6

Sources of Health Outcome Information

This chapter addresses the fifth subtask in the committee’s charge, to characterize the quality and completeness of the information in the records related to health outcomes, including cancer occurrence and cause of death, of the exposed veterans. It begins with an overview of the well-documented health outcomes associated with exposure to ionizing radiation and other major chemicals and metals that military veterans may have encountered during 1942–1947 while at one of the Manhattan Project sites listed in the statement of task. As discussed in Chapter 5, military personnel at each site had different types of jobs and duties, which would lead to different types, frequencies, and durations of exposures to radioactive, chemical, or combined forms of waste. Therefore, they had different risks for developing adverse health outcomes. See Chapter 3, Figure 3-2, for an overview of the processes and sites that may have resulted in exposures to ionizing radiation, chemicals, and combined exposures.

After overviews of recognized health outcomes from Manhattan Project–related exposures, this chapter describes the sources the committee identified that might provide health outcome records for this specific population and methodologic considerations for their use. Using these sources and methods assumes that a roster of veterans who served at the sites of interest during 1942–1947 can be created. The sources are categorized as those that include the general U.S. population (not limited to military personnel) and those with military-specific information. The committee determined that the military sources are fewer and more difficult to access, while the broader non-military sources have been used successfully to obtain health outcomes information for numerous published epidemiologic investigations.

This chapter describes the types of data available from these sources, how these data are collected, processes for accessing them, and limitations of each for use in an epidemiologic study.

OVERVIEW OF HEALTH OUTCOMES OF RADIOLOGICAL AND CHEMICAL EXPOSURES

Chapter 5 provides an overview of some examples of Manhattan Project exposures to both radiological and chemical sources, with a focus on internal and external exposures to ionizing radiation and exposures to a variety of chemicals. The committee did not identify any epidemiologic studies that specifically examined the health outcomes of these exposures among military personnel who were part of the Manhattan Project overall or at the sites of interest. Therefore, it considered reviews of epidemiologic studies conducted in populations with known exposure to the same hazards to better understand what health outcomes might be relevant to a future study. These studies included those on Japanese atomic bomb survivors, Chernobyl cleanup workers, nuclear industry and health care workers, and other populations with occupational exposures; reviews by the World Health Organization’s International Agency for Research on Cancer (IARC) and the United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR); and toxicology profiles from the Agency for Toxic Substances and Disease Registry (ATSDR).

The effects of internal radiation exposure depend in part on the chemical form and route of intake. The health outcomes are largely associated with the tissues and organs where the incorporated radionuclide accumulates and energy deposition occurs. Similarly, for external radiation exposure, effects will depend on the portion of the body irradiated. A number of systematic reviews have discussed the carcinogenic effects of exposure to ionizing radiation (e.g., UNSCEAR, 2006, 2017, 2019). IARC examined cancers associated with ionizing radiation and concluded there was sufficient evidence for a causal association between X-ray and gamma radiation and cancers of the salivary gland, esophagus, stomach, colon, lung, bone, skin (basal cell and melanoma), breast, bladder, kidney, brain and central nervous system, thyroid, and leukemia (El Ghissassi et al., 2009; Richardson et al., 2018). Based on the International Nuclear Workers Study, IARC also found evidence of positive associations with exposure to ionizing radiation for cancers of the rectum, pancreas, ovary, peritoneum, larynx, pleura, and testis (IARC, 2025). Leuraud et al. (2024), in an update to the International Nuclear Workers Study, identified a positive association between long-term exposure to low dose ionizing radiation and mortality due to leukemia (excluding chronic lymphocytic leukemia), both chronic and acute myeloid leukemia, myelodysplastic syndrome, and multiple myeloma. Additional

evidence for radiation dose–cancer associations comes from a variety of studies, including long-term follow-up of the Japanese survivors of the atomic bombings of Hiroshima and Nagasaki, Chernobyl cleanup workers, patients and health care professionals exposed to ionizing radiation used for diagnostic and therapeutic purposes, and nuclear industry workers (Kamiya et al., 2015; Little, 2009; Wakeford, 2009).

Cancer may not be the only adverse health outcome of concern for military veterans exposed to ionizing radiation while working on the Manhattan Project. A growing body of evidence suggests that exposure to ionizing radiation may also be associated with an increased risk for certain nonmalignant diseases, including cardiovascular disease, noncancer respiratory diseases, and digestive diseases (Little et al., 2021, 2024). Evidence of such associations has been found among Japanese atomic bomb survivors and in some studies of nuclear industry workers and Chernobyl cleanup workers (Hatch and Cardis, 2017; Little, 2009; Wakeford, 2009). Such outcomes might be available in mortality records. Recognized noncancer effects are also associated with exposure to acute high doses of ionizing radiation, such as skin erythema and nausea (ATSDR, 1999; Stewart et al., 2012; UNSCEAR, 1988). However, given that the veteran population was thought to have had low, chronic exposures rather than high, acute exposures, the committee did not further consider effects known to arise from high doses.

