|Home | About | Journals | Submit | Contact Us | Français|
While rural residents are more likely to be diagnosed with more advanced cancers and die of cancer, little is known about rural-urban disparities in self-reported health among survivors.
We identified adults with a self-reported cancer history from the National Health Interview Survey (2006–2010). Rural-urban residence was defined using US census definitions. Logistic regression with weighting to account for complex sampling was used to assess rural-urban differences in health status after accounting for differences in demographic characteristics.
Of the 7,804 identified cancer survivors, 20.8% were rural residents. This translates to a population of 2.8 million rural cancer survivors in the US. Rural survivors were more likely than urban survivors to be non-Hispanic white (p<.001), have less education (p<.001) and lack health insurance (p<.001). Rural survivors reported worse health in all domains. After adjustment for sex, race/ethnicity, age, marital status, education, insurance, time since diagnosis, and number of cancers, rural survivors were more likely to report fair/poor health [OR=1.39, 95%CI=1.20–1.62, psychological distress [OR=1.23, 1.00–1.50], ≥2 non-cancer comorbidities [OR=1.15, 1.01–1.32], and health-related unemployment [OR=1.66, 1.35–2.03].
We provide the first estimate of the proportion and number of US adult cancer survivors who reside in rural areas. Rural cancer survivors are at greater risk for a variety of poor health outcomes, even many years after their cancer diagnosis, and should be a target for interventions to improve their health and well-being.
Rural residents have higher cancer mortality than urban residents, with the largest disparities observed for lung, colorectal, prostate, and cervical cancers1. Disparities associated with rural residence have also been documented in cancer diagnosis and treatment2–5, but little is known about disparities that may persist into the post-treatment survivorship period. A majority of survivors are now expected to live at least five years after their cancer diagnosis6, increasing the importance of monitoring long-term health and well-being years after treatment is completed. Although estimates of the number of rural cancer survivors have not been published, 21% of the general population resides in rural areas7, defined by the United States (US) Census as residence outside of areas with an urban core of at least 50,000 persons and densely settled contiguous areas8.
Rural residents may face challenges in accessing medical care and necessary support services due to extended and sometimes difficult travel and a limited number of health care facilities9, 10. Further, rural residents in general tend to be older, poorer, less educated, less likely to have insurance, and more likely to encounter transportation challenges11–13, exacerbating health disparities. The few studies that have examined quality of life, symptoms, or mental health among rural survivors9, 14, 15 are limited by patient samples recruited from a single state or region. Thus we know little about the population characteristics and health status of rural cancer survivors in the US. The lack of information about rural cancer survivors hampers public health planning, allocation of medical resources, and the development of interventions to target this potentially vulnerable population of survivors.
To address these limitations, the goal of this study was to estimate the number of adult cancer survivors who reside in rural areas of the United States and to describe their self-reported health status relative to urban survivors. Compared to non-rural survivors, we expected that rural survivors would report more medical comorbidities and greater psychological distress and would be more likely to report being in fair/poor health and being unable to work because of a health condition.
We analyzed data from the National Health Interview Survey (NHIS)16, an annual in person survey of 30–40,000 households that provides a representative sample of the US civilian, non-institutionalized population. The NHIS is administered by trained census workers and utilizes a complex sampling framework involving clustering, stratification, and multistage sampling; Black, Hispanic, and Asian persons are oversampled. Basic demographic information is collected on all members of the household. Information about cancer history is asked on the Adult questionnaire administered to one randomly chosen adult per family. Data were combined across the years 2006–2010 to increase power for rural-urban subgroup analyses of cancer survivors. During these years the final response rate for the Adult survey ranged from to 60.8% to 70.8%. Data collection for the NHIS was approved by the National Center for Health Statistic (NCHS) Research Ethics Review Board (ERB), which also approves protocols for use of restricted data that are not publicly available. Analysis of deidentified data from the survey is exempt from the federal regulations for the protection of human research participants; local approval is not required.
