key: cord-0686712-mcifraou
authors: Nunez, I.; Belaunzaran-Zamudio, P. F.; Caro-Vega, Y.
title: Delay of molecular SARS-CoV-2 testing and turnaround time in Mexico and Colombia
date: 2021-06-24
journal: nan
DOI: 10.1101/2021.06.16.21259046
sha: 9c51f23a3e6559bc2ee178a3906c99afa97c3e03
doc_id: 686712
cord_uid: mcifraou

Objective: To quantify the delay in SARS-CoV-2 real time polymerase chain reaction (RT-PCR) testing and test result reporting in Mexico and Colombia, and to assess the relation between margination status and these delays. Methods: We quantified time in days from symptom onset until testing (latency one) and delay in test results report (latency two) using freely available country-wide open data from Mexico and Colombia. Directed acyclic graphs were built to determine which associations were appropriate to assess. Stratification by margination status, state and hospitalization status was used to determine corresponding associations. Results: In almost all the study period latency two was longer than latency one. Median latency one was 3 (IQR 0-6) days and latency two 7 (IQR 4-11) days in Colombia, while in Mexico they were 3 (IQR 1-5) days and 4 (IQR 3-6) days. In Colombia, worse margination status prolonged latency two. In Mexico, a lower number and percentage of point-of-care (POC) tests in areas with worse margination. Conclusion: POC tests must be used as a widespread means to reduce latency two, and until then should be prioritized in areas with longer latency two. An unequal distribution of this resource should be avoided.

We classified individuals as early testers if they were tested for COVID-19 within the first 3 days inefficient if it took longer. Finally, total latency was classified as optimal if it was shorter than 5 1 0 7 days, regular if it was between 6 and 10 days long, or inadequate if it was more than 10 days. A recent systematic review and meta-analysis reports that virus transmissibility begins two days 1 0 9 before symptoms onset up to 9 days after tit (11). Thus, we used this interval of time (extending 1 1 0 until 10 days after symptom started for practicality) to define patients as "infectious". We used rolling means to describe the length of latencies one, two, and total latency during the 1 1 4 study period, and proportions to describe the frequency of early, late and very late testers (latency 1 1 5 one); efficient and inefficient test results reporting (latency two), and optimal, regular, inadequate 1 1 6 total latency. We calculated the amount of time that would be saved with point-of-care tests, 1 1 7 which would eliminate latency two, dividing the total amount of latency two and the number of 1 1 8 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) 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 June 24, 2021. ; https://doi.org/10.1101 /2021 doi: medRxiv preprint 6 tested patients. We stratified the analysis by margination status, which was calculated differently 1 1 9

for Colombia and Mexico. For Colombia we included the "multidimensional poverty index" 1 2 0 (MPI), which is calculated yearly using a national representative household census (12, 13). It 1 2 1 evaluates five key aspects: education, childhood and youth, health, work, and living place (13). A 1 2 2 score of zero (no deprivation) and one (deprivation) is then calculated for each household and the 1 2 3

proportion of deprived households sampled by country and state (the smallest specified For Mexico, we calculated the "margination index" (MI), a metric developed by the Mexican 1 2 6

Consejo Nacional de Población (National Population Council, CONAPO) also calculated for 1 2 7 each household sampled in a national census. The smallest specified geographic area for which 1 2 8 this index is calculated is a "locality". It is constructed by the percentage of people that have each 1 2 9 of nine socio-economic characteristics for that geographical area: older than 15 years and 1 3 0 illiterate, older than 15 years and incomplete elementary school, no sewer system, no electricity, 1 3 1 no tubing water, household overcrowding, dirt floor, less than 5,000 habitants in the locality, and is not available in the most recent census, so we will substitute it for "percentage of people above 1 3 4 age 12 that are not economically active or currently attending school". Each variable weights 1/9, 1 3 5 and with a score that ranges from 0 to 100 (higher score indicating higher margination). The 2020 1 3 6 census did not include the MI, so it was manually calculated using the information on individual identified up to municipality (multiple localities join to form a municipality), so a mean score was calculated for each municipality using the score of the localities that conform each 1 4 1 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) 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 June 24, 2021. ; https://doi.org/10.1101/2021.06.16.21259046 doi: medRxiv preprint municipality, using the total population of each locality as a weight. We stratified patients by 1 4 2 MPI or MI quintiles.

