A roadmap for precision medicine research recruitment: empirical assessment of the public’s willingness to participate

Abstract

Aim:

Precision medicine research recruitment poses challenges. To better understand factors impacting recruitment, this study assessed hypothetical willingness, public opinions of and familiarity with precision medicine research.

Materials & methods:

Adult attendees (n = 942) at the 2017 Minnesota State Fair completed an electronic survey.

Results:

Few respondents had heard of ‘precision medicine’ (18%), and familiarity came mostly from media (43%). Fifty-six percent expressed hypothetical willingness to participate in precision medicine research. Significant predictors of willingness were: comfort with unconditional research; perceiving precision medicine research as beneficial, trustworthy and confidential; having a graduate degree; comfort with self- but not family-participation; and familiarity with precision/personalized medicine.

Conclusion:

This study identified predictors of hypothetical willingness to participate in precision medicine research. Alternative recruitment strategies are needed.

Precision medicine, which is widely expected to transform medical practice [1], incorporates consideration of an individual’s genetic, environment and lifestyle information in order to help prevent and treat disease [2]. Precision medicine research studies are under way across the world to build the evidence base for successful integration into clinical care [3]. The NIH’s ‘All of Us’ program (previously known as the Precision Medicine Initiative [PMI]) comprises the largest US effort to date, with an aim of recruiting at least 1 million individuals nationwide and an emphasis on including underrepresented populations [4].

Large cohort recruitment

An essential element of successful precision medicine research is the recruitment of cohorts of hundreds of thousands of participants. Large datasets are necessary in order to have sufficient statistical power to detect significant associations among genetic, environment, and disease factors [5]. Reports of recruitment difficulties for research studies such as ‘All of Us’ and ‘BabySeq’, which incorporate genomic sequencing, highlight the importance of determining barriers to successful recruitment. The ‘All of Us’ study aims to recruit one million or more individuals who are asked to consent to blood draws, genomic testing [6], access to their medical records, and wearing devices to monitor their health [7]. As of early 2019 [8] ‘All of Us’ had recruited over 200,000 participants [7]. Potential participants have expressed questions about what precision medicine is, how the research impacts clinical care, and how their data will remain private and secure, especially given that access to their medical record is requested [9,10]. ‘BabySeq’ has also experienced recruitment challenges, with only 7% of families approached agreeing to participate [11]. Those who declined cited being uninterested in research, burdensome study logistics (e.g., blood draw), and privacy/insurability as the main reasons [11,12]. The Clinical Sequencing Exploratory Research (CSER) Consortium (and subsequent Clinical Sequencing Evidence-Generating Consortium) reported comparable concerns regarding their efforts to recruit participants for studies of the associations between genomic sequence and various diseases and has worked to engage diverse populations through developing culturally and linguistically appropriate consent forms with innovative technologies [13–15]. Recruitment challenges are ubiquitous in scientific research, but precision medicine research faces additional potential barriers. Innovative recruitment approaches are needed to optimize recruitment and include a broad and representative cohort of participants.

Lack of familiarity with precision medicine is a known barrier to recruitment for precision medicine research [16–19]. A 2018 survey by GenomeWeb and the Personalized Medicine Coalition, an industry group whose goal is to promote understanding and adoption of personalized medicine, found that only 15% of respondents reported having heard the term ‘precision medicine’ [19]. Another potential barrier is changes in terminology from ‘personalized’ medicine to ‘precision’ medicine [20]. Recruitment of representative samples presents yet another barrier. Recruitment for precision medicine research occurs mainly through healthcare facilities or crowdsourcing from online volunteers. Cohorts are therefore often composed of convenience samples who are receiving clinical care, and/or have access to internet services and electronic devices [21]. Crowdsourcing capitalizes on technology’s ability to reach numerous potential participants quickly and inexpensively but results in higher participation by female, White, younger, married, and college-educated individuals [22,23].

The All of Us program has used a variety of techniques to increase diversity in their recruitment. For instance, recruitment is occurring both through healthcare provider organizations and through a direct volunteer program that engages people through the internet or community-based efforts [24]. All of Us plans to recruit through community libraries and blood banks, among other sites, to increase diversity [25,26]. The use of community-based recruitment is an attempt to reach populations previously underrepresented in medical research when researchers draw primarily from healthcare provider organizations [21]. Although numerous efforts are under way to diversify study participants, multiple investigators have expressed concern that precision medicine studies could lead to increased disparities in healthcare if their efforts to recruit diverse cohorts are unsuccessful [27–29]. Enrollment of diverse and representative cohorts is essential to successful development of precision medicine and integration into clinical care.

Purpose of the study

Prior investigations have identified a number of individual factors that comprise potential barriers to recruitment. To date, however, no single study has investigated the relative contribution of these factors to the public’s willingness to participate in precision medicine research. Accordingly, the purpose of this study is to assess the impact of the public’s familiarity and opinions (attitudes and perceptions) regarding precision medicine research and selected demographic characteristics on their ‘hypothetical willingness’ to participate in precision medicine studies. This multiple variable study adds to prior research through the simultaneous consideration of several possible barriers to recruitment for individuals sampled outside the healthcare system.

The current investigation was conducted in the summer of 2017, right before the official launch of All of Us. We thus sampled public attitudes prior to intensive efforts to recruit individuals for this major federal program. We also surveyed individuals unselected for healthcare concerns and outside of a healthcare setting, to assess broad public willingness to participate in precision medicine research. The findings will assist researchers in tailoring recruitment strategies to match the needs and desires of potential participants in the general public, with the goal of increasing recruitment for precision medicine research [30].