As noted in Chapter 5, ionizing radiation was not the only hazard present at these sites. Military personnel were also potentially exposed to a variety of chemicals and heavy metals. Some radiological hazards, such as uranium, also posed a risk based on their nonradiological toxic effects. Some of these agents are carcinogenic, including many of the volatile organic compounds (such as benzene and trichloroethylene), polychlorinated biphenyls, formaldehyde, hexavalent chromium, and beryllium. Many of these exposures are also associated with noncancer health outcomes. For example, exposure to uranium is most often associated with damage to the kidneys, but has been associated with potential damage to the brain, heart, liver, and other organ systems, and exposure to beryllium, asbestos, phosgene, nickel, and vanadium may cause respiratory disease (ATSDR, 2023; Darnton, 2023; Rathod et al., 2023; Zhang et al., 2009). ATSDR has developed toxicology profiles that describe the adverse health effects associated with exposure to many of the chemical and radiological hazards veterans may have been exposed to at the Manhattan Project sites of interest (ATSDR, 1999, 2004a,b, 2010, 2013, 2023, 2024; Wilbur et al., 2008). These profiles can be used to develop a list of possible health outcomes for use when reviewing cause of death and other health outcome information for the veteran population. The adverse health outcomes in the ATSDR profiles are similar for many radiological and chemical hazards. As described in Chapter 3, the sites with the largest military

presence during 1942–1947 were responsible for multiple steps in the nuclear weapons manufacturing process, and each step represented an opportunity for exposure to numerous radiological and chemical hazards. For example, some Manhattan Project sites may have had both beryllium and plutonium present. As a result, workers and veterans may have had overlapping exposures to such hazards that may have similar health outcomes.

Discerning all possible health outcomes associated with exposures to the various radiological and chemical hazards veterans likely encountered at the Manhattan Project sites is complicated by the nature of the health effects of these hazards. Chemicals that independently produce outcomes resembling those of ionizing radiation or have similar adverse outcome pathways may amplify radiation effects. Conversely, some chemicals may shield healthy tissue from damaging properties induced by radiation (Cléro et al., 2021; Streffer and Müller, 1984). Therefore, the effects of each hazard must be considered alone and in combination. Some metals, such as cadmium are associated with health effects that are similar to those from exposure to ionizing radiation, such as kidney cancer and cardiovascular diseases (Xu et al., 2025). For other exposures, such as beryllium, health effects similar to radiation exposure, such as lung fibrosis, depend on dose and exposure levels. Generalization from one element to another, one compound to another, and even one isotype to another is improbable and results in substantial challenges when attempting to define risks and biologic consequences of these complex exposures.

POTENTIAL SOURCES AND METHODS TO IDENTIFY HEALTH OUTCOMES

In an ideal epidemiologic investigation, researchers would prospectively follow participants over time and ascertain the incidence of disease, capturing information on both new diagnoses and fatal outcomes. However, given the historic nature of this feasibility assessment (active-duty military personnel of the 1940s) and the fragmented nature of historical records associated with the Manhattan Project, retrospective assessment of health outcomes is the only option. In the United States, identifying health outcomes that occurred in the past—especially 80 or more years ago—is challenging, in part because people have differing access to medical care, medical records are not centralized, and relatively few disease diagnoses are required to be reported to local, state, or national public health agencies. Some epidemiologic studies attempt to contact study participants to ascertain health outcomes by self-report. However, such an approach is impractical here, given that less than 1% (approximately 58,000) of World War II veterans were living in 2024, and it is unknown if any of them participated in the Manhattan Project (VA, 2025). Without the ability to contact participants,

researchers must often use passive sources of information, such as administrative records, that were not collected for research purposes.

This section describes the most relevant potential sources of information for retrospective ascertainment of health outcomes posed in the statement of task, beginning with general (not military specific) sources, followed by sources that are specific to military personnel. The committee considered a wide range of potential sources, including the National Death Index (NDI), state-based death certificates, the Social Security Administration (SSA), cancer registries, the Department of Energy (DOE) Energy Employees Occupational Illness Compensation Program Act (EEOICPA), Centers for Medicare & Medicaid Services (CMS), the Oak Ridge Associated Universities (ORAU) Comprehensive Epidemiologic Data Resource (CEDR), military personnel records, and VA medical and benefits records. For each source, a description is provided of years of data captured, population, types of information included, notable strengths and limitations, process to access, and cost, if available. The sources described in this chapter are not exhaustive but are those that the committee determined are the most likely to have health outcomes data for this specific veteran population.

Sources Not Limited to Military Personnel

Death records are standardized and typically more reliable sources of cause of death information than many other sources. This section discusses sources of cause of death information that encompass the entire U.S. population. While not specific to military veterans, these sources are most likely to contain cause of death data for them. However, no one single source of cause of death information exists for veterans or civilians in the United States.

