National Hospital Care Survey Demonstration Projects: Opioid-involved Emergency Department Visits, Hospitalizations, and Deaths

2020·0.67 MB·English

by  Spencer, Marianne Rose, Flagg, Lee Anne, Jackson, Geoff, DeFrances, Carol, Hedegaard, Holly

National Health Statistics Reports
Number 141 (cid:132) June 16, 2020
National Hospital Care Survey Demonstration Projects:
Opioid-involved Emergency Department Visits,
Hospitalizations, and Deaths
by Merianne Rose Spencer, M.P.H., Lee Anne Flagg, Ph.D., Geoff Jackson, M.S.,
Carol DeFrances, Ph.D., and Holly Hedegaard, M.D., M.S.P.H.
Abstract Introduction
Objective—This report demonstrates the utility of linking the restricted-use 2014 Decision-makers such as clinicians,
National Hospital Care Survey (NHCS), 2014–2015 National Death Index (NDI), and policymakers, and researchers need
2014–2015 Drug-Involved Mortality (DIM) data to study opioid-involved emergency comprehensive data on opioid-involved
department (ED) visits, hospitalizations, and mortality within 1 year post-discharge. emergency department (ED) visits,
Example research questions and unweighted results are presented. Results are not hospitalizations, and deaths to identify
nationally representative. and examine strategies to reduce
Methods—Patient records from the 2014 NHCS with sufficient identifying morbidity and mortality from opioid-
information were linked to the 2014–2015 NDI and DIM data. Visits were considered involved overdoses and other related
opioid-involved if they had International Classification of Diseases, Ninth Revision, harms. However, national statistics on
Clinical Modification (ICD–9–CM) diagnosis codes 304.00–304.02, 304.70–304.72, opioid-involved hospital encounters and
305.50–305.52, 760.72, 965.00–965.02, 965.09, 970.1, or E850.0–E850.2 in any subsequent outcomes among the same
diagnosis or external cause of injury code field. Opioid-involved drug overdose deaths individual patient are limited.
were deaths with an International Classification of Diseases, 10th Revision (ICD–10) The National Center for Health
underlying cause-of-death code of X40–44, X60–64, X85, or Y10–Y14 and a multiple Statistics (NCHS) received support
cause code of T40.0–T40.4 or T40.6. from the Department of Health and
Results—In the 2014 NHCS, there were 15,495 patients with an opioid-involved Human Services’ Office of the Secretary
ED-only visit and 24,059 patients with an opioid-involved hospitalization. Of the Patient-Centered Outcomes Research
20,962 patients with an opioid-involved hospitalization eligible to be linked to NDI, Trust Fund (PCORTF) (1) to link several
1,805 died (9%) within 1 year of discharge. Of these deaths, 341 (19%) resulted from data sources. The first data source, the
a drug overdose. Of drug overdose deaths, 243 (71%) involved an opioid, where National Hospital Care Survey (NHCS),
12% died within 30 days post-discharge, 19% within 31–90 days, and 69% within collects ED, inpatient, and outpatient
91–365 days. Opioids most frequently mentioned included heroin (46%), fentanyl department administrative claims or
(20%), oxycodone (13%), methadone (12%), and morphine (12%). These categories electronic health records (EHR) data
are not mutually exclusive because a death may involve more than one drug. For from a sample of participating hospitals
approximately 22% of patients who died of an opioid-involved drug overdose in 2014, (2). These data are not nationally
their last ED-only visit or hospitalization was opioid-involved. representative. The second source,
Conclusion—While the NHCS data are not nationally representative, these the National Death Index (NDI), is a
unlinked and linked National Center for Health Statistics data allow for exploratory centralized database of death records
analyses of ED visits, hospitalizations, and associated mortality outcomes. on file in jurisdictional vital records
offices for all deaths occurring in
Keywords: health care • mortality • drug overdose • National Hospital Care Survey
the United States, as well as the U.S.
U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES
Centers for Disease Control and Prevention
National Center for Health Statistics
NCHS reports can be downloaded from: https://www.cdc.gov/nchs/products/index.htm.
Page 2 National Health Statistics Reports (cid:132) Number 141 (cid:132) June 16, 2020
military overseas (3). NCHS established NHCS are categorized using the International
NDI to aid epidemiologists and other Classification of Diseases, 10th Revision
NHCS collects data from inpatient
health and medical investigators with (ICD–10) for underlying and multiple
discharges and in-person visits to ED
their mortality ascertainment activities. causes of death (6). Drug overdose
and outpatient departments, including
These records are compiled annually deaths are identified using underlying
hospital-based ambulatory surgery
and include detailed information on the cause-of-death ICD–10 codes X40–X44
centers. The initial sampling frame was
underlying and multiple causes of death. (unintentional), X60–X64 (intentional
constructed in 2011 using the 2010
The third source, the Drug-Involved self-harm or suicide), X85 (homicide),
IQVIA (formerly Quintiles and IMS
Mortality (DIM) restricted-use data, was and Y10–Y14 (undetermined intent).
