RESEARCH ARTICLE

Geographic Variation in Opioid and Heroin Involved
Drug Poisoning Mortality Rates
Christopher J. Ruhm, PhD
Introduction: An important barrier to formulating effective policies to address the rapid rise in U.S.
fatal overdoses is that the specific drugs involved are frequently not identified on death certificates.
This analysis supplies improved estimates of state opioid and heroin involved drug fatality rates in
2014, and changes from 2008 to 2014.
Methods: Reported mortality rates were calculated directly from death certificates and compared to
corrected rates that imputed drug involvement when no drug was specified. The analysis took place
during 2016–2017.

Results: Nationally, corrected opioid and heroin involved mortality rates were 24% and 22%
greater than reported rates. The differences varied across states, with particularly large effects in
Pennsylvania, Indiana, and Louisiana. Growth in corrected opioid mortality rates, from 2008 to
2014, were virtually the same as reported increases (2.5 deaths per 100,000 people) whereas changes
in corrected heroin death rates exceeded reported increases (2.7 vs 2.3 per 100,000). Without
corrections, opioid mortality rate changes were considerably understated in Pennsylvania, Indiana,
New Jersey, and Arizona, but dramatically overestimated in South Carolina, New Mexico, Ohio,
Connecticut, Florida, and Kentucky. Increases in heroin death rates were understated in most states,
and by large amounts in Pennsylvania, Indiana, New Jersey, Louisiana, and Alabama.
Conclusions: The correction procedures developed here supply a more accurate understanding of
geographic differences in drug poisonings and supply important information to policymakers
attempting to reduce or slow the increase in fatal drug overdoses.
Am J Prev Med 2017;](]):]]]–]]]. & 2017 American Journal of Preventive Medicine. Published by Elsevier Inc.
All rights reserved.

INTRODUCTION
he U.S. is “experiencing an epidemic of drug
overdose (poisoning) deaths,” with fatal drug
poisonings rising 137% from 2000 to 2014.1
Increases in poisoning deaths, around 90% of which
are now caused by drugs, were the most important source
of the growth in the all-cause mortality rates of nonHispanic whites aged 45–54 years occurring between
1999 and 2013.2,3 The involvement of opioids in these
deaths has received particular attention, including a
White House Summit in August 2014.1,4–8
The rapid rise in fatal drug poisonings justifies the
concerted efforts undertaken to reduce them, including
establishing prescription drug monitoring programs;
restricting the ability of pain clinics and online pharmacies to dispense oxycodone and other controlled

T

substances; and developing abuse-deterrent formulations
of some prescription drugs.9–13 The federal Comprehensive Addiction and Recovery Act of 2016 (S. 524)
supports expansions of drug diversion programs (reducing the criminality of low-level drug violations),
medication-assisted treatments, and naloxone administration for opioid overdoses.
These efforts have been partially but not completely
successful.1,5,13–16 An important barrier to formulating
From the Frank Batten School of Leadership and Public Policy, University
of Virginia, Charlottesville, Virginia
Address correspondence to: Christopher J. Ruhm, PhD, Frank Batten
School of Leadership and Public Policy, University of Virginia,
235 McCormick Road, P.O. Box 400893, Charlottesville VA 22904-4893.
E-mail: ruhm@virginia.edu.
0749-3797/$36.00
https://doi.org/10.1016/j.amepre.2017.06.009

& 2017 American Journal of Preventive Medicine. Published by Elsevier Inc. All rights
reserved.

Am J Prev Med 2017;](]):]]]–]]] 1

 2

Ruhm / Am J Prev Med 2017;](]):]]]–]]]

the most effective policies to deter the dangerous use of
prescription pharmaceuticals, while avoiding the potential substitution with other harmful legal or illegal drugs,
is the lack of reliable information on the drugs causing
fatal overdoses. This occurs when no specific drug is
identified on the death certificates, as happened in one
fifth to one quarter (depending on the year) of drug
deaths.17 This leads to an underestimate of the involvement rates of specific drugs at a point in time and
additional errors when measuring changes in drug
involvement rates across time. The lack of specificity
on the reporting of drug involvement has previously been
recognized, and initial efforts have been made to analyze
the resulting errors.17–20 Less attention has been given to
the inaccuracy of existing information on geographic
variations in drug involvement; previously reported
results examining such differences are likely to be
misleading because the specificity in reporting drug
involvement varies across locations.21,22
This analysis supplies improved estimates of state-level
opioid and heroin involved drug fatality rates in 2014, as
well as changes in these rates occurring between 2008 and
2014. Specifically, reported drug involved death rates
are calculated directly using information from death
certificate reports, as has commonly been done previously, and these are then compared with corrected rates
that imputed information on drug involvement in cases
where no drug category was identified on the death
certificate. The corrected rates provide the best currently
available information on geographic variation in
opioid and heroin involved fatality rates. Disparities
between the corrected and reported rates indicate errors
resulting from previous analyses that rely only on the
latter.

METHODS
The outcomes analyzed were opioid and heroin involved drug
fatality rates by state, per 100,000 people, in 2014 and changes in
these rates from 2008 to 2014. Counts of drug deaths among U.S.
residents were obtained from the 2008 and 2014 Centers for
Disease Control and Prevention Multiple Cause of Death (MCOD)
files and were analyzed in 2016–2017.23 The MCOD data provided
information from death certificates on a single underlying cause of
death, up to 20 additional causes, and limited demographic data.
The cause-of-death information were categorized using four-digit
ICD-10 codes with data also provided on age, race/ethnicity,
gender, year, and weekday and place of death.24 The public use files
lack geographic identifiers; however, information on the county
and state of residence are available under restricted conditions and
were obtained for use in this study. Drug poisoning deaths include
ICD-10 underlying cause of death codes X40–X44, X60–X64, X85,
and Y10–Y14. The IRB for the Social and Behavioral Sciences at
the University of Virginia reviewed this project and determined
that it did not involve human subjects.

