Weekly US COVID-19 Cases and Deaths
- Rows
- 10,380
- Columns
- 10
- Data range
- 2020-01-23 to 2023-05-11
- Auto-refresh
- Re-checks its source every 7 days
Weekly US COVID-19 cases and deaths by state/jurisdiction. Time-series mortality data, government source.
AI analysis
Dataset Analysis: Weekly US COVID-19 Cases and Deaths
1. Dataset Overview and Value
This dataset originates from the CDC's public data portal (data.cdc.gov), a highly authoritative government source for epidemiological surveillance data. It captures weekly COVID-19 case and death counts at the US state/jurisdiction level, spanning 173 distinct time periods across 60 jurisdictions — slightly more than 50 states, likely including territories such as Puerto Rico, Guam, and Washington D.C. With 10,380 rows representing the intersection of ~173 weeks × 60 jurisdictions, this is a well-structured panel dataset. Its time-series, multi-jurisdictional nature makes it exceptionally well-suited for correlation studies examining how mortality trends relate to demographic, socioeconomic, healthcare infrastructure, or policy variables across geographies and time.
2. Data Quality Observations
Data quality is notably strong at first glance: zero null cells across all 10 columns and across all 10,380 rows is a remarkable completeness score for a real-world epidemiological dataset. Duplicate row counts are pending final recomputation (Phase D), so that figure should be confirmed before any aggregation analysis — even a small number of duplicates in a cumulative totals column like totcases could meaningfully distort trends. One quality concern worth flagging is the newdeaths column, which has a minimum value of -3,442 — negative deaths are almost certainly retrospective corrections or data reconciliation artifacts rather than true values, and the 1,148 outliers in this column (the highest outlier count of any field) suggest systematic reporting irregularities that must be handled carefully before modeling. Similarly, newhistoriccases and newhistoricdeaths have extreme skewness (both ~64) with medians of 0, indicating these are sparse correction fields rather than analytically dense columns.
3. Key Column Distributions
The totcases column tells a stark story of geographic disparity: with a mean of 832,604 but a median of only 300,937, the distribution is heavily right-skewed (skew = 3.87), with Q1 at ~29,005 and Q3 at ~1,000,660 — meaning large states like California, Texas, and Florida are pulling the upper tail dramatically, producing 914 outliers. The newcases column is even more extreme (skew = 12.0, 1,018 outliers), reflecting the sharp surge-and-recede pattern of COVID waves — its mean of 10,083 versus a median of 3,102 confirms that most weeks in most states were relatively quiet, punctuated by intense outbreak periods. For mortality, totdeaths shows a mean of 10,865 against a median of 4,215 (skew = 2.74), while newdeaths has a mean of only 109 with a median of 35, but a max of 4,915 — confirming that weekly death spikes were severe but episodic. The newhistoriccases column is analytically marginal, with a mean near zero (6.24), a median of 0, and only 9 distinct values — it likely represents late-reported historical corrections and should be excluded from primary trend analysis.
4. Recommended Join Key Columns
The natural composite join key is state + startdate (or equivalently enddate), which together uniquely identify each weekly observation per jurisdiction. The state column (60 distinct values, 0 nulls) aligns well with standard US geographic identifiers such as FIPS codes or postal abbreviations, though a lookup/crosswalk table may be needed depending on how jurisdictions are labeled. For temporal joins, startdate is preferable to dateupdated as the join anchor, since dateupdated reflects data refresh cadence rather than the epidemiological reporting period. For state-level cross-sectional joins (e.g., against Census or election data), state alone serves as the key after aggregating across weeks.
5. Suggested Pairing Datasets for Correlation Discovery
Several dataset categories would unlock powerful correlations with this data:
- Vaccination rates by state/week (also available from CDC) — the most direct causal complement, enabling analysis of whether vaccination uptake correlates with reductions in newdeaths and case fatality rate over time - US Census demographic data (age distribution, population density, income, race/ethnicity by state) — to explore whether mortality rates correlate with demographic vulnerability factors - Hospital capacity and ICU utilization data (HHS Protect) — to test whether healthcare system strain predicts excess mortality spikes - State-level policy tracking data (e.g., Oxford COVID-19 Government Response Tracker or Kaiser Family Foundation policy datasets) — to assess whether mask mandates, lockdowns, or reopening timelines correlate with case and death trajectories - Flu/respiratory illness surveillance data (CDC FluView) — to contextualize COVID mortality seasonality and identify potential co-infection or displacement effects
Columns
- date_updated (date)
- state (string)
- start_date (date)
- end_date (date)
- tot_cases (integer)
- new_cases (integer)
- tot_deaths (integer)
- new_deaths (integer)
- new_historic_cases (integer)
- new_historic_deaths (integer)