To determine whether confounding could impair the validity of stratification by multidimensional 1 4 4 poverty index /margination index, we constructed two directed acyclic graphs (DAGs), one for beginning in November 15 th , 2020, the date in which these tests started to be performed. Colombia did not have the data to make this calculation. (15). Data used in the analysis is freely available in respective government websites (9, 10), and 1 5 3 code used in the analysis will be made available in the final version. The study was approved by Only datasets from Mexico and Colombia met inclusion criteria (9, 10). Open data included 1 5 9 different information for each country. Only confirmed cases were available for Colombia, while 1 6 0 all tested people were available for Mexico. People were registered in the Colombian database 1 6 1 from March 2 nd 2020, and the study period had to be limited until July 27 th because antigen tests 1 6 2 began to be reported starting July 28 th with no distinction between positive cases by RT-PCR or 1 6 3 antigen tests. Mexico does not provide a testing nor a diagnosis date. Testing date was assumed 1 6 4 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) 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 June 24, 2021. ; https://doi.org/10.1101/2021.06.16.21259046 doi: medRxiv preprint 8 to be that in which an individual first appeared in a database. As these individuals are included 1 6 5 with a "pending result" in the result variable, we consider this to be reasonable. In Colombia, median latency one was 3 days long (IQR 0-6), latency two 7 days (IQR 4-11), and 1 7 7 total latency 12 days (IQR 7-15). In Mexico the corresponding values were 3 days (IQR 1-5), 4 1 7 8 days (IQR 3-6), and 8 days (5-12). Latency periods by country and margination quintiles are 1 7 9

shown in Table 1 . Over time, latency one remained stable during the study period in Mexico. In 1 8 0 Colombia, it had an initial peak at the beginning of the study period but decreased soon and 1 8 1 remained stable time in Colombia (Figure 1) . Length of latency one was similar across 1 8 2 margination status in both countries while latency two was markedly higher in places with worse 1 8 3 margination in Colombia, but not in Mexico (Table 1, Appendix Figure 2 ). Latency two was much longer than latency one in both countries, causing the lengthiest delay in 1 8 7 tests result reports. While there was a tendency towards the reduction of latency two in Mexico 1 8 8 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) 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 June 24, 2021. ; https://doi.org/10.1101/2021.06.16.21259046 doi: medRxiv preprint over time (interrupted by a huge peak in October-November 2020), the length of latency two 1 8 9 increase over time in Colombia (Figure 1) . We observed a clear tendency towards a lengthier 1 9 0 latency two in the lowest quintiles of margination status indicators in Colombia but not in 1 9 1

Mexico. The number and proportion of performed tests that were point-of-care in Mexico was 1 9 2 lower in the worst margination quintiles across the study period, even if by the end the percentage Mexico (6.9 per person) from testing to result .

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The pandemic has caused a delay in healthcare provision in patients suffering non-covid related test turnaround might depend also on geographic access but we assume that mainly on laboratory We used this framework to describe the length of time between symptoms onset, testing and 2 1 0 sample processing, and test result report for SARS-CoV-2 and its association with socioeconomic 2 1 1 status in Colombia and Mexico. We observed that the median time between symptoms onset and 2 1 2 tests results report exceeds that of infectiousness in both countries. There is an overall poor use of 2 1 3 resources that depends to a great extent in a delay in tests results reporting after sample collection 2 1 4 (latency two period). Most people were early testers, more so in Colombia than in Mexico, with 2 1 5 considerable heterogeneity according to state. This supports the fact that testing strategies are not 2 1 6 "one fit all", and should be tailored according to each regions requirement. The presence of "local 2 1 7 epidemics" as a way of expressing how different geographic areas are impacted differently, has 2 1 8 been previously described (5, 17, 18 ). Interestingly we found longer latency two in areas with less margination in Mexico, but longer in Colombia. This could be due to the higher use of POC tests in Mexico, and thus we are only 2 2 1 observing RT-PCR tested patients, as well as the sole use of RT-PCR tests in Colombia during 2 2 2 the selected study period. Since latency one was short and remained constant over time and while important to test more people, are less likely to have an impact in reducing delays in the 2 2 5 diagnostic process, as the main component of the delay is the prolonged latency two.