For drug poisoning mortality, the death certificate also lists one
or more drugs involved as immediate or contributory causes of
death, included separately as ICD-10 “T-codes.” T-codes 40.0–40.4
and 40.6 indicate the involvement of opioids and T-code 40.1
refers to heroin. However, for around half of overdose fatalities,
unspecified drugs, medicaments, and biologicals (ICD-10 code
T50.9) were mentioned, and this was the only designation in one
fifth to one quarter of cases (depending on the year). Data from
2014 were analyzed because this was the latest year with available
information in the MCOD files at the time of initial analysis.
Changes from 2008 to 2014 were examined because the specificity
of drug reporting increased fairly steadily over this period.
Numbers of deaths were converted to mortality rates per
100,000 people, using population data from the Surveillance
Epidemiology and End Results (SEER) program.25

Statistical Analysis
Reported mortality rates were defined using only information
contained on death certificates, and so did not attribute drug
involvement in fatal overdoses where only unspecified drugs (Tcode 50.9) were mentioned. Corrected rates were obtained by
using information from death certificate reports where at least one
specific drug category was identified to impute drug involvement
for cases where drug involvement categories were left unspecified.
The imputations were done using the following procedure. First,
year-specific probit models were estimated, by maximum likelihood, for the sample of fatal overdoses where at least one drug
was specified on the death certificate. The dependent variables in
these models were equal to 1 if opioids or heroin, respectively, were
mentioned and to 0 if not. Dichotomous individual explanatory
variables included sex, race categories (white, black, other nonwhite), Hispanic origin, marital status (currently married versus
not), education categories (high school dropout, high school
graduate, some college, college graduate), age categories (r30,
31–40, 41–50, 51–60, 61–70, 71–80, 480 years), day-of-the week
indicators, location of death categories (hospital inpatient, hospital
outpatient/ED, dead on arrival at hospital/ED, home, other), and
interactions between sex X race/ethnicity. The following 2010-year
county level characteristics were also controlled for: poverty rates,
education shares (same categories as above), percentage of households headed by females, median income, population per square
mile and its square, and physicians per 1,000 people. Data sources
and additional details for these variables are provided in the
Appendix (available online). Second, predicted probabilities of
opioid or heroin involvement were imputed, using the probit
estimates, for deaths where only unspecified drugs were mentioned. Corrected mortality rates were then calculated using the
predicted values in these cases and reported involvement otherwise
to estimate the total number of deaths involving opioids or heroin
and then dividing by population.
All analyses were conducted using STATA, version 14.

RESULTS
Fatal overdoses nationwide to U.S. residents were 36,450
in 2008 and 47,055 in 2014. Residents of foreign
countries dying in the U.S. were excluded from the
analysis (residence is defined by the place where the
www.ajpmonline.org

 Ruhm / Am J Prev Med 2017;](]):]]]–]]]

decendent actually resided, not by citizenship or legal
status).
A specific drug was not identified for 19.5% of fatal
overdoses in 2014 and 25.4% in 2008. These patterns
varied dramatically across states. For instance, in 2014, a
drug category was mentioned for over 99% of drug
poisoning deaths in Rhode Island, Connecticut, and New
Hampshire, but only around half the time in Pennsylvania, Indiana, Mississippi, Louisiana, and Alabama. Rates
of non-reporting generally tended to be low in parts of
the Northeast and the West, but high for much of the
South. However, there were exceptions. For example,
specificity of reporting was relatively low in Pennsylvania, New Jersey, and the upper Mountain states (Montana, Idaho, and Wyoming), but quite high for several
South Atlantic states (particularly Virginia, West Virginia, and South Carolina). Further details are available
in Appendix Table 1 and Appendix Figure 1 (both
available online). As mentioned, reported drug involvement rates understate the true rates, in both absolute and
relative terms, in states that frequently list only the
unidentified drug category on overdose death certificates.
Changes over time in the specificity of reporting also
deserve attention. For example, in several states where a
specific drug was identified in over 90% of 2014 overdose
deaths, considerably less detail was provided in 2008.
Particularly noteworthy are Connecticut, New Mexico,
South Carolina, and South Dakota, which had 15 to 34
percentage point increases in reporting rates between
2008 and 2014, with specific drug categories identified in
494% of cases in the later year. A number of other states
(Ohio, Tennessee, Georgia) also had substantial trend
increases in the specificity of reporting, but with drug
involvement still unidentified for 410% of drug fatalities
in 2014. These changes are mapped in Appendix Figure 2
(available online).
Table 1 shows reported and corrected 2014 opioid and
heroin involved overdose death rates per 100,000 people,
as well as differences between the two. States are ordered
from largest to smallest disparities between the corrected
and reported rates, with state mortality rate rankings
included in parentheses.
Nationally, corrected opioid involved mortality rates
were 24% greater than reported rates in 2014 (11.2 vs 9.0
per 100,000) and those for heroin were 22% more (4.0 vs
3.3 per 100,000). However, the differences were much
bigger in Pennsylvania (opioids¼108%; heroin¼107%),
Indiana (opioids¼103%; heroin¼89%), Louisiana (opioids¼125%; heroin¼103%), Alabama (opioids¼108%;
heroin¼61%),
and
Mississippi
(opioids¼107%;
heroin¼139%). In absolute terms, corrected opioid death
rates exceeded reported rates per 100,000 by 9.2 in
Pennsylvania (heroin¼4.2), 7.3 in Indiana (heroin¼2.3),
] 2017

