B. Statistical Methods
1. Universe and Respondents Selection
The universe for the state prisons collection is all 50 state departments of correction, and BJS collects data on the complete universe. For local jails, the universe is all local jails nationwide. For the law enforcement collection, the universe is the roughly 18,000 law enforcement agencies throughout the country.
For the prison and jail collections, BJS uses the universe as respondents, rather than drawing samples from which to collect the data. There are several reasons for this. The original DICRA legislation required that all state prisons and local jails report data on deaths in custody, and this established the basis for collecting data from each prison and each jail in the country. Throughout the years of the collection, BJS has had universal participation by state prison departments and almost universal participation (99% in any given year) by local jails. A major goal of the program is to track changes in deaths over time within locations and to compare mortality between locations.
Sampling to obtain the data on state prison deaths would mitigate BJS’ ability to achieve its goals, would not appreciably reduce burden, and would greatly diminish the quality and utility of the data. For state prisons, the goal of the program is to produce state-level information on mortality in prisons, not simply to generate a national estimate of prison mortality. The only option for sampling would be to sample death records within state, but given the relative rarity of some important categories of deaths (e.g., suicide and homicide) sampling strategies would either have to be very complicated and potentially burdensome on the respondents’ end, and BJS would have to move from nearly complete enumeration of deaths to estimates.
The goals for the local jail portion of the DCRP are similar to the state-prison portion: track changes within locations over time and make comparisons between locations, as well as produce national data on mortality in jails. Sampling local jails would not appreciably reduce burden but would diminish the quality of the data. A sampling strategy that would allow for precise national estimates would include with certainty the largest jails in the country plus regional jails (those that serve multiple counties). These 300 jails provide the majority of deaths; so under an efficient sampling plan, their burden would not be reduced.
Sampling smaller jails would also not appreciably reduce burden or increase the accuracy of results. During a given year, approximately 85% of jails report zero deaths, and their reporting burden is quite small—the annual summary form. However, the same jails do not report zero deaths from year to year. Hence, sampling to capture the year-to-year variation would require either a fairly large panel of small jails (to capture the year-to-year variation in the number with no deaths) or different cross-sections from year-to-year. A large panel would not appreciably reduce burden but would reduce accuracy. A cross-section design could allow for greater accuracy in estimates of deaths in jails with small numbers of deaths, but would involve greater administrative burden and possibly increase respondent burden for sample respondents that fall in and out of the sample from year-to-year.
Further, while local jails are typically administered at a county level, in the field there is interest in state-level total counts of jail inmate deaths. As there are on average 60 jails per state, sampling to produce state-level estimates would require taking most of the jails in state, thereby practically eliminating any reduction in burden. Consequently, sampling jails to obtain mortality data does not provide cost or burden savings and would like reduce the accuracy of information. Finally, jails are now accustomed to the collection and have adapted their information systems to readily provide the data we seek.
To determine the universe of jails, BJS uses data from the 2006 Census of Jail Facilities, and updates the roster of jails during the collection to reflect closures and opening of new facilities. During the upcoming period, BJS plans—through its data collection agent RTI—to conduct further work to establish the universe of jails before it conducts its next census of jails. RTI will verify the existing universe of jails by comparing it to several sources of information, including the American Correctional Association’s (ACA) professional directory of jails and jail jurisdictions, and the American Jail Association’s directory. Further, RTI will compare the roster of jails in the DCRP to updated data that RTI obtains from its work on the National Inmate Survey under BJS Prison Rape Elimination Act surveys.
For the law enforcement portion of DCRP, BJS has identified state-level respondents who collect and report to BJS the data on law enforcement deaths. BJS opted for the route of 50 state reporters due to lack of resources to contact approximately 18,000 state and local law enforcement agencies yearly, especially when the vast majority have no deaths to report. To assess the quality of the data obtained from the state respondents, BJS has undertaken a comparative analysis of law enforcement homicide data in the DCRP with justifiable homicide data from the FBI’s Supplementary Homicide Reports. Both systems have individual-level mortality data that contain identifiers that allow for matching records at the level of the individual agency that provided the data. While all law enforcement homicides are not necessarily justifiable homicides, the two concepts are similar enough for purposes of determining whether the sources of data—that is the agencies reporting the deaths—are the same. Preliminary analysis shows that the two sources have comparable data on the characteristics of decedents. Future work will focus on understanding differences in reporting between the two sources.
