Rural-Urban Differences in Health Behaviors and Implications for Health Status among US Cancer Survivors

Abstract

Purpose

Rural US adults have increased risk of poor outcomes after cancer, including increased cancer mortality. Rural-urban differences in health behaviors have been identified in the general population and may contribute to cancer health disparities, but have not yet been examined among US survivors. We examined rural-urban differences in health behaviors among cancer survivors and associations with self-reported health and health-related unemployment.

Methods

We identified rural (n=1,642) and urban (n=6,162) survivors from the cross-sectional National Health Interview Survey (2006–2010) and calculated the prevalence of smoking, physical activity, overweight/obesity, and alcohol consumption. Multivariable models were used to examine the associations of fair/poor health and health-related unemployment with health behaviors and rural-urban residence.

Results

The prevalence of fair/poor health (rural 36.7%, urban 26.6%), health-related unemployment (rural 18.5%, urban 10.6%), smoking (rural 25.3%, urban 15.8%), and physical inactivity (rural 50.7%, urban 38.7%) was significantly higher in rural survivors (all p<.05); alcohol consumption was lower (rural 46.3%, urban 58.6%), and there were no significant differences in overweight/obesity (rural 65.4%, urban 62.6%). All health behaviors were significantly associated with fair/poor health and health-related unemployment in both univariate and multivariable models. After adjustment for behaviors, rural survivors remained more likely than urban survivors to report fair/poor health (OR= 1.21, 95%CI 1.03–1.43) and health-related unemployment (OR= 1.49, 95%CI 1.18–1.88).

Conclusions

Rural survivors may need tailored, accessible health promotion interventions to address health compromising behaviors and improve outcomes after cancer.

Addressing health disparities associated with rural residence has been identified as an important public health priority in the United States (US) both generally [1, 2] and for cancer outcomes specifically [3]. Differences in health behaviors may contribute to poorer health among rural residents. Adults living in rural areas are more likely to be overweight or obese and physically inactive or insufficiently active compared to urban residents [4–9]. Rural residence has also been linked with higher smoking prevalence [10–12]. Rural-urban differences in alcohol use are present at both extremes; rural residents are more likely to be current or lifetime abstainers, but rural residents who drink are more likely to have a current alcohol disorder and to exceed recommended daily limits[13].

Numerous studies have documented that many cancer survivors do not meet healthy lifestyle recommendations [14–20]. Maintenance of a healthy weight and physical activity have been linked with better outcomes for cancer survivors, including lower risk of recurrence and improved survival in several studies [for reviews see 21–24]. Continued smoking after cancer diagnosis has been associated with poorer outcomes, particularly for smoking related-cancers [e.g., 25–27]. Although alcohol use is also associated with increased risk of several cancers[28], less is known about the risks and benefits of alcohol consumption after cancer[21, 24]. If survivors choose to consume alcohol, they are generally advised to drink no more than a moderate amount [21, 29]. Meeting health behavior recommendations and exercise are also associated with better health-related quality of life after cancer [14, 30, 31].

It is not known whether differences in health behaviors observed between rural and urban adults without cancer are also present among cancer survivors and if so, whether these differences in modifiable risk factors are related to poorer health outcomes among rural survivors. There are an estimated 2.8 million cancer survivors in the US who reside in rural areas and are at increased risk for poor health and increased psychological distress relative to their urban counterparts [32]. In a study of Canadian rural and small town breast cancer survivors, Vallanace and colleagues [33] found that 31% met exercise recommendations, a prevalence very similar to other predominantly urban samples. In contrast, Rogers and colleagues [34] reported that only 19% of their sample of rural US breast cancer survivors met exercise recommendations. One Australian study found that urban colorectal cancer survivors were significantly less likely to maintain or achieve sufficient physical activity post-diagnosis compared to rural survivors [35]. With this exception, very few studies have directly compared health behaviors between rural and urban cancer survivors. Many interventions have been designed to positively influence health behaviors after cancer[19, 23]. If health behaviors contribute significantly to rural-urban survivorship disparities, then delivery of behavioral interventions targeting these modifiable risk factors to rural cancer survivors should be prioritized.