Government Solutions) hospital database
developed through a collaborative project Deaths involving specific drugs or drug
for hospitals meeting eligibility criteria
between NCHS and the Food and Drug classes are identified using multiple
as noninstitutional, nonfederal hospitals
Administration (FDA) using the National cause-of-death ICD–10 codes
with six or more inpatient staffed beds
Vital Statistics System (NVSS) restricted- T36–T50.8.
in the 50 states and the District of
use mortality files. This source provides
Columbia. From the sampling frame of
information on the specific drugs DIM restricted-use data
6,622 hospitals, a stratified list of 1,000
involved in a death, based on information
hospitals was created based on hospital The DIM restricted-use data, also
provided on the death certificate by the
bed size, type of hospital, and urban or known as the National Vital Statistics
medical certifier—typically a medical
rural designation and then split into two System Mortality Drug Overdose
examiner or coroner in the case of drug
samples: a base sample of 500 hospitals (NVSS-M-DO) data, include information
poisoning deaths. The DIM data may be
and a reserve sample of 500 hospitals. on the specific drugs involved in a
accessed through the NCHS Research
The base sample was fielded for data death. The methods for preparing this
Data Center (4).
collection, and the reserve sample was file were developed collaboratively by
The NHCS data linked to these
used to replenish the base sample if NCHS and FDA and involve extracting
mortality files, while not nationally
more hospitals were needed. In 2013, 81 information from the literal text on death
representative, have the potential to
hospitals with 500 or more staffed beds certificate records obtained through the
lend insights into hospital use and drug
were moved from the reserve sample NVSS (7). The literal text is the verbatim
overdose deaths involving opioids.
into the base sample. The 2014 sample text provided by the medical certifier—
Researchers can study patients presenting
consists of 581 hospitals: 506 acute care usually a medical examiner or coroner in
to hospitals due to an opioid-involved
hospitals and 75 other specialty hospitals, the case of a drug overdose death—about
event and examine their care either until
including children’s, psychiatric, long- the cause(s) and circumstances of death.
death or the end of the study period.
term acute care, and rehabilitation The methodology for developing
Conversely, researchers can also look at
hospitals. For the 2014 NHCS data the DIM data involved searching the
the population of individuals who died
collection, 94 of the 581 sampled literal text from three fields of the death
of an opioid-involved drug overdose in a
hospitals provided inpatient department certificate (the cause-of-death sequence
given timeframe and examine the patterns
claims data and 83 hospitals provided from Part I, significant conditions
and types of care they received before
ED claims data. Participating hospitals contributing to death from Part II, and
death.
were asked to provide data on all ED and the “Describe how injury occurred” box)
This report provides examples of the
inpatient encounters in the 2014 calendar using search terms for specific drugs.
types of research questions and analyses
year. The unweighted total number of Search terms included generic drug
possible using the linked data on opioid-
encounters was approximately 4.5 million names, brand names, common usage or
involved ED visits and hospitalizations
ED visits. With a response rate of 14.3% street names, abbreviations, metabolites,
and opioid-involved drug overdose
for hospitals submitting ED data and misspellings, and other variations. Each
deaths.
unweighted results, the 2014 NHCS data search term was mapped to a “principal
are not nationally representative. More variant,” the overarching label assigned
Methods information on NHCS methodology is to a drug, a drug class, or exposure not
published elsewhere (5). otherwise specified (7).
Data sources
NDI
Data linkage processes
These analyses use data linked
NDI is a centralized database of
across three sources: the 2014 NHCS, A description of the linkage of the
death record information on file in
the 2014–2015 NDI, and the 2014–2015 2014 NHCS to the 2014–2015 NDI has
jurisdictional vital records or statistics
DIM data. Data are restricted use and been published elsewhere (8). To link
offices and maintained by NCHS (3).
can be accessed through the Research to the DIM restricted-use data, which
These data can be used to identify each
Data Center at https://www.cdc.gov/rdc/ includes the specific drugs mentioned
person who has died in the United States
b1datatype/dt122.htm. in the literal text of death certificate
and U.S. military overseas and the
records (i.e., Part I, Part II, and Describe
cause(s) and manner of death. All deaths
how injury occurred), only those NHCS
National Health Statistics Reports (cid:132) Number 141 (cid:132) June 16, 2020 Page 3
patient records that were eligible for and cause-of-death codes X40–X44, X60– Virginia); “Midwest” includes the
successfully linked to the 2014–2015 X64, X85, and Y10–Y14 in combination north-central area of the country (i.e.,
NDI could be linked to the 2014–2015 with at least one multiple cause-of- Illinois, Indiana, Iowa, Kansas, Michigan,
DIM files. NHCS records that were death code T40.0–T40.4 or T40.6 Minnesota, Missouri, Nebraska, North
already linked to the 2014–2015 NDI (includes opium, heroin, other natural Dakota, Ohio, South Dakota, and
were linked to the 2014–2015 DIM files and semisynthetic opioids, methadone, Wisconsin); and the “Northeast” includes
using a unique ID consisting of the year other synthetic opioids, and unspecified the mid-Atlantic and New England areas
of death, jurisdiction of death, and death narcotics). Deaths that were not due to a (i.e., Connecticut, Maine, Massachusetts,
certificate number. drug overdose had an underlying cause New Hampshire, New Jersey, New
code other than X40–X44, X60–X64, York, Pennsylvania, Rhode Island, and
X85, or Y10–Y14. The most frequent Vermont) (11).
