Variations in state-level SARS-COV-2 testing recommendations in the United States, March-July 2020

Abstract

Background:

Testing recommendations for COVID-19 in the United States varied by state and over time in the spring and summer of 2020.

Methods:

We compiled data about COVID-19 testing, cases, and deaths, and excess pneumonia + influenza + COVID-19 deaths to assess relationships between testing recommendations, per capita tests performed, epidemic intensity, and excess mortality during the early phase of the COVID-19 pandemic in the United States.

Results:

As of July 2020, 16 states recommended testing asymptomatic members of the general public. The rate of COVID-19 tests reported in each state correlates with more inclusive testing recommendations and with higher epidemic intensity. Higher per capita testing was associated with more complete reporting of COVID-19 deaths, which is a fundamental requirement for analyzing the pandemic.

Conclusions:

Reported deaths due to COVID-19 likely represent an undercount of the true burden of the pandemic. Coordinated, consistent guidelines for COVID-19 testing should be a high priority for state and national health systems.

Background

SARS-CoV-2, the virus that causes COVID-19, was declared a global pandemic by the World Health Organization in March 2020.¹ Worldwide, responses to the pandemic have varied and had varying success. In the United States, each state determined testing guidelines, lockdown procedures, and reopening procedures individually.² The Centers for Disease Control and Prevention also maintained testing recommendations and guidelines, but states and localities were largely responsible for their own policies.³ The COVID-19 pandemic in the United States is ongoing, costing around 150,000 lives by the end of July 2020.⁴

Testing for COVID-19 has been a continual hurdle in the United States.⁵ Supply chain issues,⁶ doubts about testing accuracy and methodology,⁷ and conflicting guidance from federal, state and local officials about testing have all interfered with testing efforts.⁸ State governments and health departments established testing guidelines in March 2020, which were sometimes augmented by local governmental health departments.⁹ The turnaround time for test results has been strained,¹⁰ and resources for individuals to sustain their household resources while remaining isolated or quarantined while awaiting the result of a test have not been widely available.¹¹

Here we describe the changing recommendations for COVID-19 testing to the general public in the United States in the spring and summer of 2020 and the consequences of these changes for obtaining accurate statistics about the pandemic.

Methods

The objective of this analysis was to establish whether varying the state-level COVID-19 testing guidelines had any impact on the actual number of tests performed, and whether the number of tests performed was related to the rate of recognizing COVID-19 deaths as a cause of excess mortality (reported COVID-19 deaths/excess pneumonia + influenza+ COVID-19 (PIC) deaths). COVID-19 testing guidelines were documented from the websites of the departments of health in each of the 50 states and the District of Columbia once monthly from March-July 2020 (Supplemental Table 1). We ranked recommendations to tests by the general public (as opposed to members of specific named risk groups) by assigning 0= criterion not specified, 1= accessible to people with a negative influenza test, 2= accessible to people with symptoms consistent with COVID-19 illness, 3= accessible to anyone, including asymptomatic people.

Correlations between testing recommendation score, tests per capita, and percentage of positive tests were analyzed with Spearman correlation tests.

We used a generalized additive model (negative binomial link) to assess the completeness of death reporting. The outcome variable was the number of reported COVID-19 deaths in each state and month. The number of excess PIC deaths was included as an offset term, so this effectively models the ratio of reported COVID-19 deaths/excess PIC deaths. The per capita testing rate in each state and month was included as a penalized spline. Uncertainty in the estimates of excess PIC deaths were ignored in these analyses, and excess PIC deaths were constrained to be greater than or equal to the reported COVID-19 deaths. The model was fit using the package “mgcv” in R.

We also evaluated relationships between the number of COVID-19 tests performed per capita, testing recommendations score, and epidemic intensity (defined as excess PIC deaths¹² per capita) using a negative binomial regression model. Total COVID-19 tests performed in each state was the outcome variable, offset by the log of the total population of each state, with the testing recommendations score (categorical) and the log of epidemic intensity as the independent variables.

The number of excess PIC deaths per state and the reported number of COVID-19 deaths were obtained from a repository describing excess deaths during the COVID-19 outbreak (https://github.com/weinbergerlab/excess_pi_covid).¹² Final data were obtained August 15, 2020. Positive tests and the overall number of tests in each state were collected from the website of The COVID Tracking Project (https://covidtracking.com/). Population data were taken from the United States Census Bureau’s 2019 estimates. R (R: A language and environment for statistical computing, version 3.6.1. R Foundation for Statistical Computing, Vienna, Austria, https://www.R-project.org/.) was used for statistical analyses.