3

7.0 in Louisiana (heroin¼2.4), 6.0 in Alabama
(heroin¼1.5), and 4.1 in Mississippi (heroin¼1.1). Conversely, reported and corrected mortality rates were
almost identical for most New England States.
State mortality rankings also changed substantially.
For instance, Pennsylvania had the 32nd highest reported
opioid mortality rate and the 20th highest reported
heroin mortality rate, but ranked 7th and 4th based on
corrected rates. Similarly, Indiana’s rankings moved from
36th and 29th to 15th and 19th, and Louisiana’s from
40th and 31st to 21st and 20th. In 19 states, corrected and
reported opioid rankings differed by at least five places
and in eight states this occurred for heroin.
Figure 1 shows that corrected rates (on the right)
yielded a more coherent geographic pattern than
reported rates (on the left). Specifically, the corrected
death rates demonstrate that opioid involved mortality
was concentrated in the Mountain States, Rust Belt, and
Industrial North—extending to New England—and
much of the South, whereas heroin deaths were particularly high in the Northeast and Rust Belt, but less so in
the South or Mountain States. The results were less
apparent when using reported rates, because high mortality in states such as Pennsylvania and Indiana were
concealed by a frequent lack of specificity about drug
involvement on death certificates.
Table 2 focuses on changes in reported and corrected
opioid and heroin involved mortality rates between 2008
and 2014. States are ordered from largest to smallest
disparities between reported versus corrected changes,
with state rankings again shown in parenthses. For the
entire U.S., the growth in opioid involved drug deaths
was essentially the same when using reported rather than
corrected rates (2.5 vs 100,000 in both cases), whereas the
trend increase in heroin involved mortality was underestimated by around 18% (2.3 vs 2.7 per 100,000).
The errors were often much larger, and in varying
directions, at the state level. For instance, reported rates
substantially understated the rise in opioid mortality—by
1.5 to 3.1 per 100,000—in Pennsylvania, Indiana, New
Jersey, and Arizona, and drastically overstated it—by 1.7
to 3.0 per 100,000—in Connecticut, Ohio, New Mexico,
and South Carolina. Rank orderings of corrected versus
reported rates changed by five places or more in ten
states. On the other hand, the corrected and reported
changes in rates differed by less than 0.2 per 100,000 in
18 states.
When based on death certificate reports, growth in
heroin mortality was generally understated by substantial
amounts (1.0 to 3.2 per 100,000) in Pennsylvania,
Indiana, New Jersey, Louisiana, and Alabama, and
moderately (by 0.5 to 0.9 per 100,000) in nine other
states. Increases in heroin involved mortality were

 Ruhm / Am J Prev Med 2017;](]):]]]–]]]

4

Table 1. Opioid and Heroin Involved Drug Poisoning Death Rates by State, 2014
Any opioid
State
Pennsylvania
Indiana
Louisiana
Alabama
Kentucky
Mississippi
Michigan
Wyoming
New Jersey
Delaware
Idaho
Arizona
Montana
Florida
Missouri
Arkansas
Colorado
Ohio
Kansas
Tennessee
California
Georgia
Texas
Hawaii
Nebraska
Alaska
Wisconsin
Illinois
Minnesota
Oklahoma
North Carolina
Nevada
South Carolina
Oregon
Iowa
Washington
New York
West Virginia
New Mexico
Washington DC
Maryland
Utah
North Dakota
South Dakota
Massachusetts
Virginia
New Hampshire
Maine
Vermont

Heroin

Reporteda

Correcteda

Difference

State

Reporteda

Correcteda

Difference

8.5 (32)
7.0 (36)
5.6 (40)
5.6 (41)
16.5 (7)
3.8 (50)
10.6 (19)
9.2 (27)
8.1 (33)
13.3 (11)
4.8 (45)
8.8 (30)
5.2 (43)
7.0 (34)
11.5 (16)
5.8 (38)
9.7 (23)
18.2 (5)
6.0 (37)
13.2 (13)
5.2 (42)
7.0 (35)
4.3 (46)
4.2 (48)
3.0 (51)
10.3 (20)
10.9 (17)
9.4 (26)
5.8 (39)
12.9 (14)
9.7 (22)
13.2 (12)
10.7 (18)
8.6 (31)
5.1 (44)
9.5 (25)
8.8 (29)
29.9 (1)
19.3 (4)
9.6 (24)
15.4 (9)
15.5 (8)
4.2 (47)
3.9 (49)
16.9 (6)
9.1 (28)
22.4 (2)
12.9 (15)
10.2 (21)

17.8 (7)
14.3 (15)
12.6 (21)
11.6 (26)
20.8 (3)
8.0 (43)
14.7 (14)
13.3 (19)
11.7 (25)
16.8 (9)
8.2 (41)
12.1 (22)
8.5 (39)
9.7 (34)
14.0 (17)
8.3 (40)
12.0 (24)
20.5 (4)
8.1 (42)
15.1 (12)
6.9 (44)
8.6 (38)
5.9 (46)
5.5 (48)
4.3 (50)
11.5 (27)
12.1 (23)
10.4 (30)
6.7 (45)
13.8 (18)
10.6 (29)
14.0 (16)
11.3 (28)
9.2 (37)
5.6 (47)
10.0 (32)
9.2 (36)
30.3 (1)
19.7 (5)
9.9 (33)
15.8 (11)
15.8 (10)
4.5 (49)
4.1 (51)
17.1 (8)
9.3 (35)
22.6 (2)
13.0 (20)
10.4 (31)