During the upcoming collection period, BJS also intends to explore the prospects of developing an in-depth collection on law enforcement-related deaths in the largest police agencies (the largest 25 or 50) nationwide, which account for a large portion of deaths and arrests nationwide. This collection will explore issues related to police use of force (lethal and otherwise), critical incidents that could lead to use of force, and law enforcement homicides. Such a collection would enhance BJS’ understanding of mortality in large police departments and supplement the national-level DCRP collection of law enforcement deaths. BJS will create or use opportunities to meet with stakeholders to discuss this collection—such as the annual Justice Research and Statistics Association Conference, the meetings of the International Association of Chiefs of Police, and meetings with the Police Executive Research Forum. These efforts will focus on determining the feasibility of such a focused collection, that is, the level of participation, the costs and burden, as well as consequences of such a collection for the DCRP law enforcement collection. If through these efforts BJS concludes that such a collection is feasible, BJS will request clearance for it through a separate submission.
2. Procedures for Information Collection
In January of each year the respondent is notified by letter (see attachment E for an example) about how and when to report any deaths that may occur in the coming year. Respondents are asked to submit quarterly reports, in combination with a form for each death detailing the incident characteristics. (2009 collection will request a single report for all four quarters, given the current postponement of the collection activity). Some jurisdictions prefer to submit the incidents on an on-going basis as they occur, while others collect and submit the data for the entire year at one time.
At this time there is a web submission option for jails and prison systems (see section A.3 “Use of Technology”) hosted through a contractor. During the upcoming cycle, BJS’ data collection agent will implement edit checks into the web-based collection instrument that will alert respondents about data entry errors, incorrect totals based on subtotals, and item non-response. There is no current web option for law enforcement agencies, as this portion is conducted in-house and the server is not able to accommodate the necessary specifications and protections to ensure the confidentiality of the information submitted.
Data are due within 30 to 60 days following the end of the quarter; in practice, however, follow-up contacts don’t begin until the Spring following the collection year since some jurisdictions submit all four quarters at once. Reminder phone calls/emails are placed/sent to non-respondents. Respondents submitting partial or incomplete data are contacted on an on-going basis as data come in. BJS staff and collection agents generally have existing relationships with respondents, so scripts for calls and emails are not necessary.
For the upcoming collection years, the contractor will implement systems to track respondents based on time to complete, number of contacts, and other information that can be used to develop strategies to reduce time and error in data collection. For example, in conducting follow-up calls, staffing strategies that have proven effective in other large surveys will be employed, such as organizing respondents into regions and assigning a project team member to each region to serve as an agency liaison.
3. Methods to Maximize Response
BJS has achieved consistently high rates of response to the state prison and local jail inmate death collections. In each year since collection began, BJS has collected death records from all 50 state departments of correction, and at least 99% of local jail jurisdictions. While the linkage of these reports to VOI/TIS grant funds certainly aided these response rates from state prison authorities during 2001-2006 (see Part A, Section 1 “Necessity of Information” for a discussion of the Death in Custody Reporting Act of 2000 and the relationship of this data collection to federal funds), it should be noted that the vast majority of local jail jurisdictions had no obligations under these grant programs. In any event, response rates have been maintained at these levels since the expiration of the grant programs in 2006.
BJS believes that our efforts to minimize the reporting burden have been integral to achieving high rates of voluntary response. In an effort to minimize respondent burden, the data collection plan allows for the local jail respondents to submit data by mailing their reply in a postage-paid envelope or by fax, in addition to the internet-based reporting system described in section A.3 “Use of Information Technology”. Furthermore, given the greater volume of prison inmate deaths, BJS has developed a computer format in which State prison administrators can submit their prison inmate death records electronically. During 2008, 16% of State prison death records were submitted in electronic files using this method, while another 75% were submitted using the Web-based reporting instrument.
The data collection forms submitted for this collection involve data that are available from the current record-keeping practices of correctional systems and law enforcement jurisdictions. Data are available from incident logs and reports of coroners or medical examiners. To minimize burden the data collection forms contain only those items that may be known to the correctional and law enforcement administrators. In addition, the burden has been minimized by use of a close-ended, pre-coded format.