Analysis
underlying causes of death other than
Opioid-involved ED visits and drug overdose were grouped by ICD–10
Examples of research
hospitalizations were identified by chapter (6). ICD–10 chapters that
applying an algorithm previously occurred for less than 4% of deaths were questions
developed utilizing NHCS data, but combined into the category, “All other
with minor modifications (9). NHCS ICD–10 chapters.” There are several types of research
records were flagged as opioid- Analyses examining the time from approaches that can be taken with these
involved if they had any mention of the the date of hospital discharge to when data. NHCS data can be used alone
International Classification of Diseases, the death occurred were arranged into (unlinked) to examine characteristics
Ninth Revision, Clinical Modification three time interval categories: 1) death of patients and visits within a certain
(ICD–9–CM) codes 304.00–304.02, occurring within 30 days post-discharge, year. As described previously, NHCS
304.70–304.72, 305.50–305.52, 2) deaths occurring within 31 and 90 days data can also be linked with NDI and
760.72, 965.00–965.02, 965.09, 970.1, post-discharge, and 3) deaths occurring other data sources, such as the DIM
or E850.0–E850.2 in any diagnosis within 91 and 365 days post-discharge. files, to conduct cohort-type analyses
or external cause code field. These Urbanization levels, based on where patients are identified from
include codes for past or present opioid hospital location from NHCS, are ED visits or hospitalizations and then
(including opium, heroin, methadone, delineated in accordance with the 2010 followed to assess mortality outcomes,
and other opiates and related narcotics) Office of Management and Budget or characterized by cause of death and
use, abuse, dependence, and poisoning; standards for defining metropolitan analyzed to examine patterns in hospital
narcotics affecting the fetus or newborn; statistical areas (MSA), which are areas care prior to death. Example research
and poisoning by opiate antagonists containing a large population nucleus questions and results are presented for
(e.g., naloxone). Codes for opioids together with adjacent communities these various approaches. Please see
causing adverse effects in therapeutic having a high degree of economic and Appendices I–V for visualizations of the
use (E935.0–E935.2, E940.1) were not societal integration, as determined respective populations of interest for each
included. by the U.S. Census Bureau (10). The of the various types of research questions
Newborns (identified using ICD–9– “urban” and “suburban” categories are and results.
CM diagnosis codes V30–V39) and those large central metro and fringe metro
who died during hospitalization were counties, respectively, of an MSA with a Example analyses of the 2014
excluded from analyses. ED visits that population of 1 million or more. “Small
NHCS
resulted in the patient being admitted to metro” consists of MSAs with less than 1
the hospital were excluded from analyses million population and “rural” consists of
Patient-level: What are the age, sex,
of ED-only visits but were included in areas that are nonmetropolitan.
and expected source of payment
analyses of hospitalizations. Numbers The four regions, based on hospital
distributions for patients who had
and percentages were calculated to location from NHCS, are based on
an opioid-involved ED-only visit or
illustrate several types of research the U.S. Census Bureau’s definitions:
hospitalization in the 2014 NHCS?
questions that can be explored with these “West” includes the Pacific and
data. Due to the low response rate of Mountain states (i.e., Alaska, Arizona, In the 2014 NHCS, there were
NHCS, the results presented in this report California, Colorado, Hawaii, Idaho, 15,495 patients who had an opioid-
are not nationally representative. Montana, Nevada, New Mexico, Oregon, involved ED-only visit and 24,059
Drug overdose deaths were identified Utah, Washington, and Wyoming); patients who had an opioid-involved
using the ICD–10 underlying cause-of- “South” includes the south-central and hospitalization (Figure 1). Visits are
death codes X40–X44 (unintentional), southeastern parts of the country (i.e., mutually exclusive with patients who
X60–X64 (intentional self-harm or Alabama, Arkansas, Delaware, District of were admitted to the hospital through
suicide), X85 (homicide), and Y10–Y14 Columbia, Florida, Georgia, Kentucky, the ED being counted only in the
(undetermined intent). Drug overdose Louisiana, Maryland, Mississippi, North hospitalization group. Among patients
deaths involving an opioid were Carolina, Oklahoma, South Carolina, with an opioid-involved ED-only visit,
identified using ICD–10 underlying Tennessee, Texas, Virginia, and West about 57% were male and 43% were
Page 4 National Health Statistics Reports (cid:132) Number 141 (cid:132) June 16, 2020
female. Among patients with an opioid-
Male Female involved hospitalization, 52% were male
and 48% were female. For further detail
70
see Appendix I.
60 57 About 18% of patients with an
52 opioid-involved ED-only visit were aged
50 48 15–24 years, 31% were 25–34, 19% were
43
35–44, 18% were 45–54, and 14% were
nt 40 55 and over (Figure 2 and Appendix I).
e
c
Per 30 For patients with an opioid-involved
hospitalization, about 11% were aged
20 15–24 years, 23% were 25–34, 17% were
35–44, 21% were 45–54, and 26% were
10 55 and over.
