key: cord-0832880-am9bdw5t authors: Stubbs, C. W.; Springer, M.; Thomas, T. S. title: The Impacts of Testing Cadence, Mode of Instruction, and Student Density on Fall 2020 COVID-19 Rates On Campus date: 2020-12-09 journal: nan DOI: 10.1101/2020.12.08.20244574 sha: 286aaccd2b680d81d1a1ccd70388d118b913e54a doc_id: 832880 cord_uid: am9bdw5t We analyzed the COVID-19 infection rate among undergraduate students at 9 colleges and Universities in the greater Boston area and 4 comparison schools elsewhere, from Fall 2020. We assessed whether the cumulative rate of infection is dependent on the mode of instruction (in-person, hybrid, or remote), on the number and density of dorm-resident undergraduates, and/or on COVID-19 testing cadence. We limited our analysis to institutions that have implemented at least weekly PCR testing of dormitory-resident undergraduates. Our primary conclusions are that (i) the fraction of students that succumbed to a COVID-19 infection up through Nov 22, 2020 shows no correlation with either the total number of students on campus, or the fractional occupancy of dormitories, (ii) remote instruction vs. hybrid instruction has no significant measurable impact on cumulative infection rate, and (iii) there is evidence that testing 2 or 3 times per week is correlated with lower infection rates than weekly testing. These data are consistent with a hypothesis of students predominantly acquiring infection off-campus, with little community transmission within dormitory housing. This implies good student compliance with face mask and social distancing protocols. the needs of their community. The testing performed by the 23 higher education community comprises a significant fraction 24 of all the COVID-19 testing undertaken in the Commonwealth 25 of Massachusetts. 26 To mitigate viral spread, high cadence PCR testing of 27 campus-residential students is accompanied by prompt isola-28 tion of those individuals whose samples contain SARS-COV2 Table 1 presents the information we compiled from institu-42 tional web sites and press releases. We determined the total In order to factor out variability in the underlying local rate, 58 we selected a variety of Boston-area institutions * as well as 59 a few large and small institutions in other urban and rural 60 areas. Since we are particularly interested in investigating the 61 possible impact of testing cadence, we limited our selection to 62 schools that have at least once per week PCR tests (unpooled) 63 for their residential undergraduates. The numbers in Table 1 being obscured by variations in the surrounding regional cir-89 cumstances. If all the acquired infections were statistically independent, 91 i.e. were the result of off-campus interactions in the commu-92 nity, the fractional incidence in the student groups should 93 be independent of both total population size and dormitory 94 density. If, on the other hand, one were to see a strong depen-95 dence on dormitory density, that might support a hypothesis 96 of transmission due to congregate housing and proximity. The interval in question spans 100 days. An estimate 98 of the COVID-19 prevalence in the Boston-area residential 99 undergraduates, expressed as new cases per day per 100,000 100 people, is then our normalized f , the cumulative number of 101 infections per thousand students over the 100 day period. The 102 mean of f for the Boston area schools is < f >= 16 ± 3 new 103 cases per 100,000 person-days. The uncertainty was computed 104 by taking the standard deviation of the Boston-area f values 105 and dividing by where DOF is remaining 106 degrees of freedom, This is consistent with the mean case rate 107 of 10.8 per 100,000 reported (24) for Middlesex County, MA 108 over the same period. The local colleges contribute a very small 109 fraction of the countywide case totals, and don't distort that 110 statistic. Since the on-campus testing campaigns would detect 111 asymptomatic cases, who are presumably under-represented 112 in the regional case rate estimates, one might expect the case 113 rate determined from campus surveillance to exceed the local 114 regional rate, due to higher completeness. Making a correction 115 for selection effects in the regional case rate requires better 116 data than is currently available. Occupancy Density. Figure 1 shows the fraction f of dorm-119 resident students who tested positive over the study interval vs. 120 total number of students in residence on each campus. Figure 121 2 shows how f depends on dormitory occupancy density. . CC-BY-NC 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. The copyright holder for this preprint