Methodology

How OpiEye collects, processes, and presents drug overdose data.

Data last refreshed: July 2026. CDC provisional data is updated quarterly. OpiEye refreshes its data within 30 days of each CDC release. Coverage: 3,147 counties, 104 quarterly periods (2000–2025, through Q4).

What This Data Covers

OpiEye shows all drug overdose deaths at the county level. According to 2024 CDC data (NCHS Data Brief 549), 68% of drug overdose deaths involve opioids and 60% involve fentanyl or other synthetic opioids specifically. The county-level data from the CDC does not break out by drug type, so the map shows total drug overdose deaths.

Why Our Numbers May Differ from Other Sources

If you look up overdose statistics and find different numbers, common reasons include:

Provisional versus finalized data. Our recent data (2024 onward) is provisional: death certificates are still being processed. Final counts are typically 5–15% higher.
12-month rolling versus calendar year. The CDC publishes 12-month rolling counts ending each month. We convert these to quarterly estimates by dividing by four, which smooths seasonal variation.
Different base years. Per capita rates depend on which population estimate is used. We use year-specific Census Bureau estimates.

United States Data Sources

2020 onward: CDC Vital Statistics Rapid Release (VSRR) Provisional County-Level Drug Overdose Death Counts (dataset gb4e-yj24). These are 12-month rolling counts converted to quarterly estimates. Counties with fewer than 10 deaths are suppressed by the CDC for privacy.

2003–2019: National Center for Health Statistics model-based estimates (dataset rpvx-m2md). These use a Bayesian spatial-temporal model that smooths rates across neighboring counties. Individual county figures are approximate; the model pulls small counties toward their neighbors' averages.

2000–2002: NCHS Drug Poisoning Mortality by County, categorical edition (dataset p56q-jrxg). Same Bayesian methodology as rpvx-m2md, but covering the pre-2003 period. Rates are published in 16 categorical ranges rather than as exact values. OpiEye maps each range to its midpoint. Individual county figures are approximate to within about ±1 death per 100,000.

Population: Census Bureau Population Estimates Program, Vintage 2025 county estimates (co-est2025-alldata.csv). Each quarter uses the population estimate for its corresponding year.

Emergency department visits: Estimated at 3–6 times the death count based on published epidemiological ratios. Primary source: Casillas et al., Injury Prevention 2024;30(2):114–124 (PMC10958315). For synthetic opioids (fentanyl), the ratio inverts: fatal overdoses exceed nonfatal ED visits. The 3–6x estimate should not be cited as measured data.

CDC VSRR: data.cdc.gov/NCHS/.../gb4e-yj24
NCHS Mortality 2003–2021 (exact rates): data.cdc.gov/NCHS/.../rpvx-m2md
NCHS Mortality 1999–2016 (categorical): data.cdc.gov/NCHS/.../p56q-jrxg

County Medical Examiner and Coroner Data

For 213 counties, OpiEye sources deaths directly from the local medical examiner, coroner, or state vital-statistics office (the primary authority that investigates each death) instead of CDC provisional data, which lags 6–12 months and undercounts large urban counties. Rows from these feeds are marked (ME) on the map and in county panels, with the exact source named in each county's panel.

Cook County, IL: Medical Examiner Case Archive (updated daily)
San Diego County, CA: Medical Examiner Cases (monthly)
Santa Clara County, CA: Medical Examiner-Coroner total drug deaths (nightly)
San Francisco, CA: OCME preliminary unintentional overdose deaths (monthly aggregate)
Sacramento County, CA: Coroner drug deaths (5-year rolling window)
Allegheny County, PA: Medical Examiner fatal accidental overdoses (daily)
Connecticut (8 counties): Office of the Chief Medical Examiner accidental drug deaths, by county of residence (annual)
Pennsylvania (66 counties): Department of Health overdose deaths by county, residence-based annual death-certificate counts divided evenly across quarters (Allegheny County uses its Medical Examiner case-level feed instead)
Virginia (133 localities): Department of Health overdose deaths by locality, residence-based annual counts divided evenly across quarters

Methodological caveats: Most county ME feeds are occurrence-based (deaths investigated in that county, including non-residents), while CDC counts by decedent residence, so the two differ slightly. Most ME feeds count accidental drug deaths only, while CDC counts all intents (accident, suicide, undetermined), so ME counts can run roughly 10–15% lower by definition. The most recent months of any ME feed are provisional: cases pending toxicology are added over time, so recent counts revise upward. Provisional periods are marked (ME, prov.).