Materials & methods

Sample & procedures

This project was approved by the University of Minnesota Institutional Review Board. Participants were recruited from a population of attendees of the 2017 Minnesota State Fair. Recruitment for a survey study of ‘opinions on medical research’ occurred over a 5-day period. Eligibility criteria were: English speaking, able to read, and at least 18 years old. All participants received a University of Minnesota drawstring bag upon completion of the survey.

Instrumentation

The research team considered past published survey instruments used to assess familiarity with and willingness to participate in precision medicine research [31]. The team iteratively developed a 75-question online survey to assess familiarity with, attitudes and perceptions about, and hypothetical willingness to participate in precision medicine research. The survey included a number of questions modified from prior studies of the public’s familiarity, attitudes and perceptions about, and willingness to participate in activities that included the Precision Medicine Initiative, biobank research, biomedical research, genetics research, and genetic counseling [31–34].

Participants were first asked if they had heard the term ‘precision medicine’ or the term ‘personalized medicine’ (Scale: Yes/No/Not Sure). Follow-up questions asked how familiar they were with each term (4-point scale: Little or no familiarity to Very familiar), and how they heard about each term (Checklist). Next, participants read a definition of precision/personalized medicine and an example of a precision medicine research study.

Public attitudes on precision medicine research were determined through items on the topics of privacy, consent, specimen storage, family involvement, participant recruitment, trust, and return of results. To minimize order effects, the presentation of items in this section was randomized. Participants indicated whether they were comfortable with, not comfortable with, or not sure about statements provided regarding precision medicine research. Opinions on privacy were assessed by addressing personal identifiers being included or removed from biological samples for research, access to medical records, and access to genetic information by research groups and/or a national database (e.g., My genetic information is sent to a national database). Preferences between broad consent versus study-by-study consent, when to opt out of research, how participants’ engagement changes by disease type were also included (e.g., I can give written permission each time my genetic information is used in a study). Participants’ opinion on family involvement, a critical component of research studies on genetic aspects of a disease, was assessed through participants’ desire to have involvement of children, involvement of extended family members and/or participant only involvement (e.g., My child is also participating in the study). Attitudes towards recruitment in a precision medicine research study were determined through two statements concerning an individual’s preference for participation with or without a healthcare provider’s recommendation (e.g., A doctor or healthcare provider recommends me to participate in a study). Lastly, attitudes toward storage of genetic information were assessed by two statements about how long samples would be maintained (e.g., My DNA is stored only while I am living).

Then participants indicated their hypothetical willingness to participate in precision medicine research (4-point scale: Little or not willing to Very willing) for five different medical conditions representing future precision medicine investigations to be conducted at the University of Minnesota: cancer, a bad reaction to a prescription medication, dementia, heart disease and depression. To minimize order effects, the scenarios in this section were presented randomly.

Participants next reported their perceptions of precision medicine research by indicating their agreement with statements about precision medicine (4-point scale: Disagree to Agree). Participants’ perception of precision medicine research was assessed through items addressing the purpose of precision medicine research, alignment of personal values with research, privacy, impact of research results, who benefits from the research, and disparities in precision medicine research (e.g., Findings from precision medicine would help people improve health). Items also addressed whether individuals perceive that results from precision medicine research would be confidential, if these types of studies are upsetting or lead to family conflict, and if respondents perceive that precision medicine research benefits family members, drug companies, individuals with rare medical conditions, and individuals with a family history of a disease (e.g., Findings from precision medicine research would be too upsetting for me). Items also addressed the public’s perception of how precision medicine research projects could influence disparities. Specifically, these items assessed the public’s perception of the effect of precision medicine research on health insurance, employment, discrimination, disparities in healthcare for people of different socioeconomic statuses and disparities in healthcare for different groups (e.g., Findings from precision medicine research would cause certain groups of people to be discriminated against). Finally, attitudes toward return of results from precision medicine research were ascertained through items addressing in what circumstances results should or should not be returned (e.g., I would only want to know precision medicine research findings about my health if there is a treatment). Then participants noted their overall willingness to participate in a precision medicine study (Scale: Yes/No/Not sure), and the reason for their response (Open-ended).

The final section of the survey contained demographic questions (e.g., age, gender, race, ethnicity, education, relationship status, and income), and questions about participants’ personal medical history, family medical history, and experience with medical research and genetic testing. Piloting of a draft of the survey with six individuals from the public (Age range: 20–70 years) resulted in minimal changes and yielded the final version.

Data analysis

Preliminary analyses

We excluded surveys from analysis if they had more than 20% missing data. A multiple imputation procedure provided estimations for items with missing responses for surveys that had less than 20% missing data (352 surveys), a procedure supported by Schlomer et al. (2010) to increase the statistical power of the linear and logistical regression analyses [35]. For each missing response, an estimate was completed ten-times using separate regressions with a random error term added for every estimation [35], using the Amelia package for R [36], with the combined analysis following Rubin’s rules [37]. For the linear and logistical regressions combined, a total of 949 (2.14% of overall data) values were imputed with an average of 2.64 imputations per person (range: 1–9).