Thus, the purpose of our study was to 1) examine rural-urban differences in health behaviors in a population-based sample of cancer survivors, 2) assess whether health behaviors were associated with fair or poor self-reported health and unemployment due to health, and 3) evaluate whether rural-urban differences in health status persisted after controlling for health behaviors. Relative to urban survivors, we hypothesized that the prevalence of smoking, physical inactivity, and overweight/obesity would be higher among rural cancer survivors, and alcohol use lower. We also hypothesized that these behaviors would be independently associated with self-reported health and unemployment due to health. Finally, we hypothesized that controlling for health behaviors would reduce or eliminate the effect of rural-urban residence on self-reported health and unemployment due to health.

Methods

Setting

We conducted a cross-sectional study using data from the 2006–2010 National Health Interview Survey (NHIS), a continuous nationally representative survey of the non-institutionalized, civilian US population [36]. The NHIS is conducted by the National Center for Health Statistics (NCHS), which is part of the Centers for Disease Control and Prevention (CDC). Overrepresentation of African Americans, Asians, and Hispanics is accomplished through the use of a complex sampling design. Both the questions and sampling design were identical from 2006–2010, so we pooled the data across these years to maximize the number of rural cancer survivors. The possibility of duplicate participation across years is extremely unlikely, thus yearly samples are considered to be independent. Trained US Census Bureau interviewers administer the survey in participants' homes. The conditional response rates (number completed divided by the number eligible) for the sample adult survey ranged from 74.2–81.4%, with final response rates ranging from 60.8%– 70.8% [36].

Participants

We included all cancer survivors, defined as adults (18 years or older) who reported ever being told by a doctor or health care professional that they had cancer. Consistent with other NHIS-based studies of cancer survivors[37, 38] and Surveillance Epidemiology and End Results (SEER) practices, we excluded individuals reporting exclusively “unknown” skin or non-melanoma skin cancers.

Variables

Health behaviors

Self-reported health behaviors included leisure-time physical activity, alcohol use, smoking status, and maintenance of a healthy body weight, as assessed with NHIS questions. To maintain consistency with the 2008 federal physical activity guidelines [39], respondents were categorized as inactive (no leisure-time aerobic activity that lasted at least 10 minutes), insufficiently active (aerobic activities for at least 10 minutes but less than 150 minutes per week), or sufficiently active (moderate-intensity leisure-time physical activity at least 150 minutes or vigorous-intensity leisure-time physical activity at least 75 minutes per week, or an equivalent combination). Alcohol use categories over the lifetime included: abstainer (less than 12 drinks ever), former (12 or more drinks ever, but none in the past year), and current (at least one drink in the past year). Current alcohol use was sub-categorized as infrequent (1–11 drinks in past year), light (≤3 drinks per week in past year), moderate (4–14 drinks per week for men and 4–7 drinks per week for women), and heavy (more than 14 drinks per week for men, and more than 7 drinks per week for women). Lifetime smoking status was categorized as never (less than 100 cigarettes total), past smoker (at least 100 cigarettes total, but not currently smoking), and current smoker (smokes some days or every day). Current body mass index was calculated based on self-reported height and weight. Following the definitions used by the CDC[40], respondents were categorized as low or normal weight (BMI < 25.0 kg/m²), overweight (BMI 25.0–29.9 kg/m²), or obese (BMI 30.0 kg/m² or more).

Health Status

Respondents rated their current health as excellent, very good, good, fair, or poor, based on a single item derived from the Medical Outcomes Short Form Survey (SF-36)[41]. Respondents who reported being currently unemployed and specified that it was due to health reasons, or those who reported being disabled, were classified as unemployed due to health reasons.

Rural-urban residence

We defined rural-urban residence status based on the US Office of Management and Budget's 2003 Rural/Urban Continuum Codes(RUCC) Codes [42]. These codes distinguish metropolitan counties by population size and nonmetropolitan counties by degree of urbanization and adjacency to metropolitan areas. Urban residence was classified as counties encompassing metropolitan statistical areas (RUCC codes 1–3), and rural residence included counties adjacent or not adjacent to metropolitan areas (RUCC codes 4–9).