While the map shows all drug overdose deaths, opioids are the primary driver of the crisis. According to 2024 CDC data:

79,384 total drug overdose deaths
54,045 (68%) involved an opioid (ICD-10 codes T40.0–T40.4, T40.6)
47,735 (60%) involved fentanyl or other synthetic opioids (T40.4)

Opioid-involved deaths include heroin (T40.1), fentanyl and synthetic opioids (T40.4), prescription opioids like oxycodone and hydrocodone (T40.2), methadone (T40.3), opium (T40.0), and other unspecified opioids (T40.6).

Source: NCHS Data Brief 549, "Drug Overdose Deaths in the United States, 2023–2024"

The crisis has been easing since its 2022 peak of roughly 108,000 deaths. CDC provisional data for the 12 months ending February 2026 counts 67,531 total drug overdose deaths, with 42,703 (63%) involving an opioid and 36,317 (54%) involving fentanyl or other synthetic opioids. Provisional counts can shift slightly as records are finalized. Site-wide figures (the About page and the live counter on the map) use this provisional 12-month number.

Source: CDC VSRR Provisional Drug Overdose Death Counts, data.cdc.gov/NCHS/.../xkb8-kh2a

International Comparisons

International data comes from each country's national health authority. These comparisons are approximate because countries define and count opioid deaths differently:

Factor	How It Varies
What counts as an opioid	Some countries count only heroin; others include methadone, prescription opioids, and fentanyl analogues
Cause of death attribution	Some require opioids as the underlying cause; others count any mention on the death certificate
Intent	Some include suicides involving opioids; others count only accidental poisonings
Reporting system	Medical examiner or coroner (United States, Canada, Australia), police records (Germany, Italy), national drug monitoring (France), or statistical estimates (European Union)
Completeness	Some systems are acknowledged as non-exhaustive. Some use minimum counts (Italy police-sourced)

Example: Canada reports 17.7 opioid deaths per 100,000 versus the United States at 24.0. But Canada counts "apparent opioid toxicity deaths," while the United States counts deaths with specific ICD-10 opioid codes as a contributing cause. The definitions are close but not identical, so the exact gap between countries is uncertain even though the general pattern is reliable.

European opioid counts are derived from the European Union Drugs Agency (EUDA) by multiplying total drug-induced deaths by the opioid involvement percentage. This introduces rounding and assumes the percentage is consistent across regions within each country.

When you click a country on the map, OpiEye shows the specific definition used for that country so you can judge the comparison yourself.

Data Quality Tiers

Label	Time Period	What It Means
CDC Provisional	2024 onward	Death certificates still being processed. Counts will increase as investigations complete. Expect 5–15% undercount.
CDC Estimated	2020–2023	Derived from 12-month rolling counts divided by four. Seasonal variation is smoothed out.
NCHS Modeled	2003–2019	Statistical estimates that smooth rates across neighboring counties. Good for geographic patterns; less reliable for individual small counties.
NCHS Categorical	2000–2002	Same NCHS modeling, but rates published in 16 categorical ranges instead of exact values. Precise to within about ±1 death per 100,000.

Suppressed counties (fewer than 10 deaths reported to the CDC) are filled with the most recent available modeled estimate. Approximately 462 counties are affected.

Known Limitations

Seasonal patterns are hidden. Drug overdose deaths peak in winter months, but our quarterly conversion distributes deaths evenly.
Small counties are less reliable. Modeled estimates smooth toward regional averages.
Emergency visits are synthetic. The 3–6x multiplier is based on Casillas et al. 2024. The ratio has been declining over time and inverts for fentanyl.
Population estimates. Per capita rates use Census Bureau year-specific estimates, which are themselves estimates with margins of error.
International definitions are not equivalent. Cross-country comparisons show the general scale of the crisis but should not be treated as precise rankings.

How to Cite This Data

OpiEye aggregates and visualizes public data. When citing specific numbers, reference the original source:

United States county data: CDC VSRR Provisional County-Level Drug Overdose Death Counts
United States 2003–2019: NCHS Drug Poisoning Mortality by County, Model-Based Estimates (rpvx-m2md)
United States 2000–2002: NCHS Drug Poisoning Mortality by County, Categorical Rates (p56q-jrxg)
International data: See each country's source listed in the country detail popup

OpiEye is a public health tool, not a primary data source. Our methodology is transparent so you can evaluate the data yourself.