Examining Rural-Urban Differences in Fatalism and Information Overload: Data from Twelve NCI-designated Cancer Centers

Abstract

Background.

Rural populations experience a disproportionate cancer burden relative to urban populations. One possibility is that rural populations are more likely to hold counterproductive cancer beliefs such as fatalism and information overload that undermine prevention and screening behaviors.

Methods.

Between 2016 – 2020, twelve U.S. cancer centers surveyed adults in their service areas using online and in person survey instruments. Participants (N = 10,362) were designated as rural (n = 3,821) or urban (n = 6,541). All participants were 18 and older (M = 56.97, SD = 16.55), predominately non-Hispanic White (81%), and female (57%). Participants completed three items measuring cancer fatalism (“It seems like everything causes cancer,” “There’s not much you can do to lower your chances of getting cancer,” “When I think about cancer, I automatically think about death”) and one item measuring cancer information overload (“There are so many different recommendations about preventing cancer, it’s hard to know which ones to follow”).

Results.

Compared to urban residents, rural residents were more likely to believe that (1) everything causes cancer (OR = 1.29, 95% CI: 1.17–1.43), (2) prevention is not possible (OR = 1.34, 95% CI: 1.19–1.51), (3) there are too many different recommendations about cancer prevention (OR = 1.26, 95% CI: 1.13–1.41), and cancer is always fatal (OR = 1.21, 95% CI: 1.11–1.33).

Conclusions.

Compared to their urban counterparts, rural populations exhibited higher levels of cancer fatalism and cancer information overload.

Impact:

Future interventions targeting rural populations should account for higher levels of fatalism and information overload.

1. INTRODUCTION

In 2020, there are an estimated 1.8 million new cancer cases and 606,250 cancer deaths in the U.S.¹ Evidence shows that avoiding tobacco use, maintaining a healthy weight, eating a healthy diet, and staying physically active are among the most effective strategies for reducing cancer risk.² In addition, the early detection of certain cancer types (colorectum, breast, cervix and lung) through screening has the potential to reduce cancer mortality by detecting cancer in earlier and more treatable stages.³ Beliefs people hold about cancer may be barriers to practicing these cancer prevention behaviors.⁴

The Health Information National Trends Survey (HINTS) is a nationally representative data collection effort funded by the National Cancer Institute (NCI) that tracks public beliefs and perceptions about cancer.^5–6 From 2003 – 2019, HINTS has consistently demonstrated that a large number of U.S. adults hold potentially counterproductive beliefs about cancer prevention and screening behaviors.^4,6 For example, the 2019 HINTS survey demonstrated that 69.3% of U.S. adults agree that “It seems like everything causes cancer” and 71.1% agree that “There are so many recommendations about preventing cancer, it’s hard to know which ones to follow.”⁷ Researchers have demonstrated⁸ that the former is an indicator of fatalistic thinking – the belief that a person can do little to prevent or treat an illness⁹ – and the latter indicates information overload – feeling overwhelmed by the amount of information about a topic.¹⁰ Several studies have shown that both beliefs are negatively related to the performance of cancer prevention and screening behaviors.^4,10–15

Explicating the meaning and impact of these counterproductive beliefs is a priority for cancer research and implementation.^4,8,14,16 An emerging question is whether cancer beliefs vary along rural-urban or geographic divides.¹⁷ Cancer incidence and mortality vary along both divides with rural populations from particular geographic regions, such as the southeastern U.S. and Appalachia, carrying a disproportionate cancer burden.^17–21 This burden could foster, or be fostered by, underlying fatalism or information overload.²² Indeed, the interrelations among burden, belief, rurality, and geography are complex.²³ For instance, rural populations have significant access barriers to healthcare^24–26 which increases cancer burden and may leave rural populations feeling that they lack sufficient resources to manage health issues. Relatedly, fatalism and overload could reduce interest in and utilization of prevention infrastructure. Consistent with this idea, rural adults are less likely to engage in health behaviors (sufficient sleep, not smoking, no or moderate alcohol consumption, normal BMI, and physical activity) compared to adults residing in urban areas (27% vs. 31% reporting at least four of the five behaviors, respectively).²⁷ Thus, fatalism and overload may function as both a byproduct of resource inequities and a barrier to using resources.⁸

Why might rural populations exhibit higher levels of fatalism and overload? Research on transactional explanations of stress has demonstrated that when individuals have or perceive a lack of resources, they feel stressed and initiate a coping process wherein stressors are either reduced, revised, avoided and/or devalued.^28–32 If fatalism and information overload are more prominent in rural individuals and/or particular underserved populations, then it could be an indicator of attempts to cope with resource-related stress.²⁸ Fatalistic thinking reduces the value of having access to resources as “there’s not much you can do,” “everything causes cancer,” and cancer equates to “death.”^7–9 Likewise, information overload revises inaction into a response to over-abundant recommendations.⁸