For patients with an opioid-involved
0
ED-only visit, the expected source of
Patients with opioid-involved Patients with opioid-involved
ED-only visit hospitalization payment for their last opioid-involved
ED-only visit was approximately 51%
NOTES: ED is emergency department. Sex is based on information from the last ED-only visit or hospitalization. There
Medicaid, 12% Medicare, and 25%
were 15,495 patients with opioid-involved ED-only visits and 24,059 patients with opioid-involved hospitalizations. Visits
are mutually exclusive. ED visits that resulted in the patient being admitted to the hospital were included in private insurance (Figure 3). For patients
hospitalizations. Newborns were excluded from analysis. Data are not nationally representative.
with an opioid-involved hospitalization,
SOURCE: NCHS, National Hospital Care Survey, 2014.
the expected source of payment for their
last opioid-involved hospitalization was
Figure 1. Sex distribution among patients with an opioid-involved emergency department-
only visit or hospitalization: National Hospital Care Survey, 2014
Under 15 15–24 25–34 35–44 45–54 55–64 65 and over
35
31
30
25
23
21
20 19
ent 18 18 17 17
c
er
P
15
11 11
10 9
5
3
0.9 1
0
Patients with opioid-involved ED-only visit Patients with opioid-involved hospitalization
NOTES: ED is emergency department. Age is based on information from the last ED-only visit or hospitalization. There were 15,495 patients with opioid-involved ED-only visits and 24,059
patients with opioid-involved hospitalizations. Visits are mutually exclusive. ED visits that resulted in the patient being admitted to the hospital were included in hospitalizations. Newborns were
excluded from analysis. Data are not nationally representative.
SOURCE: NCHS, National Hospital Care Survey, 2014.
Figure 2. Age distribution among patients with an opioid-involved emergency department-only visit or hospitalization: National Hospital
Care Survey, 2014
National Health Statistics Reports (cid:132) Number 141 (cid:132) June 16, 2020 Page 5
Medicaid Medicare Private Other Unknown
60
51
50
44
40
nt
e
c 30
er
P 25 25
22
20
12
10
5 5 5
2
0
Patients with opioid-involved ED-only visit Patients with opioid-involved hospitalization
NOTES: ED is emergency department. Expected source of payment is based on information from the last ED-only visit or hospitalization. There were 15,495 patients with opioid-involved
ED-only visits and 24,059 patients with opioid-involved hospitalizations. Visits are mutually exclusive. ED visits that resulted in the patient being admitted to the hospital were included in
hospitalizations. Newborns were excluded from analysis. Data are not nationally representative.
SOURCE: NCHS, National Hospital Care Survey, 2014.
Figure 3. Distribution of expected source of payment among patients with an opioid-involved emergency department-only visit or
hospitalization: National Hospital Care Survey, 2014
approximately 44% Medicaid, 25%
Medicare, and 22% private insurance. For Urban Suburban Small Metro Rural
additional detail see Appendix I.
70
Visit-level: What are the geographic 60 58
distributions for opioid-involved
and nonopioid-involved ED-only 50 48
visits?
nt 40
e
Visit-level analyses are presented to c
demonstrate the geographical burden of Per 30 26
23 23
opioid-involved ED-only visits. There
20 18
were 18,857 ED-only visits that were
opioid-involved and 3,730,913 ED-only 10
visits that were not opioid-involved (see 2 4
Appendix I). Of the opioid-involved 0
ED-only visits in 2014, 58% occurred Opioid-involved ED-only visits Nonopioid-involved ED-only visits
in urban areas, 23% in suburban areas,
NOTES: ED is emergency department. There were 18,857 opioid-involved and 3,730,913 nonopioid-involved ED-only
18% in small metropolitan areas, and visits. ED visits that resulted in the patient being admitted to the hospital were excluded from the analysis. Data are not
nationally representative.
2% in rural areas. Among the ED-only
SOURCE: NCHS, National Hospital Care Survey, 2014.
visits that were not opioid-involved, 48%
occurred in urban areas, 23% in suburban Figure 4. Urban or rural classification for opioid-involved and nonopioid-involved
areas, 26% in small metropolitan areas, emergency department-only visits: National Hospital Care Survey, 2014
and 4% in rural areas (Figure 4).
Among the opioid-involved ED- the Midwest, 22% in the South, and 9% opioid-involved, 35% occurred in the
only visits occurring in the 2014 NHCS, in the West (Figure 5 and Appendix I). Northeast, 20% in the Midwest, 29% in
48% occurred in the Northeast, 21% in Among the ED-only visits that were not the South, and 17% in the West.
Page 6 National Health Statistics Reports (cid:132) Number 141 (cid:132) June 16, 2020
Example analyses of the
Northeast Midwest South West 2014 NHCS linked to the
60 2014–2015 NDI and the
2014–2015 DIM data
50 48
How many patients with an opioid-
40 involved hospitalization died within
35
nt 1 year post-discharge?
ce 30 29
er Of the 24,059 patients with an
P 21 22 20 opioid-involved hospitalization, 20,962
20 17
(87%) were eligible for linkage to NDI
9 and alive at discharge. Among these
10
20,962 patients, 1,805 (9%) had a date of
death and 19,157 (91%) were presumed
0
alive within 1 year post-discharge. Two
Opioid-involved ED-only visits Nonopioid-involved ED-only visits
records had a date of death but did not
NOTES: ED is emergency department. There were 18,857 opioid-involved and 3,730,913 nonopioid-involved ED-only have an underlying cause of death. Of
visits. ED visits that resulted in the patient being admitted to the hospital were excluded from the analysis. Data are not
nationally representative. the remaining 1,803 patients who died,
SOURCE: NCHS, National Hospital Care Survey, 2014. 341 (19%) died from a drug overdose
and 1,462 (81%) died from a cause other
Figure 5. Census regions for opioid-involved and nonopioid-involved emergency
than drug overdose (Figure 6). Of the 341
department-only visits: National Hospital Care Survey, 2014
patients who died of a drug overdose, 243
(71%) of the deaths involved an opioid
(see also Appendix II).