9.2
7.3
7.0
6.0
4.3
4.1
4.1
4.1
3.6
3.5
3.4
3.4
3.3
2.7
2.5
2.5
2.3
2.3
2.2
2.0
1.7
1.6
1.6
1.3
1.3
1.2
1.2
1.1
0.9
0.8
0.8
0.7
0.7
0.6
0.5
0.4
0.4
0.4
0.4
0.4
0.4
0.4
0.3
0.3
0.2
0.2
0.2
0.2
0.1

Pennsylvania
Louisiana
Indiana
New Jersey
Delaware
Alabama
Michigan
Kentucky
Mississippi
Wyoming
Florida
Ohio
Montana
Idaho
Arizona
Colorado
Arkansas
Missouri
Kansas
Georgia
Tennessee
Hawaii
California
Texas
Alaska
Washington DC
Nebraska
Wisconsin
Illinois
Minnesota
South Carolina
Nevada
North Carolina
Iowa
Maryland
New York
Oregon
Oklahoma
West Virginia
Washington
Massachusetts
North Dakota
Utah
New Hampshire
New Mexico
Virginia
South Dakota
Vermont
Connecticut

3.9 (20)
2.3 (31)
2.6 (29)
4.7 (16)
5.8 (8)
2.5 (30)
5.3 (12)
5.2 (15)
0.8 (43)
1.7 (36)
1.7 (35)
10.4 (1)
0.3 (47)
0.7 (44)
2.9 (25)
2.9 (26)
0.2 (49)
5.5 (11)
0.7 (46)
1.5 (38)
2.3 (32)
0.9 (42)
1.5 (39)
1.6 (37)
3.4 (22)
5.6 (9)
0.3 (48)
4.7 (17)
5.5 (10)
1.8 (34)
1.3 (40)
2.3 (33)
2.7 (28)
1.2 (41)
5.2 (14)
4.2 (18)
3.1 (23)
0.7 (45)
8.8 (2)
4.1 (19)
7.0 (5)
0.1 (51)
3.7 (21)
7.4 (4)
6.7 (6)
3.0 (24)
0.2 (50)
5.3 (13)
8.3 (3)

8.1 (4)
4.7 (20)
5.0 (19)
6.4 (10)
7.3 (6)
4.1 (23)
6.6 (9)
6.3 (12)
1.8 (40)
2.8 (32)
2.6 (34)
11.2 (1)
1.1 (46)
1.4 (42)
3.7 (26)
3.6 (27)
0.9 (47)
6.0 (13)
1.2 (45)
2.0 (37)
2.7 (33)
1.3 (44)
1.8 (39)
2.0 (38)
3.7 (25)
6.0 (14)
0.6 (49)
5.0 (18)
5.8 (15)
2.1 (36)
1.5 (41)
2.5 (35)
2.9 (31)
1.4 (43)
5.4 (16)
4.3 (21)
3.3 (28)
0.8 (48)
9.0 (2)
4.2 (22)
7.1 (7)
0.2 (51)
3.8 (24)
7.5 (5)
6.7 (8)
3.1 (29)
0.3 (50)
5.3 (17)
8.4 (3)

4.2
2.4
2.3
1.7
1.6
1.5
1.3
1.1
1.1
1.0
0.8
0.8
0.8
0.8
0.7
0.7
0.7
0.5
0.5
0.5
0.4
0.4
0.4
0.4
0.3
0.3
0.3
0.3
0.3
0.3
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.0

(continued on next page)

www.ajpmonline.org

 Ruhm / Am J Prev Med 2017;](]):]]]–]]]

5

Table 1. Opioid and Heroin Involved Drug Poisoning Death Rates by State, 2014 (continued)
Any opioid
State
Connecticut
Rhode Island
U.S.

a

Reported

Corrected

14.6 (10)
19.4 (3)
9.0

14.7 (13)
19.5 (6)
11.2

Heroin
a

Difference
0.1
0.1
2.2

State
Maine
Rhode Island
U.S.

Reporteda

Correcteda

Difference

2.9 (27)
6.3 (7)
3.3

2.9 (30)
6.3 (11)
4.0

0.0
0.0
0.7

Note: Table shows 2014 state mortality rates per 100,000 of drug poisoning deaths involving opioids or heroin. Drug poisoning deaths included ICD–
10 underlying cause of death codes: X40–X44, X60–X64, X85, and Y10–Y14. Any opioid included ICD–10 T-codes 40.0–40.4 and 40.6; heroin
included ICD–10 T-code 40.1. Reported mortality rates were based on mentions of the specified drugs on death certificates. Corrected rates used
mentions on death certificates for fatalities where at least one specific drug category was identified. In cases where only unspecified drugs were
mentioned (ICD–10 code T50.9), opioid or heroin involvement was imputed. This was done using predicted values from a probit model, estimated by
maximum likelihood, where the dependent variable was any opioid or heroin involvement (two separate models) and the model covariates included
sex, race (white, black, other non-white), Hispanic origin, marital status, education categories (high school dropout, high school graduate, some
college, college graduate), seven age categories, day-of-the week indicators, location of death (hospital inpatient, hospital outpatient/ED, dead on
arrival at hospital/ED, home, other), interactions between sex and race/ethnicity and with the following 2010-year county characteristics: poverty
rates, share of persons aged Z25 years with the four education levels described above and of households headed by females, median income,
population per square mile and its square, and physicians per 1,000. Difference is the corrected rate minus the reported rate and the table sorts
states in descending order of this difference.
a
Data are shown as mortality rates per 100,000 (Rank). Rank refers to the state ranking, from highest to lowest death rate.

overestimated when using reported rather than corrected
rates in only five states, with the largest disparity being a
relatively modest 0.4 per 100,000 difference in New
Mexico. Corrected and reported mortality rate change
rankings deviated by at least five places in eight states and
differed by less than 0.2 per 100,000 in 25 states.
Figure 2 maps the changes, from 2008 to 2014, in state
opioid and heroin involved mortality rates. The largest
difference between the reported (left-side) and corrected
(right-side) trends is that the latter more clearly demonstrate that relatively modest increases in heroin involved
deaths (0.1 to 0.5 per 100,000) were largely restricted to
four geographically contiguous states—Montana, North

Dakota, South Dakota, and Nebraska—with more rapid
growth occurring in all other locations. By contrast, the
uncorrected estimates misleadingly suggest that similarly
small increases in heroin mortality rates also extended
west to Idaho and California, as well as south through
Oklahoma, Arkansas, and Texas.

