Estimation of Individual Probabilities of COVID-19 Infection, Hospitalization, and Death From A County-level Contact of Unknown infection Status

230 words 12 Manuscript: 2747 words 13 References: 822 words 14 Tables: 3 15 Figures: 2 16 17 * Corresponding Author 18 19 Version Date: June 23, 2020 20 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 24, 2020. ; https://doi.org/10.1101/2020.06.06.20124446 doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.

CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 24, 2020. ; https://doi. org/10.1101org/10. /2020 4 estimates as data on setting specific infection incidence rates, susceptibility and 66 secondary attack rates permit.

68
Here, we contribute to COVID-19 risk assessment by demonstrating a method to 69 estimate the individual probabilities of acquiring infection, being hospitalized, and dying 70 in U.S. Counties. We identify areas of available and future knowledge that could make 71 risk assessment more precise and context specific.

75
Our objective is to estimate the probability of acquiring COVID-19 infection from a 76 "contact" with a random individual of unknown infection status. We conceptualize this 77 probability under steady state conditions (e.g. no epidemic growth or decline) as a 78 function of individual susceptibility, the current reported case incidence, accounting for 79 undetected infection, the share of infection transmission occurring without a known 80 contact, the chance of transmission per contact (e.g., the secondary attack rate), and 81 the duration of infectiousness, accounting for pre-symptomatic transmission.

83
We use the formulae below to compute probabilities of infection, confirmed infection, 84 hospitalization, and death.

88
(3) P hospitalization | contact = CHR × P confirmed-infection | contact 89 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 24, 2020. We describe the parameters used in Table 1 and explain their sources below. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 24, 2020. The prevalence of susceptibility to COVID-19 is unknown. Pre-existing immunity due to 98 previous COVID-19-related coronaviral infections is plausible but speculative. Reliable 99 estimates for the proportion of the population who have acquired immunity is unknown 100 but non-zero. We have conservatively estimated the prevalence of susceptibility to be 101 95%.   Limited data is available on the share of reported infections arising without a known 118 contact. As of this date, the US CDC has not included any statistics on this attribute of 119 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 24, 2020. ; https://doi.org/10.1101/2020.06.06.20124446 doi: medRxiv preprint 7 confirmed infections. The State of Oregon currently publicly reports infections without a 120 known contact a varying between 30 and 50%%. Because of limited national data, we 121 use 100% for this parameter; we leave the parameter in the equation for the purpose of 122 future applications. handshaking, embraces, wearing a mask or indoor ventilation). We do not consider 131 contacts to be short-term events, such as passing by a person on the street. Contact 132 within households involves habitual and typically unprotected close physical. We 133 understand that attack rates will vary across such diverse exposure settings. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 24, 2020. We assume an average plausible attack rate across all settings of exposure based on 143 these range of estimates to be 10% in the absence of more setting and activity specific 144 data. We acknowledge that this estimate may overestimate the attack rate for a non-145 household contact and underestimate it for a household contact. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 24, 2020. ; https://doi.org/10.1101/2020.06.06.20124446 doi: medRxiv preprint 9 relationship between total infections and confirmed infections is fixed and defined by the 165 parameter alpha above.  CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

189
Among US Counties with populations greater than 500,000 people (N= 1224), for the 190 week ending June 13, 2020, the median observed county-level daily case incidence is   Table 2 lists 201 probabilities for other age groups.

202
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

206
hospitalization and fatality per contact as a function of daily case incidence. Figure 2 207 Illustrates the estimated number of hospitalizations and fatalities per 1 million contacts 208 in a subset of analyzed US counties with populations greater than 1.5 million.

210
We found good concordance between the estimated weekly hospitalization rates and

215
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

228
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 24, 2020.  CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 24, 2020. ; https://doi.org/10.1101/2020.06.06.20124446 doi: medRxiv preprint 14 As mentioned above, systematic data on the share of confirmed infections arising 251 without a known contact is not currently available. Appling this fraction to the model 252 would reduce our estimates based on the assumption of a 100% share.  CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 24, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 We have assumed a high fraction of the population remains susceptible as the perception. Exploring methods to communicate risk and the concordance of perceived 295 risk and risk probabilities would be an appropriate subject for further work.

296
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 24, 2020.   CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 24, 2020. ; https://doi.org/10. 1101/2020 Return to community workplace and social life will require individuals to be comfortable 321 with their personal risk of acquiring COVID-19 infection. Estimates on the individual 322 probabilities of infection, hospitalization and death may contribute to a more accurate 323 risk perception. Systematically collected and publicly reported data on infection 324 incidence by, for example, the geographic setting of exposure, residence type, whether 325 a case had a known exposure, and would allow more precise estimation than those 326 possible with currently available public data. Calculation of secondary attack rates by 327 setting and prevalence of seropositivity would further improve these estimates.  CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 24, 2020. ; https://doi.org/10. 1101/2020