Report comparing Bureau of Justice Statistics and Bureau of Justice Assistance mortality death collections

MEMORANDUM TO: Robert Sivinski
Office of Statistical and Science Policy
Office of Management and Budget
THROUGH:

Doris J. James, Acting Director, BJS
Allen Beck, Senior Statistician, BJS
Richard Kluckow, Unit Chief, Corrections Unit, BJS

FROM:

E. Ann Carson, Statistician, BJS

SUBJECT:

Report comparing Bureau of Justice Statistics and Bureau of Justice
Assistance mortality death collections, to fulfill Terms of Clearance for
OMB Control Number 1121-0249

DATE:

May 11, 2021

In June 2019, the Bureau of Justice Statistics (BJS) received clearance from the Office of
Management and Budget (OMB) to conduct its Mortality in Correctional Institutions (MCI) data
collection for an additional two years, through June 2021 (OMB Control Number 1121-0249.
The Terms of Clearance (TOC) for the MCI, issued June 30, 2019, required BJS to assess the
overlap and quality of data collected by the Bureau of Justice Assistance (BJA) when it began
collecting data of deaths in prisons and jails in response to the Deaths in Custody Reporting Act
of 2013. The attached report fulfills the TOC condition for BJS’s 2019 MCI collection.
There are three overarching factors that contextualize the findings in the attached report. First,
the period of comparisons of the BJS and BJA data collections covers three months: October,
November, and December of 2019. These are the only three months when both agencies
concurrently administered the two data collections on deaths in the custody of state prisons
and local jails. Secondly, the comparison months represent the first three months of BJA’s data
collection effort, while BJS has 20 years of experience collecting these data. Thirdly, the two
agencies have very distinct functions. BJS is a principal federal statistical agency with a more
than 40-year history of collecting, analyzing and disseminating reliable statistics on crime and
justice. BJA provides leadership and services in grant administration and criminal justice policy
development to support local, state, and tribal law enforcement in achieving safer
communities.

While BJA continues to develop and enhance its data collection to produce complete and
accurate statistics over time, this report provides a snapshot of the initial efforts in comparison
to the long-term MCI collection.
Key findings for October through December 2019:
• States reported 1,246 deaths in state prisons and local jails to BJS and 744 deaths to BJA
during the three-month period, or 59.7% of the deaths reported to BJS.
•

Six states, including California and New Jersey, did not report in-custody deaths to BJA
because they did not accept Justice Assistance Grants (JAG) for Fiscal Year (FY) 2020.
These six states reported 169 deaths to BJS’s MCI collection, or 13.6% of the MCI total
number of deaths.

•

When compared to MCI, BJA’s data collection did not capture any state prison deaths in
11 states or any local jail deaths in 12 states and the District of Columbia, which were
reported to BJS.

•

BJA’s data collection included only 38.9% of local jail deaths and 66.3% of state prison
deaths that occurred during the three-month period when compared to MCI.

•

Fifty-six of the deaths reported to BJA as having occurred in the process of arrest had
actually occurred in local jails and state prisons when compared to MCI.

•

BJA’s collection did not obtain full date of birth and had high rates of missing data on
key elements: date of facility admission (41.8%), description of death (23.8%), Hispanic
origin (19.7%), cause of death (13.8%), name of reporting agency/facility (8.9%), and
year of birth (5.5%).

•

The high rate of missing data limits statistical analysis of age of decedent, cause of
death, and time from admission to death, and it limits the use of the data to obtain
death certificates through the National Death Index.

•

BJS was able to matched 87 of the 117 local jail deaths (74.4%) and 601 of the 627 state
prison deaths (95.5%) in the BJA collection to records in the MCI collection following
additional data processing, including reconciling misspelled names, dates of death, and
years of birth.

•

When considered as a portion of all MCI records, the overall match rate of BJA’s data
was 34.5% for local jails and 72.8% for state prisons. For cause of death, the two
collections had a higher match rate: 76.4% of state prison deaths and 74.7% of local jail
records.

2

An Assessment of the Overlap and Quality of Mortality Data Collected by
the Bureau of Justice Statistics and Bureau of Justice Assistance
May 2021
Bureau of Justice Statistics
In June 2019, the Bureau of Justice Statistics (BJS) received clearance from the Office of
Management and Budget (OMB) to conduct BJS’s Mortality in Correctional Institutions
(MCI) data collection for an additional two years, through June 2021 (OMB Control
Number 1121-0249). This permitted BJS to collect data on deaths in state prisons and
local jails that occurred during calendar years 2018 and 2019.
As part of the MCI clearance, OMB imposed the following terms of clearance (TOC) on
BJS:
Clearance for the Mortality in Correctional Institutions collection is
granted for 2 years from approval date. BJS will assess the overlap and
quality of the data once BJA begins its collection of prison and jail deaths
in 2020, and the need for BJS to continue MCI, revise the collection, or
discontinue it in its entirety. This decision will be based on an assessment
of the quality and completeness of the data collected by BJA, including
whether BJA central state respondents are able to obtain sufficient
participation from local jails on a quarterly basis.
The following assessment fulfills this TOC condition.
Background
BJS is a principal federal statistical agency that was established in 1979 to collect,
analyze and report reliable statistics on crime and justice. BJS began collecting data on
deaths occurring in local jails in 2000 and on deaths in state prisons in 2001, in response
to the Death in Custody Reporting Act of 2000 (DICRA; P.L. 106-297). BJS has continued
to collect these data for the last 20 years, including when the law expired in 2006 and
was reauthorized in 2014 (P.L. 113-242). Throughout the 20-year period, BJS has
maintained an average annual 98% response rate for local jails and a 100% response
rate for state departments of corrections. These response rates are impressive, given
the voluntary nature of participation in the MCI collection.
In 2016, the Department of Justice (DOJ) decided to place more emphasis on the section
of P.L. 113-242 that concerned non-compliance with the data collection. Per the law,
states that did not report on a quarterly basis individual-level data on deaths occurring
in local jails, in state prisons, or in the process of arrest, could be penalized up to 10% of
the DOJ-administered Justice Assistance Grants (JAG) awards. The DOJ determined that
the Bureau of Justice Assistance (BJA) should manage collection of the data pursuant to
3