As less than 1% of World War II veterans were alive in 2024, cause of death is a key outcome for those at the specified Manhattan Project sites between January 12, 1942, and August 15, 1947. Most major sources of mortality data require identifying information beyond a name to conduct a search for cause of death. Military service records typically contain this information, including date of birth and identification numbers necessary for researchers to obtain accurate search results from mortality data sources. However, such information would only be available if military records were first obtained to create a comprehensive roster of individuals (see Chapter 4 for further discussion of a process to develop such a roster). A best practice is to use multiple sources to verify individuals and their vital status. After overviews of sources of death data, general sources of health outcome data, including cancer registries, are briefly described.

National Death Index

NDI was established by the Centers for Disease Control and Prevention (CDC) National Center for Health Statistics (NCHS) as a central resource for epidemiologists and other health and medical researchers to determine whether participants in their studies have died (CDC, 2024a). This database is considered the most complete source of death information in the United States, containing data from all death certificates submitted by the 50 states, the District of Columbia, Puerto Rico, and the U.S. Virgin Islands since 1979 (CDC, 2024b). Applications to access NDI data are considered by an independent review committee (NCHS, 2024). The duration of the approval process varies but is typically at least 2–3 months.

NDI searches have a base fee and a fee per subject and search year, which varies depending on the type of search and level of detail requested. Researchers must submit a minimum amount of information for each subject that includes at least one of the following combinations:

• First name, last name, and Social Security number;
• First name, last name, and birth month and year; or
• Social Security number, date of birth (day, month, and year), and sex (CDC, 2024b).

NDI has additional requirements for search requests, such as encryption and submission formatting, that are described in greater detail in its user’s guide. NDI will return date of death, state of death, and death certificate number if a match is found in the requested file years. Researchers seeking cause of death must request an NDI Plus search, which links NDI data with the corresponding annual NCHS multiple cause of death dataset, a public use dataset based on death records processed by NCHS that has been released annually since 1968 (NCHS, 2024). Cause of death and additional contributing causes of death are noted using ICD-9 or ICD-10 codes (World Health Organization’s International Classification of Disease 9th or 10th Revision), depending on which system was in use when the death occurred (NCHS, 2024).

Researchers using NDI to obtain death data for a veteran population would need to consider several potentially limiting factors. If, as described in Chapter 4, a near complete roster of individuals with the needed identifiers could be assembled, NDI will only be applicable to those who died in or after 1979, which is 32 years after the end of the Manhattan Project. Members of the military were mostly young adults during 1942–1947, with more than 90% aged 37 years or younger in 1945 (Smith, 1947). Life expectancy for a 37-year-old at that time was 55.5 for men and 63.9 for women. For those who were 18 years old (minimum age of enlistment

without parental consent) in 1942, life expectancy was 63.8 for men and 71.0 for women (SSA, n.d.a). Therefore, a large proportion of deaths may predate the NDI data. Additionally, the military identification number for individuals who served prior to 1968 is different from their Social Security number, which was not usually included in their military records (Boice et al., 2020). NDI data does not generally include military status, nor do state-level death certificates uniformly report veteran status.¹

Another limitation of NDI data is that multiple cause of death data are incomplete before 1983, which may affect the completeness of mortality outcomes for some veterans. NDI search results yield exact and possible matches; in the latter, the requested data satisfy at least one of seven NDI matching criteria. Researchers will have to consider the time and staff resources needed to evaluate a possible match for validity, using their own criteria or NDI’s suggested assessment criteria, to determine whether the possible match represents the person of interest.

Social Security Administration Death Data

SSA uses death reports received from family members, funeral homes, financial institutions, postal authorities, states, and other federal agencies for its death information files (SSA, n.d.b). A complete SSA death information file includes an individual’s Social Security number (if available); first, middle, and last name; birth and death dates; and state death certificate (SSA, n.d.b). While federal law restricts access to the full file of death information to certain federal and state agencies, researchers can submit a request for information about fact of death, that is, whether study participants are alive or deceased (SSA, n.d.b, n.d.c). Researchers’ requests must include each subject’s Social Security number; first, middle, and last name; full date of birth; and sex (SSA, n.d.b). SSA will provide researchers with one of the following:

• Date of death and state where a claim was filed or the state of residence when the death occurred (if available),
• Presumption that an individual is alive if SSA records have sufficient information to support this determination, or
• A finding of unknown status if SSA does not have a record of death or enough information to determine the individual is alive.

Search fees are not dependent on the number of individuals in a request. Instead, requesters must reimburse SSA for all costs associated with finding

___________________

¹ Personal communication, Lillian Ingster, director, National Death Index, National Center for Health Statistics, February 28, 2025.

and providing vital status information. SSA will not provide copies of death certificates or information from those documents beyond date of death and state where death occurred. Additionally, SSA notes on its website that “it is important to note our records are not a comprehensive record of all deaths in the country” (SSA, n.d.b). However, other than noting that the file only includes individuals who were issued Social Security numbers, the site does not elaborate on what deaths are not included (SSA, n.d.b).

