Clinical trial recruitment in rural South Carolina : a comparison of investigators ’ perceptions and potential participant eligibility

Introduction: Participation in clinical trial (CT) research can help decrease health disparities in rural communities. The purpose of this study was to examine the perceptions of principal investigators (PIs) regarding CT participation barriers and recruitment efforts in rural South Carolina, USA and to assess the actual pool of potential CT participants in rural and urban South Carolina. The ultimate goal was to evaluate the fit between PIs’ perceptions and the pool of eligible participants in rural South Carolina. Methods: An online survey was conducted with 119 CT PIs from South Carolina’s five main academic medical centers located in urban areas of the state, for a response rate of 31%. Secondary data analyses were also conducted using data from government health insurance plans, including the 2009 South Carolina Medicaid, the 2009 State Health Plan (SHP) data, and census data from the 2005–2009 American Community Survey (ACS). Both parametric and non-parametric statistics were used to analyze survey and secondary data. Results: Principal investigators perceived greater recruitment barriers in rural areas than in the general population. They indicated having difficulty finding CT participants in rural areas compared to the general population (t=–2.985, p=0.004). Rural residents were significantly more likely to be perceived as lacking knowledge and understanding about CT than the general public (t=–2.105, p=0.038), having significantly lower literacy than the general public (t=–2.058, p=0.043), lacking information about available CTs (t=–2.913, p=0.005), and having limited accessibility to trial sites compared to the general population (t=–4.380, p=0.000). Patients’ insurance coverage, however, was not found to be a significant barrier for CT participation (t=0.418, p=0.677). Secondary data variables were aligned with these barriers. Data revealed that rural residents have slightly lower educational attainment than urban citizens (t=5.384, p=0.000), and more people live below poverty level in rural areas (23%) than