the law because BJS, as a federal statistical agency, may not collect data for law
enforcement purposes. BJA is not under similar requirements to collect data for
statistical purposes only. (The Report of the Attorney General to Congress Pursuant to
The Death in Custody Reporting Act, December 16, 2016, is located at
https://www.justice.gov/archives/page/file/918846/download).
OMB determined that, since the overlap in BJS’s MCI and BJA’s Deaths in Custody
Reporting Act data collection (DCRA) was significant, only one agency should burden the
public with its request for data. OMB encouraged BJS to work with BJA to ensure that all
necessary data elements were collected by BJA to allow BJS to publish annual statistical
reports describing deaths in local jails and state prisons. Because BJA required time to
develop and implement the DCRA protocol and web-based data entry system, BJS
agreed to collect MCI data through the end of calendar year 2019. BJA would begin
collecting the data in the first quarter of fiscal year 2020 (starting October 1, 2019). This
schedule resulted in an overlap of data collection by the two agencies of three months
(October through December 2019).
This assessment compares BJS’s collection of mortality data from prisons and jails,
which has been administered for 20 years, to BJA’s new data collection covering three
months. While BJS is a statistical agency, BJA provides leadership and services in grant
administration and criminal justice policy development to support local, state, and tribal
law enforcement in achieving safer communities.
BJA has made a concerted effort over the past year to improve its collection of mortality
data. BJS assisted in this effort by mailing information regarding the new reporting
agency to all of BJS’s respondents, explaining the new process individually via email or
over the phone, and discussing issues with the data-collection elements and the
processes by which BJA collects data and follows up on non-response and data quality.
BJA is undertaking enhancements to its data collection (see attached Appendix A), and
anticipates significant improvements in response rates and quality of data over time.
Quality and Completeness of the Data Collected by BJA
Per P.L. 113-242, BJA must collect death data from states on a quarterly basis, to include
deaths that occurred in state prison, in local jails, and in the process of arrest (also
known as arrest-related deaths, or ARD). If individual states do not comply, the law
allows the Attorney General to withhold up to 10% of the states’ JAG awards. JAG grants
awarded directly to local agencies are not subject to any penalties for non-compliance
under P.L. 113-242. During the development of the MCI data collection, BJS observed
the difficulties with states assembling data from individual law enforcement agencies,
local jails, and state department of corrections. Few of the individual law enforcement
agencies would be directly affected by the law’s penalties on the state grant awards.
Additionally, the linkage of death reporting to JAG funding means that states that do not
apply for or accept JAG awards are under no obligation to provide death data to BJA.
4

When comparing the data collections, BJS found that states that did not accept JAG
awards did not report deaths to BJA that were reported to BJS.
During the implementation of the MCI data collection, BJS concluded that quarterly
reporting could lead to survey fatigue, diminished unit response rates, and, particularly
in the case of deaths that require autopsies, poor response rates on items such as cause
of death. Since autopsies can take months to complete, quarterly reporting
requirements may force states to leave the cause of death field blank. BJA has
instructed states to submit a system helpdesk ticket to update cause of death when that
information becomes available; however, BJS found a higher percentage of unknown
causes of death in BJA‘s data than in BJS’s data, which BJS attributes to the incomplete
reporting due to the quarterly reporting requirement.
During the initial implementation of its data collection, BJA incorporated the DCRA
reporting into its existing Performance Measurement Tool (PMT) tracking system. BJS
recommended that BJA adopt a data collection platform that allowed for the uploading
of multiple records in a single database. Based on BJS’s 20-year data collections, some
states have tens of deaths per quarter while other states have hundreds of in-custody
deaths per quarter. While this option was not available for the first quarter of BJA’s data
collection, BJA enacted this change in the second quarter of FY 2020.
The PMT does not have an internal data error identification tool. As a result, BJS did
extensive cleaning of the BJA data prior to analyzing the data, including removing more
than 200 duplicate death records. BJS found that multiple decedents’ first and last
names were recorded in the wrong columns and determined that the PMT system did
not generate automatic missing data prompts to respondents for key data elements,
such as year of birth, location of death, and date of facility admission.
BJA is actively working to improve their cleaning and validation processes; however,
these efforts were not undertaken on the first quarter of FY 2019 data. Many of the
data entry errors could have been caught with systematic data checks. For example,
Florida reported execution as the cause of death for 110 of its 114 submitted records for
the second quarter of FY 2020. This is clearly incorrect, given that 25 persons were
executed across all states in 2018 and 22 executions in 2019 (advance count), based on
BJS’s capital punishment data collection which has been ongoing since 1930.
The BJA data collection form
BJA designed its data collection form (Appendix B) to obtain the information specified in
section 2 of P.L. 113-242 (https://www.congress.gov/113/plaws/publ242/PLAW113publ242.htm):
(b) Information Required. --The report required by this section
shall contain information that, at a minimum, includes-5