To date, only three studies have examined whether rural/urban differences are related to cancer fatalism and information overload. An analysis of 2007 HINTS data found that rural adults had higher levels of three cancer beliefs (i.e., everything causes cancer, not much you can do, there are so many recommendations),²² however, a follow-up analysis of 2011–2017 HINTS data found no relationship.³³ Likewise, Vanderpool and colleagues³⁴ found no rural/urban difference in cancer beliefs among Appalachian populations in three states (Kentucky, Ohio, and Pennsylvania). The current study extends this research by examining differences in rural-urban cancer beliefs using data collected from 12 NCI geographic service areas, or catchment areas. This effort is part of NCI’s Population Health Assessment in Cancer Center Catchment Areas initiative which provided funding to cancer centers between 2016 and 2020 to “acquire, aggregate, and integrate data from multiple sources to better define and describe their catchment areas.”^17(141) The robust data collected across multiple regions provides a unique opportunity to test for rural-urban differences.

2. METHODS

2.1. Data source and survey methods

The current manuscript reports data collected as part of a multi-phase effort funded by NCI and implemented by twelve cancer centers. The primary goal of the research was to survey cancer center catchment areas to identify barriers and opportunities to facilitate cancer prevention and screening. Several working groups were also formed to lead projects focused on underserved populations; the current manuscript is a byproduct of the Rural Working Group which designed and implemented identical measures across all data collection sites with the goal of examining rural-urban differences. Data collection occurred from 2016 – 2020 with four centers – University of Pittsburgh Medical Center (UPMC), University of Kentucky (UK), Ohio State University (OSU), and Indiana University (IU) and eight centers – University of Utah (UU), Virginia Commonwealth University (VCU), University of Virginia (UVA), University of Minnesota (UM), University of Alabama at Birmingham (UAB), Oregon Health & Science University (OHSU), University of Kansas Cancer Center (KUCC), and the Fred Hutch/University of Washington (UW) Cancer Center Consortium. The surveys were completed by mail, telephone, or online. At each site, protocols were reviewed, approved, and monitored by local institutional review boards.

2.2. Study population

The study population (N = 10,362) represents 12 sites across the U.S. The sample size of each site, sampling designs, and catchment areas are reported in Table 1. Rural areas were over-sampled by most grantees to increase the representativeness of the rural population.

Table 1.

Summary survey design features of surveys implemented by the 12 grantees.

Cancer Institute Grantee	Probability Sample Design	Oversampling Strategy	Inclusion/Exclusion	Catchment Areas
Round 1 (Data collection 2016–2018)
Indiana University (IU) Melvin & Bren Simon Cancer Center (n= 970)	Address-based probability sample selected from IU Health system database, stratification by Race, Location, Sex, and Age (18–49, 50–75)	None	Age 18 and over residing in one of 34 counties with higher cancer mortality rates than the state	State of Indiana
Ohio State University Comprehensive Cancer Center (n=360)	Marketing Systems Group (MSG) sample	Ordered listing of addresses to target rural and Appalachian	Age 21–74, residing within Ohio	State of Ohio
University of Kentucky Markey Cancer Center (n=734)	Address-based sample obtained through Marketing Systems Group (MSG) (stratums 1–4)	Oversampled for RUCC 1–3 (urban) and RUCC 4–5, undersampled RUCC 6–7	Age 18 and older, Appalachian Kentucky resident, English speaking	54 counties in KY designated as Appalachian
University of Pittsburgh Hillman Cancer Institute (n=981)	Dual frame RDD sample	None	Age 18 or older; residing within the Hillman Cancer Institute catchment area	29 counties in western PA
Round 2 (Data collection 2018–2020)
Fred Hutch/University of Washington Cancer Consortium (n=1,052)	Stratified random sample of adults ages 18 and older. Mailing addresses purchased from commercial vendor; mailed personalized survey invitation to an online survey. Paper copy of survey mailed upon request.	None	Age 18 or older; residing within the Fred Hutchinson Cancer Research Center catchment area	13 counties in western Washington State.
University of Virginia Emily Couric Clinical Cancer Center (n=598)	Stratified address-based sample; and Random Digit Dialing (Cell)	None	Age 18 or older; residing within the Emily Couric Clinical Cancer Ctr catchment area	Western Virginia
O’Neal Comprehensive Cancer Center at the University of Alabama at Birmingham (n=751)	Stratified address-based random sample,	Oversampled for high-risk individuals	Surveys targeted at adults over age 50 years.	State of Alabama
The University of Kansas Cancer Center (n=1,364)	Stratified random sample from three health systems and simple random sample from one health system files	Two strata based on RUCC code and 5 strata based on county population size with oversampling in RUCC 7–9 and the smaller counties.	Age 18 and older, seen at the academic health center in the prior year.	State of Kansas and 18 counties in Missouri
Oregon Health & Science University Knight Cancer Institute (n=773)	Address-based sample	Two strata based on RUCC code (1–3 and 4–9) with oversampling in RUCC 4–9 counties.	Age 18 or older; residing within the Knight Cancer Institute catchment area	State of Oregon
University of Minnesota Masonic Cancer Center (n=717)	Survey Sampling International (SSI) mail survey	5 sampling regions, four rural, 1 urban. Oversampled rural region for a 1.7:1 ratio of ‘rural’ to urban sample	Age 18 and older, 1 adult per household, individual with nearest birthday, residing within the Masonic Cancer Center catchment area	State of Minnesota
University of Utah Huntsman Cancer Institute (n=1,199)	Simple random sample	Two strata based on RUCC code (1–3 and 4–9) with oversampling in RUCC 4–9 counties.	Age 18 or older, residing in Utah	State of Utah
Virginia Commonwealth University Massey Cancer Center (n=895)	Address-based probability sample purchased from commercial vendor. Mai-based survey (a survey link was provided for those who prefer online)	None	Age 18 or older; residing within the VCU Massey Cancer Center catchment area	Eastern Virginia