Socio-Demographics and Clinical Variables

Socio-demographic variables included age, gender, geographic region, race/ethnicity, marital status, education, and health insurance coverage. Geographic areas included the four US census regions (Northeast, Midwest, South, and West). We categorized education as less than high school, high school graduate or general equivalency diploma, or some college or more. Health insurance coverage was categorized as public only, private with or without public, or neither. Public only health insurance included Medicare only, Medicaid, military, other government health care coverage, and other state-sponsored health care. Clinical variables included number of cancers (1 or >1) and time since first cancer diagnosis (in years). We recoded cancer type to group esophagus, liver, lung, pancreas, or stomach cancers as “short-survival” cancers (all have a 5-year relative survival of less than 25%[43]) and cervix, uterus, and ovary as gynecologic cancers.

Missing data

Data were missing for less than 1% of participants for the demographic characteristics, cancer variables, self-reported health, and employment. Rates of missing data for the health behaviors were 1% (smoking), 3% (alcohol use), 4% (BMI), and 8% (physical activity). Only participants with no missing data were included in the analyses.

Statistical Analysis

Analyses were conducted using SAS, version 9.2 (SAS Institute, Inc., Cary, NC), and p<0.05 was considered statistically significant. All analyses took into account the NHIS complex survey design and weighted sampling probabilities [36]. We conducted a series of logistic regression analyses to identify rural-urban differences in health behaviors (BMI, smoking, physical activity, and alcohol use) both unadjusted and adjusted for age, gender, marital status, race/ethnicity, education, geographic region, number of cancers, and time since first cancer diagnosis. We also adjusted for health insurance coverage in models for self-reported health, but not unemployment due to health, given the relationship of employment with health insurance coverage. We used multiple logistic regression models to examine the association between the individual health behaviors and dichotomized health status outcomes: self-reported health (fair or poor vs. good, very good, or excellent health) and unemployed due to health (yes vs. no). Finally, we used multiple logistic regression models to assess the contribution of health behaviors to rural-urban differences in health status. Exploratory analyses, revealed no significant interactions between rural-urban status and any of the health behaviors

Results

Sample Characteristics

The NHIS sample included 1,642 survivors who resided in a rural county and 6,162 survivors who resided in an urban county. Demographic, clinical and health status characteristics are summarized in Table 1 by rural-urban status. Rural survivors were significantly more likely to be Non-Hispanic, white, less educated, and uninsured. There were also differences in geographic region of residence, with rural survivors more likely to live in the Midwest or South. Cancer characteristics were similar between the groups, except for the higher prevalence of gynecologic cancers among the rural survivors. There were also significant rural-urban differences in self-reported health and unemployment due to health.

Table 1.

Characteristics of Rural and Urban Cancer Survivors from the National Health Interview Survey (2006–2010)