(1) the name, gender, race, ethnicity, and age of the
deceased;
(2) the date, time, and location of death;
(3) the law enforcement agency that detained, arrested, or
was in the process of arresting the deceased; and
(4) a brief description of the circumstances surrounding the
death.
Many of the items overlap those collected in MCI, although BJS did not have a free-text
description of the circumstances of the death. Instead, BJS designed questions to
standardize the reporting on topics such as whether the decedent was undergoing an
autopsy, where in the facility a person died (general housing unit, segregation unit, in
transit, etc.), and whether the decedent had received specific medical care since
admission, including medication and overnight stays in a mental health unit. BJS also
asked respondents to indicate whether illness-related deaths were due to pre-existing
medical conditions or conditions that developed after admission to prison or jail.
While the BJA free-text descriptions of the circumstances surrounding a death can yield
additional information not captured in the MCI questions, the free-text fields are
difficult to analyze and can have high rates of item missingness. Of the 892 unique
death records reported to BJA from October to December 2019, 212 records (almost
24%) had some form of “Unknown,” “N/A,” or other missing value in the text
description field. Also, while BJA collected structured data on the name, address, and
type of the facility where the person died, the location within the facility where the
death occurred was collected only in the free-text field.
In this report, table 1 presents the number of missing records for key data elements in
the BJA collection. These include all responses of “Don’t know,” “Unknown,” “N/A,”
and any other indication of no response. Except for the text description of death, the
count of missing records does not include those labelled as “unknown pending
autopsy.” The high rates of missing data on date of arrest and date of admission to the
facility where the death occurred are of particular concern. These variables are
important for calculating the time served in prison or jail before death, and provide key
data for understanding risk of death by cause. Dates of arrest or admission were
missing for more than 40% of BJA death records submitted from October to December
2019.

6

Table 1. Number and percent of BJA DCRA death records with missing values for key
data elements, October 1 to December 31, 2019
Data element
Sex
Race
Hispanic origin
Unknown race AND unknown Hispanic origin
Decedent’s first name
Decedent’s last name
Year of birth
Date of death
Date of arrest or facility admission
Time of death
Location of death
Name of facility/agency
Cause of death/a
Text description of death circumstances/b

Number of
records

0
85
176
32
5
5
49
0
373
2
75
79
123
212

Percent of total
records (n=892)
0.0 %
9.5
19.7
3.6
0.6
0.6
5.5
0.0
41.8
0.2
8.4
8.9
13.8
23.8

Note: Analysis based on records reported to BJA for unique deaths that occurred between October 1
and December 31, 2019.
a/Includes all records for which the cause of death was reported as unknown or pending autopsy.
b/Includes all records for which the text description of death field contained “N/A,” “Unknown,” “None
provided,” “*,” “No information to report,” or a description of other missing data elements (e.g., “Time
of death not reported”). This does not include records for which the death description was reported as
“unknown pending results from a medical examiner or autopsy.”

BJS’s MCI data collection separates accidental cause of death into three separate
categories: accidental alcohol or drug intoxication, accidental injury to self, and
accidental injury by other. In contrast, BJA’s collection combines all accidental deaths, so
intoxication deaths cannot be disaggregated. BJS found that in 2016 intoxication deaths
accounted for 10% of all deaths in local jails, compared to less than 2% for other
accidents. The number of intoxication deaths in local jails is of special interest to jail
administrators and policymakers. Intoxication deaths have more than tripled since 2000.
Because inmates who die from intoxication spend a median of one day in jail before
death, this statistic informs jail administrators of the need to rapidly recognize and
intervene in situations where an inmate is intoxicated upon admission to jail. Such
intervention could significantly reduce the number of deaths.
BJA did not request full date of birth (only year of birth) or social security number (SSN)
on each decedent, but in the future, should collect these data elements to improve data
quality and enhance the utility of the data. During BJA’s development of the DCRA
collection form, BJS requested these data elements be added, in order to submit the
records to the Centers for Disease Control and Prevention (CDC) to match to the
National Death Index (NDI). The NDI includes all causes of death on a decedent’s death
7

certificate and can provide multiple medical conditions that contributed to the death.
The NDI uses the SSN for matching records, but in lieu of SSN, it requires full name and
date of birth. If the BJA death data could be linked to NDI, data with unknown or
unreported causes of death can be reconciled for the BJS comparison and assessment of
the data collection.
Direct comparison of BJS and BJA data
BJA provided a file of all reported deaths in the PMT as of mid-September 2020. After
limiting the data to those deaths occurring between October 1 and December 31, 2019,
there were 1,102 records. Of these records, 210 were duplicates. Of the remaining 892
unique death records, 204 had been classified by BJA as not having occurred in one of
the correctional facilities named on the form. (See figure 1). BJS initially interpreted
these as ARD reported by local law enforcement agencies. Upon further investigation,
BJS found that 56 of these deaths matched individual decedents in state prisons or local
jails in BJS’s MCI collection. They were reclassified to match the MCI location of death.
Since MCI does not include ARD deaths, the remaining 144 BJA ARD deaths were
removed from the subsequent analysis. In addition, MCI does not include statesanctioned executions in its counts of deaths in custody, so these records were also
removed from the BJA comparison file.

