United States COVID-19 Community Levels by County
- Rows
- 206,332
- Columns
- 12
- Data range
- 10001 to 9015
- Auto-refresh
- Re-checks its source every 30 days
Time-series data of COVID-19 community levels (Low, Medium, High) used for public health recommendations by county.
AI analysis
Dataset Analysis: US COVID-19 Community Levels by County
1. Dataset Overview and Value This dataset, sourced from the CDC's public data API at [data.cdc.gov](https://data.cdc.gov/api/views/3nnm-4jni/rows.csv), represents a time-series of COVID-19 community-level classifications (Low, Medium, High) tracked across US counties over 64 distinct reporting periods. With 206,332 rows spanning 3,224 unique county FIPS codes and 56 state/territory values, it offers near-complete national coverage including territories beyond the 50 states. The CDC's authorship makes this a high-credibility public health source, and the 64 date snapshots suggest roughly weekly or bi-weekly reporting cadence spanning one to two years of the pandemic's managed phase. Its combination of case rates, hospitalization metrics, and bed utilization makes it exceptionally valuable for multi-dimensional correlation studies linking community transmission to healthcare system stress.
2. Data Quality Observations Overall data quality is strong. Total null cells across all 12 columns number only 812 out of ~2.5 million total cells (~0.03%), indicating a well-maintained administrative dataset. The three columns carrying the most nulls — covidinpatientbedutilization (346, 0.2%), covidhospitaladmissionsper100k (214, 0.1%), and covid-19communitylevel (206, 0.1%) — are noteworthy because they are the analytically critical columns. Their missingness likely clusters around specific rural counties or early reporting periods where hospital-level data collection was incomplete, which could introduce systematic bias if not accounted for. The single null in countypopulation is negligible. Duplicate row counts are pending Phase D Parquet recomputation and should be validated before any aggregation, as time-series datasets from API sources are prone to ingestion duplicates across overlapping pull windows. The covid-19communitylevel column's 7 distinct values (versus the expected 3) warrants inspection — likely artifacts include "N/A", blank strings, or legacy classification labels that should be harmonized before categorical analysis.
3. Key Column Distributions The three core epidemiological metrics reveal telling distributional patterns. covidcasesper100k is the most volatile: mean of 113.5 versus a median of 77.4 signals strong right skew (skew=17.7), with 10,531 outliers and a max of 13,017.8 — likely reflecting acute surge periods in specific counties. covidhospitaladmissionsper100k similarly skews right (skew=3.24, mean=7.2 vs. median=5.9, max=185.3) with 6,844 outliers, suggesting episodic regional hospital crises. covidinpatientbedutilization is comparatively well-behaved (skew=1.75, mean=3.03%, median=2.6%, IQR of 1.3–4.2%) but still has 6,072 outliers reaching a maximum of 36%, which could represent small critical-access hospitals overwhelmed during peaks. countypopulation is dramatically skewed (skew=13.6, mean=102,920 vs. median=26,118, with 28,287 outliers), reflecting the well-known US population concentration in a handful of large urban counties — a critical confound to control for in any per-county analysis. healthserviceareapopulation follows a similar pattern (skew=3.77, 25,408 outliers), consistent with major metro-anchored hospital catchment areas.
4. Recommended Join Key Columns countyfips is the primary recommended join key — it is integer-typed, has zero nulls, and with 3,224 distinct values it aligns well with standard US Census FIPS codes, enabling joins to virtually any federal or state administrative dataset. The combination of countyfips + dateupdated forms a composite natural key for time-aligned joins, which is essential when correlating against other time-series sources (e.g., vaccination rates, mobility data). state (56 distinct, zero nulls) serves as a reliable secondary join for state-level aggregations, though its 56 values confirm inclusion of territories (Puerto Rico, Guam, etc.) that may lack counterpart records in some join targets. healthserviceareanumber provides a useful intermediate geographic grain between county and state for healthcare-system-centric joins.
5. Recommended Companion Datasets for Correlation Analysis Several dataset pairings would unlock high-value correlations. CDC Vaccination Coverage by County (also on data.cdc.gov) would allow direct testing of whether higher vaccination rates correlate with lower community level classifications and reduced hospitalization metrics. US Census Bureau ACS demographic data (joinable via countyfips) would enable analysis of how age structure, poverty rates, and racial composition correlate with case severity and bed utilization — important for equity analysis. CMS Medicare/Medicaid hospital capacity data joined on healthserviceareanumber could reveal whether pre-existing hospital capacity predicts bed utilization spikes. Google or Apple COVID-19 Mobility Reports joined on FIPS + date would test whether reduced mobility preceded or followed community level escalations, a classic leading-indicator analysis. Finally, BLS unemployment or economic impact data by county and date would support socioeconomic burden studies correlating pandemic severity with labor market disruption.
Columns
- county_fips (integer)
- county (string)
- state (string)
- county_population (integer)
- health_service_area_number (integer)
- health_service_area (string)
- health_service_area_population (integer)
- covid_inpatient_bed_utilization (decimal)
- covid_hospital_admissions_per_100k (decimal)
- covid_cases_per_100k (decimal)
- covid-19_community_level (string)
- date_updated (date)