		Rural n=1642 (Weighted %,SE)		Urban n=6162 (Weighted %, SE)
Age, years	<50	20.8	1.1	21.2	0.6
	50–64	31.7	1.4	30.4	0.6
	65–79	34.2	1.3	32.4	0.7
	80+	13.2	1.0	15.9	0.6
Sex	Male	38.9	1.4	39.6	0.7
	Female	61.1	1.4	60.4	0.7
Marital Status	Married/Living with Partner	63.5	1.4	62.8	0.8
	Not Married	36.5	1.4	37.2	0.8
Race/Ethnicity	Non-Hispanic, White	90.6	0.9	81.9	0.6
	Hispanic	2.7	0.6	6.5	0.4
	Non-Hispanic, Black	4.5	0.6	8.4	0.5
	Asian Pacific Islander	0.4	0.2	2.4	0.2
	Other	1.8	0.4	0.8	0.1
Education	< High School	23.4	1.3	14.5	0.6
	High School or GED	34.5	1.5	28.5	0.7
	Some College	42.2	1.7	57.0	0.8
Geographic Region	Northeast	10.9	2.1	21.1	1.1
	Midwest	30.3	2.6	22.5	1.0
	South	42.1	2.3	34.7	1.2
	West	15.8	2.1	21.8	0.8
Health Insurance	None	10.1	1.0	5.7	0.4
	Public only	33.3	1.5	27.4	0.7
	Private (with or without public)	56.6	1.8	66.9	0.8
Cancer Type (s)#	Breast only	17.1	1.0	19.8	0.6
	Gynecologic only	19.3	1.0	15.4	0.6
	Prostate only	13.0	0.9	14.5	0.5
	Colorectal only	6.9	0.9	6.9	0.4
	Melanoma only	7.4	0.8	8.8	0.5
	Hematologic only	5.4	0.6	5.3	0.4
	Short-survival	6.5	0.7	6.4	0.4
	Other single cancer	16.2	1.3	15.9	0.6
	Multiple cancers	8.2	0.7	7.1	0.4
Number of Cancers	1	89.9	0.9	90.7	0.5
	>1	10.1	0.9	9.3	0.5
Time since	<2	14.2	1.0	14.8	0.6
First Cancer	2–5	25.3	1.4	27.5	0.7
Diagnosis, years	6–9	16.5	1.1	16.0	0.6
	10+	44.1	1.4	41.7	0.8
Overall Health*	Excellent	9.2	0.7	13.9	0.6
	Very Good	21.8	1.0	26.0	0.6
	Good	32.3	1.3	33.5	0.7
	Fair	23.4	1.2	18.5	0.6
	Poor	13.3	1.0	8.1	0.4
Employment	Retired	38.9	1.8	42.8	0.9
	Employed	31.1	1.7	38.6	0.8
	Unemployed due to health	18.5	1.2	10.6	0.5
	Unemployed for another reason	11.5	1.0	7.9	0.4

^#Gynecologic cancers include cervix, uterus, and ovary. Short-survival cancers included esophagus, liver, lung, pancreas, or stomach.

^*Overall health was measured by a single item from the MOS SF-36. SE= standard error

Health Behaviors by Rural-Urban Status

Current smoking was reported by 25.3% of rural cancer survivors, compared to only 15.8% of urban survivors (Table 2). Rural survivors were also more likely to report being physically inactive (50.7% compared to 38.7% of urban cancer survivors). In contrast, rural survivors were more likely to report lifetime abstinence from alcohol and less likely to report moderate or heavy drinking. There were no significant rural-urban differences in body mass index, although rural survivors did have a slightly higher prevalence of obesity.

Table 2.

Health Behaviors of Rural and Urban Cancer Survivors from the National Health Interview Survey (2006–2010)

	Rural N=1642 (n, Weighted %, SE)			Urban N=6162 (n, Weighted %, SE)			p-value*
Body Mass Index							.10 (.23)
<25.0	549	34.5	1.30	2210	37.4	0.75
25.0–29.9	533	34.8	1.43	2048	34.7	0.74
30.0+	489	30.6	1.38	1653	27.9	0.69
Smoking							<.001 (.001)
Never	675	39.9	1.49	2862	46.2	0.80
Past	555	34.8	1.55	2228	38.0	0.81
Current	393	25.3	1.22	995	15.8	0.62
Leisure-time Physical Activity #							<.001 (<.001)
Sufficiently Active	263	18.8	1.45	1220	22.8	0.73
Insufficiently active	448	30.5	1.41	2140	38.5	0.84
Inactive	755	50.7	1.91	2349	38.7	0.88
Alcohol Use ±							<.001 (<.001)
Lifetime abstainer	393	23.1	1.36	1240	18.4	0.59
Former	497	30.6	1.39	1427	23.0	0.65
Current Infrequent	210	13.6	1.16	860	14.3	0.56
Current Light	296	19.8	1.09	1422	24.7	0.75
Current Moderate	134	8.0	0.83	748	13.8	0.57
Current Heavy	71	4.9	0.61	308	5.8	0.42

^*First p-values are unadjusted; second p-values in parentheses are adjusted for age, gender, marital status, race/ethnicity, education, geographic region, health insurance, cancer type, and time since first cancer diagnosis.