Adults with a self-reported history of cancer other than non-melanoma skin or “skin-type unknown” cancers comprised the survivor sample. We excluded these skin cancers to be consistent with National Cancer Institute Surveillance Epidemiology and End Results (SEER) cancer survivor estimates that do not include basal and squamous cell skin cancers and because the treatment, surveillance, and survivorship of this type of cancer is likely very different from other cancer sites. Cancer site and date of diagnosis for up to three cancers are available for each person. We calculated time since cancer diagnosis by subtracting age at first cancer diagnosis from age at interview. Self-reported age, education, race/ethnicity, marital status, sex, income, and employment status are contained in the Person file.
The NHIS rural-urban residence code is based on the US census metropolitan area definitions that became effective in June 2003. Rural residences were defined as those located outside of urbanized areas and urbanized clusters (defined as core census block groups or blocks that have a population density of at least 1,000 people per square mile and surrounding census blocks that have an overall density of at least 500 people per square mile)8. Rural-urban residence is a restricted variable; therefore these data were accessed through the NCHS Research Data Center.
Self-reported health was assessed via a single global health question derived from the Medical Outcomes Short Form Survey Instrument17 (SF-36): “Would you say your health in general is excellent, very good, good, fair, or poor?” Psychological distress was assessed via the Kessler K-618, a brief screening scale designed for epidemiologic studies with a possible range of 0–24. Consistent with prior population research19, 20, we used a three level categorization of psychological distress[no probable (0–7), mild-moderate (8–12), and serious (13–24)]. These categories correspond to 87.5%, 8.5%, and 3.9% of the general adult population20 and have been shown to have good correspondence with Structured Clinical Interview for DSM-IV anxiety and mood diagnoses21. For logistic regression models we modeled the likelihood of reporting mild/moderate/serious distress (8–24) compared to none (0–7). We chose this approach rather than focusing on the clinical cutpoint of >12 because we were interested in identifying survivors at risk for poor outcomes because of subclinical or moderate distress, in addition to those with a likely serious anxiety or mood disorder. Consistent with prior approaches22–24, a non-cancer comorbidity score was calculated by summing five conditions which are ascertained in the NHIS (hypertension, heart disease, stroke, diabetes, and lung disease). Heart disease included coronary heart disease, angina, myocardial infarction, or any other heart condition, and lung disease included chronic bronchitis, emphysema, and current asthma. Adults who report that they are not working or looking for a job and are temporarily unable to work for health reasons or disabled were considered unemployed due to health. Data was not available on the specific medical condition leading to unemployment.
Chi-square tests were used to assess rural/urban differences in sociodemographic characteristics and to assess unadjusted rural/urban differences in each health status outcome. We then defined binary versions of all outcome variables (fair or poor health, K-6 score greater than seven, two or more non-cancer comorbidities, and unemployment due to health) and used multivariable logistic regression models to assess rural/urban differences in each dichotomized health status outcome after adjustment for demographic factors (age, sex, race/ethnicity, and marital status), cancer variables (type, number of cancers, and time since diagnosis), and socioeconomic (SES) resources (education and health insurance coverage). We used education rather than income in models due to significant collinearity between the variables (p-value <. 0001, X2) and the large amount of missing income data. We assessed the impact of the covariates in a hierarchical fashion, first adjusting for the demographic factors, then the demographic and cancer factors, and finally the demographic, cancer, and SES factors. All of the statistical analyses were conducted using the SURVEYFREQ, SURVEYMEANS, and SURVEYLOGISTIC procedures in SAS, version 9.2 (SAS Institute, Inc., Cary, NC), which incorporated strata and cluster information as well as sampling weights to account for the complex survey design of the NHIS. Cancer history was included as a domain to obtain estimates of the effects for cancer survivors. As a sensitivity analysis, we also examined health-related unemployment models that excluded retired persons. There was very little change in the parameter estimates, so we report models that include all survivors.
For the years 2006–2010, a total of 9958 adults reported a history of cancer. We excluded 2154 persons who reported a history of exclusively non-melanoma skin or “unknown” skin cancers, for a total survivor sample of 7804.