DISCUSSION
Current death certificate data are problematic for understanding the drug poisoning epidemic, with a particular
issue being the frequency with which no specific drug is
identified. This results in an underestimate of the

Figure 1. Reported and corrected 2014 overdose death rates (per 100,000).
] 2017

 Ruhm / Am J Prev Med 2017;](]):]]]–]]]

6

Table 2. Changes in Opioid Analgesic and Heroin Involved Drug Poisoning Death Rates by State, 2014 versus 2008
Any opioid
State
Pennsylvania
Indiana
New Jersey
Arizona
Michigan
Idaho
Wyoming
Alabama
Missouri
Alaska
Washington DC
Oklahoma
Iowa
North Carolina
Nevada
Louisiana
Maryland
New Hampshire
Hawaii
Texas
Illinois
Vermont
Minnesota
Arkansas
Mississippi
Kansas
Massachusetts
Oregon
New York
West Virginia
Wisconsin
Delaware
Nebraska
North Dakota
Virginia
California
Montana
Rhode Island
Tennessee
South Dakota
Georgia
Washington
Colorado
Maine
Utah
Kentucky
Florida
Connecticut
Ohio

Heroin

Reporteda

Correcteda

Difference

State

Reporteda

Correcteda

Difference

3.7 (17)
2.2 (28)
4.4 (14)
0.9 (35)
4.2 (16)
0.1 (42)
1.9 (29)
1.7 (31)
3.5 (18)
–2.5 (51)
4.2 (15)
1.8 (30)
0.9 (36)
0.6 (37)
–1.8 (49)
2.7 (24)
6.4 (8)
15.2 (1)
0.2 (41)
0.4 (38)
2.7 (23)
1.1 (34)
1.5 (32)
–1.0 (48)
0.3 (40)
2.5 (25)
7.7 (6)
–0.5 (44)
3.0 (20)
9.8 (3)
4.4 (13)
5.9 (9)
1.4 (33)
–0.7 (45)
2.9 (21)
0.3 (39)
–2.0 (50)
6.8 (7)
5.5 (11)
0.0 (43)
2.8 (22)
–1.0 (47)
2.4 (26)
5.0 (12)
2.3 (27)
8.0 (5)
–0.9 (46)
8.1 (4)
11.1 (2)

6.8 (6)
4.2 (15)
5.9 (10)
2.4 (26)
5.1 (12)
0.9 (36)
2.6 (23)
2.3 (27)
4.0 (17)
–2.0 (49)
4.6 (14)
2.1 (28)
1.1 (33)
0.9 (37)
–1.6 (47)
2.8 (21)
6.5 (7)
15.4 (1)
0.2 (40)
0.4 (38)
2.7 (22)
1.1 (35)
1.5 (30)
–1.0 (45)
0.2 (41)
2.4 (25)
7.6 (4)
–0.6 (43)
2.9 (19)
9.6 (2)
4.2 (16)
5.7 (11)
1.1 (34)
–1.1 (46)
2.5 (24)
–0.2 (42)
–2.5 (51)
6.3 (9)
4.8 (13)
–0.8 (44)
2.0 (29)
–1.8 (48)
1.4 (31)
3.9 (18)
1.3 (32)
6.9 (5)
–2.2 (50)
6.4 (8)
9.3 (3)

3.1
2.0
1.6
1.5
0.9
0.8
0.7
0.7
0.5
0.5
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0.1
0.1
0.0
0.0
0.0
0.0
0.0
–0.1
–0.1
–0.1
–0.1
–0.2
–0.2
–0.2
–0.2
–0.3
–0.4
–0.4
–0.5
–0.5
–0.5
–0.7
–0.8
–0.8
–0.8
–1.0
–1.0
–1.0
–1.1
–1.2
–1.7
–1.8

Pennsylvania
Indiana
New Jersey
Louisiana
Alabama
Michigan
Delaware
Wyoming
Kentucky
Idaho
Mississippi
Arizona
Arkansas
Montana
Missouri
Ohio
Washington DC
Colorado
Kansas
Florida
Tennessee
Texas
Wisconsin
Hawaii
Alaska
Minnesota
Illinois
Georgia
Nebraska
Iowa
Nevada
North Carolina
Maryland
Oklahoma
Oregon
New Hampshire
West Virginia
California
Vermont
New York
Massachusetts
Virginia
Washington
North Dakota
South Dakota
Utah
Rhode Island
Maine
South Carolina

2.7 (20)
1.8 (30)
3.4 (15)
2.0 (26)
2.4 (22)
3.1 (17)
4.8 (8)
1.5 (32)
4.9 (7)
0.5 (44)
0.7 (39)
1.9 (29)
0.2 (48)
0.0 (51)
3.5 (13)
8.4 (1)
4.6 (11)
1.9 (27)
0.4 (46)
1.2 (35)
2.1 (24)
0.6 (43)
3.5 (12)
0.6 (41)
2.4 (23)
1.7 (31)
4.7 (9)
1.3 (33)
0.3 (47)
0.9 (38)
1.1 (36)
2.0 (25)
3.4 (16)
0.5 (45)
0.7 (40)
6.6 (3)
6.9 (2)
0.6 (42)
4.6 (10)
3.0 (19)
6.1 (4)
1.9 (28)
3.1 (18)
0.1 (49)
0.1 (50)
1.2 (34)
5.9 (5)
2.4 (21)
1.1 (37)