SSA also compiles a file of information that excludes state death records, known as the Death Master File, for the Department of Commerce’s National Technical Information Service (SSA, n.d.b), which sells it to other agencies and private organizations, such as financial institutions and credit companies, but it is generally not provided to researchers (SSA, n.d.b; NTIS, n.d.). The Death Master File does not contain cause of death information, which makes it less useful than SSA’s other service for researchers to obtain death data.

The information requirements for SSA epidemiologic search requests could limit usefulness for matching individuals’ military records with death data. As noted, military records for those who served 1942–1947 usually did not include Social Security numbers, which could pose challenges for searching SSA records (Boice et al., 2020). Between 1978 and 1980, deaths were not routinely recorded in SSA files, which could further affect the use of SSA data for this veteran population (Cowper et al., 2002). However, if individual official military personnel files were obtained from the National Archives at St. Louis in creating a roster, they contain service numbers that can be linked to Social Security numbers which can then be used to request SSA data. Fact, date, and state of death obtained from SSA are useful for conducting an NDI Plus search (which returns cause of death for known deaths) and verifying vital status from disparate sources, such as VA records.

State-Level Death Certificates

In the United States, vital statistics, including death, have been recorded locally since before the Revolutionary War and can help fill the gap in death information that predates inclusion in NDI (Hetzel, 1997). Death certificates are registered locally, usually at the state or territory level (or the city level for New York City and Washington, DC), by one of the 57 reporting jurisdictions–maintained registries and report vital statistic data to the National Vital Statistics System (NCHS, n.d.b; NRC, 2009). Each state or jurisdiction has its own laws governing release of death certificates and processes for requesting those records. Researchers must contact each jurisdiction’s vital records office directly to request death certificates. Death certificate fees vary and are $5–$30 dollars each as of 2024 (NCHS, n.d.b).

Assuming that a veteran roster can be compiled, researchers could obtain fact of death and state of death from SSA and then seek death certificates state by state for deaths before 1979. A certified nosologist could then code cause of death from the death certificates. This process is labor intensive but has been successfully used for numerous retrospective epidemiologic investigations.

The committee learned that ORAU has collected over 88,000 pre1979 death certificates as part of health and mortality studies of DOE workers (Golden, 2024). The ORAU database includes digitized date of death and underlying cause of death data, including for individuals who worked at Hanford, Oak Ridge (K-25, X-10, Y-12, and Tennessee Eastman Company), Los Alamos, and Mallinckrodt Chemical Works during 1942–1947 but generally not military status. Assuming it is possible to create a roster, Ashley Golden from ORAU told the committee that researchers could use it to match ORAU death certificate information. While the ORAU death certificate database is not a comprehensive source, this would reduce the time- and resource-intensive demands associated with requesting individual death certificates from multiple states and having them coded for cause of death, and it would capture cause of death information that predates NDI.

Limitations of Death Certificates

All sources that provide specific cause of death have at least some limitations related to coding. Most sources obtain cause of death information from the ICD codes that indicate underlying and contributing cause of death on the death certificate. ICD codes and guidance for their use for cause of death have been revised several times with each revision increasing specificity (Foreman et al., 2016). State laws also vary in whether a treating physician, medical examiner, or coroner is responsible for entering cause of death information on death certificates. Coroners are elected at the county level and not required to have a formal medical education, which may affect ICD code specificity (Skopp et al., 2017). Coding for multiple causes of death has improved the sensitivity of death certificate–based ascertainment of diseases that may not be fatal but contributed to death (such as some pulmonary diseases). Since 1967, the NCHS Mortality Medical Data System has sought to improve accuracy of ICD codes entered on death certificates through the use of software programs to automate cause of death data classification, entry, and retrieval (NCHS, n.d.a).

As individuals age, the most common causes of death change, which may confound the process of connecting this directly to exposures at a Manhattan Project site. For example, a younger person who died in a car crash might have developed cancer had they lived longer.

Cancer Registries

U.S. cancer registries collect and maintain detailed surveillance data, including patient demographics, date of diagnosis, primary tumor site, stage at diagnosis, and treatment information, and follow up for vital status (White et al., 2017). The first hospital cancer registry was established in 1926, the first state cancer registry (Connecticut) in 1935, and the first national cancer registry was established in 1971 (NCI, 2023). However, several of the states that correspond with the Manhattan Project sites in the statement of task (e.g., Tennessee, Washington, Missouri) did not have state cancer registries during 1942–1947 or the years immediately after. U.S. cancer registries only became widespread, and federal-level registries established, after the National Cancer Act of 1971, which mandated data collection and reporting (White et al., 2017).

Virtual Pooled Registry-Cancer Linkage System and State Cancer Registries

VPR-CLS is a collaborative effort of the North American Association of Central Cancer Registries, the National Cancer Institute, at least 48 U.S. central cancer registries, and researchers to provide a more centralized approach for research studies to link to cancer registry data (NCI, n.d.b). The oldest registries available via VPR-CLS are Connecticut (from 1935), Utah (from 1966), and New Mexico (from 1969) (NAACCR, 2025a). However, most registries date back only to the late 1980s or 1990s, indicating a lack of nationwide coverage for several decades after the period of interest. Before 1992, when the National Program of Cancer Registries was established, 10 states did not have a state-level cancer registry (CDC, 2024a).

