Causal diagramming for assessing human system risk in spaceflight.

For over a decade, the National Aeronautics and Space Administration (NASA) has tracked and configuration-managed approximately 30 risks that affect astronaut health and performance before, during and after spaceflight. The Human System Risk Board (HSRB) at NASA Johnson Space Center is responsible for setting the official risk posture for each of the human system risks and determining-based on evaluation of the available evidence-when that risk posture changes. The ultimate purpose of tracking and researching these risks is to find ways to reduce spaceflight-induced risk to astronauts. The adverse effects of spaceflight begin at launch and continue throughout the duration of the mission, and in some cases, across the lifetime of the astronaut. Historically, research has been conducted in individual risk "silos" to characterize risk, however, astronauts are exposed to all risks simultaneously. In January of 2020, the HSRB at NASA began assessing the potential value of causal diagramming as a tool to facilitate understanding of the complex causes and effects that contribute to spaceflight-induced human system risk. Causal diagrams in the form of directed acyclic graphs (DAGs) are used to provide HSRB stakeholders with a shared mental model of the causal flow of risk. While primarily improving communication among those stakeholders, DAGs also allow a composite risk network to be created that can be tracked and configuration managed. This paper outlines the HSRB's pilot process for this effort, the lessons learned, and future goals for data-driven risk management approaches.

Levels of evidence for human system risk evaluation.

NASA uses a continuous risk management process to seek out new knowledge of spaceflight-induced risk to human health and performance. The evidence base that informs the risk assessments in this domain is constantly changing as more information is gleaned from a continuous human presence in space and from ongoing research. However, the limitations of this evidence are difficult to characterize because fewer than 700 humans have ever flown in space, and information comes from a variety of sources that span disciplines, including engineering, medicine, food and nutrition, and many other life sciences. The Human System Risk Board (HSRB) at NASA is responsible for assessing risk to astronauts and communicating this risk to agency decision-makers. A critical part of that communication is conveying the uncertainty regarding the understanding of the changes that spaceflight induces in human processes and the complex interactions between humans and the spacecraft. Although the strength of evidence grades is common in the academic literature, these scores are often not useful for the problems of human spaceflight. The HSRB continues to update the processes used to report the levels of evidence. This paper describes recent updates to the methods used to assign the level of evidence scores to the official risk postures and to the causal diagrams used by the HSRB.

Updates to the NASA human system risk management process for space exploration.

This paper describes updates to NASA's approach for assessing and mitigating spaceflight-induced risks to human health and performance. This approach continues to evolve to meet dynamically changing risk environments: lunar missions are currently being designed and the ultimate destination will be Mars. Understanding the risks that astronauts will face during a Mars mission will depend on building an evidence base that informs not only how the humans respond to the challenges of the spaceflight environment, but also how systems and vehicles can be designed to support human capabilities and limitations. This publication documents updates to the risk management process used by the Human System Risk Board at NASA and includes changes to the likelihood and consequence matrix used by the board, the design reference mission categories and parameters, and the standardized evaluation of the levels of evidence that the board accepts when setting risk posture. Causal diagramming, using directed acyclic graphs, provides all stakeholders with the current understanding of how each risk proceeds from a spaceflight hazard to a mission-level outcome. This standardized approach enables improved communication among stakeholders and delineates how and where more knowledge can improve perspective of human system risks and which countermeasures can best mitigate these risks.

Viability of internal comparisons for epidemiological research in the US astronaut corps.

This study aims to determine whether astronauts who have not flown in space can provide an unbiased comparison to astronauts who have flown in space when analyzing long-term health outcomes such as incidence of chronic disease and mortality. Various propensity score methods failed to achieve good balance between groups, demonstrating that even with sophisticated rebalancing methods the group of non-flight astronauts cannot be demonstrated to be an unbiased comparison group for examining the effect of the hazards of spaceflight on incidence and mortality from chronic diseases.

Validating Causal Diagrams of Human Health Risks for Spaceflight: An Example Using Bone Data from Rodents.