5.9 (5)
3.5 (17)
4.7 (12)
3.2 (20)
3.4 (19)
4.1 (14)
5.5 (7)
2.3 (28)
5.5 (8)
1.1 (38)
1.2 (35)
2.5 (25)
0.7 (45)
0.5 (48)
3.9 (15)
8.7 (1)
4.9 (10)
2.3 (29)
0.7 (44)
1.4 (33)
2.4 (26)
0.8 (42)
3.7 (16)
0.8 (41)
2.6 (24)
1.9 (32)
4.9 (11)
1.4 (34)
0.4 (49)
1.0 (39)
1.2 (36)
2.1 (30)
3.5 (18)
0.6 (47)
0.7 (43)
6.6 (3)
6.9 (2)
0.6 (46)
4.7 (13)
3.1 (21)
6.1 (4)
1.9 (31)
3.1 (23)
0.1 (50)
0.1 (51)
1.2 (37)
5.8 (6)
2.3 (27)
0.9 (40)

3.2
1.8
1.3
1.2
1.0
0.9
0.8
0.7
0.6
0.6
0.6
0.5
0.5
0.5
0.4
0.4
0.3
0.3
0.3
0.3
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.0
0.0
0.0
0.0
0.0
–0.1
–0.1
–0.2

(continued on next page)

www.ajpmonline.org

 Ruhm / Am J Prev Med 2017;](]):]]]–]]]

7

Table 2. Changes in Opioid Analgesic and Heroin Involved Drug Poisoning Death Rates by State, 2014 versus 2008 (continued)
Any opioid
State
New Mexico
South Carolina
U.S.

a

Heroin

Reported

Corrected

3.2 (19)
5.8 (10)
2.5

0.4 (39)
2.8 (20)
2.5

a

Difference
–2.8
–3.0
–0.1

State
Connecticut
New Mexico
U.S.

Reporteda

Correcteda

Difference

5.2 (6)
3.4 (14)
2.3

5.0 (9)
3.1 (22)
2.7

–0.3
–0.4
0.4

Note: Table 1 Note provides more detail. Table shows changes, between 2008 and 2014, in state mortality rates per 100,000 of drug poisoning
deaths involving opioids or heroin.
a
Data are shown as mortality rates per 100,000 (Rank). Rank refers to the state ranking, from highest to lowest death rate.

involvement of specific drugs in fatal overdoses (but not
in the overall number of drug fatalities), which is sometimes substantial. For instance, mortality rates calculated
using imputed data on specific drugs where such
information was lacking on death certificates suggest
that in 2014 opioid and heroin involved death rates were
understated by more than half in Pennsylvania (8.5 vs
17.8 per 100,000 for opioids and 3.9 vs. 8.1 per 100,000
for heroin).
Direction of the corresponding errors is theoretically
ambiguous when examining mortality trends. This is
because increased specificity of reporting will cause
growth to be overstated but rising drug death rates will
lead to a potentially offsetting underestimate. In practice,
uncorrected death certificate reports accurately indicated
the average growth, between 2008 and 2014, in opioid
involved death rates but understated the increase in
heroin mortality. The size of any errors again varied
dramatically across states. Increases in opioid involved
mortality rates were overstated by at least 0.5 per 100,000

in 15 states and underestimated by an equivalent amount
in nine states. Growth in heroin mortality rates was
understated by 0.6 per 100,000 or more in 12 states, but
was rarely overestimated. The corrections also often
substantially changed the state rankings of opioid and
heroin involved mortality rates. The most striking
example was Pennsylvania where reported 2014 opioid
death rates ranked 32th compared to 7th for corrected
rates. The corresponding change was from 20th to 4th for
heroin and, when looking at increases between 2008 and
2014, from 17th to 6th and 20th to 5th.
Additional training and standardization in states with
low specification rates may be helpful for obtaining
accurate information on drug involvement in fatal overdoses, particularly because this is a bigger problem when
death certificates are completed by coroners rather than
medical examiners and in states without centralized
oversight.26 Others have also recommended adding detail
to death certificates on the drugs involved, toxicology
levels, and ICD categories, as well as m distinguishing ore

Figure 2. Reported and corrected change in overdose death rates (per 100,000), 2014 versus 2008.
] 2017

 8

Ruhm / Am J Prev Med 2017;](]):]]]–]]]

carefully between cases where a given drug is the cause of
mortality versus those where it was detected but was not
a major contributor.27,28
Until such information becomes available, correction
methods like those developed here are needed to provide
more accurate estimates of drug involvement in fatal
overdoses occurring at a point in time. Moreover, even
with improvements in reporting, these or similar procedures will be necessary for investigating mortality trends,
because greater specificity on death certificates in later
(but not earlier) years introduces additional errors into
the estimates of changes over time.
Understanding the inaccuracies resulting from the lack
of specificity of drug involvement on death certificates is
also important because federal policies often target states
believed to have especially severe opioid or heroin
problems.29,30 More fundamentally, geographic disparities in drug poisoning deaths are substantial and a
correct assessment of them is almost certainly a prerequisite for designing policies to address the fatal drug
epidemic. For example, some researchers have suggested
that factors such as economic insecurity, poverty, and
low levels of education explain the decline in life expectancy for some groups of non-Hispanic whites, which has
been substantially driven by increased drug fatality
rates.3,31,32 Developing such policies may be complex,
however, because a preliminary analysis did not reveal an
obvious relationship between differences in reported
versus corrected opioid or heroin death rates, and
previously used measures of state-level legal restrictions
on controlled substances or Medicaid coverage for
medications treating opioid use disorders.33,34 Nevertheless careful, empirical investigation of these possibilities cannot be performed without accurate information
on how drug fatality rates differ across geographic
locations.