8

Figure 1. BJA and BJS death records, October 1 to December 31, 2019
BJS MCI data
1,246 unique death records

BJA DCRA data
892 unique death records

94 local jail

594 state prison
204 “None of the
above” (ARD)

627 state prison*

144 ARD

117 local jail

301 local jail

945 state prison

87 local jail matches
601 state prison matches
*Executions deleted.
Sources: Bureau of Justice Statistics, Mortality in Correctional Institutions 2019; Bureau of Justice Assistance,
Deaths in Custody Reporting Act 2019.

The 744 deaths reported to BJA as having occurred in state prisons, local jails, external
hospitals, and other or unknown locations were included in the comparison with BJS
data on state prison and jail deaths. Using the reporting agency and details in the text
field that allowed for a description of the death, BJS categorized as many unknown and
other deaths into jail or arrest-related deaths as possible.
Six states (California, Connecticut, Delaware, New Jersey, Utah, and West Virginia) had
not accepted their JAG awards as of September 2020, and therefore were not required
to report mortality data to BJA. These states contributed 169 deaths (13.6%) to the BJS
total count of deaths in state prisons and local jails between October 1 and December
31, 2019. California is one of a few states with a centralized reporting system for law
enforcement agencies, local jails, and state prisons, which would make reporting to
DCRA relatively straightforward.
State departments of corrections (DOCs) and local jails reported 1,246 deaths between
October and December 2019, to BJS’s MCI collection. At the time of this analysis, BJS
was still collecting data on deaths that occurred in calendar year 2019 from a few local
jails.
9

Table 2 summarizes the state-specific count of records submitted to BJA and BJS,
disaggregated by type of facility. For the deaths reported to BJA, inmates who died in
external hospitals or other or unknown locations were combined with those who died in
local jails. Note that this is a comparison of the raw count of records, not a count of
individuals matched between the BJA and BJS collections.
Table 2. Number of deaths reported by state, facility type, and data collection agency,
October to December 2019
Total

State
Total
Total without 6 non-reporting BJA states
Alabama
Alaska/a
Arizona
Arkansas
California/b
Colorado
Connecticut/a,b
Delaware/a,b
District of Columbia
Florida
Georgia
Hawaii/a
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire

BJA
744
744
30
0
0
17
/
0
/
/
0
83
44
5
10
26
19
3
0
18
0
5
2
13
1
1
5
23
0
4
12
0

BJS
1,246
1,077
42
1
32
19
138
23
2
3
1
116
52
1
9
47
20
4
10
23
42
5
13
14
32
1
23
24
5
3
9
1

Local jails
BJA BJS
117 301
117 252
3
8
0
~
0
9
1
3
/
43
0
8
/
~
/
~
0
1
0
33
9
13
0
~
4
3
7
15
5
5
0
1
0
1
6
6
0
10
1
1
2
3
2
3
1
6
1
1
1
4
4
5
0
1
1
0
2
2
0
1

State prisons
BJA
BJS
627
945
627
825
27
34
0
1
0
23
16
16
/
95
0
15
/
2
/
3
~
~
83
83
35
39
5
1
6
6
19
32
14
15
3
3
0
9
12
17
0
32
4
4
0
10
11
11
0
26
0
0
4
19
19
19
0
4
3
3
10
7
0
0
10

Total
State
New Jersey/b
New Mexico
New York
North Carolina
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island/a
South Carolina
South Dakota
Tennessee
Texas
Utah/b
Vermont/a
Virginia
Washington
West Virginia/b
Wisconsin
Wyoming

BJA

/
0
28
31
37
23
12
45
1
15
4
29
157
/
2
33
3
/
0
3

BJS
12
9
31
33
44
28
12
54
1
27
4
40
152
6
2
28
17
8
20
3

Local jails
BJA BJS
/
2
0
4
0
4
4
6
4
10
1
4
1
4
8
18
0
~
4
6
0
0
0
11
28
23
/
1
0
~
13
8
3
7
/
3
0
3
1
1

State prisons
BJA
BJS
/
10
0
5
28
27
27
27
33
34
22
24
11
8
37
36
1
1
11
21
4
4
29
29
129
129
/
5
2
2
20
20
0
10
/
5
0
17
2
2

~ Not applicable.
/Data not reported.
a/BJS defines state as a combined prison/jail system, and reports all deaths as having occurred in prison.
b/State did not provide data because it had not accepted a JAG award as of November 1, 2020.
Sources: Bureau of Justice Statistics, Mortality in Correctional Institutions 2019; Bureau of Justice Assistance, Deaths
in Custody Reporting Act 2019.