^#Inactive (no leisure-time aerobic activity that lasted at least 10 minutes), insufficiently active (aerobic activities for at least 10 minutes but less than 150 minutes per week), or sufficiently active (moderate-intensity leisure-time physical activity at least 150 minutes or vigorous-intensity leisure-time physical activity at least 75 minutes per week).

^±Lifetime abstainer (less than 12 drinks in lifetime), former (12 or more drinks in lifetime, but none in the past year), current infrequent (1–11 drinks in past year), current light (≤3 drinks per week in past year), current moderate (4–14 drinks per week for men and 4–7 drinks per week for women), and current heavy (≥14 drinks per week for men, and ≥7 drinks per week for women).

Associations between Health Behaviors and Health Status

In separate adjusted models, all four health behaviors were significantly associated with both fair/poor self-reported health and unemployment due to health (Table 3). The odds of being in fair/poor health and unemployed due to health were significantly greater for survivors who currently smoked compared to never smokers (OR= 1.76, 95% CI: 1.46–2.13 and OR= 1.93, 95%CI: 1.50–2.49, respectively). Obese survivors were more likely to have fair/poor health and to be unemployed due to health compared to those in the normal BMI range (OR = 1.50, 95% CI: 1.28–1.75 and OR= 1.68, 95% CI: 1.35–2.09, respectively). Similarly, the odds of having fair/poor health or being unemployed due to health were approximately two and one-half times greater for inactive survivors compared to those who were sufficiently active. Survivors who currently consumed alcohol were less likely to report fair/poor health and to be unemployed due to health, and this relationship did not vary much by the amount of alcohol consumed (ORs from .38–.67).

Table 3.

Association between Health Behaviors and Health Status Outcomes (adjusted for selected covariates) for Cancer Survivors from the National Health Interview Survey (2006–2010)

	Fair/Poor Self-Reported Health OR (95% CI)	Unemployed Due to Health OR (95% CI)
Body Mass Index
<25.0	ref	ref
25.0–29.9	0.93 (0.80–1.08)	1.11 (0.90–1.39)
30.0+	1.50 (1.28–1.75)	1.68 (1.35–2.09)
Smoking
Never	ref	ref
Past	1.12 (0.97–1.29)	1.05 (0.83–1.32)
Current	1.76 (1.46–2.13)	1.93 (1.50–2.49)
Leisure Physical Activity #
Sufficiently Active	ref	ref
Insufficiently active	1.14 (0.90–1.44)	1.29 (0.94–1.76)
Inactive	2.56 (2.07–3.16)	2.72 (2.02–3.66)
Alcohol Use ±
Lifetime abstainer	ref	ref
Former	1.05 (0.88–1.25)	1.25 (1.00–1.57)
Current Infrequent	0.61 (0.50–0.76)	0.67 (0.49–0.91)
Current Light	0.44 (0.36–0.54)	0.45 (0.34–0.61)
Current Moderate	0.44 (0.33–0.57)	0.38 (0.26–0.55)
Current Heavy	0.47 (0.32–0.68)	0.52 (0.33–0.82)

Each model was adjusted for age, gender, marital status, race/ethnicity, education, geographic region, health insurance (self-reported health only), cancer type, and time since first cancer diagnosis. Odds ratios that are statistically significant, p<.05, are shown in bold.

^#Inactive (no leisure-time aerobic activity that lasted at least 10 minutes), insufficiently active (aerobic activities for at least 10 minutes but less than 150 minutes per week), or sufficiently active (moderate-intensity leisure-time physical activity at least 150 minutes or vigorous-intensity leisure-time physical activity at least 75 minutes per week).

^±Lifetime abstainer (less than 12 drinks in lifetime), former (12 or more drinks in lifetime, but none in the past year), current infrequent (1–11 drinks in past year), current light (≤3 drinks per week in past year), current moderate (4–14 drinks per week for men and 4–7 drinks per week for women), and current heavy (≥14 drinks per week for men, and ≥7 drinks per week for women).