Limitations
This analysis is subject to limitations. First, the imputations contain potential errors and were only used in cases
where no drug was specified. However, fatal overdoses
frequently involve combinations of drugs, leading to a
downwards biased estimate of drug involvement in cases
where information on some, but not all, drug categories
was provided.17 Second, the imputations assume that
data on the drugs involved in overdose deaths were
“missing at random,” meaning that the probability of a
missing value varies only with the characteristics controlled for in the imputation process.35 More comprehensive controls could provide different results as might
the use of multiple rather than single imputation
procedures, although a recent related analysis indicated
that almost identical results were obtained using single

versus multiple imputation.36 Third, the reporting of
specific drug involvement on death certificates may
sometimes be inaccurate. For instance, heroin use is
sometimes attributed to morphine or codeine and overdose deaths may be misclassified as being due to nondrug causes, or vice versa.37,38 Also, some ICD-10 codes
lack specificity. For instance, T-code 40.6 refers to
poisoning by “unspecified” narcotics that will sometimes,
but not always, include opioids.

CONCLUSIONS
Notwithstanding these caveats, the corrected mortality
rates developed here almost certainly provide a more
accurate understanding of geographic differences by state
in opioid or heroin involved drug poisoning death rates
than the raw information contained on death certificates.
These or similar methods should be used in related future
analyses.

ACKNOWLEDGMENTS
The author, Christopher J. Ruhm, was responsible for all aspects
of this paper including: designing the study, acquiring the data,
performing and interpreting the analysis, and writing up the
results.
No financial disclosures were reported by the author of
this paper.

SUPPLEMENTAL MATERIAL
Supplemental materials associated with this article can be
found in the online version at https://doi.org/10.1016/j.
amepre.2017.06.009.

REFERENCES
1. Rudd RA, Aleshire N, Zibbell JE, Gladden RM. Increase in drug
and opioid overdose deaths – United States, 2000-2014. MMWR
Morb Mortal Wkly Rep. 2016;64(50–51):1378–1382. https://doi.org/
10.15585/mmwr.mm6450a3.
2. Warner M, Chen LH, Makuc DM, Anderson RN, Miniño AM. Drug
poisoning deaths in the United States, 1980–2008. NCHS Data Brief, no
81. Hyattsville, MD: National Center for Health Statistics. 2011. http://
dx.doi.org/10.1073/pnas.1518393112.
3. Case A, Deaton A. Rising morbidity and mortality in midlife among
white non-Hispanic Americans in the 21st century. Proc Natl Acad
Sci U S A. 2015;112(49):15078–15083. https://doi.org/10.1073/pnas.
1518393112.
4. CDC. Vital Signs: Overdoses of prescription opioid pain relievers United States, 1999–2008. MMWR Morb Mortal Wkly Rep. 2011;60
(43):1487–1492.
5. CDC. Vital Signs: Risk for overdose from methadone used for pain
relief – United States, 1999-2010. MMWR Morb Mortal Wkly Rep.
2012;61(26):493–497.
6. Volkow ND, Frieden TR, Hyde PS, Cha SS. Medication-assisted
therapies – tackling the opioid overdose epidemic. N Engl J Med. 2014;
370(22):2063–2066. https://doi.org/10.1056/NEJMp1402780.

www.ajpmonline.org

 Ruhm / Am J Prev Med 2017;](]):]]]–]]]
7. Jones CM, Logan J, Gladden RM, Bohm MK. Vital signs: demographic
and substance use trends among heroin users—United States, 2002–
2013. MMWR Morb Mortal Wkly Rep. 2015;64:719–725.
8. Hardesty C. White House Summit on the Opioid Epidemic. Washington DC:
Office of National Drug Policy, June 19, 2014. www.whitehouse.gov/blog/
2014/06/19/white-house-summit-opioid-epidemic Accessed June 27, 2017.
9. CDC. Opioids drive continued increase in drug overdose deaths. 2013. www.
cdc.gov/media/releases/2013/p0220_drug_overdose_deaths.html.
10. Finklea KM, Bagalman E, Sacco LN. Prescription drug monitoring programs.
Washington, DC: Library of Congress, Congressional Research Service.
2013. www.hsdl.org/?abstract&did=728239. Accessed June 27, 2017.
11. Rannazzisi JT. Testimony for “Curbing Prescription Drug Abuse in
Medicare”. Hearing before the Committee on Homeland Security and
Governmental Affairs, United States Senate, 113th Congress, 2013. www.
gpo.gov/fdsys/pkg/CHRG-113shrg82571/pdf/CHRG-113shrg82571.pdf
Accessed June 27, 2017.
12. Kirschner N, Ginsburg J, Sulmasy LS, Health and Public Policy
Committee of the American College of Physicians. Prescription drug
abuse: executive summary of a policy position paper from the
American College of Physicians. Ann Intern Med. 2014;160(3):
198–200. https://doi.org/10.7326/M13-2209.
13. Paulozzi LJ, Kilbourne EM, Desai HA. Prescription drug monitoring
programs and death rates from drug overdose. Pain Med. 2011;12
(5):747–754. https://doi.org/10.1111/j.1526-4637.2011.01062.x.
14. Gugelmann HM, Perrone J. Can prescription drug monitoring programs help limit opioid abuse? JAMA. 2011;306(20):2258–2259. https:
//doi.org/10.1001/jama.2011.1712.
15. Li G, Brady JE, Lang BH, Giglio J, Wunsch H, DiMaggio C.
Prescription drug monitoring and drug overdose mortality. Inj
Epidemiol. 2014;1(1):1–8. https://doi.org/10.1186/2197-1714-1-1.
16. Johnson H, Paulozzi L, Poruncznik C, Mack K, Herter B. Hal Jonson
Consulting, Division of Disease Control and Health Promotion,
Florida Department of Health. Decline in drug overdose deaths after
state policy changes – Florida, 2010-2012. MMWR Morb Mortal Wkly
Rep. 2014;63(26):569–574.
17. Ruhm CJ. Drug involvement in fatal overdoses. SSM Popul Health.
2017;3:219–226. https://doi.org/10.1016/j.ssmph.2017.01.009.
18. Slavova S, O’Brien DB, Creppage K, et al. Drug overdose deaths: let’s
get specific. Public Health Rep. 2015;130(4):339–342. https://doi.org/
10.1177/003335491513000411.
19. Vestal C. Stateline: getting better data on which drugs are killing
people. Pew Charitable Trusts. August 19, 2016. www.pewtrusts.org/
en/research-and-analysis/blogs/stateline/2016/08/19/getting-better-da
ta-on-which-drugs-are-killing-people. Accessed November 1, 2016.
20. Ruhm CJ. Drug poisoning deaths in the United States, 1999-2012:
a statistical adjustment analysis. Popul Health Metr. 2016;14:2. https:
//doi.org/10.1186/s12963-016-0071-7.
21. Rossen LM, Khan D, Warner M. Trends and geographic patterns in
drug-poisoning death rates in the U.S., 1999–2009. Am J Prev Med.
2013;45(6):e19–e25. https://doi.org/10.1016/j.amepre.2013.07.012.
22. Rossen LM, Bastian B, Warner M, Khan D, Chong Y. Drug poisoning
mortality: United States, 1999–2014. National Center for Health
Statistics. 2016. https://blogs.cdc.gov/nchs-data-visualization/drug-poi
soning-mortality/. Accessed November 2, 2016.