To be considered for linkage with VPR-CLS, a study must be able to provide identifying information for at least 75% of its population. This includes a Social Security number, date of birth, and either first or last name; or first name, last name, date of birth, gender, and at least one of the last four digits of the Social Security number, phone number, or full address (NAACCR, 2025b). Thus, VPR-CLS would only be useful if researchers have these identifying elements. Also, while it has generally broad coverage for cancer cases in the United States, it does not link to every U.S. cancer registry, and data years available vary by registry. This increases the likelihood of missed cases.

The National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results (SEER) program collects data from population-based cancer registries that cover almost 50% of the U.S. population, with some registries dating back as far as 1973 (NCI, n.d.a). SEER includes registries in New Mexico and Utah and the Seattle-Puget Sound registry, which may capture cancer incidence data from some—but not all—veterans who served

at Manhattan Project sites during the 1942–1947 era. However, cancer data obtained from SEER will have the same limitations as that available from VPR-CLS in terms of completeness. The core SEER registries are required to participate in VPR-CLS activities, which streamlines the work-flow for researchers seeking to obtain cancer incidence data (Penberthy and Friedman, 2024).

Other Sources

CEDR has analytic datasets from epidemiologic studies of employees of DOE or its contractors at several Manhattan Project sites of interest that include demographics, exposure information, follow-up dates, and health outcomes. However, the data are deidentified (Golden, 2024). These datasets include measures of both radiologic and chemical exposure for workers collected by researchers. While this information is not specific to veterans, they often worked alongside civilian workers (Golden, 2024; Seidel, 1993). Researchers would have difficulty matching any health outcomes information to a roster without a link to the original data. Therefore, as stated in Chapter 5, CEDR datasets would likely be more useful for determining types of exposures than health outcomes for a given site (see Appendix D for a list of studies in CEDR).

Depending on an individual’s post–Manhattan Project career path, some health outcome information may be available through DOE sources for those who returned to civilian life and worked at DOE facilities, such as worker screening programs or compensation programs, such as EEOICPA. EEOICPA does not specifically cover service members on active duty. It would only cover them if they worked as a DOE employee and met the criteria outside of their military service. Further investigation determined that EEOICPA and other worker screening programs are unlikely to contain information on individual Manhattan Project veterans because these programs began in the early 1990s and are limited to claimants and certain health outcomes only (22 radiogenic cancers, chronic beryllium disease, beryllium sensitivity, or chronic silicosis) determined to be related to DOE work at covered sites (Lewis, 2024; NIOSH, 2024). Only a subset of Manhattan Project veterans had careers in DOE, and not all of them would have filed an EEOICPA claim. The records contain no mechanism or indicator of military or veteran status. Therefore, even if EEOICPA records could be made available for research (they are not), they could not be used to create a roster or provide health outcomes information.

The Department of Labor (DOL) is the primary administrator of EEOICPA, including claims intake, adjudication, and compensation and medical bill payments. Under the umbrella of the EEOIPCA program are two helpful resources: special exposure cohorts (SECs; evaluated by

NIOSH) and Site Exposure Matrices (SEMs; maintained by DOL). An SEC is defined for a specific site, the type(s) of workers at that site, and the period when work was performed. It is used to help determine compensation for the 22 specific cancers when radiation dose reconstruction is not possible; it has no additional health information (NIOSH, 2024). SEM is a database of toxic substances and potential health outcomes at DOE facilities (see Chapter 5). However, it is not time period specific, nor does it contain data on individual workers, so it cannot be used for an epidemiologic study (DOL, n.d.; IOM, 2013). A 2013 Institute of Medicine review of SEM found several issues with how health outcomes were identified. As of May 2024, SECs exist for Oak Ridge for the Clinton Engineer Works, gaseous diffusion plant (K-25), National Laboratory/X-10, S-50 thermal diffusion plant, and Y-12; Hanford; Los Alamos National Laboratory; and Dayton Project.²

CMS data could be useful for identifying incident disease and other conditions not well captured by mortality records (such as mental health conditions). However, CMS datasets available to researchers begin in the 1990s (Research Data Assistance Center, 2025). Therefore, CMS data likely will not capture the vast majority of adverse health outcomes in the population of interest.

The committee heard presentations from DOE records management offices at Oak Ridge, Hanford, and Los Alamos to identify any examples of deidentified medical records they may have from 1942 to 1947. In general, DOE sites appear to have limited records regarding relevant health outcome information, and these are most often accident reports. Jennifer Hamilton stated that the Oak Ridge Consolidated Service Center Records Management Library has records for DOE and contractor employees that include medical records from that period (Hamilton, 2024). However, these records have not been digitized or examined but could possibly contain an indicator of military or veteran status (Hamilton, 2024). Records managers from Hanford told the committee that they did not identify any medical records for members of the military who served there. However, they have a collection of documents from a study that was never completed that may include such medical records (see Chapter 4, “Lists of Veterans Compiled by Researchers”) (Milligan and Zaback, 2024).