Note: The analysis was restricted to the probability sample components of the grantees’ studies. ABS: address-based sample.

2.3. Outcome measures

Four HINTS survey items related to cancer beliefs and perceptions were used as outcome measures: (1) “It seems like everything causes cancer;” (2) “There’s not much you can do to lower your chances of getting cancer;” (3) “There are so many different recommendations about preventing cancer, it’s hard to know which ones to follow;” and (4) “When I think about cancer, I automatically think about death.”⁵ Response options were identical across all four items: 1 = strongly agree, 2 = somewhat agree, 3 = somewhat disagree, and 4 = strongly disagree. HINTS has included three of these items (items 1–3) in twelve of the thirteen data collection cycles whereas the fourth item has only been included in three cycles.⁷ To better understand typicality for these items, we examined the response patterns for each cycle. Across all data collection cycles, the median option selected for each item is: item 1 (2 = somewhat agree), item 2 (3 = somewhat disagree), item 3 (2 = somewhat agree), and item 4 (2 = somewhat agree).

Past research has occasionally combined these items into a single composite score,³³ however, other work has demonstrated that the four items represent three distinct factors: prevention-focused cancer fatalism (items 1 and 2), cancer information overload (item 3), and treatment-focused cancer fatalism (item 4).⁸ Confirmatory factor analysis reported multiple sub-dimensions within the items, and, similar to past studies,³³ the items showed a poor reliability score when combined (Cronbach’s alpha <.60), thus a single composite score was not used in the analysis. Prior to modeling, outcome measures were re-coded to 1 = agree (by combining strongly agree and somewhat agree) and 2 = disagree (by combining strongly disagree and somewhat disagree).

2.4. Covariates

Using validated questionnaires from population-based surveys (HINTS, Behavioral Risk Factor Surveillance System, National Health Interview Survey), 12 cancer centers’ surveys assessed a range of socio-demographics and behavioral variables including: age, gender, race/ethnicity, income, education, employment, primary source of health care coverage, marital status, cost barrier to medical care, and smoking status. Among the core measures adopted by nearly all the cancer centers, these measures were considered the most useful in understanding cancer beliefs and perceptions.³⁴

2.5. Predictor measures

The key predictor variable for this analysis was rural/urban status. To create the rurality/urbanicity variable, we linked FIPS codes with the 2013 rural-urban continuum codes (RUCC) developed by the United States Department of Agriculture (USDA) with nine levels. We dichotomized the variable as urban for RUCC 1–3 and rural for RUCC 4–9.³⁵

2.6. Statistical analysis

We used bivariate cross tabulations and chi-square tests to examine whether differences in demographics and cancer beliefs between the rural and urban areas were evident. All significance tests were two-sided with a significance level of .05. We also examined whether the interaction terms between the rurality/urbanicity variable and other predictors had significant effects on outcome measures.

We then conducted a series of logistic regression models to identify predictors that explained the most variance in each of the cancer belief outcome measures by performing a backward variable selection approach. Each model started with all candidate variables; non-significant variables were eliminated until all remaining variables had p-values smaller than 0.01. Site was retained in all models to demonstrate robustness of the models across all twelve sites. Site was selected instead of Census region because the former is a more precise representation of the sampling differences.