Introduction: Participation in clinical trial (CT) research can help decrease health disparities in rural communities.The purpose of this study was to examine the perceptions of principal investigators (PIs) regarding CT participation barriers and recruitment efforts in rural South Carolina, USA and to assess the actual pool of potential CT participants in rural and urban South Carolina.The ultimate goal was to evaluate the fit between PIs' perceptions and the pool of eligible participants in rural South Carolina.Methods: An online survey was conducted with 119 CT PIs from South Carolina's five main academic medical centers located in urban areas of the state, for a response rate of 31%.Secondary data analyses were also conducted using data from government health insurance plans, including the 2009 South Carolina Medicaid, the 2009 State Health Plan (SHP) data, and census data from the 2005-2009 American Community Survey (ACS).Both parametric and non-parametric statistics were used to analyze survey and secondary data.Results: Principal investigators perceived greater recruitment barriers in rural areas than in the general population.They indicated having difficulty finding CT participants in rural areas compared to the general population (t=-2.985,p=0.004).Rural residents were significantly more likely to be perceived as lacking knowledge and understanding about CT than the general public (t=-2.105,p=0.038), having significantly lower literacy than the general public (t=-2.058,p=0.043), lacking information about available CTs (t=-2.913,p=0.005), and having limited accessibility to trial sites compared to the general population (t=-4.380,p=0.000).Patients' insurance coverage, however, was not found to be a significant barrier for CT participation (t=0.418,p=0.677).Secondary data variables were aligned with these barriers.Data revealed that rural residents have slightly lower educational attainment than urban citizens (t=5.384,p=0.000), and more people live below poverty level in rural areas (23%) than In the USA, people living in rural areas report poorer health status than urban residents [1][2][3][4][5] .This health disparity is often linked to rural residents' lack of access to healthcare services, shortages of healthcare professionals, lower rates of health insurance, problems working with third-party payers, and low socioeconomic and educational status [6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21] .For example, individuals living in rural areas experience a higher prevalence of chronic diseases such as obesity, heart disease, cancer, and diabetes than people in urban areas 2,[21][22][23][24][25][26] .
US states like South Carolina (SC), with large rural populations, are particularly vulnerable to rural-urban health disparities, with rural residents facing poorer overall health, engaging less in healthy behaviors, and lacking access to preventive services and care for chronic diseases such as diabetes 27 .Nearly 35% of the population in South Carolina lives in rural areas (Fig1) 27,28 .This represents more than 1.5 million people living in small and remote locations across the state who may be lacking needed medical care 28 .
A possible solution for providing health care to these medically underserved communities is through clinical trials (CTs) 29,30 .Clinical trials are biomedical or health-related research studies intended to add to medical knowledge 31 .CTs can offer patients the most advanced medical treatments and screening options, and can help participants such as those in rural areas obtain the medical care they need 30,31 .Despite potential 'therapeutic misconceptions', where potential disadvantages related to CT participation are overlooked 32 , CTs can help medically underserved populations access novel medical care and reduce health disparities 33,34 .Nevertheless, CT participation is particularly low among medically underserved communities (ie people living in rural areas) 6,9- 11,18,19,34-37 .
A significant reason for low participation in CTs among rural communities is the lack of access to healthcare facilities 27,38,39 .The low socioeconomic status of rural individuals may also make them more sensitive to additional costs and time often required in CT participation 40,41 .As some trials may not be covered by insurance, some individuals may not be able to afford to participate 7,8,18,42,43 .In South Carolina, 25 of 46 counties are designated as medically underserved areas (MUAs) (Fig1) 27 .The median per capita income in 2005 for rural counties in the state was $23,344, about $11,000 less than the national average of $34,471 27 .
Another key barrier to CT accrual in rural areas is limited knowledge and awareness about CTs [42][43][44][45][46][47] .The study by Friedman et al 48 on knowledge and perceptions of CTs found that rural residents share important misperceptions about CTs, which may be fueled by some lack of knowledge and understanding.Further, Lara et al 46 found a correlation between education level and knowledge about CTs, with the highly educated being more knowledgeable and willing to participate in medical research.According to US Census data, nearly 30% (29.4%) of adults in rural South Carolina lack a high school diploma 49 .These less educated rural individuals may be less knowledgeable about CTs 50,51 and, therefore, less likely to participate.
Although CT participation is low in rural areas 18,35,52 , no study has explored the proportion of rural residents who are eligible to participate in CTs.CT participants must meet specific inclusion criteria to qualify for a study.These criteria vary by trial, and are often based on factors such as age, gender, race, health insurance status, type and stage of disease, previous treatment history, and other medical conditions 30,31,53 .
Pre-existing medical conditions based on a clinical risk grouping (CRG), a health-status classification system often utilized by insurance companies and government-funded programs 54 , are another important exclusion criterion for participation in CTs 39,53 .Depending on the CT, an individual may not be allowed to participate if he or she has multiple comorbidities and disabilities.
Regardless of the type of trial, CT principal investigators (PIs) are ultimately responsible for accruing patients into their CTs.To the authors' knowledge, there are no studies that assess what CT PIs think about their accrual of patients in rural and urban areas or studies that explore PIs' perceived barriers to CT participation.The aim of this study is to address this gap in the literature by examining the perceptions of PIs regarding CT participation barriers and recruitment efforts in rural areas of South Carolina.A secondary data analysis of eligible South Carolina CT participants was also conducted to assess the potential pool of CT participants in rural and urban areas of South Carolina.The ultimate goal was to compare PIs' perceptions of eligible participants to the actual pool of participants.

Methods
This study used both a survey and secondary data analysis to examine CT participation and eligibility.First, an online survey was developed and used to assess the PIs' perceptions about CT participation.PIs at the five large primary research hospitals in South Carolina were approached because the trials conducted at these institutions are systematically documented and conducted by South Carolina-based researchers.
Next, analyses of existing data from government programs like South Carolina Medicaid, the South Carolina State Health Plan (SHP), and the US Census, allowed the authors to describe a number of household-and individual-level characteristics of importance to CT eligibility in South Carolina.Medicaid provides an important source of insurance coverage for medically underserved populations 55 .In South Carolina, 8% of people in the state, or more than 350 000 individuals, are insured by Medicaid, including citizens in rural areas.Since rural populations are frequently categorized among the medically underserved, Medicaid data were used for these analyses.
Another 445 000 people in South Carolina are covered by the SHP (10%) (D.Dickerson, pers.comm., 2011).As individuals covered by the SHP are often considered representative of the state's population, for the purposes of this research these data were extrapolated to the rest of the non-Medicaid population (A.Brock Martin, pers.comm., 2013).The SHP, coupled with Medicaid, covers 18% of the state's population for health care (not included are other programs such as Medicare-age 65+ (5%), Veterans Affairs (VA) health insurance (8%), other private insurance (55%) for which data were unavailable, and the uninsured (17%) (D Dickerson, pers.comm., 2011) 56 .While this is a limitation of the present study, the representative nature of the SHP data helps validate results.Finally, the census data were used to complement the Medicaid and SHP data for specific criteria relating to clinical trial participation (eg education and language).