For the law enforcement part of the collection, BJS identified data providers in each state and the District of Columbia; however, three states’ providers—Georgia, Maryland, and Montana—never submitted records. Forty-seven states and the District of Columbia participated in the collection at some point, but the participation of some states varied by year. A total of 43 states participated in 2003, 45 states participated in both 2004 and 2005, and 43 states participated in 2006. The District of Columbia submitted records in all four years. In over 30 states, the reporting office also served as a state criminal justice Statistical Analysis Center (SAC).
From ongoing follow-up contacts with providers, BJS has learned that a main reason for non-response is the absence of a centralized recording system for law enforcement deaths in states, with results in imposing the burden of assembling statewide data solely for the DCRP. To maximize response rates with the state providers, BJS has undertaken the following activities:
(1) Included DCRP law enforcement death data reporting as an eligible activity towards which the SACs can use Criminal Justice System grants. (Annually, BJS makes grants to the SACs, which are required to describe the purposes for which they will use funding. DCRP data collection is an allowable category.)
(2) BJS has attended national conferences of Association of State Uniform Crime Reporting Programs to discuss the DCRP and encourage sharing of information between the UCR program and the DCRP state providers.
(3) Annually, BJS receives from the FBI a copy of the Supplementary Homicide Reports data. The DCRP law enforcement collection should include reports for the justifiable homicides that are also included in the SHR. BJS has analyzed the SHR reports and compared them with the DCRP law enforcement homicides data. During the upcoming collection cycle, BJS will identify SHR cases that may not be reported by DCRP respondents and forward this information with DCRP respondents to assist them in reconciling their reports with the SHR reports. (Note that as reported above in Section A. 4. “Efforts to Identify Duplication,” while there is some overlap in the universe of cases for the SHR and DCRP law enforcement homicides, the DCRP obtains additional information about these types of deaths, as well as data on other deaths occurring in the process of arrest that are beyond the scope of the SHR.)
More generally, DCRP respondents have indicated that they appreciate being informed about how their reports of law enforcement homicides compare with the SHR reports. They have reported that, at a minimum, even if they are unable to gather data on all reports included in SHR, they are better prepared to comment on discrepancies between their reports and SHR reports should they receive media requests.
By using these methods and working closely with individual jurisdictions, BJS hopes to increase law enforcement participation in the collection to 100% in subsequent years.
4. Test of Procedures or Methods
BJS has conducted the DCRP for several years with high rates of response and has specific plans to further improve participation; current procedures and proposed initiatives are described in Sections B.2 and B.3 above.
5. Consultation Information
The Corrections Statistics Unit of BJS takes responsibility for the overall design and management of the activities described in this submission, including data collection procedures, development of the questionnaires, and analysis of the data. BJS contacts include:
Christopher Mumola, Program Manager
Deaths in Custody Reporting Program
Bureau of Justice Statistics
810 Seventh Street, NW
Washington, DC 20531
(202) 353-2132
Margaret Noonan, Statistician
Deaths in Custody Reporting Program
Bureau of Justice Statistics
810 Seventh Street, NW
Washington, DC 20531
(202) 353-2060
William Sabol, Ph.D., Chief
Corrections Statistics Unit
Bureau of Justice Statistics
810 Seventh Street, NW
Washington, DC 20531
(202) 514-1062
The Project Director at RTI International for the state prison and local jail collections is:
Christopher Ellis, Survey Director
RTI International
3040 Cornwallis Road
Research Triangle Park, NC 27709
(919) 541-6480
C. Attachments
Data Collection Instruments (NPS-4, NPS-4A, CJ-9, CJ-9A, CJ-10, CJ-10A, CJ-11,
and CJ-11A)
BJS authorizing legislation, confidentiality regulations (Title 42, USC, Sections 3732,
3735 and 3789g)
C. The Death in Custody Reporting Act of 2000 (PL 106-297)
D. Public comments received during the 60-day period and BJS responses
E. Sample cover letter announcing data collection
F. BJS online data tables of DCRP data
G. BJS publication of DCRP data: Medical Causes of Death in State Prison, 2001-2004
| File Type | application/vnd.openxmlformats-officedocument.wordprocessingml.document |
| Author | pricel |
| File Modified | 0000-00-00 |
| File Created | 2021-02-03 |