In total, states reported 744 deaths from October to December 2019 to BJA, 59.7% of
the 1,246 deaths reported to BJS in the same timeframe. When the six states not
reporting to BJA because they have not accepted their JAG awards are removed from
the BJS total, the BJA records represent 69.1% of those received by BJS.
From October to December 2019, a total 117 deaths in local jails were reported to BJA,
which overall represent 38.9% of the 301 reported to BJS (and 46.4% when the six nonreporting states are removed). State prison deaths are better represented in BJA’s DCRA
data collection: 627 deaths in state prisons were reported to BJA in the first quarter of
FY 2020, representing 66.3% of the 945 reported to BJS and 76.0% when the six nonreporting states are removed.
In addition to the six states that had not accepted JAG grants, nine states and the
District of Columbia reported zero deaths in state prisons and local jails to BJA from
11

October to December 2019. The states included Alaska, Arizona, Colorado, Kansas,
Louisiana, Montana, New Hampshire, New Mexico, and Wisconsin. Each of these nine
states reported at least one death in state prison and/or local jail to MCI. Overall, they
represented 144 of the 1,246 deaths reported to BJS during this period.
Considering jail and prison deaths separately, twelve states and the District of Columbia
reported zero local jail deaths to BJA, while reporting at least one jail death to BJS from
October to December 2019. There were no reported prison deaths in the BJA data from
11 state departments of corrections over the same three months, while only 2 of these
departments of corrections (Minnesota and New Hampshire) also reported zero deaths
to BJS. Together, these jurisdictions reported 87 jail deaths and 152 state prison deaths
to MCI.
In the first quarter of FY 2020, BJA records from four states included a greater number
of deaths in local jails than were reported to MCI. Idaho and Nebraska each had one
more death in the BJA collection than in the MCI, while Texas and Virginia each had five
additional deaths. BJS attempted to reconcile these deaths, but the BJA records for each
lacked entries for location of death, name of the agency reporting the death, and any
additional information in the text description of the death that would allow BJS to follow
up with the agencies for more details. There is no way to confirm whether individuals
could have been released from jail prior to their death. However, if that were the case,
those deaths would not fit under BJS’s definition of a death in custody.
Similarly, five states reported more prison deaths to BJA than to BJS. These records
contained more contextual information than the jail records, including the facility name
and address. Pennsylvania and New York each had one more reported death in BJA’s
DCRA data than in MCI. Nevada and Oregon each had three more state prison deaths
reported to BJA than to BJS, and Hawaii had four additional deaths. BJS is contacting the
DOCs to obtain additional information on these deaths. It appears that several of these
individuals died in community medical facilities; however, it is unknown whether they
had been officially released from prison prior to their death.
Table 3 shows the distribution of sex, race and Hispanic origin, and cause of death for
decedents in local jails and state prisons for each data collection. With the exception of
Hispanic decedents in local jails, the sex and race/Hispanic origin distributions are
similar between the collections. The BJA collection, however, shows a lower percentage
of illness and drug or alcohol intoxication deaths in both state prisons and local jails than
those reported to BJS, and a far higher percentage of deaths with unknown or
unresolved causes of death. While BJA state respondents are expected to update the
causes of death in the PMT system once autopsy findings are complete, this had not
occurred for a large number of deaths from the first quarter of FY 2020, even after nine
months. BJA is working to implement a data cleaning system and provide more technical
assistance to the state respondents, which may ameliorate some of these issues.
12

Table 3. Characteristics of deaths by facility type and data collection agency, October 1
to December 31, 2019
Characteristic
Total deaths, Q1, FY 2020
Sex
Male
Female
Race/Hispanic origin
White/a
Black/a
Hispanic
American Indian/Alaska Native/a
Asian/a,b
Two or more races/a
Unknown/c
Cause of death
Illness
Drug/alcohol intoxication/d
Other accidents
Suicide
Homicide
Unknown/e
Other

Local jail deaths
BJA
BJS
117
301
100 %
100
81.2 %
82.7
18.8
17.3
100 %
100
58.1 %
58.1
29.1
27.9
5.1
11.0
1.7
1.0
0.9
1.3
0.0
0.0
5.1
0.3
100 %
100
35.0 %
45.8
6.8
15.3
6.0
2.3
29.1
26.6
2.6
2.0
17.1
3.3
3.4
4.7

%
%
%
%

%
%

State prison deaths
BJA
BJS
627
945
100 %
100 %
96.2 %
96.3 %
3.8
3.7
100 %
100 %
57.3 %
56.1 %
30.0
30.1
10.0
11.6
0.8
1.4
1.3
0.7
0.3
0.1
1.0
0.0
100 %
100 %
65.1 %
75.7 %
2.2
6.5
1.1
0.7
7.7
8.3
3.2
4.1
14.0
3.8
6.7
1.0

Note: Comparisons are of unmatched records from BJA and BJS mortality collections for deaths occurring
between October 1 and December 31, 2019.
a/Not of Hispanic origin.
b/Includes Asians, Other Pacific Islanders, and Native Hawaiians.
c/Includes other races and missing race responses.
d/BJA does not disaggregate intoxication deaths from other accidental deaths. BJS examined the
accidental deaths reported to BJA and recoded those that suggested intoxication deaths. Since not all
deaths included additional contextual details, the BJA percentages may be an underestimate.
e/Includes unknown and missing causes of death due to incomplete autopsy results.
Sources: Bureau of Justice Statistics, Mortality in Correctional Institutions 2019; Bureau of Justice
Assistance, Deaths in Custody Reporting Act 2019.