Rural-Urban Status, Health Behaviors, and Health Status

In unadjusted models, rural survivors were significantly more likely to be in fair or poor health and to be unemployed due to health (OR= 1.58, 95% CI: 1.35–1.85 and OR=1.90, 95% CI: 1.55–2.33, respectively). We observed reduction in the odds of poor health outcomes associated with rural-urban status after adjustment for the sociodemographic, cancer-related variables, and the four health behavior variables, but the risk associated with rural residence remained significant. The odds of reporting fair/poor health remained 21% greater for rural survivors compared to urban survivors (OR=1.21, 95% CI: 1.03 – 1.43) after adjustment (Table 4). Rural survivors were also more likely to report being unemployed due to health reasons (OR=1.49, 95% CI: 1.18 – 1.88).

Table 4.

Correlates of Health Status Outcomes in Cancer Survivors from the National Health Interview Survey (2006–2010)

Variables		Fair or Poor Self-Reported Health n=6.790 OR (95%CI)	Unemployed Due to Health n=6,801 OR (95%CI)
Residence	Rural	1.21 (1.03 – 1.43)	1.49 (1.18 – 1.88)
Age, 10 years		1.01 (0.96 – 1.08)	0.72 (0.68 – 0.78)
Sex	Male	1.17 (0.96 – 1.42)	1.17 (0.90 – 1.53)
Marital Status	Not Married	0.98 (0.85 – 1.13)	1.40 (1.15 – 1.70)
Race/Ethnicity	Minority	1.32 (1.10 – 1.58)	1.39 (1.10 – 1.75)
Education	< High School	2.86 (2.26 – 3.63)	2.85 (2.01 – 4.02)
	High School of GED	1.53 (1.24 – 1.88)	1.77 (1.26 – 2.49)
	Some College	1.45 (1.16 – 1.81)	1.45 (1.02 – 2.06)
	4 year degree+	Ref	Ref
Geographic Region	Northeast	Ref	Ref
	Midwest	0.79 (0.62 – 0.99)	0.80 (0.56 – 1.13)
	South	0.88 (0.72 – 1.09)	1.02 (0.77 – 1.37)
	West	0.82 (0.65 – 1.02)	1.05 (0.75 – 1.46)
Health Insurance	None	1.58 (1.17 – 2.13)	---
	Public Only	1.99 (1.71 – 2.32)	---
	Private	Ref	---
Cancer Type (s)	Breast	0.59 (0.44 – 0.80)	0.68 (0.44 – 1.05)
	Gynecologic	0.57 (0.42 – 0.79)	0.62 (0.39 – 1.00)
	Prostate	0.56 (0.39 – 0.82)	0.39 (0.23 – 0.67)
	Colorectal	0.69 (0.47 – 1.00)	0.81 (0.48 – 1.35)
	Melanoma	0.40 (0.27 – 0.58)	0.60 (0.36 – 1.01)
	Hematologic	1.13 (0.75 – 1.70)	1.05 (0.63 – 1.75)
	Short-survival	1.44 (1.04 – 2.01)	1.81 (1.15 – 2.85)
	Other single	0.64 (0.46 – 0.90)	0.88 (0.56 – 1.38)
	Multiple cancers	Ref	Ref
Time since First Cancer Diagnosis, years	<2	1.43 (1.16 – 1.76)	0.85 (0.64 – 1.15)
	2 – 5	1.07 (0.89 – 1.27)	1.10 (0.85 – 1.42)
	6 – 9	1.00 (0.80 – 1.24)	0.84 (0.62 – 1.13)
	10+	Ref	Ref
BMI	<25.0	Ref	Ref
	25.0 – 29.9	0.98 (0.83 – 1.17)	1.16 (0.93 – 1.46)
	30.0+	1.47 (1.25 – 1.74)	1.63 (1.27 – 2.10)
Smoking	Never	Ref	Ref
	Past	1.28 (1.09 – 1.51)	1.12 (0.86 – 1.46)
	Current	1.90 (1.53 – 2.37)	2.08 (1.58 – 2.73)
Physical Activity	Sufficiently Active	Ref	Ref
	Insufficiently active	1.14 (0.90 – 1.44)	1.21 (0.89 – 1.66)
	Inactive	2.22 (1.78 – 2.76)	2.21 (1.65 – 2.97)
Alcohol Consumption	Lifetime abstainer	Ref	Ref
	Former	1.08 (0.89 – 1.31)	1.14 (0.88 – 1.48)
	Current Infrequent	0.64 (0.51 – 0.81)	0.63 (0.44 – 0.90)
	Current Light	0.49 (0.39 – 0.60)	0.47 (0.35 – 0.64)
	Current Moderate	0.49 (0.36 – 0.67)	0.43 (0.29 – 0.63)
	Current Heavy	0.46 (0.31 – 0.69)	0.53 (0.32 – 0.87)