Drug overdose
involving an opioid
14% For patients who died of a cause
other than drug overdose, what
Drug overdose not
involving an opioid were the causes?
5%
Of the 1,462 patients who had at
least one opioid-involved hospitalization
and died of a cause other than drug
Cause other than overdose within 1 year post-discharge,
drug overdose
29% of the deaths were due to neoplasms
81%
(cancer), 22% were due to diseases of
the circulatory system, 9% were due to
diseases of the respiratory system, and
9% were due to infectious or parasitic
diseases (Figure 7 and Appendix II).
NOTES: Patients with an opioid-involved hospitalization who died within 1 year post-discharge is based on information
from the last opioid-involved hospitalization. This 1,803 represents the number of linkage-eligible patients with an What are the most common time
opioid-involved hospitalization who died within 1 year post-discharge and had an underlying cause of death. Emergency intervals from date of discharge to
department visits that resulted in the patient being admitted to the hospital were included in hospitalizations. Newborns
and those who died during the hospitalization were excluded from analysis. Two records that did not have an underlying date of death for patients who had
cause of death were also excluded from analysis. Data are not nationally representative. an opioid-involved hospitalization
SOURCES: NCHS, 2014 National Hospital Care Survey data linked to the 2014–2015 National Death Index.
and died within 1 year post-
discharge?
Figure 6. Percentages of patients with an opioid-involved hospitalization who died within
1 year post-discharge due to a drug overdose involving an opioid, a drug overdose not Of the 1,462 patients with an opioid-
involving an opioid, or some other cause: National Hospital Care Survey, 2014, linked to
involved hospitalization who died within
National Death Index, 2014 and 2015
1 year post-discharge due to a cause other
than drug overdose, 22% died within 30
days post-discharge, 22% within 31–90
days, and 55% within 91–365 days. Of
the 98 patients with an opioid-involved
hospitalization who died within 1 year
post-discharge due to a drug overdose not
involving an opioid, 25% died within 30
National Health Statistics Reports (cid:132) Number 141 (cid:132) June 16, 2020 Page 7
Neoplasms 29
Diseases of the
22
circulatory system
Diseases of the
19
respiratory system
Certain infectious and
9
parasitic diseases
Diseases of the
8
digestive system
External causes
6
of mortality
All other ICD–10
18
chapters
0 5 10 15 20 25 30
Percent
1Bar length may not appear equal due to rounding.
NOTES: ICD–10 is International Classification of Diseases, 10th Revision. Patients with an opioid-involved hospitalization who died within 1 year post-discharge is based on information
from the last opioid-involved hospitalization. The denominator (n = 1,462) is the number of patients who died within 1 year post-discharge of a cause other than drug overdose. Emergency
department visits that resulted in the patient being admitted to the hospital were included in hospitalizations. Newborns and those who died during the hospitalization were excluded from
analysis. Two records did not have an underlying cause of death and were also excluded from analysis. Figures may not add to 100 because of rounding. Data are not nationally
representative.
SOURCES: NCHS, 2014 National Hospital Care Survey data linked to the 2014–2015 National Death Index.
Figure 7. Most frequent underlying causes of death by ICD–10 chapter for patients with an opioid-involved hospitalization who died within
1 year post-discharge due to a cause other than drug overdose: National Hospital Care Survey, 2014, linked to National Death Index, 2014
and 2015
days post-discharge, 16% within 31–90 For patients who died of a drug only visit in 2014. Among those who died
days, and 59% within 91–365 days. Of overdose in 2014, what were their of another cause or were presumed alive,
the 243 patients with an opioid-involved patterns of hospital care use in the more than 99% did not have an opioid-
hospitalization who died within 1 year 2014 NHCS? involved ED-only visit (Table C and
post-discharge due to a drug overdose Appendix III).
Among the 1,014 patients with ED-
involving an opioid, 12% died within 30 Similarly, among the 568 patients
only visits who died of a drug overdose
days post-discharge, 19% within 31–90 with hospitalizations who died of a drug
in 2014, 83% did not have an opioid-
days, and 69% within 91–365 days overdose in 2014, 70% had no opioid-
involved ED-only visit, 13% had one
(Table A and Appendix II). involved hospitalizations, 25% had one
opioid-involved ED-only visit, and 4%
opioid-involved hospitalization, and
had more than one opioid-involved ED-
For patients with an opioid-
involved hospitalization who died
within 1 year post-discharge of a Table A. Time interval from date of discharge for last opioid-involved hospitalization to date
of death, by cause of death
drug overdose involving an opioid,
what were the specific opioids Death due to Death due to Death due to
cause other than drug overdose not drug overdose
identified on the death certificate?
drug overdose involving an opioid involving an opioid
Time interval
Of the 243 patients with an opioid-
(post-discharge) Number Percent Number Percent Number Percent
involved hospitalization who died within
1 year post-discharge of a drug overdose Within 30 days . . . . . . . . . . . . . . . . . 326 22.3 24 24.5 28 11.5
Within 31–90 days . . . . . . . . . . . . . . 327 22.4 16 16.3 47 19.3
involving an opioid, 46% involved
Within 91–365 days . . . . . . . . . . . . . 809 55.3 58 59.2 168 69.1
heroin, 20% involved fentanyl, 13%
Total . . . . . . . . . . . . . . . . . . . . . . . . . 1,462 100.0 98 100.0 243 100.0
involved oxycodone, 12% involved
methadone, and 12% involved morphine. NOTES: Emergency department visits that resulted in the patient being admitted to the hospital were included in hospitalizations.