Principal investigator survey
An e-survey using Qualtrics, a subscription-based survey website for researchers in industry and academia (http:// www.qualtrics.com), was administered to PIs at South Carolina's five main academic medical centers located in urban areas of the state: Spartanburg Regional Healthcare System and Gibbs Cancer Center, Greenville Hospital System University Medical Center, Medical University of South Carolina, University of South Carolina and University of South Carolina School of Medicine, and Palmetto Health Hospital.The survey was open for 3 weeks during June and July of 2011.The survey link was sent to 382 potential survey participants identified through a variety of tracking methods (see Tanner et al 57 for detailed information on survey development and data collection).A convenience sample of 119 CT investigators completed the survey, for a response rate of 31% 58 .Respondents could only click and access the survey once, which prevented the same person from completing the survey multiple times 59 .The majority (83%) of respondents were PIs, while the others were project managers, recruiters, research associates, nurses, or 'other'.Because of their work with CTs, all respondents were included in this study and are referred to as PIs throughout this article.
The survey instrument was composed of 33 items, which were developed based on extensive literature on CT participation 6,8,9,12,36,43,46,47 .The survey instrument included a combination of open-and closed-ended questions to gather information about the volume and scope of CT research taking place at South Carolina's main academic medical centers, the extent to which rural residents in South Carolina are represented in clinical research, and PIs' perceptions of barriers to CT recruitment.All responses were captured automatically via the website 59 .The survey instrument was approved by the university's Institutional Review Board.
Although the survey, as a whole, addressed a wide range of issues pertaining to CT participation and recruitment, for the current study the researchers focused on several variables that specifically aligned with the secondary data.These six items touched on PIs' perceptions of the barriers to CT participation in the general population and in rural communities, and included: • difficulty finding potential clinical trial participants • patients' lack of awareness and knowledge about CTs • lack of literacy • limited access to trial sites • lack of insurance coverage • lack of information about available trials.
For each item, respondents were asked to rate their level of agreement on a five-point Likert scale that ranged from 'strongly disagree' to 'strongly agree'.No incentives were provided to survey respondents.As described elsewhere 57 , parametric statistics including paired-sample t-tests were used to analyze the survey data.Analyses were generated using the Statistical Package for the Social Sciences v20 (SPSS Inc.; http://www-01.ibm.com/software/analytics/spss/ products/statistics). Next, data from the US Census (2005-2009) American Community Survey (ACS) were used and compiled at the individual and household levels.While some of these data are available online (http://www.census.gov/acs/www),data sets were received for education, language, poverty level, and access to a vehicle through the ORS.The rationale for the assessment of these variables is explained in Figure 2.