Logistic regression was used to obtain odds ratios and 95% confidence intervals for the association between rural residence and each cancer belief separately adjusting for potential confounding such as age, race/ethnicity, sex, income, education, marital status, smoking, health care coverage, and employment.

3. RESULTS

Across all 12 centers, data were obtained from 3,821 individuals residing in rural, RUCC 4–9, areas and 6,541 individuals residing in urban, RUCC 1–3, areas (see Table 2 for differences by rural/urban geographic location). The percent of rural respondents ranged from 5.7% (Fred Hutchinson/UW Cancer Consortium) to 82.6% (UK). The sample was 82.7% non-Hispanic white, though centers from the South (UAB and VCU) had a higher percent of non-Hispanic Black participants, particularly in rural locations (39.4% and 22.0% respectively) than other centers (see Table 3 for differences by region). While it was not within the scope of this paper to evaluate differences in actual screening practices, it was important to have an understanding of what percent of total respondents were of an age or gender to be eligible for cancer screenings. Of all participants, 21.9% were eligible for cervical cancer screening (age 21 to 65 and female), 24.5% were eligible for mammography (over age 50 and female), and 49.3were eligible for colorectal cancer screening (over age 50). While 9.6% reported that they were current smokers, we did not collect adequate data to determine eligibility for low-dose CT lung cancer screening. Across nearly all demographic measures, significant differences between rural and urban respondents were observed. Compared to urban respondents, those from rural areas were older, white, more likely to be female, lower income, less educated, more likely to be retired and receiving Medicare, married, currently smoker, and classified as obese according to their BMI.

Table 2.

Demographic characteristics for all sites by rural vs. urban area.

	Urban (RUCC 1–3) n=6541		Rural (RUCC 4–9) n=3821		p-value (chi-square)
	n	%	n	%
Age (years)
18–34	795	13.31	309	9.46	<.0001
35–44	879	14.72	326	9.98
45–64	2277	38.12	1176	35.99
65+	2022	33.85	1456	44.57
Sex
Male	2815	43.52	1551	41.07	0.016
female	3654	56.48	2225	58.93
Race / Ethnicity
Non-Hispanic White	5162	80.29	3258	86.33	<.0001
Non-Hispanic Black	541	8.42	246	6.52
Hispanic	212	3.30	68	1.80
Other	514	7.99	202	5.35
Income
Less than $20,000	700	13.18	737	21.88	<.0001
$20,000 – $49,999	1090	20.52	1023	30.37
$50,000 – $99,999	1813	34.14	1053	31.25
Greater than $100,000	1708	32.16	556	16.50
Education
High school or less	1199	18.66	1268	33.63	<.0001
Post-high school training	1744	27.14	1211	32.12
College graduate	3483	54.21	1291	34.24
Health Care Coverage
Uninsured	319	5.38	172	5.51	<.0001
Medicare	1583	26.72	1231	39.46
Medicaid	291	4.91	204	6.54
Private Insured & Employee based	3394	57.28	1313	42.08
Other	338	5.71	200	6.41
Marital Status
Married /Living as Married	3603	62.95	2292	64.55	<.0001
Divorced, Widowed, Single	2121	37.06	1259	35.45
Trouble paying for medical care
Yes	652	11.24	427	11.81	0.40
No	5150	88.76	3188	88.19
Body Mass Index
Normal/Underweight (<25 kg/m2)	1633	33.92	781	27.89	<.0001
Overweight (25–29.9 kg/m2)	1615	33.54	1003	35.82
Obese (>30 kg/m2)	1567	32.54	1016	36.29
Smoking Status
Current	609	9.55	477	12.87	<.0001
Former/ Never	5771	90.45	3228	87.13
Employment
Employed	2980	52.21	1302	36.91	<.0001
Not employed	2728	47.79	2225	63.09

Table 3.

Demographic characteristics for all sites by Census regions.