Deaths may involve more than one drug Newborns and those who died during the hospitalization were excluded from analysis. Two records did not have an underlying
cause of death and were also excluded from analysis. Data are not nationally representative.
(Table B and Appendix II). SOURCES: NCHS, 2014 National Hospital Care Survey data linked to the 2014–2015 National Death Index.
Page 8 National Health Statistics Reports (cid:132) Number 141 (cid:132) June 16, 2020
6% had more than one opioid-involved For the patients who died of a drug or hospitalization in 2014 was not
hospitalization in 2014. Among those overdose involving an opioid in opioid-involved, the distribution of
who died of another cause or were 2014 and whose last 2014 NHCS principal diagnoses for the last ED-only
presumed alive, around 98% did not have ED-only visit or hospitalization visit or hospitalization was ill-defined
an opioid-involved hospitalization (Table was not opioid-involved, what conditions (19%), injury and poisoning
D and Appendix IV). were the most frequent principal (17%), mental disorders (15%), and
diagnoses? other conditions (49%) (Table F and
For patients who died of a drug Appendix V).
For the 545 patients who died of
overdose involving an opioid in
a drug overdose involving an opioid
2014 and had at least one ED-only
in 2014 and whose last ED-only visit
visit or hospitalization, was the
last 2014 NHCS visit prior to death
Table B. Specific opioids identified on death certificates among patients with an opioid-
opioid-involved?
involved hospitalization who died within 1 year post-discharge of a drug overdose involving
an opioid
Patients who died of a drug
overdose involving an opioid in 2014 Drug Number1 Percent
and had at least one ED-only visit or
Heroin. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 45.7
hospitalization in 2014 were examined.
Fentanyl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 19.7
Of the 698 patients who died of a drug Oxycodone. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 12.8
overdose involving an opioid in 2014 Methadone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 11.9
Morphine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 11.5
and who had at least one ED-only visit or
Hydrocodone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 4.1
hospitalization in 2014, the last ED-only Other opioids2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 8.6
visit or hospitalization prior to death The term “opioid” or “opiate” was noted,
but no specific drug was named. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 9.5
was opioid-involved for 22% of those
decedents (Table E and Appendix V). 1Categories are not mutually exclusive. A death may involve more than one opioid (e.g., a death involving both hydrocodone and
heroin would be counted in both categories). Deaths may also involve drugs other than opioids.
2Includes oxymorphone, codeine, hydromorphone, buprenorphine, and tramadol. For each of these drugs, there were fewer than 10
deaths that mentioned this drug as involved in the death.
NOTES: Emergency department visits that resulted in the patient being admitted to the hospital were included in hospitalizations.
Newborns and those who died during the hospitalization were excluded from analysis. Two records did not have an underlying
cause of death and were also excluded from analysis. Data are not nationally representative. There were 243 decedents with an
opioid mentioned as involved in the death.
SOURCES: NCHS, 2014 National Hospital Care Survey linked to the 2014–2015 National Death Index and to the 2014–2015 Drug-
Involved Mortality.
Table C. Number and percentage of patients by number of opioid-involved ED-only visits and mortality outcome in 2014
Died of a drug overdose Died of another cause Presumed alive
Characteristic Number Percent Number Percent1 Number Percent
No opioid-involved ED-only visits . . . . . . . . . . . . . . . . . . . . . . 845 83.3 31,799 99.4 2,352,763 99.4
One opioid-involved ED-only visit. . . . . . . . . . . . . . . . . . . . . . 133 13.1 178 0.6 12,185 0.5
More than one opioid-involved ED-only visit. . . . . . . . . . . . . . 36 3.6 26 0.1 2,007 0.1
Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,014 100.0 32,003 100.0 2,366,955 100.0
1Figures may not add to 100.0 because of rounding.
NOTES: ED is emergency department. ED visits that resulted in the patient being admitted to the hospital were included in hospitalizations. Two records that did not have an underlying cause of death
were excluded from analysis. Data are not nationally representative.
SOURCE: NCHS, 2014 National Hospital Care Survey data linked to the 2014–2015 National Death Index.
Table D. Number and percentage of patients by number of opioid-involved hospitalizations and mortality outcome in 2014
Died of a drug overdose Died of another cause Presumed alive
Characteristic Number Percent Number Percent Number Percent
No opioid-involved hospitalizations. . . . . . . . . . . . . . . . . . . . . 395 69.5 60,844 98.6 905,170 97.9
One opioid-involved hospitalization . . . . . . . . . . . . . . . . . . . . 139 24.5 675 1.1 16,507 1.8
More than one opioid-involved hospitalization . . . . . . . . . . . . 34 6.0 198 0.3 3,149 0.3
Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 568 100.0 61,717 100.0 924,826 100.0
NOTES: Emergency department visits that resulted in the patient being admitted to the hospital were included in hospitalizations. Newborns and those who died during the hospitalization were
excluded from analysis. Two records did not have an underlying cause of death and were also excluded from analysis. Data are not nationally representative.