The next step in the preliminary data analysis involved conducting factor analyses for data reduction purposes. Data reduction allows researchers to maintain a level of statistical power sufficient for detecting significant associations between study variables. Factor analysis groups items according to those which respondents tend to answer similarly. Each group of items constitutes a factor to which the researchers assign a label representing the concept reflected across the items. In the present study, separate exploratory factor analyses using principal axis factoring with promax rotation were conducted on the 19 items assessing participant attitudes and the 20 items assessing perceptions, respectively.

Multiple variable analyses

We conducted separate linear regression analyses to identify significant predictors of willingness to participate in each of the five precision medicine research study scenarios (cancer, bad reaction to a prescription medication, dementia, heart disease, depression), and a logistic regression analysis to determine which variables were significant predictors of general willingness to participate in precision medicine research. These types of multiple variable analyses yield models that allow researchers to determine the relative unique and combined contribution of several predictors to the variance in the dependent variables (in this case, willingness). The Akaike Information Criteria (AIC) assessed the model fit for the data in each logistic regression.

Qualitative data

We used interpretive content analysis methods (as described in Giarelli and Tulman [38]) to analyze comments to an open-ended question asking participants to explain their response to the question, “Would you be willing to participate in a precision medicine study?” One investigator (KM) reviewed the responses, grouping them based on their conceptual similarity and assigned each group a label representing the common theme reflected therein. Another investigator (PMV) served as data auditor, reviewing each response and assigned theme. They engaged in discussion to reach concordance on any disagreements.

Results

Sample demographics

Nine-hundred fifty-three individuals began the survey. Of these, 11 either did not meet inclusion criteria (n = 6) or stopped the survey before completion (n = 5). Thus, the final sample consisted of 942 participants. Participants who responded to most, but not all questions, were included in the analysis. The total number of respondents for each question varies.

Table 1 contains a summary of respondent demographics. A majority were female (59.5%), identified themselves as White (87.2%), and were not of Hispanic, Latino, or Spanish ancestry (89.7%). Age varied, with 39.5% less than 50-years-old, half (50.1%) between the ages of 50 and 69, and 10.4% were 70 years old or older. A majority of the participants were married (60.4%) and highly educated, with over half (59.7%) having at least a Bachelor’s degree. About two-thirds (66.4%) reported an annual household income of USUS$50,000 or greater, and 36.2% did not have children. Table 1 also contains Minnesota census data for comparison to the present study sample. Comparison of the study data and Minnesota Census data indicates that African American, American Indian, Alaskan Natives and individuals with Hispanic, Latino or Spanish ancestry were under represented. The current study sample also comprised an older and wealthier population, which likely reflects the survey recruitment setting.

Table 1.

Demographics of study participants.

Demographic	Current study sample n = 942		State of Minnesota (2010 Census) n = 5,303,925
	n	%	n	%
Gender:
– Female	562	59.9	2,671,793	50.4
– Male	366	39.0	2,632,132	49.6
– Prefer not to answer	10	1.1
– Other	1	0.1
Age:
– <20	33	3.7	1,431,211	26.9
– 20–29	135	14.9	728,337	13.7
– 30–39	82	9.1	671,090	12.7
– 40–49	107	11.8	759,107	14.4
– 50–59	224	24.8	751,284	14.2
– 60–69	229	25.3	482,345	9.1
– 70–79	89	9.8	273,971	5.2
– 80 and above	5	0.6	206,580	3.9
Education:†
– <9 years of school	0	0.0	158,858	5.1
– Some High School	4	0.4	222,487	7.0
– High School Graduate or GED	74	7.9	912,672	28.8
– Some College, no degree	144	15.3	759,153	24.0
– Associate’s degree	93	9.9	243,093	7.7
– Bachelor’s degree	326	34.6	605,210	19.1
– Master’s degree	184	19.5	171,023	5.4
– Professional school degree	44	4.7	63,444	2.0
– Doctorate degree	52	5.5	28,405	0.9
– Other	14	1.5
– Prefer not to answer	7	0.7
Relationship status:†
– Married	567	60.4	2,172,676	56.3
– Never Married	224	23.9	1,083,369	28.1
– Divorced	88	9.4	336,648	8.7
– Widowed	36	3.8	224,775	5.8
– Separated	9	1.0	40,287	1.0
– Prefer not to answer	15	1.6
Income:†
– <US$10,000	20	2.1	127,955	6.7
– US$10,000–$14,999	16	1.7	102,205	5.4
– US$15,000–$24,999	26	2.8	216,089	11.4
– US$25,000–$34,999	36	3.8	234,300	12.4
– US$35,000–$49,999	69	7.3	322,529	17.0
– US$50,000–$74,999	150	16.0	424,867	22.4
– US$75,000–$99,999	142	15.1	228,834	12.1
– US$100,000–$149,999	178	18.9	156,565	8.3
– US$150,000–$199,999	85	9.0	40,734	2.1
– US$200,000 or more	69	7.3	42,131	2.2
– Prefer not to answer	149	15.9
Race:
– White	819	87.2	4,524,062	85.3
– Asian	44	4.7	214,234	4.0
– Multiracial	33	3.5	125,145	2.4
– Black or African American	12	1.3	274,412	5.2
– Hawaiian or Other Pacific Islander	4	0.4	2,156	0.0
– American Indian or Alaska Native	3	0.3	60,916	1.1
– Other	9	1.0	103,000	1.9
– Prefer not to answer	15	1.6
Ethnicity:
– Not Hispanic, Latino or Spanish	786	89.7	5,053,667	95.3
– Hispanic, Latino or Spanish	17	1.9	250,258	4.7
– Prefer not to answer	73	8.3
Children:
– No children	329	36.2	No data available
– ≥1 child	581	63.8

n varies as not every participant answered every question; some percentages do not total to 100 due to rounding.