Matching individual BJA and BJS death records
BJS performed three rounds of matches for state prison deaths and local jail deaths,
each with slightly less stringent variable requirements, as summarized in Table 4.
After the initial rounds of matching, all remaining unmatched local jail and state prison
BJA death records were combined with the BJA ARD death records and visually
13

compared to the remaining unmatched BJS records for each state. A number of
decedents in the two data collections were clearly the same individuals, but their death
records contained misspelled names, inverted first and last names, or mismatched
facility types, dates of death, or years of birth. In the case of mismatched dates of death,
dates were usually off by one or two days. In the case of mismatched years of birth, the
most common difference was a missing value in DCRA. A number of deaths coded in
DCRA as having occurred in the process of arrest had occurred in a state prison or local
jail as reported in the MCI collection.
Because the MCI is the existing collection with a 20-year history, it was considered the
data collection of record in terms of correcting misspelled names, dates, and facility
types. In all, BJS was able to reconcile mismatches between the two collections in 20 last
names, 7 first names, 28 dates of death, 4 years of birth, and 70 facility types.
Table 4. BJA-BJS matching process for local jail and state prison deaths, October 1 to
December 31, 2019
Match

1

Variables used in
match

State, last name,
first name, date of
death, year of birth
2
State, last name,
first initial of first
name, date of
death, year of birth
3
State, last name,
date of death
TOTAL MATCHES

Local jail matches
Before
After
record
record
reconciliation
reconciliation
54
77

State prison matches
Before
After
record
record
reconciliation
reconciliation
383
470

2

2

98

101

8

8

24

30

64

87

505

601

Note: Reconciliation involved visual comparison of all BJA and BJS records unmatched after the initial three
rounds of matching. Arrest-related deaths were included in the BJA unmatched group to capture any
misclassified state prison or local jail deaths. See Figure 1.
Sources: Bureau of Justice Statistics, Mortality in Correctional Institutions 2019; Bureau of Justice Assistance,
Deaths in Custody Reporting Act 2019.

BJS matched 87 of the 117 local jail deaths (74.4%) and 601 of the 627 state prison
deaths (95.5%) to records in the MCI collection. When considered as a portion of MCI
records, the match rates drop to 34.5% for local jails and 72.8% for state prisons, even
when records from the six states that had not accepted JAG grants were removed from
the BJS collection.
The match rates varied by state. Table 5 shows the rate of BJA and BJS matched death
records by type of facility as a percentage of the BJS record totals. Eighteen states
matched 100% of the state prison death records between the BJA and BJS collections,
including Texas, which reported 129 state prison deaths to both BJA and BJS for the first
14

quarter of FY 2020. Florida had 82 of 83 state prison deaths match between the BJA and
BJS collections, but did not submit any deaths from local jails to BJA.
Table 5. Death record matches by facility type and data collection agency, October 1 to
December 31, 2019
Local jails

State

Total
Total w/out 6 states
Alabama
Alaska/a
Arizona
Arkansas
California/b
Colorado
Connecticut/a,b
Delaware/a,b
District of Columbia
Florida
Georgia
Hawaii/a
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey/b
New Mexico
New York
North Carolina
Ohio
Oklahoma

BJA
117
117
3
0
0
1
/
0
/
/
0
0
9
0
4
7
5
0
0
6
0
1
2
2
1
1
1
4
0
1
2
0
/
0
0
4
4
1

BJS
301
252
8
~
9
3
43
8
~
~
1
33
13
~
3
15
5
1
1
6
10
1
3
3
6
1
4
5
1
0
2
1
2
4
4
6
10
4

Matched
87
87
3
~
0
1
/
0
/
/
0
0
4
~
3
4
4
0
0
5
0
1
1
2
1
0
1
4
0
0
2
0
/
0
0
3
1
0

Percent
matched
28.9
34.5
37.5
:
0.0
33.3
/
0
/
/
0.0
0.0
30.8
:
100.0
26.7
80.0
0.0
0.0
83.3
0.0
100.0
33.3
66.7
16.7
0.0
25.0
80.0
0.0
:
100.0
0.0
/
0.0
0.0
50.0
10.0
0.0

State prisons

%
%

BJA
627
627
27
0
0
16
/
0
/
/
0
83
35
5
6
19
14
3
0
12
0
4
0
11
0
0
4
19
0
3
10
0
/
0
28
27
33
22

BJS
945
825
34
1
23
16
95
15
2
3
0
83
39
1
6
32
15
3
9
17
32
4
10
11
26
0
19
19
4
3
7
0
10
5
27
27
34
24

Matched
601
601
27
0
0
16
/
0
/
/
0
82
32
1
6
16
13
2
0
11
0
4
0
11
0
0
4
19
0
3
7
0
/
0
27
27
31
21

Percent
matched
63.6
72.8
79.4
0.0
0.0
100.0
/
0.0
/
/
:
98.8
82.1
100.0
100.0
50.0
86.7
66.7
0.0
64.7
0.0
100.0
0.0
100.0
0.0
:
21.1
100.0
0.0
100.0
100.0
:
/
0.0
100.0
100.0
91.2
87.5
15