	Northeast n=982		Midwest n=3411		West n=3030		South n=3135		p-value (chi-square)
	n	%	n	%	n	%	n	%
Age (years)
18–34	176	18.07	204	8.49	508	16.83	227	7.57	<.0001
35–44	123	12.63	265	11.03	556	18.42	277	9.24
45–64	384	39.43	826	34.39	994	32.94	1306	43.58
65+	291	29.88	1107	46.09	960	31.81	1187	39.61
Sex
Male	461	46.95	1310	39.10	1346	44.66	1312	42.64	<.0001
Female	521	53.06	2040	60.90	1668	55.34	1765	57.36
Race / Ethnicity
NH White	900	92.98	2844	84.92	2444	81.25	2377	78.27	<.0001
NH Black	35	3.62	305	9.11	22	0.73	430	14.16
Hispanic	-	-	76	2.27	140	4.65	69	2.27
Other	33	3.41	124	3.70	402	13.36	161	5.30
Income
Less than 20,000	168	20.72	526	16.57	169	7.89	610	22.73	<.0001
20,000 – 49,999	200	24.66	881	27.76	387	18.06	690	25.71
50,000 – 99,999	257	31.69	1093	34.44	802	37.42	741	27.61
100,000+	186	22.93	674	21.24	785	36.63	643	23.96
Education
High school or less	367	37.60	829	24.95	370	12.29	964	31.70	<.0001
Post-high school trainings	257	26.33	1025	30.85	909	30.19	809	26.60
College graduate or higher	352	36.07	1468	44.19	1732	57.52	1268	41.70
Health Care Coverage
Uninsured	41	4.24	145	6.27	173	5.78	140	4.77	<.0001
Medicare	208	21.53	917	39.65	751	25.08	998	34.04
Medicaid	74	7.66	86	3.72	136	4.54	219	7.47
Private Insured & Employee based	599	62.01	1052	45.48	1752	58.50	1364	46.52
Other	44	4.55	113	4.89	183	6.11	211	7.20
Marital Status
Married/Living as married	-	-	2128	63.43	2038	67.66	1820	59.44	<.0001
Not married	-	-	1227	36.57	974	32.34	1242	40.56
Trouble paying for medical care
Yes	97	9.90	441	13.26	256	11.47	298	9.74	<.0001
No	883	90.10	2885	86.74	1975	88.53	2763	90.26
Body Mass Index
Normal/Underweight (<25 kg/m2)	-	-	618	30.40	1123	37.81	714	25.72	<.0001
Overweight (25–29.9 kg/m2)	-	-	688	33.84	951	32.02	1040	37.46
Obese (>30 kg/m2)	-	-	727	35.76	896	30.17	1022	36.82
Smoking Status
Current Smoker	159	16.70	409	12.25	171	5.81	366	12.08	<.0001
Former Smoker or never smoke	793	83.30	2930	87.75	2774	94.19	2665	87.92
Employment
Employed	-	-	1414	42.44	1626	54.04	1299	42.62	<.0001
Not employed	-	-	1918	57.56	1383	45.96	1749	57.38
Urban / Rural Status
Urban (RUCC 1–3)	741	75.54	1742	51.55	2340	77.38	1718	57.69	<.0001
Rural (RUCC 4–9)	240	24.46	1637	48.45	684	22.62	1260	42.31

Note. Census regions included the following sites: Northeast (UPMC, VCU, UVA), Midwest (OSU, IU, UM), West (UU, OHSU, UW), and South (KU, UAB, KUCC). Not all demographic variables were measured by sites in the Northeast region.

For the four beliefs, the median response options were somewhat agree (everything causes cancer), somewhat disagree (“there’s not much you can do to prevent cancer”), somewhat agree (“too many recommendations”), and somewhat agree (“think about death”) (see Table 4). Compared to urban participants, rural participants were significantly more likely to respond that they agree or strongly agree with all four beliefs (p<.0001).

Table 4.

Responses to belief questions by rural or urban residence.

Belief	Overall N=10558		Urban (RUCC 1–3) n=6541		Rural (RUCC 4–9) n=3821		p-value
	n	%	n	%	n	%
Everything Causes Cancer
Strongly/Somewhat agree	5755	56.53	3366	53.34	2273	61.52	<.0001
Strongly/Somewhat disagree	4426	43.47	2944	46.66	1422	38.48
Strongly agree	1432	14.07	796	12.62	595	16.11	<.0001
Somewhat agree	4323	42.46	2570	40.73	1678	45.41
Somewhat disagree	2672	26.24	1746	27.67	891	24.11
Strongly disagree	1754	17.23	1198	18.98	531	14.37
Not Much You Can Do
Strongly/Somewhat agree	2217	23.24	1182	20.1	1001	28.15	<.0001
Strongly/Somewhat disagree	7322	76.76	4698	79.9	2555	71.85
Strongly agree	491	5.15	253	4.3	228	6.41	<.0001
Somewhat agree	1726	18.09	929	15.8	773	21.74
Somewhat disagree	3817	40.01	2372	40.34	1409	39.62
Strongly disagree	3505	36.74	2326	39.56	1146	32.23
Too Many Recommendations
Strongly/Somewhat agree	7487	73.65	4470	71.03	2878	77.87	<.0001
Strongly/Somewhat disagree	2678	26.35	1823	28.97	818	22.13
Strongly agree	2299	22.62	1290	20.5	961	26	<.0001
Somewhat agree	5188	51.04	3180	50.53	1917	51.87
Somewhat disagree	1855	18.25	1242	19.74	585	15.83
Strongly disagree	823	8.1	581	9.23	233	6.3
Think About Death
Strongly/Somewhat agree	5594	54.84	3297	52.14	2186	59.05	<.0001
Strongly/Somewhat disagree	4607	45.16	3026	47.86	1516	40.95
Strongly agree	1859	18.22	996	15.75	808	21.83	<.0001
Somewhat agree	3735	36.61	2301	36.39	1378	37.22
Somewhat disagree	2933	28.75	1891	29.91	996	26.9
Strongly disagree	1674	16.41	1135	17.95	520	14.05