As Mexico started using antigen tests during the late second half of 2020, we explored if these 2 2 7

were being performed preferably in vulnerable areas. Even if by the end of the study period the greatly, as only 18.4% of all POC tests were performed in the lower two quintiles. This is solely because of latency two, three times as much. Our study has several limitations. It is observational and used repurposed data, and as such it is 2 4 6 difficult to ascertain the precision of initial symptom date and diagnosis date, but both countries 2 4 7 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Colombia does not share information on every tested individual is also a limitation, as positive 2 4 9 individuals in that country could arrive at a different time to testing, even if our exploratory 2 5 0 analysis in Mexico does not show this. Selection bias might also present, since both countries do 2 5 1 only limited testing, with Mexico testing only one of 10 ambulatory patients and all hospitalized. We are unable to make conclusions regarding symptomatic patients who do not search for care evidently influence either people's ability to get tested and the laboratory's speed to give results, 2 5 5 but this is precisely our point. These variables are not accounted for in current testing strategies at 2 5 6 both of these countries, and low testing efficiency is a side effect of this.

Our study also has several strengths. It conveys information of two countries and a large number information. Thus, diagnostic delays can be adequately quantified. Also, our use of DAGs makes 2 6 1 our thought process transparent on estimating variable effect on testing delays. The low efficiency of RT-PCR observed in our study supports points to the need of improve the 2 6 3 efficiency of sample processing and test results reporting. Antigen tests for epidemiological 2 6 4 COVID-19 surveillance might contribute to reduce the time between sample collection and test 2 6 5 result delivery. Our results also indicate that efforts and resources should be more heavily 2 6 6 invested in high margination areas and populations, which would make resource allocation more 2 6 7 efficient. (which was not certified by peer review) 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 June 24, 2021. ; https://doi.org/10. 1101 /2021 

Classification of total latency, n (%): --Optimal

All rights reserved. No reuse allowed without permission.(which was not certified by peer review) 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 June 24, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 Negative Results for SARS-CoV-2 Surveillance in Mexico. Rev Invest Clin. https://www.gob.mx/salud/documentos/datos-abiertos-bases-historicas-direccion-general- Protecci-n-Social/Casos-positivos-de-COVID-19-en-Colombia/gt2j-8ykr/data All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. tema/pobreza-y-condiciones-de-vida/pobreza-y-desigualdad/pobreza-monetaria-y- CoV-2 Seroprevalence in the US as of September 2020. JAMA Intern Med. 2020; E1- (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Mexico of non-communicable diseases and SARS-CoV-2: factors associated with 3 4 2 increased case-fatality rates. Int J Epi. 2021; dyab008. Second quintile All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. until result availability; Total latency: time from symptom onset until result availability.3 6 0 2 "Early testers" were tested within the first three days of symptoms, "late testers" from the 3 6 1 fourth day until the eighth day, and "very late testers" afterwards.

3 Defined as those that were reported within two days from testing date.

4 Total latency was classified as "optimal" if it was shorter than 5 days, "regular" if it was 3 6 4 between 6 and 10 days long, or "inadequate" if it was more than 10 days. All rights reserved. No reuse allowed without permission.(which was not certified by peer review) 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 June 24, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 3 7 0 All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. reporting of test result.

All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. status quintile in Mexico (7-day rolling means). Quintile five represents the worst margination (which was not certified by peer review) 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 June 24, 2021. ; https://doi.org/10. 1101 /2021