All four health behaviors were significantly associated with self-reported health and unemployment due to health in the final multivariable logistic models (see Table 4). Obese, but not overweight survivors were significantly more likely to report fair/poor health and to be unemployed for health reasons, as were current smokers and those survivors who were physically inactive. Former smokers were also significantly more likely to be in fair/poor health. A reduced risk of fair/poor health and unemployment due to health associated with consuming any amount of alcohol remained after adjustment for the other health behaviors and rural-urban status.

Discussion

Our results mirror those reported in the general rural-urban health behavior literature and extend these findings to the context of cancer survivorship. Specifically, we documented a higher prevalence of current smoking and physical inactivity among rural compared to urban cancer survivors, as well as lower alcohol consumption. Body mass index did not differ. Furthermore, four modifiable risk factors- body mass index, physical activity, smoking, and alcohol consumption - were all associated with both self-reported health and unemployment due to health among cancer survivors. Rural-urban differences in these outcomes were attenuated as hypothesized, but persisted despite adjustment for the health behaviors. The high rates of smoking and lower participation in leisure-time physical activity we observed among rural survivors are very concerning. The prevalence of smoking among rural survivors (25%) is substantially higher than that in urban survivors in our study (16%) and the general population of US survivors [18, 15]. Similarly, half of our rural cancer survivors reported no physical activity, compared with only 31.5% of all survivors from recent population estimates [18]. The cross-sectional nature of this study precludes us from knowing whether the differences observed among survivors are exclusively due to rural-urban differences prior to cancer or whether differential rates of quitting smoking or resumption of physical activity after cancer are involved. Although several studies have examined health behavior change after a cancer diagnosis [e.g., 44–46], to our knowledge none have specifically examined rural-urban differences in positive or negative health behavior changes after cancer. It will be important for future studies to address this knowledge gap.

However, we found no significant difference between urban and rural survivors for body mass index, unlike previous studies reporting an increased prevalence of obesity among rural adults[4, 47],. The prevalence of obesity in our sample (30.6% rural and 27.9% urban) is slightly higher than in general US population studies (27.4% rural, 23.9% urban)[4], but similar to the overall and regional prevalence ranges reported in a recent study of survivors [18].

Similar to general population health surveys [13], we observed that urban survivors were more likely than rural survivors to consume alcohol (infrequent through heavy use). Although rates of heavy drinking were similar in rural and urban survivors, we did not assess alcohol use disorders, which are more prevalent among rural adults, particularly in the Midwest[13]. Rural-urban differences in this study may be due to cultural/community traditions of alcohol abstinence and religious beliefs/norms, particularly in the rural South [13, 48].

Consistent with other studies among cancer survivors, we observed that health behaviors were significantly associated with patient-reported health outcomes [e.g., 14, 30, 49, 50]. To our knowledge, the association between health behaviors and health-related unemployment has not been previously described in cancer survivors. Other studies have linked both smoking and inadequate physical activity to recurrence, second cancers, and poorer survival [e.g., 23, 51, 52, 21, 26, 53, 54].