As part of the risk management plan for human system risks at the US National Aeronautics and Space Administration (NASA), the NASA Human Systems Risk Board uses causal diagrams (in the form of directed, acyclic graphs, or DAGs) to communicate the complex web of events that leads from exposure to the spaceflight environment to performance and health outcomes. However, the use of DAGs in this way is relatively new at NASA, and thus far, no method has been articulated for testing their veracity using empirical data. In this paper, we demonstrate a set of procedures for doing so, using (a) a DAG related to the risk of bone fracture after exposure to spaceflight; and (b) four datasets originally generated to investigate this phenomenon in rodents. Tests of expected marginal correlation and conditional independencies derived from the DAG indicate that the rodent data largely agree with the structure of the diagram. Incongruencies between tests and the expected relationships in one of the datasets are likely explained by inadequate representation of a key DAG variable in the dataset. Future directions include greater tie-in with human data sources, including multiomics data, which may allow for more robust characterization and measurement of DAG variables.

Contrapositive logic suggests space radiation not having a strong impact on mortality of US astronauts and Soviet and Russian cosmonauts.

Space travelers are exposed to unique forms of ionizing radiation that pose potentially serious health hazards. Prior analyses have attempted to quantify excess mortality risk for astronauts exposed to space radiation, but low statistical power has frustrated inferences. If exposure to deep space radiation were causally linked to deaths due to two particular causes, e.g., cancer and cardiovascular disease, then those cause-specific deaths would not be statistically independent. In this case, a Kaplan-Meier survival curve for a specific cause that treats deaths due to competing causes as uninformative censored events would result in biased estimates of survival probabilities. Here we look for evidence of a deleterious effect of historical exposure to space radiation by assessing whether or not there is evidence for such bias in Kaplan-Meier estimates of survival probabilities for cardiovascular disease and cancer. Evidence of such bias may implicate space radiation as a common causal link to these two disease processes. An absence of such evidence would be evidence that no such common causal link to radiation exposure during space travel exists. We found that survival estimates from the Kaplan-Meier curves were largely congruent with those of competing risk methods, suggesting that if ionizing radiation is impacting the risk of death due to cancer and cardiovascular disease, the effect is not dramatic.

Mortality Among International Astronauts.

INTRODUCTION: Research on the mortality of space explorers has focused exclusively on U.S. astronauts and Soviet and Russian cosmonauts. However, other nations have organized space programs over the last 40 yr and the European Space Agency, the Canadian Space Agency, the China National Space Administration, and the Japan Aerospace Exploration Agency all offer an opportunity for further study of the mortality of space explorers.METHODS: We used biographical and vital data abstracted from public sources for European, Canadian, Chinese, and Japanese astronauts. Using general population mortality rates from the Human Mortality Database and mortality rates derived from the cohort of U.S. astronauts, we computed standardized mortality ratios.RESULTS: The groups displayed different preferences in selection of astronauts. As there were no deaths in any of the four groups, the point estimates for standardized mortality ratios were all 0. However, the European cohort experienced a statistically significant reduction in all-cause mortality risk in comparison to the European general population as well as in comparison to U.S. astronauts.DISCUSSION: The healthy worker effect predicts that all study cohorts should have lower all-cause mortality risk in comparison to their general populations. The general population of Japan has mortality rates low enough that any reduction in mortality risk may remain undetectable in the Japanese cohort. Continued surveillance of these populations in the coming decades will make them a useful addition to the evidence base for astronaut mortality.Reynolds RJ, Day SM. Mortality among international astronauts. Aerosp Med Hum Perform. 2019; 90(7):647-651.

Mortality of US astronauts: comparisons with professional athletes.

OBJECTIVE: Studies of mortality among US astronauts are complicated by the healthy worker effect, which predicts lower mortality for astronauts than the general population based solely on the ability to become and remain an astronaut. We attempt to evaluate astronaut mortality risk while accounting for the healthy worker effect. METHODS: We compare mortality rates of male US astronauts with those of professional athletes from Major League Baseball and the National Basketball Association between January 1, 1960 and May 31, 2018. RESULTS: Both athlete cohorts and astronauts had significantly lower-than-expected mortality in comparison with the general population. For the overall study period, there were no significant differences in all-cause mortality rates between astronauts and athletes. Astronauts were at greater risk of death from external causes (SMR=583; 95% CI 377 to 860) and reduced risk of death from cardiovascular disease (SMR=39; 95% CI 18 to 73) and all natural causes (SMR=67; 95% CI 47 to 93). CONCLUSIONS: The data presented here do not support increased mortality for astronauts due to unique exposures received in space. The mortality experience of astronauts as compared with professional baseball and basketball players should be re-examined periodically as part of the ongoing surveillance of astronaut mortality in years to come.

The effect of competing risks on astronaut and cosmonaut mortality.

