Biostatistics In Action

NYC Neighborhoods COVID-19 Dashboard

Dashboard by Qixuan and Shing 
By Qixuan Chen

The NYC Neighborhoods COVID-19 Dashboard was developed by Professor Qixuan Chen and MS students Ziqi Zhou, Mengyu Zhang, Yuanzhi Yu, and Yuchen Qi in the Biostatistics Department. The Dashboard uses data from the NYC Department of Health and Mental Hygiene to track daily cases, hospitalizations, deaths, and tests for every NYC neighborhood by Zip Code and provides data visualizations of distributions and time trends for COVID cases, hospitalizations, deaths, and tests by neighborhoods and demographics.

Sorting by new cases or incidence rates in the tracker allows to quickly identify the neighborhoods with the newest cases citywide. Interactive maps of new cases and incidence rate by Zip Code enable the identification of clusters of neighborhoods with the most virus cases emerging on a daily basis. Using the 7-day moving average time trend plot of new cases by Zip Code, we can detect neighborhoods with alarming increases in COVID cases in the past weeks. Projections plots visualize the projected trends on new infections, new cases, new hospitalizations, and new deaths in the next 8 weeks. This dashboard has been playing a central role in disseminating timely local COVID information to the lay public in NYC.

DSCovR Dashboard for COVID-19

The Demographics by State COVID-19 Reporting (DSCovR) Dashboard is an interactive tool to track and visualize state-level demographics and time trends for COVID-19 cases, deaths, and policies in the US, as well as, cases and deaths internationally.  The dashboard was developed by a team from Columbia University’s Biostatistics Department: staff member (Aijin Wang), former and current MS students (Christian B. Pascual,  Lauren Franks, Courtney Johnson, Gloria Liu, and Amy Sullivan) and professors Ken Cheung and Shing Lee.  

Using data gathered from state health department websites, USAFacts.org, and the United States Census Bureau’s Annual State Resident Population Estimates, the dashboard allows for easy comparisons of time trends in the number of US new COVID-19 cases and deaths and rate per 100K, as well as the distribution of demographic characteristics (age, sex, and race) by state.  In addition, case and death data from The World Health Organization (WHO), and country-level population data were obtained from the US Census Bureau’s International Database, the dashboard allows for time trend comparisons in the number of US new COVID-19 cases and deaths and rate per 100K across countries. With the re-opening in the summer, in September 2020 information on state-level reopening policies for each US state was obtained from a combination of state government press releases and executive orders. 

The dashboard is supported by the Irving Institute for Clinical and Translational Research, the National Center for Advancing Translational Sciences through Grant Number UL1TR001873.

The Survival Convolution Model for Forecasting the COVID-19 Pandemic

Yuanjia Forecast Model 
By Yuanjia Wang

A survival convolution model (SurvCon) developed in the Department of Biostatistics has been helping the COVID-19 Forecast Hub and US Center of Disease Control (CDC) to forecast cumulative COVID-19 cases and deaths in the near future. The model was created by Professor Yuanjia Wang, Ph.D. student Qinxia Wang, and post-doctoral fellow Shanghong Xie in Biostatistics, together with collaborators at UNC-Chapel Hill. It is now part of the COVID-19 Forecast Hub which predicts the course of the COVID-19 pandemic in the US in real-time and provides weekly forecasts to the CDC. The methodology has been reported in Frontiers in Public Health.  The open-source software is available at the Github website COVID19BIOSTAT (Figure 1).

SurvCon uniquely combines nonparametric statistical curve fitting with underlying mechanisms of infectious disease transmissions known from research in epidemiology. 

The model accurately predicted the national-level apex of the first surge of COVID-19 cases in early April and confirmed a large effect of NPIs implemented immediately after the declaration of a national public health emergency in the US on March 13th. More recent development leverage community mobility data provided by Google and Apple to capture changes in population movements during the pandemic, incorporating census and state-level health data to adjust for confounding, and developing causal inference methods to evaluate NPI effects.

Such refined state-level analysis is expected to reveal the heterogeneity of NPI effects across key indicators of COVID-19 (e.g., the timing of NPI implementation, urban vs suburban areas, race/ethnicity, poverty levels) on transmission rates. These insights will help guide the implementation of precision public health interventions.