Note. Significant differences for dichotomized and four-level versions of belief variables. Dichotomized versions are used for subsequent multivariate analyses to ease interpretation.

These associations remained even after accounting for other demographic and lifestyle variables that were significant in the bivariate models (see Table 5). In the multivariate model, we adjusted for grantee site to account for differences in sampling strategies and site characteristics. A number of the covariates remained significant in the multivariate models. In particular, lower educational attainment (high school vs. college or more) and smoking were significantly associated with greater likelihood of agreeing with the four cancer beliefs statements. Other covariates that remained significant in at least one of the four models included age, sex, race/ethnicity, income, trouble paying for medical care, employment and marital status, though none of these was significant across all four cancer beliefs outcome measures.

Table 5.

Multivariate Models.

Everything Causes Cancer
Variable	Level	OR (95% CI)	Trend p-value
Geographic location	Rural (RUCC 4–9)	1.29 (1.17–1.43)	<0.0001
	Urban (RUCC 1–3)	Reference
Age (years)	65+	0.38 (0.31–0.47)	<0.0001
	35–64	0.65 (0.55–0.76)
	18–34	Reference
Sex	Female	1.22 (1.11–1.34)	<0.0001
	Male	Reference
Income ($)	< 20,000	1.38 (1.11–1.71)	0.003
	20,000 – 49,999	1.21 (1.03–1.42)
	50,000 – 99,999	1.25 (1.10–1.41)
	>=100,000	Reference
Education (highest completed)	<= High School	2.03 (1.76–2.34)	<0.0001
	Post-high school	1.64 (1.46–1.83)
	>= College	Reference
Insurance Status	Uninsured	0.76 (0.60–0.95)	0.005
	Medicare	0.91 (0.76–1.08)
	Medicaid	0.70 (0.55–0.89)
	Other	1.10 (0.88–1.37)
	Private Insured & Employee based	Reference
Smoking Status	Current Smoker	1.65 (1.38–1.96)	<0.0001
	Former or Never Smoke	Reference
Trouble paying for medical care	Yes	1.24 (1.05–1.46)	0.01
	No	Reference
Marital Status	Not Married	0.90 (0.81–1.01)	0.08
	Married/Living as married	Reference
Employment	Not Employed	0.90 (0.80-.02)	0.09
	Employed	Reference
Site	All Sites	0.99 (0.98–1.01)	0.33
	Fred Hutchinson	Reference
Not Much You Can Do
Variable	Level	OR (95% CI)	Trend p-value
Geographic location	Rural (RUCC 4–9)	1.22 (1.09–1.36)	0.0005
	Urban (RUCC 1–3)	Reference
Income ($)	< 20,000	2.07 (1.72–2.49)	<0.0001
	20,000 – 49,999	1.53 (1.30–1.81)
	50,000 – 99,999	1.22 (1.05–1.43)
	>=100,000	Reference
Education (highest completed)	<= High School	2.84 (2.45–3.28)	<0.0001
	Post-high school	1.67 (1.46–1.91)
	>= College	Reference
Smoking Status	Current Smoker	1.37 (1.17–1.60)	<0.0001
	Former or Never Smoke	Reference
Site	All Sites	1.01 (0.99–1.02)	0.31
	Fred Hutchinson	Reference
Too Many Recommendations
Variable	Level	OR (95% CI)	Trend p-value
Geographic location	Rural (RUCC 4–9)	1.26 (1.13–1.41)	<0.0001
	Urban (RUCC 1–3)	Reference
Sex	Female	0.89 (0.80–0.98)	0.02
	Male	Reference
Income ($)	< 20,000	1.06 (0.87–1.30)	0.01
	20,000 – 49,999	1.23 (1.05–1.44)
	50,000 – 99,999	1.20 (1.06–1.37)
	>=100,000	Reference
Education (highest completed)	<= High School	2.11 (1.80–2.47)	<0.0001
	Post-high school	1.56 (1.38–1.76)
	>= College	Reference
Smoking Status	Current Smoker	1.38 (1.14–1.68)	0.0008
	Former or Never Smoke	Reference
Marital Status	Not Married	0.88 (0.78–0.99)	0.03
	Married/Living as married	Reference
Site	All Sites	0.99 (0.98–1.01)	0.39
	Fred Hutchinson	Reference
Think About Death
Variable	Level	OR (95% CI)	Trend p-value
Geographic location	Rural (RUCC 4–9)	1.09 (0.99–1.21)	<0.0001
	Urban (RUCC 1–3)	Reference
Age (years)	65+	0.80 (0.67–0.96)	0.05
	35–64	0.90 (0.78–1.02)
	18–34	Reference
Sex	Female	0.91 (0.83–0.99)	0.02
	Male	Reference
Race/Ethnicity	Other	0.97 (0.82–1.15)	0.0006
	Hispanic	1.46 (1.12–1.91)
	Black	1.31 (1.11–1.54)
	Non-Hispanic White	Reference
Education (highest completed)	<= High School	1.74 (1.55–1.96)	<0.0001
	Post-high school	1.32 (1.20–1.46)
	>= College	Reference
Insurance Status	Uninsured	0.82 (0.66.−1.00)	0.005
	Medicare	0.91 (0.79–1.05)
	Medicaid	1.31 (1.08–1.61)
	Other	0.94 (0.77–1.14)
	Private Insured & Employee based	Reference
Smoking Status	Current Smoker	1.33 (1.15–1.54)	0.0002
	Former or Never Smoke	Reference
Trouble paying for medical care	Yes	1.45 (1.25–1.68)	<0.0001
	No	Reference
Site	All Sites	1.00 (0.99–1.01)	0.62
	Fred Hutchinson	Reference