Astronauts and cosmonauts have been reported to be at substantially lower age-specific risk of death from chronic disease (primarily heart disease and cancers) in comparison to the general populations of the United States and Russia, respectively. Yet, both groups have been at greater age-specific risk of death from external causes, mainly due to plane crashes and spacecraft accidents. In this study we tested the hypothesis that the reported reductions in mortality from natural causes result, to some degree, from survival bias created by early deaths from external causes. Statistical comparisons of baseline characteristics between cause-of-death groups showed no significant differences. Cause-specific survival curves showed no difference in long-term mortality from external causes among either astronauts or cosmonauts compared to Kaplan-Meier curves with censoring for competing causes. Cause-specific survival curves for natural causes suggested a possible upward bias in mortality estimates published thus far for both groups of space explorers. Differences in survival between Kaplan-Meier curves and the cause-specific survival curves were 7% and 5% for astronauts and cosmonauts respectively after 55 years. The data do not support the hypothesis that observed reductions in mortality from natural causes are due in whole or in part to bias created by deaths from external causes at young ages. The data imply that reports of cause-specific mortality for astronauts and cosmonauts may in fact systematically overestimate mortality rates, though these findings should be interpreted with caution as the data are thin at the extremes of follow-up time.

Mortality Rates and Excess Death Rates for the Seriously Mentally Ill.

OBJECTIVES: -To compute mortality rates and excess death rates for patients with serious mental illness, specific to categories of gender, age and race/ethnicity. BACKGROUND: -People with serious mental illness are known to be at greatly increased risk of mortality across the lifespan. However, the measures of mortality reported for this high-risk population are typically only summary measures, which do not provide either the mortality rates or excess death rates needed to construct life tables for individuals with serious mental illness. METHODS: -Mortality rates were computed by dividing the number of deaths by the amount of life-years lived in strata specific to gender, age and race/ethnicity. Age-specific excess death rates were determined as the difference between the study population rate and the corresponding general population rate in each stratum. To compute excess death rates beyond observed ages in the cohort, a method with documented reliability and validity for chronic medical conditions was used. RESULTS: -For the cohort with mental illness, mortality rates for Black and White females were mostly equal, and consistently greater than those for Hispanic females; excess death rates for females displayed a similar pattern. Among males, mortality rates were highest for Whites, with Hispanics and Blacks close in magnitude at all ages. Excess death rates for males showed more divergence between the categories of race/ethnicity across the age range. CONCLUSIONS: -Mortality rates specific to categories of gender, age and race/ethnicity show sufficient differences as to make them the preferred way to construct life tables. This is especially true in contrast to broader summary measures such as risk ratios, standardized incidence rates, or life expectancy.

Mortality Due to Cardiovascular Disease Among Apollo Lunar Astronauts.

INTRODUCTION: Recent research has postulated increased cardiovascular mortality for astronauts who participated in the Apollo lunar missions. The conclusions, however, are based on small numbers of astronauts, are derived from methods with known weaknesses, and are not consistent with prior research. METHODS: Records for NASA astronauts and U.S. Air Force astronauts were analyzed to produce standardized mortality ratios. Lunar astronauts were compared to astronauts who have never flown in space (nonflight astronauts), those who have only flown missions in low Earth orbit (LEO astronauts), and the U.S. general population. RESULTS: Lunar astronauts were significantly older at cohort entry than other astronaut group and lunar astronauts alive as of the end of 2015 were significantly older than nonflight astronauts and LEO astronauts. No significant differences in cardiovascular disease (CVD) mortality rates between astronaut groups was observed, though lunar astronauts were noted to be at significantly lower risk of death by CVD than are members of the U.S. general population (SMR = 13, 95% CI = 3-39). DISCUSSION: The differences in age structure between lunar and nonlunar astronauts and the deaths of LEO astronauts from external causes at young ages lead to confounding in proportional mortality studies of astronauts. When age and follow-up time are properly taken into account using cohort-based methods, no significant difference in CVD mortality rates is observed. Care should be taken to select the correct study design, outcome definition, exposure classification, and analysis when answering questions involving rare occupational exposures.Reynolds RJ, Day SM. Mortality due to cardiovascular disease among Apollo lunar astronauts. Aerosp Med Hum Perform. 2017; 88(5):492-496.

The relationship of life expectancy to the development and valuation of life care plans.

