Prayer for Health and Primary Care: Results From the 2002 National Health Interview Survey

Abstract

Background and Objectives

Prayer for health (PFH) is common; in 2002, 35% of US adults prayed for their health. We examined the relationship of PFH and primary care visits, with a special focus on African American women, using data from the 2002 National Health Interview Survey (NHIS).

Methods

We used chi-square analyses to compare the demographic (age group, gender, race, region, marital status, educational level, ethnicity) and health-related covariates (alcohol use, smoking status, and selected medical conditions) between individuals who did and did not pray for their health in the past year. Univariate associations between PFH and visit to primary care provider (PCP), with Mantel-Haenszel adjustment for confounding, were determined. Multivariate regression was used to determine independent factors associated with PFH and PCP visit, with SUDAAN to adjust for the clustered survey design.

Results

Subjects who prayed were more likely to be female, older than 58, Black, Southern, separated, divorced or widowed, and nondrinkers. Subjects who prayed were also more likely to have seen a PCP within the past year. Black women who prayed were also more likely to see a PCP.

Conclusions

These findings suggest that people who pray for their health do so in addition to, not instead of, seeking primary care. This finding is maintained but with a smaller effect size, in Black women.

The adoption of religion or prayer as a means of influencing one’s health is prevalent in the US population.^1–4 Several publications in the medical literature have addressed the relationship of religion or prayer on the health or well-being of individuals or groups,^5–8 and many show associations between religion or religious service attendance and general health or well-being,^9–14 while others focus on specific outcomes such as mortality,^15,16 in-hospital complications,^17,18 longevity,¹⁹ control of chronic pain,^20–22 depression,²³ self-esteem,²⁴ and coping with cancer.²⁵ Still other studies have focused on the association of religion or prayer with health behaviors such as cancer screening,^26–28 smoking,^29,30 and alcohol use.³¹

These studies have all shown a neutral or beneficial association of religion (variably outcomes and have prompted other researchers to propose mechanisms by which religion could affect health. One such mechanism is the locus of control model for health behavior.^32–34 Within the locus of control model, individuals may believe that their religion empowers them to positively influence their health through prayer (internal locus of control) or that fate is controlling their health and they must accept it (external locus of control). In the former case, we would expect prayer to be associated with other proactive health behaviors such as cancer screening and smoking cessation. In the latter case, prayer may be associated with not seeking medical services.

It is possible that within different social or cultural groups, the perceived locus of control may vary.³⁵ For example, some studies with these with African American and Latino subjects, particularly women,^36–44 have demonstrated a less consistent health benefit associated with religion than noted in previous studies. Some studies on cancer screening in urban African American women,^35,37,38 noted an association between religiousness and an external locus of control, and women who prayed more and attended religious services were less likely to obtain cancer screening. One study, however, noted an increased incidence of dental visits and women’s cancer screening tests⁴³ in African American churchgoing subjects. This work raises the question of how the locus of control model applies in different social groups and whether religion or prayer can be assumed to have the same association with health behaviors for everyone. Table 1 offers a brief summary of the locus-of-control model with specific examples relating the model to religion or prayer.

Table 1.

Locus of Control Model for Health Behavior With Religion Examples

Locus of Control	Explanation	Religious Example
Internal	Person believes that they have the power to impact or change their health through their behavior	Person prays for their health because they believe it is something they can do to help themselves
External (powerful others)	Person believes that they do not have power to impact their own health, but other people do	Person relies on their religious leader for health advice (and maybe their physician)
External (fate)	Person believes that no one can impact their health, it is all up to fate (or God/higher power)	Person prays for their health (or adopts superstitious behaviors) because they don’t believe they can do anything to help themselves

The issue of how prayer is used by patients is of particular importance for primary care physicians (PCPs), who see a large cross-section of the population as patients. If people are using prayer as a substitute for medical care, what types of patients are doing this, and what are the implications of this behavior?

We addressed these questions by examining the association between praying for one’s health and having a PCP by using a large, nationally representative database (the 2002 National Health Interview Survey [NHIS]). While other studies have suggested that people who pray are more likely to seek medical care,⁴³ we examined the question focusing specifically on primary care, using a sample large enough to examine variability among different racial and ethnic groups.

This research has two main objectives. The first is to examine the association between prayer for health (PFH), which is the act of engaging in prayer or worship specifically with the goal of improving one’s health and whether an individual visited a PCP within the last year. The second is to determine the stratum-specific association between PFH and PCP visit for African American women, since prior studies suggested that people in this group to visit their PCP if they prayed.