Like many other studies of the general adult population [55, 56], we identified a “healthy drinker” effect with survivors reporting any level of alcohol consumption being at lower risk of poor health relative to former drinkers and lifetime abstainers. Importantly for breast cancer survivors, there are mixed data linking alcohol consumption to breast cancer recurrence that should be considered when deciding whether or not to consume alcohol [21]. Nevertheless, smoking, body mass index, physical activity, and alcohol consumption do not appear to fully account for rural-urban differences in self-reported health and health-related unemployment that we observed. This persistence suggests that future research should be attuned to the role of community factors that may influence health after cancer in rural areas. These may include availability of oncology, primary care, and other health services such as rehabilitation and mental health, as well as community socioeconomic and social capital characteristics. It will also be important to examine the availability of cancer support organizations, as well as community norms regarding communication about cancer. Differences in cancer stage at diagnosis and receipt of guideline adherent cancer treatment are other factors that could account for rural-urban differences in the long-term health status of cancer survivors. Some, but not all, studies have identified rural disparities in these outcomes [57–62]. Longitudinal studies with well characterized stage and treatment data could explore how treatment disparities impact long-term outcomes among survivors.

Our results highlight the need for evidence-based, individual-level interventions to improve health behaviors and outcomes in rural cancer survivors. Survivors in urban areas may face more safety concerns associated with neighborhood crime or traffic, whereas barriers in rural areas may include a lack of places to exercise such as parks and malls [6]. Rural women also more commonly report lack of support from others to exercise, fear of injury, and interference by caregiving duties [5]. Telephone, mail, or internet interventions may be better ways to reach rural survivors who face transportation challenges to attending classes or exercise programs that are typically held in urban centers. Home-based diet and exercise programs can be effective for increasing physical activity and improving quality of life among cancer survivors generally [e.g., 63, 64], and rural women with breast cancer specifically [65]. Telephone-based counseling is also an efficacious smoking cessation treatment both in cancer survivors [66] and the general population [67, 68] and is widely available throughout the US via state tobacco quitlines. E-health or m-health interventions may also have potential in this population because of their convenience, efficacy, and reach [69]. However, when compared to urban/suburban adults, rural adults are less likely to use the internet, have a smart phone, and use their cell phone to look for health or medical information [70, 71], suggesting that they may not be suitable for all rural survivors. Videoconferencing has been used to enhance psychosocial support among rural breast cancer survivors[72], and new home-based point-to-point video conferencing technology may enhance rural survivor's ability to participate in behavioral interventions at home. In addition to approaches targeted at individual survivors, health-systems interventions to promote the identification and treatment of risk factors among cancer survivors are needed, such as survivorship care planning. Community-based ecological approaches are also recommended to target tobacco use and obesity in the general population [73] and may ultimately have a public health impact for survivors as well.

Our study sample included a large nationally representative sample of rural and urban survivors and utilized a robust classification of rural-urban residence. Thus, our results are pertinent to all cancer survivors in the US. However, an important limitation of this study is use of self-reported data to ascertain cancer history, although several studies have confirmed the validity of self-reports of cancer [74, 75]. Under-reporting of smoking and body mass index and over-reporting of physical activity are also a concern [76], but there is no reason to suspect differential reporting by rural-urban status. With this cross-sectional design we cannot assess how health behaviors change after cancer or assess the directionality of observed associations. Finally, we used the dichotomous urban-rural definition based on county-based RUCC codes based on data availability and to be consistent with prior rural disparities research. County-based codes may result in under- or over-bounding of rural areas, and dichotomous groupings may not fully recognize the heterogeneity within rural areas. Yet on balance, these limitations are unlikely to explain our overall findings.

To our knowledge, this is the first study to describe health behaviors in relation to health status across urban and rural cancer survivors in the US. In conclusion, we present data from a nationally representative sample of US cancer survivors documenting a higher prevalence of smoking and physical inactivity among rural compared to urban survivors. Survivors who live in rural areas are a vulnerable sub-population at increased risk for poor health outcomes after cancer and are more likely to have health- compromising behaviors. Health behaviors may be more easily modified at the individual level compared with other factors associated with rural-urban disparities (e.g., education, income, insurance, travel barriers). Thus they should be an important focus of efforts to reduce disparities after cancer. Behavior change interventions for cancer survivors may need to be developed or adapted to address unique needs and barriers associated with rural residence. It will be important for future studies to move beyond simple categories to recognize the heterogeneity of rural communities and to better identify the community factors that may influence rural health disparities.