] 2017

9

23. CDC. Multiple cause of death, 1999–2014. http://wonder.cdc.gov/
wonder/help/mcd.html. Accessed June 15, 2016.
24. WHO. International Statistical Classification of Diseases and Related
Health Problems, ICD-10, 10th Revised Ed. Geneva: WHO, 2012.
25. National Cancer Institute. Surveillance, Epidemiology, and End
Results Program. www.seer.cancer.gov/resources. Accessed August 4,
2016.
26. Warner M, Paulozzi LJ, Nolte KB, Davis GG, Nelson LS. State variation
in certifying manner of death and drugs involved in drug intoxication
deaths. Acad Forensic Pathol. 2013;3(2):231–237.
27. Webster LR, Dasgupta N. Obtaining adequate data to determine causes
of opioid-related overdose deaths. Pain Med. 2011;12(Suppl 2):
S86–S92. https://doi.org/10.1111/j.1526-4637.2011.01132.x.
28. Goldberger BA, Maxwell JC, Campbell A, Wilford BB. Uniform
standards and case definitions for classifying opioid-related deaths:
recommendations by a SAMHSA Consensus Panel. J Addict Dis.
2013;32(3):231–243. https://doi.org/10.1080/10550887.2013.824334.
29. Comprehensive Addiction and Recovery Act of 2016, Pub. L. No:
114-198, 130 Stat. 695. 2016. https://www.congress.gov/114/plaws/
publ198/PLAW-114publ198.pdf. Accessed November 2, 2016.
30. Office of the Press Secretary, The White House. President Obama
proposes $1.1 billion in new funding to address the prescription opioid
abuse and heroin use epidemic. J Pain Palliat Care Pharmacother. 2016;30(2):134–137. https://doi.org/10.3109/15360288.2016.
1173760.
31. Kindig DA, Cheng ER. Even as mortality fell in most U.S. counties,
female mortality nonetheless rose in 42.8 percent of counties from
1992 to 2006. Health Aff (Millwood). 2013;32(3):451–458. https://doi.
org/10.1377/hlthaff.2011.0892.
32. Meara E, Skinner J. Losing ground at midlife in America. Proc Natl
Acad Sci U S A. 2015;112(49):15006–15007. https://doi.org/10.1073/
pnas.1519763112.
33. Meara E, Horwitz JR, Powell W, et al. State legal restrictions and
prescription-opioid use among disabled adults. N Engl J Med. 2016;375
(1):44–53. https://doi.org/10.1056/NEJMsa1514387.
34. Substance Abuse and Health Services Administration. Medicaid
Coverage and Financing of Medications to Treat Alcohol and
Opioid Disorders. HHS Publication No. SMA-14-4854. Rockville,
MD: Substance Abuse and Mental Health Services Administration,
2014.
35. Gelman A, Hill J. Data Analysis Using Regression and Multilevel/
Hierarchical Models. New York: Cambridge University Press, 2006.
http://dx.doi.org/10.1017/CBO9780511790942.
36. Hollingsworth A, Ruhm CJ, Simon K. Macroeconomic conditions
and opioid abuse. NBER Work Pap Ser. 2017:23192. https://doi.org/
10.3386/w23192.
37. Mertz KJ, Janssen JK, Williams KE. Underrepresentation of heroin
involvement in unintentional drug overdose deaths in Allegheny
County, PA. J Forensic Sciences. 2014;59(6):1583–1585. https://doi.
org/10.1111/1556-4029.12541.
38. Ellis AD, McGwin G, Davis GG, Dye DW. Identifying cases of heroin
toxicity where 6-acetylmorphine (6-AM) is not detected by toxicological analyses. Forensic Sci Med Pathol. 2016;12(3):243–247. https:
//doi.org/10.1007/s12024-016-9780-2.