Results

State Testing Guidelines

From March to July 2020, recommendations of COVID-19 testing for the general public varied between states and over time (Figure 1), as did testing prioritization for identified risk groups (Supplemental Figure 1).

Figure 1.

COVID-19 testing guidelines by state. Testing recommendations for the general public was scaled as follows: 0 = criterion not included, 1 = negative influenza test required, 2 = all symptomatic people, 3 = all people, including asymptomatic people. Testing guidelines for special risk groups are shown in Supplemental Figure 1.

The earliest-available state-level data ranged from March 1 (in Michigan, Rhode Island, and Washington) to March 12 (in Massachusetts), while the remaining 47 states began reporting COVID-19 data between March 4 and March 7. March data included all March dates – April 4; April data included April 5 – May 2; May data included May 3 – June 6; June data included June 7 – July 4; July data included records from July 5 – 31.

As of July, 16 states recommended that asymptomatic members of the general public be tested for COVID-19. 9 states actively recommended against the testing of asymptomatic people; 3 of which then reversed this recommendation in July. South Dakota also recommended against seeking a test if people were experiencing symptoms, citing the lack of effective COVID-19 therapy as a contraindication for testing. Most states (35/51) changed their testing recommendations twice between March and July. Only Oklahoma did not change their testing criteria, though the state health department website specifies that private testing facilities have their own criteria.

Factors associated with higher rates of testing

Reported COVID-19 tests per capita increased steadily, though unevenly, from March to June, while the percentage of positive tests varied widely between states (Figure 2, 3). States with more permissive testing recommendations had a lower percentage of positive tests (rho −0.22, 95% CI −0.34, −0.11), reflecting the inclusion of lower risk individuals in the testing pool. The minimum percentage of positive tests for a single month was 0.8% in May in Alaska; the maximum was 47% in April in New Jersey. In a multivariate model, both more inclusive testing recommendations and higher epidemic intensity were associated with higher rates of testing.

Figure 2.

COVID-19 tests per 1000 population in the United States (A), and the percentage of positive COVID-19 tests (B), March–July 2020. Each state has a line; colors of lines are determined by geographic region per the inserted map.

Figure 3.

COVID-19 testing recommendation scores (A) and COVID-19 tests per 1000 people (B) from March–July, 2020 in the United States

Testing intensity and completeness of death reporting

The completeness of recording deaths as due to COVID-19 was related to the rate of testing. (Figure 4). This was particularly true in the first months of the pandemic when testing levels were low relative to the intensity of the epidemic; in later months, the level of testing was high and had little influence on the difference between reported COVID-19 deaths and excess PIC deaths.

Figure 4.

Association between the rate of testing for SARS-CoV-2 (per capita) and the completeness of death reporting, as measured by the ratio between reported COVID-19 deaths and excess deaths due to pneumonia/influenza/coronavirus (PIC) The line +/− 95% confidence intervals represents the trend as estimated from a generalized additive model The size of the circles is proportional to the number of excess deaths due to PIC in each state. The color depicts the month (March–July)—earlier months are darker colors, later months are lighter.

Discussion

In order to obtain accurate statistics on the COVID-19 pandemic, viral testing needs to scale with epidemic intensity. Particularly in the early months of the pandemic, restrictive state-level testing recommendations and difficulties in accessing tests resulted in inadequate testing levels.⁵ This was manifest in higher viral positivity rates and greater differences between reported COVID-19 deaths and excess PIC deaths. From March-May, testing per capita and completeness of mortality data both increased, though afterwards this apparent relationship became less clear. It is possible that other characteristics of healthcare systems in each state influenced both the testing rates and completeness of death reporting (i.e. supply chain or personnel issues). The variation in data completeness seen in July may be due to reporting delays, and thus might decrease as mortality data are updated. Expanded access to testing can play a critical role in obtaining an accurate picture of the progression of the epidemic.

This study has certain limitations. Changes in completeness of death reporting could be influenced by other factors that might also vary over time, such as state-level guidelines for coding deaths as due to COVID-19. The viral testing criteria established by state health departments are sometimes superseded by local governments or private clinics.⁹ The testing data counts tests, not individuals, so people tested multiple times will be included multiple times.

Conclusions

The patchwork of state-level testing guidelines is associated with local variations in testing intensity, and these variations have consequences for the completeness of data on key indicators of the severity and progression of the pandemic. More coordinated national guidance could strengthen the pandemic tracking and response.

List of Abbreviations

PIC

pneumonia + influenza + COVID-19

CI

confidence interval