Secondary data analysis
It is important to point out that each of these data sets (Medicaid, SHP, ACS) reports data differently.For example, the ACS data included both individual-level (education, language, and poverty level) and household-level data (access to a vehicle).For the purpose of this research, some variables were collapsed to represent 'eligible' and 'not eligible' based on the defined set of criteria.For example, education was reported in five categories from 'no high school diploma' to 'graduate degree'.Based on the determined eligibility criterion for education (high school diploma or above), this became a dichotomous variable, with percentage 'no high school' versus percentage 'high school and above'.
Language was reported in a set of categories that focused on populations that speak English, Spanish, Indo-European languages, Asian and Pacific Island languages, and other languages 'well' or 'very well'.For this variable, those who speak English 'well' and 'very well' were collapsed with or without other languages versus all those who do not speak English well.
Access to a vehicle was reported in the ACS data as having 'no access to a vehicle' versus 'access to one or more vehicles' per household.For this variable, the categories were kept intact and excluded individuals aged 65 or older.
Regarding poverty level, only those whose income was above federal poverty standards were included in this study as it was assumed that they would have access to health insurance or could cover the additional costs related to the CTs 60 .
In the SHP and Medicaid data sets, CRGs are divided into levels 1-9.Insurance companies, including governmentfunded programs, such as Medicaid, often classify the health status of individuals into CRGs, which assign each individual to a single mutually exclusive risk group based on clinical (ie pre-existing medical conditions) and demographic characteristics 54 .For this study, individuals in CRG levels 1-5 are considered as potentially eligible CT participants.Individuals assigned to these groups are one of the following: • healthy (CRG 1) • have a history of significant acute disease (CRG 2) • have a single minor chronic disease (CRG 3) • have a minor chronic disease in multiple organ systems (CRG 4) • have a single dominant or moderate chronic disease (CRG 5).
Levels 1-9 nine were therefore divided into two categorieseligible (CRG 1-5: healthy or may have up to one single dominant or moderate chronic disease), and ineligible (CRG 6-10: having comorbidities).Medicaid also categorized individuals by 'disability' or 'no disability'.For Medicaid, individuals who were categorized as having a disability were excluded from the analyses.
The age criterion was reported in each of the three data sets as follows: 18-24, 25-34, 35-44, 45-54, 55-64, and >65 years.For the purposes of this study, only age groups between 18 and 64 years, based on general age requirements for CT participation, were included.
Because the focus was on rural populations specifically, the secondary data were divided by urban and rural regions.As the data were reported by county, each county could be designated as primarily urban or rural, based on ORS designations.There are a number of definitions of 'rural' because important differences, including demographic, cultural, and economic ones, exist across and among rural places 61 .For this research, an urban or metropolitan area is defined as an area with at least 50 000 people, while a rural area can be a micropolitan, a small rural or small remote rural area composed of 50 000 people or less 27 .Finally, for the secondary data analyses, descriptive statistics, frequencies, and paired-sample t-tests were generated using Excel 2007 and SPSS.For the t-tests, the assumption of normality was assessed for the dependent variable and was not violated.

Ethics approval
All components of this study were approved by the university's Institutional Review Board; ethics approval number Pro0009909.

Principal investigator survey
As shown in Table 1, respondents perceived the barriers to recruitment to be far greater in rural areas than in the general population.Principal investigators indicated that finding potential clinical trial participants was significantly more difficult in rural areas than in the general population (t=-2.985,p=0.004).In terms of education about CTs, rural residents were significantly more likely to be perceived as lacking knowledge and understanding about the idea of CTs than the general public (t=-2.105,p=0.038), which may affect their participation in medical research.Additionally, rural residents were perceived to have significantly lower literacy or lower health literacy than the general public (t=-2.058,p=0.043), which can be a barrier to CT recruitment and participation.
In rural communities, lack of information about available trials was found to be significantly more likely to be a barrier to CT participation than in the general population (t=-2.913,p=0.005).Access was also a barrier that was considered for CT participation across the state.Patients in rural areas were also significantly more likely to be perceived as having limited accessibility to trial sites compared to the general population (t=-4.380,p=0.000).This may be because fewer CTs are held in rural areas, thus requiring patients to have access to a vehicle to participate.Additionally, insurance coverage was examined as a barrier to CT participation.According to PIs, patients' insurance coverage of CT procedures or drugs was not considered more of a barrier for rural residents than for the general public (t=0.418,p=0.677).In other words, insurance coverage may not hinder CT participation in rural areas more than in urban areas of the state.
Poverty level: This variable refers to a person's income compared to the poverty threshold appropriate for that person's family size and composition.If that person's income is less than the threshold appropriate for the family, then the person is considered below poverty line.The poverty thresholds are determined annually by the federal government.This is an important eligibility criterion as a person below poverty line may not be able to pay for a clinical trial if required for participation or may not have health insurance to cover the extra costs.
Education: Clinical trial (CT) participants must have some education (high school diploma or higher) to be able to understand CT protocols and procedures.
Language: To participate in a CT, individuals must understand CT protocols that are typically written in English and give their consent.An important eligibility criterion for CT participation, and medical research in general, may be the requirement of speaking and understanding English well or very well.
Access to a vehicle: A vehicle or some mode of transportation may be needed for CT participants in rural areas to travel to the location of the CT.Access to a vehicle is therefore an important variable to consider in terms of CT eligibility.