^†Indicates use of 2000 Minnesota Census data. Source for Minnesota State Census Data: https://factfinder.census.gov/faces/nav/jsf/pages/index.xhtml#

Familiarity with precision medicine & personalized medicine

The majority of participants, 72.3% (n = 680/940), had never heard the term ‘precision medicine’, 18% had heard the term, and 9.7% were unsure. Of those who had heard the term, 18.0% (n = 30/167) reported being familiar or very familiar with it, and the most common source was the media (n = 72), followed by a class (n = 36), a family or friend (n = 24), and personal experience or the experience of someone they know with precision medicine research (n = 17). Some individuals (n = 18) indicated having heard the term from other sources: their job, Facebook, telehealth providers, a hospital, and an online class. A few individuals endorsed more than one source. Some participants (n = 24) could not remember or were unsure where they heard the term (n = 24).

Unlike the term ‘precision medicine’, the majority of participants, 61% (n = 560/918), had heard the term ‘personalized medicine’, 30.5% had not heard the term, and 8.5% were unsure. When those who had heard the term were asked how familiar they were with it, only 19.0% (n = 105/554) reported being familiar or very familiar, and the most common source was the media (n = 265), followed by a family member or friend (n = 94), a class (n = 73) and personal experience or the experience of someone they know with personalized medicine research (n = 45). Some individuals (n = 42) indicated having heard the term from other sources: job, Facebook, the newspaper, a podcast, literature, their physician, and an online class. Some participants endorsed more than one source. A number of individuals (n = 107) indicated they could not remember or were unsure where they heard the term.

Attitudes & perceptions regarding precision medicine research

Tables 2 & 3 contain descriptive statistics for participants’ responses to individual items assessing their attitudes about and perceptions of precision medicine research. Items in these tables are grouped based on their factor loading, determined by the factor analysis described previously and reported in Supplementary Tables 1& 2.

Table 2.

Participant attitudes toward precision medicine research (n = 942).

Response	n	Comfortable	Not sure	Not comfortable
Attitude factor 1: conditional research
A doctor or healthcare provider recommends me to participate in a study	937	88.0%	6.5%	5.4%
I can give written permission each time my genetic information is used in a study	937	83.7%	7.8%	8.5%
I can decide which diseases are studied using my genetic information	934	83.0%	9.4%	7.6%
Researchers will remove my genetic information from the study should I decide to leave the study	934	82.5%	8.8%	8.7%
Personal information that can identify me is removed (e.g., name and date of birth)	939	82.2%	11.0%	6.8%
I am the only person in my family involved in the study	937	73.6%	20.1%	6.3%
My genetic information is only used by one research group	935	73.0%	16.5%	10.5%
My DNA is stored only while I am living	936	67.5%	18.4%	14.1%
A researcher does not have access to my medical records	934	63.4%	19.8%	16.8%
Attitude factor 2: unconditional research
Researchers can decide which diseases are studied using my genetic information	935	79.9%	9.5%	10.6%
A researcher can access my medical records	938	60.8%	15.0%	24.2%
My genetic information is sent to a national database	940	58.7%	15.0%	26.3%
My DNA is stored indefinitely (including after my death)	941	57.0%	17.6%	25.4%
I can give written permission once for my genetic information to be used for multiple studies	938	50.1%	16.1%	33.8%
Personal information that can identify me is included (e.g. name and date of birth)	938	49.4%	18.1%	32.5%
Researchers will use my genetic information in the study even if I decide to leave the study	939	45.2%	14.4%	40.5%
Attitude factor 3: only individual participation
I decide to participate in a study without a doctor or healthcare provider recommendation	941	67.3%	14.7%	18.1%
My extended family (e.g., parents, grandparents, aunts, uncles) are also participating in the study	938	60.2%	24.0%	15.8%
My child is also participating in the study	936	45.8%	31.3%	22.9%

Some percentages do not total to 100 due to rounding.

Table 3.

Participant perceptions of precision medicine research (n = 942).

