Data Visualization: State and County Variation in Drug Poisoning - - PowerPoint PPT Presentation

National Center for Health Statistics

Data Visualization: State and County Variation in Drug Poisoning Mortality in the United States, 1999-2015

Lauren M. Rossen, PhD, MS

Senior Service Fellow Division of Vital Statistics National Center for Health Statistics Centers for Disease Control and Prevention

AcademyHealth 2017 Annual Research Meeting June 24, 2017

Acknowledgments & Disclaimer

• Coauthors:

– Diba Khan, Margy Warner, Brigham Bastian, Yinong Chong – SOURCE: Rossen LM, Bastian B, Warner M, Khan D, and Chong Y. Drug poisoning mortality: United States, 1999–2015. National Center for Health Statistics. 2017.

• Disclaimer:

– The findings and conclusions in this paper are those of the authors and do not necessarily represent the views of the National Center for Health Statistics, Centers for Disease Control and Prevention

Background

Recent NYTimes article projected an increase of 19% from 2015 to 2016 

• Drug overdoses are a major public

health concern in the U.S.

• Drug overdose death rates increased by

~270% from 1999 to 2015

– 6.1 to 16.3 per 100,000 (age adjusted) – Provisional data suggest increases are continuing through the first half of 2016

• 17.5 per 100,000

DATA VISUALIZATION: DRUG OVERDOSE MORTALITY, 1999-2015

NCHS Data Visualization Gallery

• Drug overdose death rates for the U.S., 1999-2015

– Examine differences by

• Sex, age, race and Hispanic origin
• State
• County
• Visualization done using Tableau

– Data tables for download as .csv or through Data.CDC.gov

SOURCE: https://www.cdc.gov/nchs/data-visualization/drug-poisoning-mortality/

DRUG OVERDOSE DEATH RATES IN THE U.S., 1999-2015

• Trends by race and Hispanic origin, sex, and age group

NUMBER OF DRUG OVERDOSE DEATHS IN THE U.S.

1999  2015

NUMBER OF DRUG OVERDOSE DEATHS IN THE U.S.

NOTE:

• X-axes have dynamic scale:
• Above (1999) –
• max of 15,000
• Below (2015) –
• max of 50,000

STATE-LEVEL DRUG OVERDOSE MORTALITY, 1999-2015

In 1999

• Highest in:

– New Mexico (15.0) – Nevada (11.5) – Maryland (11.4)

• Lowest in:

– North Dakota (1.8) – Iowa (1.9) – Nebraska (2.3)

STATE VARIATION IN AGE ADJUSTED DRUG OVERDOSE DEATH RATES

In 2015

• Highest in:

– West Virginia (41.5) – New Hampshire (34.3) – Ohio & Kentucky (29.9)

• All these states had death

rates ~ 4-5 in 1999

• Lowest in:

– Nebraska (6.9) – South Dakota (8.4) – North Dakota (8.6)

Incomplete reporting

STATE VARIATION, cont.

COUNTY-LEVEL DRUG OVERDOSE MORTALITY, 1999-2015

SMOOTHING COUNTY-LEVEL DEATH RATES

• Rates are unstable for counties

with small # deaths and/or populations

• Solid sand-colored line is a large

city, other 4 counties are rural

• Data suppressed for counties with < 20 deaths
• 87% of counties have data suppressed!
• Cannot look at sub-state variation using direct

estimates (i.e., raw data)

METHODS

• Hierarchical two-stage models run in Stata using GLLAMM

(generalized linear latent and mixed models)

– Stage 1: logit model for probability of observing a death – Stage 2: log-linear model for age-adjusted death rates

• Generate empirical Bayes predictions of annual, county-level

age-adjusted death rates (AADR) for 1999-2015

Rossen, LM et al. Trends and Geographic Patterns in Drug-Poisoning Death Rates in the U.S., 1999–2009. American Journal of Preventive Medicine , 45(6):e19 - e25.

DATA VISUALIZATION: COUNTY-LEVEL ESTIMATES

• Can zoom into certain regions/counties:

DISCUSSION

DATA VISUALIZATION: COMMUNICATE HEALTH STATUS IN A SOCIAL CONTEXT

• Provide information relevant

to public health efforts at the state or local level

– Inform efforts to address drug

• verdose mortality
• Shed light on social factors

associated with drug poisoning mortality

– Urban/rural disparities – Socioeconomic factors

SOURCE: https://evergreen.data.socrata.com/stories/s/Opioid-Crisis-County/jd9m-i84d

HOW IS THIS INFORMATION BEING USED?

SOURCE: https://www.gpo.gov/fdsys/pkg/CREC-2016-05-11/pdf/CREC-2016-05-11.pdf

SOME CAVEATS

• County-level estimates have important limitations:

– Can show which areas are higher or lower than others, but should not be used to identify values for any given county or year

• We only release ‘binned’ (i.e., categorized) estimates for that reason
• Outliers will be ‘smoothed’ toward the overall or local mean

– Might not match raw data (i.e., direct estimates) – Might not capture abrupt changes (e.g., pre-post intervention)

– May not be the level of geography we want

• Is county the appropriate unit of geography to capture social context?

– Variation within counties – ‘neighborhood’ or census block level

For more information, contact CDC 1-800-CDC-INFO (232-4636) TTY: 1-888-232-6348 www.cdc.gov The findings and conclusions in this report are those of the authors and do not necessarily represent the

• fficial position of the Centers for Disease Control and Prevention.

Questions?

Email: LRossen@cdc.gov For data access, see: https://www.naphsis.org/research-requests

Extra Slides