Changing COVID-19 cases and deaths detection in Florida

Abstract

Epidemic data are often difficult to interpret due to inconsistent detection and reporting. As these data are critically relied upon to inform policy and epidemic projections, understanding reporting trends is similarly important. Early reporting of the COVID-19 pandemic in particular is complicated, due to changing diagnostic and testing protocols. An internal audit by the State of Florida, USA found numerous specific examples of irregularities in COVID-19 case and death reports. Using case, hospitalization, and death data from the the first year of the COVID-19 pandemic in Florida, we present approaches that can be used to identify the timing, direction, and magnitude of some reporting changes. Specifically, by establishing a baseline of detection probabilities from the first (spring) wave, we show that transmission trends among all age groups were similar, with the exception of the second summer wave, when younger people became infected earlier than seniors, by approximately 2 weeks. We also found a substantial drop in case-fatality risk (CFR) among all age groups over the three waves during the first year of the pandemic, with the most drastic changes seen in the 0 to 39 age group. The CFR trends provide useful insights into infection detection that would not be possible by relying on the number of tests alone. During the third wave, for which we have reliable hospitalization data, the CFR was remarkably stable across all age groups. In contrast, the hospitalization-to-case ratio varied inversely with cases while the death-to-hospitalization ratio varied proportionally. Although specific trends are likely to vary between locales, the approaches we present here offer a generic way to understand the substantial changes that occurred in the relationships among the key epidemic indicators.

Fig 1. Weekly cases, hospitalizations, and deaths among different data sources.

(a) All three case data sources are indexed using case reporting date. (b) Weekly hospital admissions were indexed by different dates: case reporting date for FDOH linelist, admission reporting date for FDOH daily report, and day of admission for HHS data. (c) Weekly deaths indexed by different dates: case reporting date for FDOH linelist, death reporting date for FDOH daily report, and day of death for CDC data.

Fig 2. Population-wide and age-stratified normalized weekly COVID-19 cases and deaths and their weekly median age by case reporting date from Mar 2020 to Apr 2021.

Cases and deaths are normalized by their corresponding mean weekly count during the first wave. All time series are derived from the State linelist and indexed according to the case reporting date. Dashed horizontal lines indicate a normalized count of 1. (a) Weekly normalized COVID-19 cases and deaths, (b) weekly median age of cases and deaths, and the age-stratified weekly normalized (c) cases and (d) deaths.

Fig 3. Case-fatality risk (CFR) over the first year of the pandemic.

(A) Weekly CFR (dots) and the 3-week moving average (line) for each age group. Note that the y-axis is log-scaled. (B) Average CFR over each wave for each age group.

Fig 4. Weekly number of SARS-CoV-2 tests reported by Florida.

Fig 5. Relationship between weekly case loads and weekly case fatality ratio, hospitalization-to-case ratio and death-to-hospitalization ratio during the third wave.

All indicators are indexed using case reporting date. Cases and deaths were derived from the state linelist. Hospitalizations were based on HHS data. We assumed that the admission date lags the case reporting date by a week, and adjusted the hospitalizations time series accordingly. Dashed lines are linear trendlines. Only the third wave is investigated since the first two waves’ hospitalization data was unreliable.

Fig 6. Comparison between excess deaths and reported COVID-19 deaths, from Mar 2020 to May 2022.

(a) Excess deaths and reported COVID-19 deaths. We calculate the excess deaths as the observed all-cause deaths minus the expected weekly all-cause deaths based on data from 2015 to Feb 2020. (b) Excess deaths not accounted for by reported COVID-19 deaths.