Response	n	Agree	Somewhat agree	Somewhat disagree	Disagree
Perception factor 1: precision medicine research is beneficial, trustworthy and confidential
Findings from precision medicine research would lead to new treatments for some health conditions	935	71.2%	25.5%	2.6%	0.7%
Findings from precision medicine research would be useful to some people with a medical condition that runs in their family.	933	69.8%	26.0%	2.4%	1.8%
Findings from precision medicine research would help people improve their health.	935	68.3%	28.2%	2.6%	0.9%
Precision medicine research is important to improve healthcare.	934	66.6%	28.9%	3.2%	1.3%
Findings from precision medicine research would help some research participants’ family members.	935	65.0%	32.0%	2.2%	0.7%
Findings from precision medicine research would help prevent some medical problems.	933	63.7%	33.7%	1.6%	1.1%
Findings from precision medicine researchers from the University of Minnesota are trustworthy.	933	62.1%	34.3%	2.8%	0.9%
Findings from precision medicine research would be confidential.	937	50.3%	33.4%	11.6%	4.7%
Perception factor 2: precision medicine research has negative social repercussions
Findings from precision medicine research would cause certain groups of people to be discriminated against.	935	11.8%	40.0%	25.8%	22.5%
Findings from precision medicine research would make healthcare inequalities worse for certain groups.	935	11.1%	39.9%	26.5%	22.5%
Findings from precision medicine research would cause some people to lose their health insurance.	936	11.1%	38.4%	27.8%	22.8%
Findings from precision medicine research would cause some people to lose their jobs.	936	7.4%	31.0%	32.6%	29.1%
Perception factor 3: precision medicine research has negative personal repercussions
I would only want to know precision medicine research findings about my health if there is a treatment.	940	16.28%	25.74%	25.11%	32.87%
Findings from precision medicine research are only useful to a small group of people with rare medical conditions.	937	9.7%	15.6%	29.6%	45.1%
Findings from precision medicine research would only benefit drug companies.	937	7.2%	21.2%	29.8%	41.8%
Findings from precision medicine research would be too upsetting for me.	936	2.8%	18.3%	31.2%	47.8%
Precision medicine research is “playing God.”	934	1.07%	8.5%	22.5%	68.0%
Perception factor 4: precision medicine research results may have additional impacts
I should be able to decide which precision medicine research findings about my health I would/would not want to know.	935	51.7%	33.6%	9.1%	5.7%
Findings from precision medicine research would benefit rich people and poor people equally.	939	23.5%	32.7%	26.7%	17.0%
Findings from precision medicine research would lead to conflict in some families.	933	22.2%	47.9%	16.8%	13.1%

Some percentages do not total to 100 due to rounding.

The principal axis factor analysis returned three factors for attitude items: Factor 1: Conditional Research refers to items where the participants have more control over the research and their information is kept private; Factor 2: Unconditional Research refers to items where participants have less control in the research and their information is not kept private; and Factor 3: Individual Participation Only refers to items where an individual is participating without a doctor recommendation and the individual’s family is not being involved in the research. There were four factors determined for perception items: Factor 1: Precision medicine research is beneficial, trustworthy and confidential; Factor 2: Precision medicine has negative social repercussions; Factor 3: Precision medicine research has negative personal repercussions; and Factor 4: Precision medicine may have additional impacts. These factors respectively comprise the attitude and perception variables used in the analyses of variables predicting willingness to participate in precision medicine research.

Willingness to participate in precision medicine research

As shown in Table 4, more than 75% of respondents reported being willing or very willing to participate in all five precision medicine research scenarios that were used to describe a study for a specific disease (range: 78% for depression to 85.8% for cancer). In contrast, when broadly asked about participation in precision medicine research without indicating a specific disease, only 55.7% (525/942) of respondents reported being willing to participate, 31% were not sure if they would be willing to participate and 13.3% indicated they were not willing to participate in a precision medicine study.

Table 4.

Willingness to participate in precision medicine for specific medical conditions.

Medical condition	Very willing	Willing	Somewhat willing	Little or not willing
Heart disease (n = 934)	52.1%	32.9%	11.6%	3.4%
Cancer (n = 936)	51.8%	34.0%	10.3%	4.0%
Dementia (n = 936)	50.0%	31.3%	14.1%	4.6%
Bad reaction to a prescription medication (n = 936)	45.5%	34.4%	14.1%	6.0%
Depression (n = 932)	45.3%	32.7%	16.1%	5.9%

Some percentages do not total to 100 due to rounding; n varies as not every participant answered every question.

Linear regression analyses: participation willingness for condition-based scenarios

An exploratory model-fitting process, using linear regression based on AIC, assessed the contribution of individual predictors to the variance in willingness to participate in each of the five specific-disease precision medicine research scenarios. For these analyses, possible predictors (variables) included: self-reported familiarity with precision/personalized medicine (Scale: 0–10, with a higher number indicating greater familiarity); age; urban or rural residence; gender; race (White or non-White); ethnicity; level of education (less than Bachelor’s degree, Bachelor’s degree or graduate degree); having children; marital status (married or not married); income (<US$50,000, US$50,00–$75,000, US$75,001–$100,000, and ≥US$100,001); number of individuals in household; self-reported religiosity; a personal history of the condition specified in the scenario; a family history of the condition specified in the scenario; the three attitude factors; and the four perception factors.

The final model for each linear regression analysis was derived by beginning with all variables included in the model and successively removing each variable that would produce the greatest drop in AIC, until removing a variable would no longer decrease the AIC. This technique allows the overall model fit rather than the significance of individual predictors to determine the final model. Therefore, each final model is the best fitting model based on the AIC for defining predictors of willingness to participate in each precision medicine scenario. Table 5 contains the final model for each scenario. The following sections describe the variables that were statistically significant predictors of willingness for one or more scenarios, and accounted for 38.4–42.1% of the variance.

Table 5.

Linear regression models of hypothetical willingness to participate in specific precision medicine studies (n = 942).