%
%

Local jails
State
Oregon
Pennsylvania
Rhode Island/a
South Carolina
South Dakota
Tennessee
Texas
Utah/b
Vermont/a
Virginia
Washington
West Virginia/b
Wisconsin
Wyoming

BJA
1
8
0
4
0
0
28
/
0
13
3
/
0
1

BJS
4
18
~
6
0
11
23
1
~
8
7
3
3
1

Matched
1
8
~
3
0
0
23
/
~
8
3
/
0
1

Percent
matched
25.0
44.4
:
50.0
:
0.0
100.0
/
:
100.0
42.9
/
0.0
100.0

State prisons
BJA
11
37
1
11
4
29
129
/
2
20
0
/
0
2

BJS
8
36
1
21
4
29
129
5
2
20
10
5
17
2

Matched
7
36
1
11
4
29
129
/
2
20
0
/
0
2

Percent
matched
87.5
100.0
100.0
52.4
100.0
100.0
100.0
/
100.0
100.0
0.0
/
0.0
100.0

: Not calculated.
~ Not applicable.
/Data not reported.
a/BJS defines state as a combined prison/jail system, and reports all deaths as having occurred in prison.
b/State did not provide data because it had not accepted JAG award as of November 1, 2020.
Sources: Bureau of Justice Statistics, Mortality in Correctional Institutions 2019; Bureau of Justice Assistance, Deaths in
Custody Reporting Act 2019.

For the most part, the demographic variables that were common between the two
collections agreed. For race and Hispanic origin, the values on matched BJA and BJS
death records agreed for 94.3% of deaths in local jails and 96.2% for deaths in state
prisons. A single state prison death record had a mismatch on sex between BJA and BJS
collections. Cause of death, however, had a slightly lower percent of agreement
between the two death collections, 76.4% for matched state prison deaths and 74.7%
for matched local jail records.
Tables 6 and 7 show the collection-specific assigned causes of death for matched
records between the BJA and BJS collections. Frequencies on the diagonal of the table
represent deaths for which the cause of death was the same on the BJA and BJS records.
Using BJS’s cause of death for comparison purposes, suicide had the highest percent
agreement in state prisons (85.7%, or 42 of the 49 BJS suicide deaths) and in local jails
(96.3%, or 26 of the 27 BJS suicide deaths). In cases in which BJS determined illness as
the cause of death, the collections were consistent in 80.6% of the matched state prison
records and 82.8% of local jail records.
The unresolved causes of death in the BJA collection were responsible for much of the
disagreement on matched records. On 81 of the 601 matched state prison death
records, the decedent was assigned an “unknown pending autopsy” cause of death in
16

the BJA data collection, while BJS records showed a known cause of death. The majority
of these deaths (55 of 81) were illness-related deaths; however, four of these records
were suicides and four were homicides. The utility of a mortality dataset is limited when
the cause of death is missing for a significant number of records. BJA will need to invest
in non-response follow-up to obtain causes for deaths in previous quarters to improve
its data collection.

17

Table 6. Comparison of matched state prison death records by cause and data collection agency, October 1 to December 31, 2019
BJA cause of death

BJS cause of death
Illness
Drug/alcohol intoxication
Other accidents
Suicide
Homicide
Unknown/b
Other
Total

Illness
378
1
0
0
0
0
11
390

Drug/alcohol
intoxication/a
0
14
0
0
0
0
0
14

Other
accidents
0
4
2
0
0
0
1
7

Suicide
2
1
0
42
0
0
1
46

Homicide Unknown/b Other
0
55
34
0
7
2
0
2
0
0
4
3
18
4
0
0
4
1
0
9
1
18
85
41

Note: Comparisons are of matched records from BJA and BJS mortality collections for deaths occurring in state prisons between
October 1 and December 31, 2019.

Total
469
29
4
49
22
5
23
601

a/BJA does not disaggregate intoxication deaths from other accidental deaths. BJS examined the accidental deaths reported to BJA
and recoded those that suggested intoxication deaths. Since not all deaths included additional contextual details, the BJA counts may
be an underestimate.
b/Includes unknown and missing causes of death due to incomplete autopsy results.
Sources: Bureau of Justice Statistics, Mortality in Correctional Institutions 2019; Bureau of Justice Assistance, Deaths in Custody
Reporting Act 2019.

18

Table 7. Comparison of matched local jail death records by cause and data collection agency, October 1 to December 31, 2019
BJA cause of death

BJS cause of death
Illness
Drug/alcohol intoxication
Other accidents
Suicide
Homicide
Unknown/b
Other
Total

Illness
29
0
0
0
0
2
0
31

Drug/alcohol
intoxication/a
0
4
0
0
0
2
1
7

Other
accidents
2
1
3
0
0
0
0
6

Suicide
0
0
0
26
0
0
0
26

Homicide Unknown/b Other
0
3
1
0
4
2
0
0
0
0
1
0
1
1
0
0
2
0
0
2
0
1
13
3

Total
35
11
3
27
2
6
3
87

Note: Comparisons are of matched records from BJA and BJS mortality collections for deaths occurring in local jails between October 1
and December 31, 2019.
a/BJA does not disaggregate intoxication deaths from other accidental deaths. BJS examined the accidental deaths reported to BJA
and recoded those that suggested intoxication deaths. Since not all deaths included additional contextual details, the BJA counts may
be an underestimate.
b/Includes unknown and missing causes of death due to incomplete autopsy results.
Sources: Bureau of Justice Statistics, Mortality in Correctional Institutions 2019; Bureau of Justice Assistance, Deaths in Custody
Reporting Act 2019.

