cdc.gov/coronavirus/2019-ncov/covid-data/covid-n... Anyway, none of this is to say that they’re definitely not doing a cherry picking exercise, but at least one should read the paper and decide if the choices they make about data sources and analysis are reasonable. Not super productive to just immediately assume bad faith, IMO. Especially since the conclusions are consistent with research from other institutions/countries
dredmorbius 5y ago
False, as is specifically noted in TFA discussing data sources. Those are: Statewide immunization databases in California are the California Immunization Registry, Regional Immunization Data Exchange, and San Diego Immunization Registry; the laboratory system is the California COVID Reporting System (CCRS). In New York, immunization information systems include Citywide Immunization Registry and the New York State Immunization Information System; the laboratory system is the Electronic Clinical Laboratory Reporting System (ECLRS). California data were matched between the immunization and case registries using a probabilistic algorithm with exact match for zip code and date of birth and fuzzy match on first name and last name. New York data were matched to the ECLRS with the use of a deterministic algorithm based on first name, last name, and date of birth. In California, person-level hospitalization data from CCRS and supplementary hospitalization reports were used to identify COVID-19–associated hospitalizations. Specifically acknowledged are: Dana Jaffe, California Department of Public Health; Rebecca Hoen, Meng Wu, New York State Department of Health; Citywide Immunization Registry Program, New York City Department of Health and Mental Hygiene. From whom those data were likely in substantial part obtained.