Of the identified cancer survivors, 20.8% (n=1,642) resided in areas defined as rural. Using population-based weights for the adult sample, we calculate that approximately 2,779,0000 US cancer survivors reside in rural areas. Table 1 shows the distribution of cancer survivors and population estimates by each category of rural-urban status. Survivors were distributed across all categories; it is estimated that approximately 304,000 survivors reside in completely rural areas (categories 8 & 9).
As expected in a sample of cancer survivors, almost half were 65 years of age or older and 60% were women. Breast and gynecologic cancers were most common in women, and prostate in men; melanoma and colorectal cancer were common in both sexes. Sixty percent of survivors were more than 5 years after their first cancer diagnosis; ten percent reported more than one cancer diagnosis.
Statistically significant differences in sociodemographic characteristics between rural and urban survivors were observed for race/ethnicity, education, poverty status, and health insurance coverage (see Table 2). Rural cancer survivors were significantly more likely to be non-Hispanic white, to report lower levels of education, to report lower incomes, and to lack health insurance coverage. Cancer types were generally similar between urban and rural survivors, with the exception of a higher percent of gynecologic cancers among rural survivors. There was a slightly higher percent of long-term survivors (≥10 years after diagnosis) among the rural survivors.
As shown in Table 3, rural cancer survivors report significantly poorer health status for all indicators. Fair/poor health was reported by 36.7% of rural survivors compared to only 26.6% of urban survivors. Non-cancer comorbidities also were more common in rural than urban survivors (36.5% with ≥ two versus 31.6%). Rural survivors also reported greater psychological distress, with 18.8% in the two highest categories compared with only 12.8% of urban survivors. Rates of unemployment due to health reasons were also higher among rural survivors (18.5% compared to 10.6% of urban survivors).
As shown in Table 4, adjustment for sociodemographic characteristics and cancer variables had little impact on the rural residence odds ratios for all four indicators of poor health. After further adjusting for health insurance coverage and education, rural survivors remained significantly more likely to report fair/poor health [OR=1.39, 95%CI=1.20–1.62, mild/moderate/severe psychological distress [OR=1.23, 1.003–1.504], ≥2 non-cancer comorbidities [OR=1.15, 1.01–1.32], and health-related unemployment [OR=1.66, 1.35–2.03].
We found that 21% of US cancer survivors reside in rural areas, representing a population of 2.8 million US survivors. This proportion is almost the same as the general adult population, which is somewhat surprising as rural residents tend to be older and cancer incidence increases with age. Rural survivors are significantly more likely than their urban counterparts to report fair or poor health, two or more non-cancer comorbidities, elevated levels of psychological distress, and unemployment due to health. The notable rural-urban health disparities observed in this study are partially, but not fully, accounted for by differences in socioeconomic resources (income, education, and health insurance coverage).
Few studies have made direct comparisons between rural and urban cancer survivors for health status or quality of life outcomes. Our results illustrating poorer heath among rural cancer survivors are consistent with prior work that found poorer mental health among rural compared to non-rural cancer survivors in Kentucky14. Another study of rural cancer survivors also reported lower physical functioning health-related quality of life (HRQOL) among rural survivors25. In contrast to our findings, one recent Australian study found similar HRQOL 12 months after treatment among urban and non-urban breast cancer survivors26. A very large majority of our survivors (94%) were more than a year after diagnosis, raising the possibility that rural-urban disparities increase in the period after treatment. Alternatively, rural-urban differences may vary across countries due to variation in health care policy, including universal insurance coverage, and/or specific characteristics of rural areas (e.g., distance to urban areas, access to health care and transportation accessibility).
Approximately 19% of the rural survivors in this sample reported mild, moderate, or severe psychological distress, compared to only 12.8% of the general rural adult population from prior studies using the Behavioral Risk Factor Surveillance System19, 24. Our finding of greater distress among rural survivors (18.8% vs 12.8%) is the opposite of the pattern observed in the general adult population where urban adults were 22% more likely than rural adults to report mild, moderate, or severe psychological distress19. There appears to be something about cancer or perhaps aging more generally that interacts with rural residence to alter the observed relationships.