Los Alamos Medical Center (LAMC) operated under a contract with the Atomic Energy Commission from 1943 to 1964. LAMC records from that era included medical files for service members posted at Los Alamos

___________________

² Special exposure cohort (SEC) documents exist for statement of task sites that were not found to have a military presence: Ames Laboratory, Metallurgical Laboratory (University of Chicago), Lake Ontario Ordnance Works, and St. Louis Airport Project Site. The Dayton Project SEC is called Monsanto Chemical Company.

before 1964.³ LAMC initially was responsible for retaining medical records for Los Alamos workers 1943–1964. In the mid-2000s, LAMC advised DOE that it could no longer do so. When DOE took possession of the files, they had been damaged by water leaks and a mouse infestation.⁴ While DOE has sorted, inventoried, and indexed the files, the committee was not able to obtain information from DOE Office of Legacy Management regarding their contents or whether they include records for military personnel who served at Los Alamos 1942–1947. The Los Alamos National Laboratory National Security Research Center does not have copies of them.⁵ If researchers were able to access these files, they might contain information relevant to identifying adverse health outcomes for the population of interest.

Military-Specific Sources of Records and Data

Individual military medical records for care received while in service and VA records, if the individual was eligible for and used VA care, are potential sources of information for adverse health outcomes, including cause of death. National Academies staff contacted the record offices at the Richland Hospital (at the Hanford site) now Kadlec Hospital, and Methodist Medical Center in Oak Ridge, which was the former Oak Ridge Army Hospital. Both offices replied that their hospitals do not keep records dating back to the 1942–1947 era and that they were uncertain whether these records were transported elsewhere or destroyed.⁶ DOE Office of Legacy Management holds the Los Alamos hospital records.

Military Records

Military records include identifying information that is necessary for conducting a search of non-military sources, such as NDI or SSA, assuming it is possible to create at least a partial roster. It is unknown how many official military personnel files would be available for the veterans in this specific population, given the millions of records destroyed by the 1973 fire at the National Personnel Records Center and National Archives at St. Louis. It is also unknown how many 1944–1946 unit records and morning reports, which could be used to identify individual veterans for a roster, were destroyed before protections for such records were put into place. Researchers seeking information from military records will need to

___________________

³ Personal Communication, Brye Anne Steeves, Director, National Security Research Center, Los Alamos National Laboratory, September 13, 2024.

⁴ Ibid.

⁵ Ibid.

⁶ Personal communication, Methodist Medical Center staff, April 22, 2024; personal communication, Kadlec Regional Medical Center staff, June 7, 2024.

use other sources, such as Department of Defense (DoD) and VA, as the National Archives and Records Administration (NARA) stated there are no duplicate or microfilm copies of these records.

Active duty and retired military personnel have been able to obtain health care at military and civilian hospitals through what was originally the DoD military health benefit since 1956 (Dolfini-Reed and Jebo, 2000). It evolved into the Civilian Health and Medical Program of the Uniformed Services, known broadly as CHAMPUS in 1966 and became TRICARE in 1996 (Dolfini-Reed and Jebo, 2000). When an individual seeks care in a civilian setting, the provider must submit a claim for reimbursement. Such claims data include information that can be useful for identifying adverse health outcomes, such as diagnosis and treatment codes. However, it is unlikely that this source would include information for the majority of veterans who served at the Manhattan Project sites. Only individuals who retired from the military in 1956 or later and enrolled in the program or were still in the military and sought care in a civilian facility would have had medical claims. Furthermore, it is unclear whether disaggregated claims data that include identifying information for individual claimants are available for the early years of the program.

Department of Veterans Affairs Sources

VA has several clinical and administrative databases that could serve as sources of information about adverse health outcomes among Manhattan Project veterans, given its role in providing health care and other benefits, such as burial services, to veterans. Researchers have used VA databases to conduct veteran mortality studies, including the Beneficiary Identification Records Locator Subsystem (BIRLS) Death File, VA Medical SAS Inpatient Datasets, VA/CMS Medicare Vital Status File, Veterans Health Administration (VHA) Corporate Data Warehouse (CDW), VA/DoD Mortality Data Repository, and VA Death Ascertainment File (Arnold, 2023; Maynard, 2022; Sohn et al., 2006).

Two of these sources draw data from only VA files: the VA Medical SAS Inpatient Datasets, which contains information from patient treatment records for care delivered at VA facilities or under the auspices of VA at non-VA facilities; and CDW, VA’s health data warehouse (Maynard, 2022). CDW may include death certificate information if that is the official source used to verify death (other accepted sources are the National Cemetery Administration and patient information from VHA facilities) (Maynard, 2022). The VA Medical SAS Datasets contain limited death information for the veterans of interest, as they do not include cause of death and date of death is only available if the death occurred in or shortly after discharge from inpatient care in a VA facility or a non-VA facility providing care

under the auspices of VA (Sohn et al., 2006). CDW only began collecting data in 2006, and the VA Medical SAS Datasets began in 1970 (Sohn et al., 2006; VA, 2020).