Predictor	Type of precision medicine study
	Cancer	Dementia	Depression	Heart disease	Prescription medication reaction
Conditional research	0.11*** (< 0.001)	0.11*** (< 0.001)	0.06* (0.02)	0.13*** (< 0.001)	0.12*** (< 0.001)
Unconditional research	0.08*** (< 0.001)	0.10*** (< 0.001)	0.12*** (< 0.001)	0.11*** (< 0.001)	0.11*** (< 0.001)
Only individual participation	0.19*** (< 0.001)	0.20*** (< 0.001)	0.21*** (< 0.001)	0.19*** (< 0.001)	0.18*** (< 0.001)
Precision medicine research is beneficial, trustworthy, and confidential	0.32*** (< 0.001)	0.31*** (< 0.001)	0.32*** (< 0.001)	0.31*** (< 0.001)	0.28*** (< 0.001)
Precision medicine research has negative social repercussions	0.06* (0.03)	0.06* (0.03)	–	–	0.07*** (< 0.001)
Precision medicine research has negative personal repercussions	−0.20*** (< 0.001)	−0.19*** (< 0.001)	−0.15*** (< 0.001)	−0.14*** (< 0.001)	−0.20*** (< 0.001)
Precision medicine research results may have additional impacts	–	–	–	0.06* (0.02)	–
Age	−0.13*** (< 0.001)	–	−0.10*** (< 0.001)	−0.04 (0.09)	–
Children	0.06* (0.05)	–	–	–	–
Personal history of condition	0.08** (0.002)	–	0.11*** (< 0.001)	0.08** (0.003)	0.04 (0.11)
Family history of condition	0.10*** (< 0.001)	0.04 (0.14)	0.07* (0.01)	0.04 (0.12)	–
Non-Hispanic ethnicity	–	−0.06*** (< 0.001)	–	−0.05* (0.05)	–
Bachelor’s degree	–	−0.04 (0.17)	–	–	0.03 (0.34)
Graduate degree	–	0.03 (0.34)	–	–	0.07* (0.03)

Values presented are β(p); empty cells indicate the predictor was dropped prior to the final model; R² values: Cancer = 0.41; Dementia = 0.41; Depression = 0.43, Heart Disease = 0.41; Prescription Medication Reaction = 0.39.

^*p < 0.05;

^**p < 0.01;

^***p < 0.001.

Attitude factors

Factor 1 (Conditional Research), Factor 2 (Unconditional Research) and Factor 3 (Only Individual Participation) were significant predictors of willingness in all five scenarios. Greater comfort with these aspects of research predicted a greater willingness to participate.

Perception factors

Factor 1 (Precision medicine research is beneficial, trustworthy and confidential) significantly predicted increased willingness in all scenarios. Factor 3 (Precision medicine research has negative personal repercussions) significantly predicted decreased willingness in all scenarios. Factor 2 (Precision medicine has negative social repercussions) significantly predicted increased willingness in three scenarios: cancer, bad reaction to a prescription medication and dementia. Factor 4 (Precision medicine research results have additional impacts) significantly predicted increased willingness for the heart disease scenario only. Factors 2 and 4 were dropped from the other scenarios before the final model.

Demographic predictors

Being younger significantly predicted increased willingness in the cancer and depression scenarios. Having children significantly predicted increased willingness in the cancer scenario. Personal history of the condition significantly predicted increased willingness in the cancer, depression, and heart disease scenarios. Family history of the condition significantly predicted increased willingness in the cancer and depression scenarios. Identifying as Non-Hispanic significantly predicted decreased willingness in the dementia and heart disease scenarios. Finally, having a graduate degree significantly predicted increased willingness for the prescription medication scenario.

Logistic regression: general willingness to participate in precision medicine research

An exploratory model-fitting, using logistic regression based on AIC, assessed significant predictors of willingness to participate in a general precision medicine research study. The possible predictors included in this analysis were the same as those included in the linear regression analyses for the disease-specific scenarios, with two exceptions - a personal history and a family history of a condition. Table 6 contains the final model. Comfort with unconditional research (Attitude Factor 2), perceiving precision medicine research to be beneficial, trustworthy and confidential (Perception Factor 1), comfort with individual participation (Attitude Factor 3), familiarity with precision/personalized medicine, and having a graduate degree significantly predicted increased willingness to participate in a general precision medicine study. The only significant predictor of decreased willingness to participate was perceiving precision medicine research as having negative personal repercussions (Perception Factor 3). These predictors accounted for about 45% of the variance.

Table 6.

Results of logistic regression analysis modeling of overall hypothetical willingness to participate (n = 942).

Variable	B	SE	z	p-value	OR [95% CI]
Intercept	0.20	0.27	0.74	0.46
Familiarity	0.19	0.04	4.69	<0.001	1.21 [1.12–1.31]
Urban	−0.35	0.24	−1.43	0.15	0.71 [0.44–1.14]
Bachelor’s degree	0.01	0.20	0.07	0.95	1.01 [0.69–1.49]
Graduate degree	0.45	0.21	2.11	0.03	1.57 [1.03–2.38]
Married	−0.30	0.17	−1.79	0.07	0.74 [0.53–1.03]
Unconditional research	0.76	0.10	7.86	<0.001	2.15 [1.77–2.60]
Only individual participation	0.42	0.09	4.46	<0.001	1.52 [1.26–1.83]
Precision medicine research is beneficial, trustworthy and confidential	0.61	0.10	6.09	<0.001	1.84 [1.51–2.24]
Precision medicine research has negative personal repercussions	−0.29	0.09	−3.22	0.001	0.75 [0.63–0.89]

Nagelkerke’s R² = 0.44; Model AIC = 932.58; Reference group for education is those with less than a Bachelor’s degree.