this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.20244574 doi: medRxiv preprint Table 2 . The scatter in the data appears to be driven by factors other 126 than total students on campus and the density of dormitory 127 occupancy. C. Dependence on Testing Cadence. Figure 4 shows the de-143 pendence of cumulative infection fraction f vs. viral test 144 cadence. Since only two of the schools in our sample carried out 146 testing three times per week, we binned the cadences into 147 once per week, and more than once per week. A two-sample t 148 test against this cadence parameter returns a p-value of 1.7%, 149 which does imply a correlation where schools that tested 2 or 150 3 times per week had fewer infections per capita than schools 151 that conducted weekly tests. This comparison was carried out 152 on the full sample. While the statistical power of this analysis is limited, the 155 results are consistent with the hypothesis that most student 156 infections were acquired outside of the dorm-residential setting, 157 with minimal community transmission within congregate on-158 campus student housing. This is consistent with anecdotal 159 contact tracing reporting on our campus. While there could have been a strong correlation of infection 161 rate with the housing variables we considered, this turned out 162 to not be so. The number of serologically-confirmed COVID-19 163 infections that manifest in the students that were sequestered 164 into quarantine due to close contacts will help determine the 165 extent to which "test+isolate+quarantine" (TIQ) strategies 166 suppressed subsequent transmission. That analysis must also 167 take into account the false-positive rate from PCR testing; the 168 frequency of this is presently unclear. Great effort should be made to minimize the latencies in the 170 TIQ timeline. This timeline starts when the students collects 171 the sample, and ends when infected students are in isolation 172 and their close contact have been placed in quarantine. Our 173 experience indicates that estimates of these latencies in earlier 174 modeling may have been over-optimistic. There is evidence (p-value of 1.7%) that a campus testing frequency of more than once per week is correlated with lower f . There were only two institutions in our sample that carried out testing three times per week, so the horizontal bars of the blue box on the right correspond to those values, and the red bar is their mean. • residential ventilation characteristics, such as cubic feet 208 per minute of HEPA-quality filtered air exchange. It is presently unclear how rapidly campus-resident pop-210 ulations will be vaccinated against COVID-19. An ongoing 211 assessment of infection rates as a function of institutional 212 policy will help inform evidence-based decisions and the agile 213 implementation of best practices. 214 ACKNOWLEDGMENTS. We gratefully acknowledge the students 215 whose conduct and adherence to public health protocols during 216 Fall 2020 helped achieve the results presented here. We thank E. 217 Carrington Gregory for suggestions on the manuscript. We also with-influx-college-students-looming-how-will-boston-keep-safe/? Real-time pcr handbook broad-institute-provides-covid-19-screening-students-faculty-and-staff-more-100-colleges-and) Meet the team behind boston university covid-19 testing lab Inside the lab that tests northeastern for the Impact of delays on effectiveness of contact tracing strategies for 237 covid-19: a modelling study Maximizing and evaluating the impact of test-trace-isolate programs Sars-cov-2, sars-cov, and mers-cov viral load dynamics, duration of viral 240 shedding, and infectiousness: a systematic review and meta-analysis Controlling covid-19 via test-trace-quarantine The timing of covid-19 transmission Brandeis covid dashboard Emerson college covid dashboard Northeastern university covid dashboard Tufts University, Tufts univ. covid dashboard Columbia univ. covid dashboard columbia-surveillance-testing-results Cornell univ. covid dashboard Amherst covid dashboard Williams covid dashboard Middlesex ma covid dashboard Assessment of sars-cov-2 screening strategies to per-272 mit the safe reopening of college campuses in the united states We review the incidence of COVID-19 infection among undergraduate students for selected colleges and universities that conducted at least weekly COVID-19 testing during the Fall of 2020. We analyzed the infection-rate dependence on number of students on campus, dormitory residential density, instructional methodology (remote vs. hybrid), and testing cadence. This compilation of outcomes can help inform policy decisions for congregate settings.