Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-16T16:19:22.918Z Has data issue: false hasContentIssue false

Using postal change-of-address data to predict second waves in infections near pandemic epicentres

Published online by Cambridge University Press:  24 March 2022

Adam Schulman
Affiliation:
Rutgers Robert Wood Johnson Medical School, New Brunswick, NJ 08901, USA
Gyan Bhanot*
Affiliation:
Department of Physics and Astronomy, Rutgers University, Piscataway, NJ 08854, USADepartment of Molecular Biology and Biochemistry, Rutgers University, Piscataway, NJ 08854, USARutgers Cancer Institute of New Jersey, New Brunswick, NJ 08903, USA
*
Author for correspondence: Gyan Bhanot, E-mail: gyanbhanot@gmail.com
Rights & Permissions [Opens in a new window]

Abstract

We propose that postal Change-of-Address (CoA) data can be used to monitor/predict likely second wave caseloads in viral infections around urban epicentres. To illustrate the idea, we focus on the tri-state area consisting of New York City (NYC) and surrounding counties in New York, New Jersey and Connecticut States. NYC was an early epicentre of the coronavirus disease 2019 (Covid-19) pandemic, with a first peak in daily cases in early April 2020, followed by the second peak in May/June 2020. Using CoA data from the US Postal Service (USPS), we show that, despite a quarantine mandate, there was a large net movement of households from NYC to surrounding counties in the period April–June 2020. This net outward migration of households was strongly correlated with both the timing and the number of cases in the second peaks in Covid-19 cases in the surrounding counties. The timing of the second peak was also correlated with the distance of the county from NYC, suggesting that this was a directed flow and not random diffusion. Our analysis shows that CoA data is a useful method in tracking the spread of an infectious pandemic agent from urban epicentres.

Information

Type
Original Paper
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
Copyright © The Author(s), 2022. Published by Cambridge University Press
Figure 0

Table 1.Summary of results for each county

Figure 1

Fig. 1.(a–c) LtoR top: Monthly Excess Change of Address (CoA) counts from New York City boroughs to all counties in New Jersey and Connecticut and 9 tri-state counties (Suffolk, Nassau, Westchester, Ulster, Dutchess, Orange, Sullivan, Rockland and Putnam) in New York State in 2019 (blue bars) and 2020 (red bars). The solid horizontal and dashed lines represent the average and one-sigma limits for Excess CoA for the twelve months in 2019. (d–f) LtoR bottom: Cumulative Excess Change of Address (CoA) counts from New York City boroughs to each county in New Jersey and Connecticut and 9 tri-state counties (Suffolk, Nassau, Westchester, Ulster, Dutchess, Orange, Sullivan, Rockland and Putnam) in New York State for the months of March–June 2020.

Figure 2

Fig. 2.(a–f) LtoR top, LtoR bottom: Seven-day averaged daily Covid-19 cases (blue circles) in the two counties with the highest population densities in each of the tri-states: Hudson and Essex in NJ, Fairfield and New Haven in CT and Nassau and Westchester in NY. The solid red curves are fits to the data for the first peak from a SIR model (Methods and [30]).

Figure 3

Fig. 3.(a–f) LtoR top, LtoR bottom: Daily cases in the second peak, obtained from data for Figures 2a–f by subtracting the SIR model predictions (solid red line) from the seven-day averaged data (blue circles). The number of cases NCS in the second peak was estimated by adding the daily case data between the black dots, which approximate the timepoints of the start and end of the second peak. The first time point was chosen as the day the second peak daily case load was approximately zero and the second time point when it tapered off to a constant value.

Figure 4

Fig. 4.(a) Bivariate linear regression fit to the model: ${\rm N}_{{\rm CS}}\sim {\rm P}^\delta {\rm E}_{{\rm CoA}}^\beta$ where NCS is the number of cases in the second peak in a given county, P is the population density in the county and ECoA is the excess households that moved into the county from NYC from March–June 2020. (b) Projection of the fit of (a) to the P axis. (c) Projection of the fit of (a) to the ECoA axis. (d) Number of days from the first peak in daily cases to the second peak for each county versus the distance by road of the county seat from NY Penn Station (miles). Note that the minimum number of days between peaks (intercept on y axis) is 13.7 days, which is possibly related to the quarantine duration mandate for infected people in the first peak and to the latency between infection and symptoms.

Supplementary material: File

Schulman et al. supplementary material

Figures S1-S3

Download Schulman et al. supplementary material(File)
File 2.5 MB