Qualitative analysis of written responses

Five hundred twenty-three individuals provided comments to explain their answer to the question, ‘Would you be willing to participate in a precision medicine study?’ Fifteen responses were not clearly stated and were unable to be categorized. The remaining comments were divided into groups based on the individuals’ original response of ‘Yes’, ‘No’ or ‘Not sure’. Some responses were complex, and therefore they were classified within multiple themes. A summary of themes and illustrative examples of participants’ comments is included in Supplementary Table 3. The most prevalent reasons for indicating ‘Yes’ for willingness to participate included: benefit to others, desire to help science/medicine, belief that research is valuable, benefit to the participant, benefit to the family, prior experience with research, and curiosity.

Some participants indicated their willingness would depend on the specifics of the study, the disease being studied, and a desire for more information before agreeing to participate. Other responses indicated concerns about privacy or discrimination, a lack of trust in the research, and fear of the results. A few responses also suggested these individuals would be willing to participate if the research would lower the cost of healthcare or be beneficial to themselves or others. Those who were unsure or not willing to participate expressed concerns such as cost, involving family members, privacy, and effort.

Discussion

The present investigation sampled a large population outside of the healthcare setting to learn more about the general public’s views of precision medicine research. The findings highlight the public’s interest in participating in precision medicine research studies, and indicate specific factors (familiarity, attitudes, perceptions and demographic characteristics) associated with greater or lesser hypothetical willingness to participate in research of this nature. This multiple variable investigation provides evidence regarding the unique contributions of factors that have been studied previously, and the findings offer new insights about recruitment for precision medicine research.

Familiarity with precision & personalized medicine

Consistent with previous studies [17,19,21], only a small percentage of the present sample had heard the term ‘precision medicine’. Use of the term ‘precision medicine’ only recently became commonplace after then-President Obama’s 2015 State of the Union address in which he announced the PMI. In the study, participants reported greater awareness of the term ‘personalized medicine’ than ‘precision medicine’. This finding may be due to the former term’s longer history [18,20]. In any case, increasing public awareness and understanding of terminology used to describe research is a necessary and important first step.

The most commonly reported source of hearing the terms precision/personalized medicine in the present study was the media. Issa et al. (2009) similarly found that most of their participants gained awareness of personalized medicine through the media [39]. These findings suggest that the media (internet, TV, magazines, etc.) may be helpful tools to increase awareness of and interest in participating. In another investigation, however, only 8% of participants said they would participate in a precision medicine study if recruited through a post on social media [30]. Therefore, while media may promote increased awareness, actual recruitment efforts may depend on the type of media. Researchers may thus need to take a multimedia approach to recruitment.

Willingness to participate in precision medicine research

This study found that 56% of individuals were willing to participate in a precision medicine study without specification of a disease scenario, which is comparable to previous research, including nationwide studies of willingness to participate in precision medicine research or biobanks [30,31,34,40–43]. Two prior investigations of smaller samples responding to a hypothetical personalized medicine study and biobank showed greater willingness (78.7 and 89%, respectively) [28,44].

In the current study, significant predictors of willingness to participate in a precision medicine research study without a specified medical condition as the focus of the research included: comfort with unconditional research (e.g., researchers can access my medical records, choose the diseases they study, identify me by name, and use my DNA indefinitely); perceiving precision medicine research as beneficial, trustworthy and confidential; having a graduate degree; familiarity with precision/personalized medicine; and comfort with only individual participation (e.g., not include my children or extended family members, participate without a doctor’s recommendation). Past research has similarly identified familiarity, education, and perception of benefits as factors influencing willingness to participate in precision medicine research [25,31,34,44].

Additionally, Kaufman et al. (2016) found younger participants were more willing to participate in precision medicine research, and Sanderson et al. (2017) found White and less religious respondents were more willing to participate in a biobank [31,34]. These findings were partially replicated in the present study, but only for specific condition-based scenarios. Younger participants were more willing to participate in cancer and depression studies, and individuals who self-identified as Hispanic, Latino, or Spanish were less likely to be willing to participate in broad precision medicine research endeavors. When considered together, the current results and those of prior studies suggest researchers should anticipate and address negative attitudes and perceptions in their recruitment efforts in order to increase willingness to participate in precision medicine research.

In the current study willingness to participate in precision medicine research for specific condition-based scenarios (cancer, a bad reaction to a prescription medication, dementia, heart disease and depression) did not vary significantly among the five conditions, but the percentages were much higher than for participation in precision medicine research without a focus on a specific disease. Non-Hispanic White participants were less willing to participate in research on dementia-related conditions. Specifically, the opinions of potential participants who are Hispanic surrounding dementia may be more optimistic and see the disease as a normal part of aging, creating better preparedness to care for and handle the condition [45]. To the best of our knowledge, condition-based scenarios have not been included in previous studies. The present results suggest incorporating condition-based scenarios may increase individuals’ willingness to participate and enhance recruitment.

Themes in willingness responses

A small group of respondents indicated a lack of willingness to participate in precision medicine research, mainly due to concerns about the time/effort required for the research, lack of interest in participation, personal fears, discomfort with the research, a desire for more information and concerns about privacy/trust. These responses mirror previous investigations of decliners for other genetic research [12,13].

Prevalent motivations for those participants who indicated a willingness to participate included: benefit to others, desire to help science/medicine, belief that research is valuable, benefit to the participant, benefit to the family, prior experience with research and curiosity. These motivations have been described in the literature, although the most prevalent reasons for willingness to participate vary across studies [30,31,34,40,44]. Appealing to these motivations may help to engage more individuals in precision medicine research.