SOURCES: NCHS, 2014 National Hospital Care Survey data linked to the 2014–2015 National Death Index.
National Health Statistics Reports (cid:132) Number 141 (cid:132) June 16, 2020 Page 9
Table E. Number and percentage of patients who died of a drug overdose involving an Interpretation of any results of this
opioid in 2014 and who had at least one ED-only visit or hospitalization in 2014 study should take into consideration the
limitations of the data and methodologies.
Characteristic Number Percent
NHCS is not currently nationally
Died of a drug overdose involving an opioid in 2014 and had at least one representative. The results of these
ED-only visit or hospitalization in 2014. . . . . . . . . . . . . . . . . . . . . . . . . . 698 100.0
analyses are not generalizable because
Last ED-only visit or hospitalization was opioid-involved . . . . . . . . . . . 153 21.9
Last ED-only visit or hospitalization was not opioid-involved . . . . . . . . 545 78.1 they are only based on the hospitals
participating in the 2014 NHCS and do
NOTES: ED is emergency department. ED visits that resulted in the patient being admitted to the hospital were included in
hospitalizations. Newborns and those who died during the hospitalization were excluded from analysis. Two records did not have an not capture events that may happen at
underlying cause of death and were also excluded from analysis. Data are not nationally representative. other non-NHCS facilities. The estimates
SOURCES: NCHS, 2014 National Hospital Care Survey data linked to the 2014–2015 National Death Index.
reported are unweighted and, therefore,
these estimates are not nationally
Table F. Principal diagnoses for patients who died of a drug overdose involving an opioid in
representative. NCHS has been making
2014 and whose last 2014 NHCS ED-only visit or hospitalization was not opioid-involved
efforts to increase hospital participation
ICD–9–CM chapter Number Percent so that national estimates can be made in
the future.
Ill-defined conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 18.7
Injury and poisoning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 16.9 The linked data only include
Mental disorders. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 14.9 hospitals participating in 2014, so any
Musculoskeletal system diseases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 8.6
hospital care received outside of 2014
Respiratory system diseases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 7.2
Circulatory system diseases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 5.5 is not included in the analyses. For the
Contact with health services. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 5.1 examples presented that select patients
Nervous system diseases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 4.8
based on mortality outcomes (e.g., drug
Skin diseases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 4.4
Digestive system diseases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 4.4 overdose deaths) and examine any prior
Other ICD–9–CM chapters1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 9.5 hospital care visits, it is important to note
Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545 100.0 that both encounters and deaths were
limited to events occurring solely in
1Includes International Classification of Diseases, Ninth Revision, Clinical Modification (ICD–9–CM) chapters: Infectious and 2014. Also, another limitation is that the
parasitic diseases, Genitourinary system diseases, Endocrine diseases, Blood diseases, Complications of pregnancy, Neoplasms,
and Congenital anomalies. length of time between hospital visits and
NOTES: ED is emergency department. ED visits that resulted in the patient being admitted to the hospital were included in death can vary. For instance, if a patient
hospitalizations. Newborns and those who died during the hospitalization were excluded from analysis. Two records did not have an
underlying cause of death and were also excluded from analysis. Data are not nationally representative. died of a drug overdose in February
SOURCES: NCHS, 2014 National Hospital Care Survey data linked to the 2014–2015 National Death Index. 2014, only hospital visits in the sample
occurring in January and February
Discussion included heroin, fentanyl, oxycodone, 2014 could be analyzed for this patient,
methadone, and morphine. Additionally, whereas if a patient died of a drug
The findings of these analyses researchers can examine patterns of overdose in December 2014, a longer
demonstrate the breadth of topics that hospital care among patients who died, period of time (i.e., January to December)
can be investigated using linked hospital such as the number of opioid-involved would be available to capture any
care and mortality data. Researchers ED-only visits or hospitalizations prior potential hospital visits. Interpretation
can examine patients’ hospital care; 30-, to death. For instance, in the example of the results should consider these
90-, and 365-day post-hospital discharge analyzing patients with hospitalizations limitations of the data. Future NHCS
mortality; and cause of death for those by mortality outcomes in 2014, this releases could be used to evaluate longer
who died. For instance, in the example study found that about 70% of those who periods of follow-up.
focusing on the 341 patients who had died of a drug overdose and over 98% The case definition for opioid-
an opioid-involved hospitalization and of those who died of another cause had involved ED-only visits and
also died due to a drug overdose within no opioid-involved hospitalizations; 6% hospitalizations consists of ICD–9–CM
1 year post-discharge, this study found of those who died of a drug overdose codes for past or present opioid (including
that 71% (243) of the deaths involved and 0.3% of those who died of another opium, heroin, methadone, and other
an opioid, where 12% died within 30 cause had more than one opioid-involved opiates and related narcotics) use, abuse,
days post-discharge, 19% within 31–90 hospitalization in 2014. While these data dependence, and poisoning; narcotics
days, and 69% within 91–365 days. are not nationally representative, these affecting the fetus or newborn; and
Furthermore, linkage to the DIM data examples highlight the novel information poisoning by opiate antagonists. Therefore,
enables researchers to identify the available and the potential for analyses cases included in these analyses are not
specific drugs involved in deaths among of hospitalizations and ED visits related limited only to patients who reported to
patients. For those who died of an opioid- to opioids and subsequent mortality the hospital because of an opioid overdose
involved drug overdose within 1 year outcomes among patients with opioid- but can include patients presenting to
post-discharge of an opioid-involved related hospital encounters, which other the hospital for other reasons if codes
hospitalization in the 2014 NHCS, data sources cannot provide. indicate a history of opioid dependence
the most commonly involved opioids
or abuse. For example, a record may be
Page 10 National Health Statistics Reports (cid:132) Number 141 (cid:132) June 16, 2020
assigned a code for opioid dependence References https://www.cdc.gov/nchs/data/
or abuse, yet the primary reason for the datalinkage/NHCS14_NDI14_15_
patient visiting the hospital may be for 1. U.S. Department of Health and Methodology_Analytic_Consider.pdf.
