Religion, Spirituality, and Health: New Considerations for Epidemiology

Abstract

Religion and spirituality are important social determinants that drive public health practice. The field of epidemiology has played a vital role in answering long-standing questions about whether religion is causally associated with health and mortality. As epidemiologists spark new conversations (e.g., see Kawachi (Am J Epidemiol. (https://doi.org/10.1093/aje/kwz204)) and Chen and VanderWeele (Am J Epidemiol. 2018;187(11):2355–2364)) about methods (e.g., outcomes-wide analysis) used to establish causal inference between religion and health, epidemiologists need to engage with other aspects of the issue, such as emerging trends and historical predictors. Epidemiologists will need to address 2 key aspects. The first is changing patterns in religious and spiritual identification. Specifically, how do traditional mechanisms (e.g., social support) hold up as explanations for religion-health associations now that more people identify as spiritual but not religious and more people are not attending religious services in physical buildings? The second is incorporation of place into causal inference designs. Specifically, how do we establish causal inference for associations between area-level constructs of the religious environment (e.g., denomination-specific church membership/adherence rates) and individual- and population-level health outcomes?

Editor’s note: A response to this commentary appears on page 759.

Religion and spirituality are important social determinants that drive public health practice (1). The field of epidemiology has long played a vital role in shaping public health research on religion and health (2). Three critical questions are raised in this research: 1) Is there an association between religion and health?; 2) Is it valid?; and 3) Is it causal? (3). Several recent studies with strong epidemiologic analyses (4, 5) established some evidence for the first 2 questions. However, recent exchanges on the topic (6, 7) illustrate that epidemiology has yet to answer the third, and new questions are yet to be asked and investigated (8).

As epidemiologists lend their expertise to advance research on this topic, now is an opportune moment to pose several additional questions that epidemiologists must grapple with going forward. The new issues raised in this commentary concern not only how we use old tools but also which new tools will be needed for analysis (9). Next, findings from epidemiologic studies should be robust enough to persuade policy-makers and practitioners to invest money in grants and interventions related to the religion-health association.

EPIDEMIOLOGY AND IDENTIFYING CHANGING RELIGIOUS AND SPIRITUAL PATTERNS

To improve causal claims of religion’s influence on health, epidemiologists will need to contend with the rising trend of people who report never attending religious services. The stakes are high, because practitioners who use our research findings might fail to produce interventions that are salient for the emerging generation. The General Social Surveys carried out by the University of Chicago’s National Opinion Research Center between 1995 and 2017 revealed sharp increases, among all racial/ethnic groups, in the proportion of the US population who report having no religious affiliation and never attending religious services (10). These trends are higher among some age groups (e.g., persons 18–34 years) but almost equivalent across the sexes. Global data reveal similar trends between 1994–2004 and 2010–2014 in the proportion of people around the world who do not attend religious services (11), with a similar epidemiologic profile as the United States with respect to age group and sex.

Parallel with the increasing trend of those who never attend religious services is an increase in the proportion of people in the United States who report being spiritual but not religious, which grew 8% between 2012 (19%) and 2017 (27%) (12). One problem with that observed correlation is that because data are cross-sectional snapshots, we cannot determine whether people who previously identified as religious are now choosing to identify as spiritual. That problem of correlation not implying causation is something the field of epidemiology has addressed before generally (13) and will therefore need to apply to research on religion, spirituality, and health in order to establish causality and recommend specific interventions.

EPIDEMIOLOGY AND ESTABLISHING UNDERLYING MECHANISMS GIVEN NEW RELIGIOUS LANDSCAPES

To strengthen causal claims about the association between religion and health given changing religious landscapes, epidemiologists will need to build on previously established mechanisms (14) but also identify and capture new ones. Some authors, for instance, have put forth the concept that social support systems within congregations are one potential explanation for the observed protective association between religious service attendance and mortality (15). While some evidence from randomized field experiments demonstrates a relationship between increased religious participation and improvements in some economic outcomes (16), there have been no similar randomized studies with a focus on public health outcomes. What, then, should epidemiologists measure? How should epidemiologists specify religious and spiritual variables (e.g., as main effects or as effect modifiers to each other) for analysis? And, if an association (causal or correlational) is found—say, between spirituality and health but not religious attendance and health—then what other causal mechanisms (beyond social support) can we put forth to explain the finding? Finally, what interventions should the epidemiologist recommend to public health practitioners if spiritual but not religious people exhibit the strongest protective causal association with a certain outcome?