4. DISCUSSION

Past research has demonstrated that many U.S. adults express fatalistic thoughts and cancer information overload.^4,10,16 The current study extended this research by demonstrating that these beliefs were more common in rural than in urban populations. Three prior studies have examined this link, but produced conflicting results.^22,33–34 The results of the present study mirror earlier findings from HINTS²² – which found rural adults were more likely to exhibit prevent-focused cancer fatalism and cancer information overload – and extend those results by also examining treatment-focused cancer fatalism in a broader, more diverse cross-section of rural communities across 12 catchment areas.

The findings have implications for researchers and practitioners working to understand and address rural health disparities. Compared to urban populations, rural populations have significant access, education, and income deficits;¹⁷ deficits which have also been linked to increased fatalism and information overload.^8–9,15,36 Indeed, the consistently strong agreement with fatalistic/overload statements and lower educational attainment found in this study corroborate these previous findings. The revelation that rural residents exhibit greater fatalism and overload further complicates the situation as interventions may need to engage structural, socio-demographic, and psychosocial factors.^37–39 Alternatively, researchers could examine whether developing materials for low-education populations is an effective means for addressing fatalism/overload in rural populations or if rural-specific, low-education approaches yield greater impact.

Identifying communication approaches tailored to rural residents that reduce cancer fatalism and information overload beliefs, or that are effective with fatalistic or overloaded populations, is needed as part of cancer prevention and screening efforts.^10,40 An effective – communication strategy likely needs to address a contradiction at the heart of these findings. On the one hand, rural adults seem overwhelmed by cancer as they perceive too many recommendations and feel that everything causes cancer. On the other hand, participants argue that there is nothing they can do and that cancer inevitably leads to death. These findings appear to conflict as it seems like cancer communication is providing too much and too little information. Since cancer communicators must convey recommendations there seems to be two basic strategies to consider: (1) finding a way to fluidly communicate multiple recommendations at once⁴¹ or (2) utilizing a technique that focuses on one recommendation at a time. Concerning the latter, researchers might consider the value of sequential request strategies. A sequential request occurs when persuasive attempts are separated across time.⁴² In compliance research, sequential requests often occur relatively close together; for example, an initial request to complete a survey about a cancer charity is followed by a second request to donate money to the charity.⁴³ But sequential requests do not have to be temporally close. Rather, they have to be topically similar so that positive response to the initial request suggests continued positive response to future requests.^42–44 From a persuasion standpoint, sequential requests trigger shifts in self-perception.⁴⁴ People use their own behavior as a guide for how to act; thus, getting someone to adopt a single preventive behavior increases the likelihood that they will initiate other, related behaviors.