another health problem or condition, such Human Services, Assistant Secretary 9. Brown AM, DeFrances C, Crane
as cancer treatment. Interpretations of the for Planning and Evaluation. Patient- E, Cai R, Naeger S. Identification
findings in this report should consider the centered outcomes research trust of substance-involved emergency
composition of this study population. fund, 2018–2020. Available from: department visits using data from
Providers may differ in their https://aspe.hhs.gov/patient-centered- the National Hospital Care Survey.
practices for diagnosing (and thus the outcomes-research-trust-fund. National Health Statistics Reports;
ICD–9–CM coding of) opioid and 2. National Center for Health Statistics. no 114. Hyattsville, MD: National
other drug-involved ED visits and National Hospital Care Survey. 2014. Center for Health Statistics. 2018.
hospitalizations. Also, a hospital facility’s Available from: https://www.cdc. Available from: https://www.cdc.
capacity to conduct toxicological testing gov/nchs/nhcs/index.htm. gov/nchs/data/nhsr/nhsr114.pdf.
on suspected drug poisonings can vary, 3. National Center for Health Statistics. 10. Ingram DD, Franko SJ. 2013 NCHS
which can lead to inconsistent reporting National Death Index. 2014–2015. Urban-rural classification scheme for
of opioid involvement across hospitals Available from: https://www.cdc. counties. National Center for Health
in the United States. Interpretation gov/nchs/ndi/index.htm. Statistics. Vital Health Stat 2(166).
of the findings should consider these 4. National Center for Health Statistics. 2014. https://www.cdc.gov/nchs/
potential biases and misclassifications. In Drug Involved Mortality data. data/series/sr_02/sr02_166.pdf.
addition, the reporting of specific drugs 2014–2015. Available from: 11. U.S. Census Bureau. Census regions
on death certificates may vary by death https://www.cdc.gov/rdc/b1datatype/ and divisions of the United States.
investigation and reporting practices in dt1229.html. Available from: https://www2.
the United States. The medical certifier, 5. Williams S, Gousen S, DeFrances census.gov/geo/pdfs/maps-data/
typically a medical examiner or coroner C. National Hospital Care Survey maps/reference/us_regdiv.pdf.
in cases of sudden and unexpected demonstration projects: Pneumonia 12. National Center for Health Statistics.
deaths (e.g., drug overdoses), relies on inpatient hospitalizations and A reference guide for completing
his or her capacity to test and collect emergency department visits. the death certificate for drug toxicity
information necessary for determining National Health Statistics Reports; deaths. Hyattsville, MD. 2019.
the causes and manner of the death. The no 116. Hyattsville, MD: National Available from https://www.cdc.gov/
infrastructure, resources, and expertise Center for Health Statistics. 2018. nchs/data/nvss/vsrg/vsrg02-508.pdf.
available for conducting and interpreting 6. World Health Organization.
autopsies and toxicological tests, as International statistical classification
well as the reporting tendencies of the of diseases and related health
medical certifier, can affect the specific problems, 10th revision (ICD–10).
drugs reported on death certificates. The 5th ed. Geneva, Switzerland. 2016.
information on the drugs involved in a Available from: https://icd.who.int/
death are limited to those written by the browse10/2016/en.
medical certifier on the death certificate. 7. Trinidad JP, Warner M, Bastian BA,
NCHS has made significant efforts Miniño AM, Hedegaard H. Using
to improve cause-of-death reporting, literal text from the death certificate
particularly in the case of drug overdoses, to enhance mortality statistics:
and the specificity of drugs reported Characterizing drug involvement
on death certificates has increased in deaths. National Vital Statistics
substantially over the past few years (12). Reports; vol 65 no 9. Hyattsville,
Even considering these limitations, MD: National Center for Health
this report demonstrates the utility of the Statistics. 2016. Available from:
newly linked data set consisting of the https://www.cdc.gov/nchs/data/nvsr/
2014 NHCS, the 2014–2015 NDI, and the nvsr65/nvsr65_09.pdf.
2014–2015 DIM data. The linked data, 8. National Center for Health Statistics.
while not yet nationally representative, The linkage of the 2014 National
can lend insights into hospital use and Hospital Care Survey to the
drug overdose deaths involving opioids. 2014/2015 National Death Index:
This information can be used to inform Methodology overview and analytic
care for patients with opioid use disorders considerations. 2018. Available from:
and drug overdose prevention efforts.