Secondary data analysis
A secondary data analysis of Medicaid, SHP, and ACS was conducted to assess whether or not the perceptions of the PIs regarding barriers to CT participation in rural communities were indeed the reality in rural South Carolina.Forty-five percent of Medicaid recipients live in rural areas and 41% of people insured by the SHP live in rural areas.When looking at CT eligibility by CRG level, Medicaid data revealed that the proportion of rural and urban residents with a CRG level between 1 and 5, or relatively healthy individuals with no comorbidities, is nearly equal.Based on these results, when looking at CRG level alone, 70% of individuals living in rural areas insured by the SHP or Medicaid are eligible for CTs.
The ACS data, which showed differences between urban and rural citizens on variables such as education, access to a vehicle, poverty, and language, were analyzed to compare these numbers to PI perceptions on lack of knowledge and information about CTs (eg relates to education level and language), accessibility to trial site (eg relates to access to vehicle and poverty level), low literacy (eg relates to education and language).As shown in Table 2, a greater percentage of South Carolina residents have a high school diploma or higher in urban areas (83%) compared to rural areas (75%), and this difference is statistically significant (t=5.384,p=0.000).The data also indicate that more individuals live above poverty level in urban areas (85%) compared to rural areas of the state (77%) (t=4.86,p=0.000).Poverty, therefore, affects more people in rural areas of South Carolina.Eighty-nine percent of citizens in rural areas have access to a vehicle compared to 94% in urban areas, and this difference was also statistically significant (t=4.609,p=0.000).The difference in terms of language (speaking English well or very well) is not statistically significant between rural and urban areas of the state (t=-1.707,p=0.095).
While it was not possible to obtain the percentage of individuals in rural areas of South Carolina who meet all eligibility criteria at the same time (ie who have a high school diploma or higher, who own a vehicle, who live above poverty level, and who speak English well or very well) because of the way the data were presented, it was possible to get a better picture of CT eligibility.Based on these analyses, a minimum of 75% of individuals in rural areas meet at least one of these CT eligibility criteria, compared to 83% or more of individuals in urban areas of the state.