The other VA databases draw data from multiple sources. The BIRLS Death File combines information from VA facilities (including VHA and VA National Cemetery Administration), family members who report a death to VA, and SSA (Maynard, 2020, 2022). BIRLS and its Death File includes information on the location of an individual’s hard copy VA file, which is important given that VHA facilities were not consistently using standardized electronic health records before 2000. The VA/CMS Medicare Vital Status file draws death information from Medicare claims data, the SSA Death Master File, and notifications from family members (Maynard, 2022). All of these databases include demographic information and date of death, which could be added to a roster to search NDI or request individual state death certificates. Researchers have noted that using NDI in conjunction with VA sources such as BIRLS, the VA Medical SAS Inpatient Datasets, and the VA/CMS Vital Status file is useful for optimizing inclusion and accuracy of death data for the veteran population (Fisher et al., 1995; Sohn et al., 2006). An investigation of adverse health outcomes among the veteran population of interest would probably need to apply that approach as well.

These sources have limitations that should be considered when designing an epidemiologic study. Death information in BIRLS is incomplete, particularly before the mid-1970s, and cause of death is generally limited to a basic classification (i.e., natural death, combat death, unknown, or other) (Cowper et al., 2002; Sohn et al., 2006). Some records for deaths before 1972 may be found in the VA Master Index, the predecessor to BIRLS (IOM, 2000). It consists of index cards (transferred to microfilm) for each veteran who applied for benefits between 1917 and 1972 (IOM, 2000). The VA/CMS Medicare Vital Status File does not include cause of death and only contains information for veterans who both registered with VA and enrolled in Medicare (Sohn et al., 2006). Additionally, the VA/CMS Medicare Vital Status File was retired in 2023 (Arnold, 2023).

VA has conducted several projects to combine its databases with SSA and NDI data to improve the accuracy of its death data. This led to development of the VA Death Ascertainment File, which combines CDW, VA administrative data, and the SSA Death Master File data (Arnold, 2022). More recently, VA, DoD, and NCHS have collaborated to develop the Mortality Data Repository as part of VA and DoD suicide prevention activities (VA, 2024). It draws from NDI, CDW, VA administrative records, and DoD administrative records. The VA Death Ascertainment File is limited in that it does not include cause of death and the earliest data it contains is from 2006 (this is data that was integrated from the VHA Vital Status File) (Arnold, 2022). The Mortality Data Repository, like other sources that

use NDI data, is limited as it does not include information for deaths that occurred before 1979 (VA, 2024).

The only source discussed in this section that does not require a researcher to be a VA employee is BIRLS. Its Death File for 1850–2010 can be accessed on Ancestry.com (Ancestry, 2025b). Additionally, except for the VA/DoD Mortality Data Repository, each VA database only contains information for veterans who enrolled in VA benefits or whose beneficiaries filed a claim with the Veterans Benefits Administration (i.e., burial benefits) (Sohn et al., 2006). BIRLS, the Medical SAS Inpatient Datasets, and the VA Death Ascertainment File are likely to be the most useful VA sources for researchers, as veteran mortality studies have successfully used these in combination with NDI.

The committee submitted an information request to the National VA History Center Archives staff to ascertain what records they may have that contain relevant health outcomes information for veterans who served at the Manhattan Project sites of interest between 1942 and 1947. Their response stated that they generally do not retain medical records and their earliest records from VHA’s Health Outcomes Military Exposures are Vietnam-era exposures to Agent Orange and other herbicides.⁷ Their response also explained that the most likely VA source for relevant health outcomes information would be the veteran’s Consolidated Medical Record. However, most of those records are retained in Federal Records Centers. They are held for 75 years after the last interaction with VA, then destroyed. As noted, any National VA History Center Archives collections pertaining specifically to Manhattan Project veterans had been transferred to NARA.⁸

In addition to sources for mortality data, VA maintains its own Central Cancer Registry, which contains records for approximately 90% of cancer cases treated in the VA system since 1995 (Zullig et al., 2012). Given the noted life expectancy for the relevant veterans, this database would include data for only a limited number of them. However, for those included, the registry contains data on era and branch of service and exposure to ionizing radiation (Zullig et al., 2024).

Commercial Sources

NARA has an agreement with Ancestry.com to digitize, index, and publish military records from World War II, including hospital admission card files 1942–1954 (Ancestry, n.d.). However, the World War II records in the collection only have Army personnel wounded in battle and treated at Army facilities, which is unlikely to include many of the veterans of interest.

___________________

⁷ Personal communication, Michael Visconage, chief historian, Department of Veterans Affairs, February 9, 2024.

⁸ Ibid.