Methods

We conducted a cross-sectional secondary data analysis using the 2002 NHIS. The survey involved 31,044 American adults (18 years or older), and Hispanics and African Americans were oversampled. The questions in the NHIS focus on self-reported assessments of health, symptoms, disease prevalence, insurance status, and degree of disability. Data are gathered in face-to-face interviews. In 2002, a supplement was added that contained questions about the use of complementary and alternative medicine (CAM) and prayer for health.

Our project was designated exempt by the Boston University Institutional Review Board. We included subjects if they had completed both the NHIS and CAM supplement interviews.

Variables

PFH was defined as answering yes to any of the PFH was following four questions: In the past 12 months, (1) Have you prayed for your own health? (2) Have you asked someone else to pray for your health? (3) Have you participated in a prayer group or prayer chain for your health? (4) Have you participated in a healing or sacramental ritual for your health?

We operationalized the PFH variable in two different ways. First, subjects were split into two groups: those who answered yes to any of the above questions and those who answered no to all. This dichotomous variable was used in all the major analyses. Second, we summed the yes answers to the above questions, creating an ordinal scale for PFH. This prayer scale allowed us to compare subjects who had used only one form of PFH with subjects who had used two, three, or four forms. The ordinal variable was included in some of the descriptive analyses for comparison purposes.

The main dependent variable was identified as having had a primary care physician visit. This was defined as a yes response to a question about a visit within the last 12 months to “the usual place where you receive routine or preventive care.”⁴⁵

We compared the following demographic covariates between subjects who prayed for health and those who did not: gender, race (white, Black, Asian, mixed, other), ethnicity (Hispanic/non-Hispanic), income, geographic region of the country (Northeast, Midwest, South, and West), educational level (up to 12th grade without graduation, high school graduate, some college or technical school), marital status (married, separated/divorced, widowed, single, living with partner), and age (born before 1944, born between 1944 and 1963, and born after 1963). These age groupings were made for comparison purposes with McCaffrey,¹ who found in her 1998 survey that the baby boom age group (born 1944–1963) was most strongly associated with PFH in a multivariate regression model.

We used a dichotomous variable for complementary and alternative medicine (CAM) use in the past 12 months, defined according to Barnes⁴⁶ as having practiced any of the following in the past 12 months: acupuncture, Ayurveda, homeopathy, naturopathy, chelation therapy, herbal products, folk medicine, megavitamin therapy, yoga, chiropractic, massage, relaxation/deep breathing/meditation, tai chi, qi gong, biofeedback, energy healing/Reiki, or hypnosis. This variable was included as a covariate in the analyses.

The following health-related covariates were also analyzed: alcohol use (current, former, never or less than 18 drinks in entire life), smoking (current, former, never), and having been diagnosed with or treated for six categories of health conditions (cardiovascular disease, lung disease, arthritis and related conditions, diabetes, depression and anxiety, and cancer). We created the health condition categories because illness burden may be an important confounder in the analysis of PFH. These conditions were categorized based on a risk-adjustment method developed by Hornbrook⁴⁷ using self-reported data. This method groups diagnoses into five categories that were found to be significant predictors of future health care utilization: cardiovascular disease (includes hypertension, high cholesterol, congestive heart failure, myocardial infarction, angina, and stroke), lung disease (includes asthma, emphysema, chronic obstructive pulmonary disease, and chronic bronchitis), diabetes (includes diabetic complications), arthritis (includes osteoarthritis, rheumatoid arthritis, gout, lupus, and other rheumatologic conditions), and depression/anxiety. We added cancer to these groups, because it may be an important potential confounder.

Statistical Analysis

We compared the distribution of demographic and health-related covariates among subjects who had and had not used PFH (both dichotomous and ordinal definitions) in the past 12 months. We used the chi-square test of independence for this comparison.

Next, we calculated crude odds ratios to determine unadjusted associations between PFH and PCP visit using the entire sample (n=31,044). To assess potential effect modification, we conducted stratified analyses. These stratum-specific odds ratios allowed us to observe any variation in the relationship between PFH and PCP visit for specific subgroups (eg, race or gender). The stratum-specific odds ratios for Black women were of particular interest and are reported separately.

Last, we assessed the relationships between PFH and PCP visit using multivariate logistic regression. Covariates were included in the final model if they changed the odds ratio for PFH by 10% or more. An additional multivariate model using the PFH ordinal variable was evaluated. SUDAAN⁴⁸ was used for all analyses to adjust for the clustered sampling design of the NHIS.