19

Conclusion
Given BJS’s more than 40-year history as a federal statistical agency and 20-year history
collecting data on deaths occurring in the custody of state prisons and local jails, BJS
anticipated that the first-time BJA deaths in custody data collection would require
enhancements over time. Based on the assessment of the first three months of BJA’s
DCRA data collection, BJS concludes that it could not be used to calculate accurate
mortality statistics for state prisons and local jails. Overall, when compared to BJS’s MCI,
BJA’s DCRA data collection included only 38.9% of local jail deaths and 66.3% of state
prison deaths that occurred between October and December 2019. BJA did not collect
any data from the six states that did not accept their JAG awards, including California,
and did not capture any state prison deaths in 11 states or any local jail deaths in 12
states and the District of Columbia. In comparison, MCI had at least one local jail or
state prison death in these jurisdictions during this time period.
The low data quality and incomplete coverage will need to be addressed to make
available accurate in-custody mortality rates in the United States for 2020 and future
years. BJS will close-out the MCI at the end of calendar year 2019, which will eliminate
the independent source to which DCRA data can be compared. The absence of highquality data and lack of national coverage will be especially missed when researchers
and policymakers try to determine the increased numbers of deaths in 2020 of prisoners
and jail inmates due to the coronavirus pandemic. As shown in the data from October to
December 2019, illness deaths reported to BJS are more likely to be reported to BJA as
having an unknown cause of death.
In addition to not meeting requirements for national coverage of local jail and state
prison deaths, DCRA data do not capture the same level of detail as MCI for location of
death, date of facility admission, and offenses for which the decedent was incarcerated.
The DCRA text description field is incomplete, inconsistent, and difficult to parse and
standardize for analytical purposes.
BJS concludes that it is difficult to draw definitive conclusions about the efficacy of a
first-time data collection with three months of data when compared to a 20-year data
collection. In the coming months, BJS will have additional aggregate death data from
2020 for jails and prisons which can be used to further assess the BJA collection. More
importantly, OJP, BJS, and BJA plan to explore alternative data collection protocols that
would provide more complete data on mortality in state prisons and local jails.
Meanwhile, BJS and BJA will continue to collaborate on implementing improvements to
the current BJA data collection.

20

Appendix A. An assessment of DCRA collection improvements, provided by the Bureau of Justice
Assistance
Improving Bureau of Justice Assistance’s Death in Custody Reporting Act Data Collection 1
As of February 2021, BJA has been assessing its Death in Custody Reporting Procedures and addressing
known issues. BJA has learned a lot during its first full year of Death in Custody Reporting Act (DCRA)
data collection. BJA is taking the following actions to improve its DCRA data collection program to
ensure it complies with the law, does not increase undue burden on reporting states, and returns
reasonable and actionable information on death occurring in jails, prisons, and during the process of
arrest.
Compliance–are state’s meeting the basic requirements of the Death in Custody Reporting Act as
reauthorized in 2014 (P.L. 113-242)?
 Developing state profiles. Based on data reported in FY 2020 and updated on a regular basis
(semiannual or annual) BJA is developing state profiles of each state describing its current
reporting status, existing capacity to act as a conduit for all entities (e.g. jails, prison, and law
enforcement), and any known issues.
 Training and Technical Assistance. BJA is working with the National Criminal Justice Association
(NCJA) to conduct outreach to State Administering Agencies about DCRA and to develop training
and technical assistance plans. Model states will be identified and used as examples for other
states to follow.
Data Accuracy–is the data that has been reported accurate and does not miss critical information?
 Data cleaning. BJA is cleaning the FY 2020 data and conducting outreach to states to address
issues that are found. The data cleaning is intended to identify outliers, unknown values, illogical
values, and duplicate records. The data cleaning process also identifies records that are left in an
“investigation pending” status that will need to be updated.
Data Validity–is the data complete?
 Collaboration with FBI’s Use-of-Force Program. BJA is working with the Federal Bureau of
Investigation’s Use-of-Force reporting program to identify reportable incidents that overlap the
purview of each program. Policies are under consideration to contact states when reportable
incidents are identified, but not reported in BJA’s DCRA program.
 Open source searches. In FY 2020, BJA piloted an open source records search on five states. This
resulted in the identification of unreported records, but more importantly, helped BJA learn
some of the barriers to reporting from a states perspective. These lessons have been considered
(see TTA above). The Bureau of Justice Statistics (BJS) learned many lessons from conducting
open source (e.g., web-searches) to identify reportable incidents that a state/locality did not
report. BJA is learning from the former BJS programs to take the lessons learned to
implemented a new open source records search. Again, policies are under consideration to
contact states when reportable incidents are identified, but not reported in BJA’s DCRA
program.

1

Last updated April 9, 2021
21

Appendix B: Bureau of Justice Assistance Deaths in Custody Reporting Act data collection form

22

23