Discussion
This study examined both sides of a situation not previously studied: PIs' perceptions of CT barriers among rural residents (survey research) and the proportion of potentially eligible participants in rural areas of the state (secondary data analyses).These analyses revealed similarities and differences between PIs' perceptions and potential participant eligibility.
Principal investigators indicated that it is most difficult to find eligible CT participants in rural areas of the state.Although the overall analyses of existing data revealed that fewer eligible participants live in rural areas of South Carolina than in urban areas (75% vs 83%), there still exists a large, untapped pool of eligible participants in rural South Carolina.With low levels of CT participation one of the most significant challenges facing medical research 8,62 , these findings suggest that rural populations represent a real opportunity for participant recruitment.Further, CTs may be an important alternative for individuals in underserved medical communities to access effective medical care 30,31 .More than one million people living in rural areas of the state could access quality medical care while contributing to medical research and scientific knowledge though their participation in CTs 30,31,33,34 .
As emphasized by results from the PI survey and secondary analyses, however, there are barriers, as perceived by respondents in the PI survey and found to be a reality in the demographic data, that hinder CT enrollment in rural communities.Accessibility to trial sites, for example, remains a barrier for CT participation in rural areas.PIs believed that rural residents have limited accessibility to trial sites compared to the general public and the secondary data suggest that rural residents may have slightly less access to a vehicle to get to the CT site (89%) than their urban counterparts (94%) (t=4.609,p=0.000).Further, because more people live in poverty in rural areas than in urban areas of South Carolina (23% vs 15%, respectively), this may affect their ability to pay for a clinical trial, have access to health insurance, purchase a vehicle, or travel to a trial site.
Principal investigators also believed that rural residents are more likely to lack knowledge and understanding about the idea of CTs than the general public.Although there was not a direct connection to 'knowledge and understanding' in the secondary data, it was possible to assess educational attainment and language.One quarter of residents in rural South Carolina lack a high school diploma, nearly 15% more than those living in urban areas of the state.On the other hand, language does not seem to be a barrier, as the data indicated that 99% of rural residents speak English well or very well.Specifically, findings suggest that rural citizens may speak the language used in clinical trials but may not understand the complicated medical jargon.This may decrease their comfort level with the idea of medical research and ultimately prevent them from enrolling in a CT 50,51,63,64 .A few limitations affect the generalizability of these findings.
For the PI survey, respondents were only recruited from South Carolina's five main academic medical centers; PIs conducting CTs from pharmaceutical companies and other settings were not included.The authors are confident, however, that this sample included the majority of PIs conducting CTs in South Carolina.In addition, the response rate of 31% for the PI survey may be considered low in research.However, this rate is fairly typical for online surveys with this specific population [69][70][71][72] .The total number of responses per question on the PI survey varied considerably.The survey, which was presented in an online format where participants could 'quit' the survey at any time, may have been too long for PIs, which created some participant burden [73][74][75] .Finally, the closed-ended survey items may not have been the most appropriate data collection method to extensively investigate the PIs' perceptions of CTs.The response options ('strongly agree' to 'strongly disagree') limited the collection of detailed comments from the PIs.Future research should consider using qualitative methods to explore the PIs' perceptions of clinical trial recruitment barriers.
For the secondary data analysis, individuals without health insurance or with private insurance were not included.While this limits the portrayal of CT eligibility in the state, the analyses conducted using ACS, Medicaid, and SHP data still provided a comprehensive overview of participant eligibility.Moreover, because of the way raw data were presented (individual and household levels), it was not possible to build on the current findings by combining all ACS data, or even combining the Medicaid and the SHP data sets to obtain a more complete representation of CT eligibility in South Carolina.Therefore, a final percentage of those potentially eligible for CTs and those who are not could not be obtained.Lastly, knowledge and education as they are used in the PI survey were not explicitly defined and categorized as they are in the secondary data analysis.Nevertheless, there was still a match between PIs' perceptions about lower levels of knowledge and education about CTs among rural residents and lower educational attainment in rural areas, as suggested by the secondary data.

Conclusions
This study was the first step in trying to provide a comprehensive representation of CT eligibility in South Carolina.It was also the first step in examining the state of CT recruitment among rural populations and the barriers to CT participation.PIs who participated in this study acknowledged the barriers they face in recruiting CT participants from rural areas.Variables examined for the secondary data analysis were aligned with these barriers, but this analysis also suggested that a large pool of eligible CT participants does reside in rural South Carolina.
Investigators' poor understanding of communities' perceptions about CTs often leads them into a practice of planning for rather than planning with underserved communities, resulting in poor CT participation in these communities 76 .Findings from the current study suggest that investigators should tailor CT educational and promotional material for rural residents by taking into account their lower educational attainment and income levels.Further, developing satellite trial sites, located in remote and rural communities, would allow those without access to a vehicle, or money for long-distance travel, an opportunity to participate in potentially life-saving research.Future research should focus on working with these rural communities to understand their medical and communication needs, as well as providing communication training to PIs to improve CT recruitment efforts in rural areas of the state.
Source: SC Rural Health Research Center.State rural plan for South Carolina.Columbia, South Carolina; 2008.

Figure 1 :
Figure 1: Rural and urban counties in South Carolina, USA 27 .

Figure 2 :
Figure 2: Rationale for selecting the following four American Community Survey (ACS) variables Three data sets were used to assess the CT eligibility of rural South Carolina populations.First, data sets describing South Carolina Medicaid recipients and those covered by the SHP were used.Data were provided by the South Carolina Office of Research and Statistics (ORS; http://ors.sc.gov).Individuals in these data sets were grouped by age and CRGs by county.The Medicaid data set compiled information on the population of South Carolina 18 years and older insured by Medicaid, excluding the optional coverage for women and infants (OCWI).The SHP data set compiled information on the population of South Carolina 18 years and older insured by the SHP.

Table 1 :
Perceived barriers to clinical trial participation by population (general public, rural residents) * Significant at p<0.05 CI, confidence interval.df, degrees of freedom.

Table 2 :
Estimated rural and urban eligibility for clinical trials