REVISED MEASURES NEEDED IN LARGE-SCALE EPIDEMIOLOGIC SURVEYS

To answer the previous questions posed, epidemiologists need not only to consider new ways to specify main, interaction, and control variables in existing religion-and-health data but also to include relevant measures in future national household surveys. Establishing causal effects will require epidemiologists to distinguish between religion and spirituality and their relationships to health outcomes.

Religion can be broadly described as a symbolic cultural system that includes a set of rituals, rites, and practices (17). It is already widely known that religion is a multidimensional construct (14), and while there is some consensus on operational definitions (e.g., frequency of worship attendance), exactly “who” the study findings apply to remains unclear (8). Analysis of the effects of spirituality in a causal framework poses similar problems for epidemiologists to resolve. Some argue that it is more practical for research if we define spirituality through a religious framework (18), which reflects a person’s understanding of the self in connection to sacred or divine power. Others posit that spirituality need not be defined according to a religious framework and that its definition should be fluid to improve inclusivity, especially among non–religiously identified people (19). Given these multiple conceptual ways to operationalize these constructs, epidemiologists will need to be clear about their intended population of inference. Space in large-scale surveys is limited, so epidemiologists will have to be specific in terms of which measures need to be included and use the literature to inform the variables they select.

SITUATING PLACE WITHIN CAUSAL INFERENCE BETWEEN RELIGION AND HEALTH

Epidemiologists have already established that place matters for health (20) and have developed methods for analyzing effects of place on health (21). Therefore, epidemiology’s contribution to cementing claims that religion is a social determinant of health should include constructs that are measured beyond the individual level (22) and that focus on macro-level factors (23). Researchers in other fields have contributed to knowledge on contextual/area-level religious measures (e.g., the prevalence of Catholics or Protestants in an area) (24). Epidemiologists can build on this work to establish causality in 3 domains: measurement, analysis, and theory development.

Measurement

There is currently no consensus on what good or reliable contextual-level religious measures are or how to analyze them in relation to health. However, epidemiology has provided overall guidelines for measurement of area-level variables so that we can avoid common epidemiologic fallacies and making incorrect inferences to populations (21). Therefore, epidemiologists can contribute to how we are to specify and interpret the associations between variables such as the percentage of people in an area who attend religious services and some health variable when there is adjustment for compositional effects of people who attend religious services. In future research, epidemiologists should identify whether there are other opportunities to assess purely contextual/area-level constructs of the religious environment that do not rely on aggregating individual-level measures. Other opportunities for epidemiologists to establish causal effects on individual health could include identifying natural events, such as changes in the number of places of worship within neighborhoods.

Analysis

Epidemiologists are already equipped with analytical tools for addressing potential problems that arise when analyzing the causal impact of contextual-level religious measures on health. For instance, epidemiologists will need to consider approaches such as cross-classified or multiple-membership multilevel models where people can be nested within their neighborhood as well as another neighborhood. Use of those models and other analytical techniques will become important for causal inference, since some people attend worship services outside of the neighborhoods where they reside (25). Epidemiologists can recommend what contextual/area-level confounders should be included in ecological or multilevel studies that aim to establish a relationship between some contextual religious variable and individual health.

Theory development

Next, epidemiology will be well poised to contribute to theory about religious effects on health through rigorous analyses on the topic that can either rule out previous associations or identify new ones that are robust. As others have noted, epidemiologists will have to ask new questions about religion and health and specify why they are asking those questions (23). Given that people relocate for job-, school-, or family-related reasons and for other opportunities, would it be reasonable to think that the religious environment in a new place of residence could have an impact on a person’s health in the new location? These are all questions that could advance previous theories or generate new ones.

CONCLUSIONS AND THE WAY FORWARD

It has long been established that there is an association between religion and health and that the association is valid. Currently, epidemiology is at the forefront of providing robust evidence showing that associations between religion and health are causal. As the field engages more with this topic, inquiry should include the impact of the changing religious landscape and inclusion of variables such as spirituality to establish causal inference. Epidemiologists are well equipped for this and therefore should also aim to establish causal inference through studies that examine relationships between contextual-level religious measures and individual health. Once systematic evidence has been developed, epidemiology will have meaningfully contributed to new theory and pursuasive findings that can inform policy and health interventions to improve population health.