From a theoretical standpoint, a key question is explicating why rural populations are more likely to exhibit fatalistic thinking and cancer information overload. One possibility is that these cancer beliefs are a form of stress-induced coping. Transactional explanations of stress posit that insufficient resources lead individuals to enact coping beliefs to manage stress.^28–29 The dominant theory from that area – psychological stress and coping theory – postulates that one outcome of coping is a cognitive reappraisal wherein individuals reduce, revise, or avoid the stressor to manage the situation.^30–32 For example, Wang and colleagues⁴⁵ conducted a review of barriers to colorectal cancer screening in rural populations. They identified that major barriers to screening were lack of access to specialists, distance to healthcare providers, screening cost, and lack of insurance, and that rural populations appeared to be more fatalistic. The findings of the current study are consistent with this logic as populations with fewer resources (rural adults) are more likely to reduce (fatalism) or revise (overload) the situation. Other research has demonstrated that individuals high in fatalism and overload are more likely to engage in cancer information avoidance.⁴⁶

Combined, these results suggest a model where resource scarcity increases stress which, in turn, triggers fatalism, information overload, avoidance, and, ultimately, reduced adherence to cancer prevention recommendations. A rigorous test of that model would require a longitudinal approach to track the relationship between resource scarcity, stress, cancer beliefs, and adherence, but researchers could also focus on individual relationships (stress increases fatalistic thinking) as a way to advance the model or to continue examining how different populations utilize fatalism/overload to make sense of their situation. For example, Drew and Schoenberg found that rural women engaged in fatalistic thinking to grapple with inadequate access to resources, reinforce self-reliance, and engage issues of privacy.³⁷ Likewise, Keeley and colleagues utilized in-depth interviews with 96 U.S. adults to identify functions of fatalistic thinking in health contexts. They found that fatalistic statements served four functions: stress relief, uncertainty management, sense making, and face saving.⁴⁷

Another important avenue of research is investigating possible communication origins of fatalism and overload. A key resource in managing health is access to quality information.⁴⁸ Health information can come from a variety of sources, including providers, public health campaigns, personal networks, and both traditional and new media. Examining the relationship among communication resources, fatalism, and information overload is useful in that these cancer beliefs seem to indirectly (fatalism) and directly (information overload) respond to the information environment.^49–53 For example, Paige and colleagues⁵⁵ found that individuals who report challenges in health information seeking are more likely to exhibit fatalistic thinking. Researchers should continue to examine whether rural and urban populations have significantly different information access and quality,^22,51,56 and if these communication disparities are related to stress, fatalism, overload.

Finally, researchers should work to rectify the findings of this study with other, related lines of research on rural-urban differences. The notion that rural adults have higher levels of overload parallels older research programs that demonstrated rural individuals were more likely to experience stimulus overload from exposure to complex stimuli.⁵⁷ A lack of background knowledge was hypothesized as a possible explanatory variable for this finding. It also parallels research documenting increased overload in other realms of rural life such as the burden of managing multiple jobs faced by rural women.⁵⁸ Role overload places stress on multiple aspects of rural life and initiates attempts to manage or cope with stressors. Researchers could explore intersections with these parallel lines including whether lack of background knowledge or other forms of overload are driving forces behind cancer information overload.

4.1. Limitations

A number of limitations need to be acknowledged. First, the current study utilized four single-item measures to assess fatalism and information overload. This is consistent with HINTS research – which allows for direct comparison – but multi-item scales exist for both constructs.^8,59 Second, the items focused on cancer-related beliefs, yet it is possible that fatalism and information overload are present for multiple health issues, or may affect risk of other health conditions with similar risk factors, such as heart disease and diabetes. Third, the surveys were cross-sectional and therefore the directionality of relationships is suggested rather than confirmed. Fourth, the data represents 12 catchment areas, but that is still a subset of all rural populations in the U.S. Fifth, the analysis does not account for larger, multilevel influences on cancer-related behaviors and beliefs, such as the presence or absence of health programs or infrastructure. Sixth, transactional explanations of stress provide a useful framework for theorizing relationships between rurality, fatalism, and overload, but researchers are still working to explicate key questions about coping response, including the creation of meaningful taxonomies of coping-related stress that account for all past findings.²⁸ Seventh, sampling frames differed by site and not all sites oversampled or stratified for rural/urban differences which resulted in unequal sample sizes for this comparison. That has a small impact on power, but rural/urban differences were observed for all four beliefs, controlling for site, which minimizes this concern.

4.2. Conclusion

Compared to urban populations, rural populations in the U.S. are more likely to exhibit cancer-related fatalism and information overload. Cancer centers and public health programs should consider whether existing resource allocation might cultivate these beliefs as well as how to position messages in light of fatalism and overload. Researchers have found that this type of thinking might be part of a larger cultural framework that fosters self-reliance and coping;³⁷ this suggests that health communicators could engage rural populations by acknowledging these feelings and working to identify new or existing resources positioned in terms that are meaningful for those grappling with fatalism and overload.