BACKGROUND: A life care plan often analyzes needs up to a person's life expectancy. Expected present value of necessary funding for such a plan is likewise based on the fixed survival time. If a client should live beyond or die before the life expectancy, a shortfall or excess of funding may seem inevitable. The life table, of which life expectancy is a summary measure, clarifies these issues. OBJECTIVES: We explain life expectancy and how it is used in tort litigation, economic calculations, and life care planning. We examine the life table, of which life expectancy is one output. We illustrate how a life table provides age-specific probabilities of death and survival, life expectancies, and median survival times, and other information and that every life expectancy must be associated with a life table. We consider the implications for life care planners, forensic economists, and others. CONCLUSIONS: Life expectancy is a summary of more detailed information provided in a life table. The full life table provides better information for planning purposes. Whether life expectancy or a full life table should be used in developing and valuing a life care plan is not well understood. A multi-disciplinary approach may help clarify these issues.

Extrapolating published survival curves to obtain evidence-based estimates of life expectancy in cerebral palsy.

Studies reporting long-term survival probabilities for cohorts of persons with cerebral palsy provide evidence-based information on the life expectancy of those cohorts. Some studies have provided estimates of life expectancy based on extrapolation of such evidence, whereas many others have opted not to do so. Here we review the basic methods of life table analysis necessary for performing such extrapolations, and apply these methods to obtain evidence-based estimates of life expectancy from several studies that do not report such estimates themselves.

Comparative Mortality and Risk Factors for Death among US Supreme Court Justices (1789-2013).

Objectives .- To compare the mortality experience of 112 justices of the US Supreme Court with that expected in the general population. To identify variables associated with mortality within this cohort. Background .- Supreme Court justices are a select occupational cohort. High socio-economic status, advanced education, lifetime appointment, and the healthy worker effect suggest lower mortality. Sedentary work, stress, and a tendency to work beyond typical retirement age may attenuate this. Methods .- Standardized mortality ratios compare the observed mortality rates of justices with those expected in age- and sex-matched contemporary general populations. Poisson regression analyzes variables associated with mortality within the cohort. Results .- From 1789 to 2013, 112 justices (108 male) contributed 2,355 person-years of exposure. Mean age (standard deviation) at appointment was 53.1 years (6.7); at retirement 69.7 years (9.9); at death (n = 100) 74.4 years (10.3); and at end of the study for those alive (n = 12) 72.1 years (11.8). Standardized mortality ratios (95% ci) were: overall 0.87 (0.70-1.05); prior to 1950 0.92 (0.61-1.33); and from 1950 to 2013 0.66 (0.42-0.99). Variables in the final Poisson model and their associated mortality rate ratios (95% ci) were: age 1.06 (1.03-1.09); calendar year 0.99 (0.99-1.00); active status 0.41 (0.25-0.68); career length 1.04 (1.01-1.07); and chief justice 1.08 (0.59-1.84). Conclusions .- Supreme Court mortality was lower than that of the general population in the period from 1950 to the present, but was on par prior to 1950. Increasing age and career length were associated with greater mortality, while active status and later calendar year with lower. These results may add to a body of knowledge that may help to develop or refine models of mortality risk in increasingly aged working populations.

Mortality among Soviet and Russian cosmonauts: 1960-2013.

INTRODUCTION: Though the mortality of U.S. astronauts has been studied repeatedly in the last 20 yr, little is known about the long-term mortality trends of Soviet and Russian cosmonauts. METHODS: Using data from 266 cosmonauts accepted into cosmonaut training from 1960 to 2013, we document the causes of death and crude death rates among cosmonauts. Using standardized mortality ratios (SMR), we compared cosmonauts to the general populations of Russia and Ukraine, and to 330 U.S. astronauts. RESULTS: Cosmonauts experienced significantly lower all-cause mortality risk compared to the general population. However, cosmonauts were at almost double the risk of all-cause mortality in comparison to U.S. astronauts (SMR = 190, 95% C.I. 154-239). Cosmonauts were also at greater risk of circulatory disease (SMR = 364, 95% C.I. 225-557) and cancer (SMR = 177, 95% C.I. 108-274) compared to U.S. astronauts. Though not statistically significant, cosmonauts experienced fewer fatal accidents (SMR = 88, 95% C.I. = 54-136) than their U.S. counterparts. DISCUSSION: Cosmonauts are at much lower risk of all-cause mortality than the general populations of Russia and Ukraine, yet are at greater risk for death by cardiovascular disease and cancer than are U.S. astronauts. This disparity may have common roots with decreases in life expectancy in Russia in recent decades. Further research is needed to understand these trends fully.