Results

Table 2 shows demographic and social characteristics of subjects who did and did not use PFH in the past year. The results shown are similar to those published by Bell et al in 2005³ with some variables classified differently for our analysis. Subjects who used at least one form of PFH were more likely to be female, Black, age 59 or over, former smokers or drinkers, live in the South, use CAM, have an income of $20,000/year or less, and have arthritis, depression/anxiety, cardiovascular disease or cancer (all P<.0001). Those who did not use PFH in the past year were more likely to be male, white, age 38 or younger, current smokers or drinkers, live in the Northeast, be non-CAM users, and be single.

Table 2.

Selected Demographic and Health Characteristics of Subjects Who Prayed and Did Not Pray for Their Health in the Past 12 Months

		Prayed in Past 12 Months (n=14,207)	Did Not Pray in Past 12 Months (n=16,837)
Gender*	Male, %	39	55
	Female	61	45
Race	White, %	76	84
	Black	16	8
	Asian	<1	<1
	Native American	3	4
	Mixed	5	3
Black female		10	4
Income	≤$20,000/year, %	22	16
	>$20,000/year	78	84
Age	38 and under, %	33	44
	39–58	39	37
	59 and over	28	19
Region*	Northeast, %	16	22
	Midwest	24	25
	South	43	33
	West	17	20
Marital status	Single, %	16	22
	Married	59	57
	Divorced/separated	12	9
CAM status	Used in past 12 months, %	41	27
	No use, past 12 months	59	73
Education level	< 12th grade, %	19	15
	High school graduate	26	26
	Any college	55	59
Health	Cardiovascular,** %	27	15
	Diabetes	9	5
	Depression***	22	11
	Arthritis****	28	15
	Cancer	10	5
Smoking	Never smoked, %	56	53
	Former smoker	25	22
	Current smoker	19	25
Alcohol use	Lifetime abstainer, %	26	19
	Former drinker	20	12
	Current drinker	54	69

All P values <.0001

CAM—complementary and alternative medicine

^*These proportions also reported in Bell et al,³ as row percents

^**Cardiovascular=hypertension, myocardial infarction, stroke, arrhythmia, congestive heart failure

^***Depression=depression or anxiety symptoms in past 6 months

^****Arthritis=osteoarthritis, rheumatoid arthritis, gout, lupus, other rheumatologic conditions

We also compared demographic and health variables as they relate to multiple forms of prayer for health used in the past year. In the past year, 21% of the general population used one form of PFH, 14% used two forms, and 8% used three or more forms. Characteristics most associated with the use of two or more forms of PFH were being Southern, Black, or mixed race and having cardiovascular disease, diabetes, depression/anxiety, arthritis, or cancer. The multivariate analysis, with prayer (yes/no) as the dependent variable, showed strong associations between PFH and race, region, marital status, CAM use, and comorbidity, even when adjusting for other variables in the model (Tables 3 and 4).

Table 3.

Selected Characteristics Associated With Use of Multiple Forms of Prayer for Health

	% Using 0	% Using 1	% Using 2	% Using 3+
General population	57	21	14	8
Southern United States	50	21	19	10
Black	39	21	24	16
Mixed race	46	25	18	11
Black female	32	22	28	18
Cardiovascular*	42	24	21	13
Diabetes	39	24	23	14
Depression**	40	23	23	14
Arthritis***	41	24	21	14
Cancer	40	24	22	14

^*Cardiovascular=hypertension, myocardial infarction, stroke, arrhythmia, congestive heart failure

^**Depression=depression or anxiety symptoms in past 6 months

^***Arthritis=osteoarthritis, rheumatoid arthritis, gout, lupus, other rheumatologic conditions

Table 4.

Logistic Regression Showing Variables Independently Associated With Prayer for Health in the Past 12 Months (total n=31,044)

	Odds Ratio	95% CI
Female*	1.71	1.59–1.85
Black	2.89	2.58–3.24
Mixed race	1.90	1.56–2.32
White	1.00
South*	1.78	1.59–1.99
Midwest*	1.39	1.24–1.56
Northeast	1.00
Married	1.38	1.23–1.54
Widowed	1.37	1.17–1.60
Divorced/separated	1.24	1.08–1.42
Single	1.00
Age >58	1.10	1.01–1.23
Age 39–58	1.00
Age <39	0.87	0.81–0.94
Cardiovascular	1.39	1.27–1.52
Diabetes	1.30	1.12–1.50
Depression	1.65	1.50–1.83
Arthritis	1.42	1.29–1.56
Cancer	1.50	1.30–1.73
CAM use in past year	1.68	1.56–1.81

CI—confidence interval

CAM—complementary and alternative medicine

^*Odds ratios for regional and gender variables also reported in Bell et al,³ using different regression model.