Of note, many respondents who indicated they were uncertain about their willingness to participate said their willingness would depend on the specifics of the study, the availability of more information, and logistical/personal considerations. They cited additional reasons such as concerns about privacy, discrimination, risks of the research, involvement of drug companies, and personal fears. Wagner et al. (2016) similarly found that almost half of their sample would like more information before agreeing to participate in a precision medicine study [30]. Overall, our data suggest that providing information about the research may be important for recruiting individuals who initially are ambivalent or unwilling to participate. Further research will help to increase understanding about the public’s desire for more information before participating and how best to incorporate this information into recruitment for precision medicine research.

Study limitations & research recommendations

Several limitations suggest caution in generalizing the results. The sample was recruited from one large public venue, making it a sample of convenience. Although the sample’s demographics are fairly comparable to state demographics for Minnesota, minority populations, those of lower socioeconomic status, and those with less education were not as well-represented compared with overall state-level census data. Findings are thus not generalizable to a national sample and further confirmation of themes is warranted. Additionally, there may be an ascertainment bias, such that those with an interest in medical research were more interested in participating in this study. Another possible limitation concerns the difference in scales used to assess willingness for the condition-based scenarios (4-point rating scale) and overall willingness to participate (‘yes, no, not sure’). The 4-point rating indicated that participants vary in their degree of willingness, and therefore, binary forced-choice options of “yes” or “no” might be misleading. At last, this study assessed hypothetical willingness, which may overestimate actual willingness. In a 2010 literature review on the difference between hypothetical and actual willingness to participate in biobank research, twelve studies found a greater actual willingness than hypothetical willingness, six studies found the reverse and four studies obtained inconclusive results [46].

Additional investigations are needed to characterize individuals who are more and less willing to participate in actual precision medicine research studies. Moreover, some authors have expressed concerns about the lack of diversity in research samples and the need to find ways to increase recruitment of minority populations for precision medicine research studies [27]. Therefore, the current study should be replicated with samples drawn from more diverse and from underrepresented populations in order to further understand factors impacting recruitment for precision medicine studies.

Implications for recruitment of precision medicine research participants

The present findings and those of prior investigations regarding barriers associated with participant involvement can help to inform recruitment for precision medicine research. Specifically, we propose a novel, two-wave recruitment process for precision medicine research. A group of people, the first wave, are those who would be more likely to participate in precision medicine research because they perceive such research as beneficial, trustworthy and confidential; are comfortable with personal data being shared without conditions; and are highly educated. Individuals who would be less likely to participate are those who have logistical concerns, personal fears, lack of trust in the healthcare system, discrimination concerns, discomfort with the research, desire for more information (e.g., about study specifics), and need to be informed about how the study is beneficial. These participants comprise a separate group, a potential second wave that are vital to diverse research efforts.

Individuals in the second wave may need additional information, alternative consent processes and forms, offer of research results, outreach on specific diseases, and the ability to choose where they participate, in order to activate different motivators to participate. All or some of these needs should be addressed in order for the second wave to feel comfortable choosing to participate in precision medicine research [47]. Some people in first-wave recruitment and in second-wave recruitment will refuse participation, and it is important to respect that refusal. Focusing recruitment strategies solely on the first group, however, may ignore important differences between these groups and the potential for second-wave recruits may take additional resources and time. However, these additional education, recruitment efforts, and alternative study designs focused on the concerns of second-wave individuals may ultimately increase participation in precision medicine research.

Conclusion

Precision medicine research requires large cohorts to gather sufficient data to power analyses on a population level [1]. As recruitment for these studies has been challenging [12], further information about barriers to participation is crucial in order to facilitate recruitment of samples that are representative of the entire population. The current study identified significant predictors of hypothetical willingness to participate in precision medicine research in a large sample obtained outside the healthcare setting. The study sample was intended to engage those who may be potential participants in future precision medicine research at the University of Minnesota. Our results suggest that recruitment approaches such as articulating the specific conditions being investigated and advertising precision medicine research through varied media may be effective as general recruitment strategies. The findings further suggest a two-wave recruitment process for precision medicine research may produce more robust samples. A first wave includes those with a high likelihood of enrolling in a precision medicine research study, while a second wave includes those individuals who will need alternative strategies in order to consider participating. Such strategies may help to inform and engage individuals, leading to the ultimate goal of recruiting large and diverse cohorts to ensure that precision medicine research benefits the entire population.

Future perspectives

In the future, clinicians will continue to tailor treatment plans based upon an individual’s genetic, environment, and lifestyle information. Over time, genetic information and the ability to predict health outcomes through precision medicine research will be more accurate and readily available to the public. Progress in precision medicine requires large-scale participation in research. Understanding the variables that effect individuals’ decisions to participate in precision medicine research will support that research and help healthcare providers deliver personalized counseling and treatment to all patients.

Executive summary.

Precision medicine research

• Precision medicine incorporates consideration of an individual’s genetic, environment, and lifestyle information in order to help prevent and treat disease.

• Research studies are necessary to build the evidence base for successful integration into clinical care.

• Recruitment for precision medicine studies has been challenging. Information about barriers to participation is crucial in order to facilitate recruitment of samples that are representative of the entire population.

• The findings of our study will assist researchers in tailoring recruitment strategies to match the needs and desires of potential precision medicine research participants in the general public.