A potentially useful NARA record collection available through Ancestry. com is the burial registers for U.S. military posts and national cemeteries for personnel buried 1862–1960. While the level of detail included in individual records varies, some include cause of death information (Ancestry, 2025a). As of October 2024, more NARA military records were digitized and searchable on Ancestry.com for the Revolutionary War through World War I eras than for the World War II era (Ancestry n.d.).

Obituaries are another potential source of information about fact of death for veterans. Families often include information about military service and may note general information about cause of death. One source of obituaries is Newspapers.com, which is affiliated with Ancestry.com. A paid membership to the site allows users to search and view obituaries published in newspapers dating back to the late 1700s by key words or an individual’s name (Newspapers.com, 2025).

SYNOPSIS

The committee searched for and examined potential sources of health outcomes data and records for the veteran population of interest. The statement of task specified outcomes of cancer occurrence and cause of death for exposed veterans. Health outcomes are routinely identified in retrospective epidemiologic investigations through passive linkage to sources of morbidity and mortality data, such as NDI and state cancer registries. If a roster of individual veterans were created and the required approval processes followed for release of such records and information, these passive sources might provide health outcomes information for those individuals. However, full names and other personal identifiers would be needed to make these linkages, as these sources do not contain an indicator of veteran status by which to search. The roster could also be used to search VA health and benefit records to determine whether individuals had ever used VA health care services and for what conditions, or if they received compensation for a particular health-related outcome.

The committee considered several sources of health records and health outcome information, including VA and state-based cancer registries, VHA clinical and administrative databases, CMS data, specific health outcomes collected as part of the EEPOICPA compensation program for DOE workers, DOL site exposure matrices, special exposure cohorts, commercial sources, such as Ancestry.com and obituaries, and databases of worker cohorts, such as CEDR, that may contain exposure and health outcome data (including death certificates). Most of these sources were established decades after the end of the Manhattan Project and lack comprehensive coverage for that population. Therefore, they cannot be used to identify health information for a study of this veteran population.

Although this was only applicable to Manhattan Project veterans who had careers with DOE or its predecessor agencies (only a subset of mostly technical Special Engineer Detachment personnel), the committee made information requests to DOE record management field offices at the three primary sites of Oak Ridge, Hanford, and Los Alamos about what health records exist at those locations and asked them to provide examples of deidentified health records. Their records managers gave invited presentations and, when possible, showed examples of site-specific health records. Select hospitals at the primary production facility locations known to exist during the Manhattan Project were contacted to learn the disposition of military and worker records. The committee learned that if these clinical records existed, they had been transferred to DOE Office of Legacy Management. However, similar to the other sources of health records that are not death certificates, these DOE site-specific records would be useful for only a subset of the population of interest and do not provide comprehensive health outcome information for a study.

One source that may contain some relevant health outcome information is the official military personnel files held by the National Archives at St. Louis, described in Chapter 4. However, it is unknown how many official military personnel files would be available for Manhattan Project veterans, given the millions of records destroyed by the 1973 fire there. This data source for health outcome information would be limited and only applicable to time in service.

The availability of military and VA administrative medical records and supplemental sources such as cancer registries is limited, particularly for 1942–1947. The committee was unable to find a comprehensive source of veteran-specific health outcome information.

Health outcome information on a small cohort of individuals (less than 10,000) in the United States is difficult to find and compile because people have differing access to medical care, medical records are not centralized, and relatively few disease diagnoses are required to be reported to local, state, or national public health agencies. Less than 1% of World War II veterans were alive in 2024, and it is unknown if any of them participated in the Manhattan Project. Therefore, death certificates will be the most complete and standardized source of health outcome (cause of death) information for the veteran population of interest. The best sources of cause of death information will be NDI for deaths in 1979 or later and state-level death certificates for deaths before 1979. Although these could provide a cause of death, underlying and contributing cause(s) may be missing or incomplete for deaths before 1979. Additionally, as individuals live longer, the most common causes of death change, which may confound the process of connecting cause of death directly to exposures at a Manhattan Project site.

The committee found that despite no comprehensive sources of health outcome records, death certificates are generally systematically available (obtained individually either through NDI or the individual states) for Manhattan Project veterans.

Given the lack of a centralized or national health system and that death certificates are the only source of systematically collected health information, a mortality study is the best approach for identifying health outcomes, as it will capture nearly all veterans of interest. No other source of health or medical records exists that is comprehensive, centralized, or collected longitudinally for all Manhattan Project veterans. The most useful sources for cause of death information for such a mortality study are NDI and state-level death certificates, assuming that a roster of individual veterans with their identifiers can be created. Given the relatively few Manhattan Project veterans overall, and the need to treat each site as separate due to the differing exposures, it is unlikely that there would be sufficient medical or health records to conduct an epidemiologic study.

Conclusion 6-1: Given the lack of health records available for individual veterans covered by the statement of task and the lack of nationwide, comprehensive cancer incidence data, the committee concludes that cancer incidence is not a feasible health outcome for an epidemiologic study of Manhattan Project veterans. However, mortality is a feasible outcome for such a study given the systematic surveillance of death in the United States.

REFERENCES