Pratt and colleagues24 established that individuals with the greatest level of psychological distress (a Kessler score of 11 or above in their study) had a 2.6 times greater hazard for mortality compared to those in the lowest category; increasing hazard of death was shown for all escalating categories of distress. Prior studies have also established the importance of comorbidity burden for cancer survivors; survivors with multiple comorbidities are much more likely to be in poor health, to report functional limitations, and to be unable to work for health reasons27. Given the linkage between poor health outcomes and comorbidities, it is concerning that more than 15% of rural cancer survivors report three or more non-cancer conditions compared to 11.5% among urban survivors.
Management of these comorbidities and cancer follow-up care in general requires access to the healthcare system. We expected the slightly higher prevalence of health insurance among urban versus rural survivors28. Consistent with prior research29, we observed that the overall prevalence of health insurance in both groups is higher than in rural communities in general28 and the US population30. This likely is due to the older age and accompanying Medicare coverage of cancer survivors. While insurance coverage is important, rural residents in general report higher perceived barriers to care31 and appear to rely more heavily on primary care than urban residents32. Rural cancer patients have been reported to travel six to ten times farther for chemotherapy and two to four times farther for radiation therapy than urban residing survivors32. Whereas ongoing care from specialists may be feasible in urban contexts, survivors in rural areas may benefit from survivorship care managed by a primary care provider and coordinated with specialists.
Unemployment among cancer survivors is a well-recognized issue and the overall rate in our population mirrors that reported in other groups of cancer survivors33, 34. Prior reports have not parsed the reason for unemployment so a unique contribution of our analysis is the attribution of more than half of the unemployment among cancer survivors to health issues. These health issues presumably include cancer-related problems and may also arise from other conditions. Another novel finding is the rural-urban disparity in both overall and health-related unemployment, which to our knowledge has not been previously reported. This finding is particularly striking given that no such disparity has existed in national unemployment data for the past two decades35. Cancer survivors may find it more difficult to maintain employment in rural settings where production, transportation and service occupations are more common than the professional, management and business occupations that predominate urban environments36. Overall these results reinforce the need to address occupational issues among cancer survivors and suggest this may be particularly important to survivors who reside in rural areas.
The primary limitation of this study was the use of self-reported rather than registry confirmed cancer history. Some studies have suggested that individuals may under-report a diagnosis of cancer relative to medical record or registry review, with greater under-reporting among men37. Female genital cancers may be particularly problematic; some studies find that endometrial and cervical cancers are underreported38 while others find overreporting37. The population estimate for the total number of US survivors suggested by these data (13.3 million), best thought of as a population estimate of the midpoint of the time interval (2008), is slightly higher than National Cancer Institute population estimates using Surveillance Epidemiology and End Results (SEER) registry data39. However, the most common cancer sites identified in our sample are the most prevalent cancers in the survivor population as estimated by SEER40. In addition, this study may have excluded very ill cancer survivors who reside in institutional settings or are too ill to respond to the study.
The primary strength of this study was the use of a population-based dataset that was designed to be representative of the US population. In addition, we utilized a robust definition of rural residence, based on US Department of Agriculture Rural Urban Commuting Area (RUCA) Codes, rather than relying on perceived rural residence or hospital/clinic location as a proxy. This will facilitate comparison with other research using the same definitions.
These data suggest that rural cancer survivors are a vulnerable population at risk for poor health outcomes after a cancer diagnosis. Future studies should examine other factors that may underlie rural-urban differences in health status, including the prevalence of health compromising behaviors and additional access to care variables. Efforts should be undertaken to identify the needs of rural cancer survivors and develop comprehensive survivorship care planning programs to address these needs.
The findings and conclusions in this paper are those of the author(s) and do not necessarily represent the views of the Research Data Center, the National Center for Health Statistics, or the Centers for Disease Control and Prevention.
This work was supported by the National Cancer Institute at the National Institutes of Health (grant number R03 CA156641-01).