After adjusting for race and geographic region, subjects who prayed for their health in the past year were 73% more likely to visit a PCP in the same year, compared with subjects who had not prayed for their health in the past year (Table 5). Using prayer for health as an ordinal variable, there was an increasing likelihood of PCP visit with multiple forms of prayer for health used in the past year. The adjusted OR for one form was 1.50 (95% CI= 1.39–1.62), for two forms it was 1.91 (95% CI=1.76–2.07), for three forms it was 2.17 (95% CI=1.91–2.46), and for four forms it was 2.03 (95% CI= 1.57–2.62). Black women were 11% more. likely to visit a PCP if they prayed for their health (OR=1.11).

Table 5.

Logistic Regression for the Relationship Between Prayer for Health (Dichotomous Classification) and Primary Care Physician Visit in the Past 12 Months (n=31,044)

	Regression Model With Dichotomous Variable Odds Ratio (95% CI)
Prayer for Health (PFH) (dichotomous)
PFH	1.73 (1.63–1.84)
No PFH	1.00
Race
Black	0.82 (0.73–0.92)
Asian	0.73 (0.49–1.08)
Native American/Alaskan Native	0.71 (0.61–0.82)
Mixed race	0.64 (0.56–0.74)
White	1.00
Region
Midwest	0.82 (0.73–0.92)
South	0.71 (0.63–0.79)
West	0.63 (0.56–0.71)
Northeast	1.00

CI—confidence interval

Discussion

In this large, nationally representative sample of American adults, PFH was associated with having had a PCP visit within the last year. The result is consistent with prior studies^{8,11,12,28,44} and supports the notion that if the locus of control model applies to prayer, then it is likely experienced as an internal locus of control, driving the subject both to pray and to visit a PCP in hopes of improving their health. It was particularly striking in this analysis that the likelihood of PCP visit increased with the use of multiple forms of prayer for health indicating that the more one engages in prayer-related behaviors, the more likely one is to visit a primary care physician. It should be noted that the use of multiple forms of prayer for health is not the same as praying more frequently, so it would be inaccurate to label this a “dose response” relationship, especially since it is impossible to consistently quantify a dose of prayer. Nonetheless, the apparent increase in likelihood of PCP visit is a new, and intriguing, finding.

Previous studies have shown mixed results on the association of prayer with certain health behaviors for African American women, though these studies were generally conducted with smaller, local samples that could have been influenced by cultural factors unique to their city or congregation. In this large, nationally representative sample, Black women were less likely than Caucasian women to attend PCP visits. However, Black women who prayed were more likely to attend PCP visits than Black women who did not pray, though the effect size of prayer was smaller in this group. These findings add to the current literature on African-American women and prayer that was discussed in the background section.^36–44 Further study is required to understand whether prayer is experienced differently or has a different meaning in regard to health in different social and cultural groups.

Limitations

Study limitations included a lack of detail on dimensions of prayer and religiosity. The National Health Interview Survey contains data collected by researchers for general purposes; our analyses were planned after data collection had already occurred. To more fully explore the relationship of prayer/religion and health care utilization, more specific data would be needed (for example, association with organized religion, attendance at services, and beliefs about religion’s effect on health or wellness).

In addition, we were not able to evaluate the relationship between illness severity and prayer for health, since NHIS does not include measures of disease severity. Due to the lack of detailed exposure data in NHIS, it is possible that residual confounding exists that we could not account for in our analyses. Nonetheless, the NHIS is an important survey, since it provides nationwide results on individuals’ health-related behavior, health-related perceptions, and health-seeking behavior.

Conclusions

Several surveys of patients^49–52 and practitioners^53–55 indicate that most patients would appreciate some discussion of religion as part of their primary care. In fact, some researchers²⁶ have proposed that patient satisfaction is greater if religion is incorporated into primary care, and researchers and educators have proposed ways to make spiritual assessment an integral part of the medical or social history.⁵⁶ Our findings that prayer for health is not only prevalent but is associated with a greater likelihood of primary care physician visits suggest that discussions of religious activity should be part of the social history for many primary care patients.