Primary Care Patient and Provider Attitudes About Population Genomic Screening, Informed Decision-making Needs, and the Potential for Chatbot Technology

Abstract

The three CDC Tier 1 conditions (Hereditary Breast and Ovarian Cancer, Lynch Syndrome, and Familial Hypercholesterolemia) are estimated to affect 2 million people in the United States. Population-based genomic screening holds promise in identifying individuals who may not know their risk. This novel study explored primary care provider and patient attitudes toward population screening and the potential for chatbot technology to support informed decision making in the primary care context, perspectives vital to inform scalable, patient-centered approaches to facilitate population screening. We conducted semistructured interviews with primary care patients (N = 20) and primary care providers (N = 9) from primary care clinics at a patient diverse academic medical center. The interviews focused on receptivity to population-level screening, information needs, and the potential for chatbot technology as a mechanism to support patient informed decision-making about testing. Patients and provider participants also reviewed a brief demo of the chatbot technology and shared their views. Interviews were recorded and transcribed. We used rapid qualitative analysis methodology to analyze interview data. Patients and providers acknowledged the benefits of population genomic screening and found the chatbot technology to be easy to navigate. Patients endorsed the utility of screening but raised concerns about data privacy and the desire for more information about the conditions and screening process. Providers gave insights into information that could be integrated to further support patients in informed decision making. Overall, chatbot technology as a facilitator of population screening is a promising approach. The results of this study can improve future efforts to ensure that chatbots and similar technology incorporate vital information to facilitate informed decisions.

INTRODUCTION

Nearly 2 million people in the United States are estimated to be at increased risk for adverse health outcomes attributable to a pathogenic genetic variant associated with one of three conditions:¹ (1) Hereditary Breast and Ovarian Cancer Syndrome, associated with an increased risk for breast, ovarian, and other cancers attributable to variants in BRCA1 or BRCA2; (2) Lynch syndrome, associated with an increased risk for colorectal, endometrial, and other cancers because of variants in mismatch-repair genes (MLH1, MSH2, MSH6, PMS2, and EPCAM); or (3) Familial hypercholesterolemia, associated with an increased risk for heart disease or stroke because of variants in LDLR, APOB, and PCSK9 leading to elevated cholesterol levels. These three conditions are designated as Tier 1 applications by the Centers for Disease Control and Prevention’s Office of Public Health Genomics for having significant potential for positive impact on public health based on evidence that early intervention reduces morbidity and mortality.¹ With current clinical screening practices based mostly on family and personal history, an estimated 90% of people in the United States with genetic predisposition to Tier 1 conditions remain unaware of their risk,² representing a missed opportunity for implementing highly actionable, risk-reducing early interventions. Population-based genomic screening is a promising approach to address this significant health care gap.³

Providers and patients have endorsed interest in broader access to this genetic risk information. In a survey among primary care providers (n = 488), 74% agreed or strongly agreed that genetic testing for common disease risk has clinical utility.⁴ In a recent survey of 746 individuals from the general population, 49% reported that they would undergo population genetic screening to learn about their cancer risk.⁵ However, despite interest, logistical challenges arise when considering expanding genetic testing for Tier 1 conditions beyond those with a known risk. For example, given the number of patients who would need support for informed decision making prior to testing, standard genetic counseling approaches would not be scalable because of limited workforce and cost.⁶ Thus, digital tools such as chatbots are a promising approach to support informed decision-making at the scale of population genomic screening.

Gia^® (Genetic Information Assistant), is a Health Insurance Portability and Accountability Act (HIPAA)-compliant clinical chatbot that facilitates virtual conversations with participants. Gia has shown promise in several contexts related to genetic testing, including obtaining informed consent from patients⁷ and delivering test results⁸ back to patients. These previous findings indicate that Gia may be a suitable tool for providing pretest education and facilitating informed decision-making in the context of population screening for Tier 1 conditions. However, further study is needed to capture patient and provider acceptability and needs in this context to guide implementation of patient-centered technologies.

To address this gap, we aimed to characterize patient and provider perspectives on the utility of chatbots for population screening in the primary care setting. We conducted interviews with primary care patients and primary care providers at primary care clinics in which a population screening research study was being planned to describe attitudes toward population level genomic screening, and to recognize the needs for informed decision-making. Further, we assessed the acceptability and potential effectiveness of a chatbot as an informed decision-making tool.

METHODS

Recruitment

We used convenience sampling to recruit UF Health Jacksonville primary care providers and patients receiving care at UF Health Jacksonville primary care clinics. We conducted 9 interviews with primary care providers, seeking representation from each clinic. We conducted 20 primary care patients given ongoing review revealed saturation in overall themes. Primary care providers were recruited through email invitations. Primary care patients were recruited directly within the clinic setting in collaboration with and via the support of their primary care provider. With the permission of the primary care provider, study coordinators attended the clinic on a specified day. Providers asked patients about their interest in participating during the visit, and those who expressed interest were connected to the study coordinator to provide consent and complete the interview.

Data Collection

Interview guides were tested and refined in mock interview sessions with staff from the health care center who would not be participating in the actual interviews and not involved in guide development. The final interview guides are available in the Appendix.

Interviews were conducted between May and July 2022. Participants were informed that their input was being sought on research that offers genetic testing to patients who are not known to be at risk for heart disease and cancer and what information is important for prospective participant informed decision-making. However, some of the terminology differed. For example, to keep the patient interview guide at an appropriate reading level, we referred to “genetic testing” for “specific types of cancer or heart disease” that may be offered to “healthy people,” whereas for providers we referred to “population screening” for “Tier 1” conditions and named the specific conditions. The framing of the interview guides also differed. The provider interview guide was focused on a population screening study that was planned for their health system that providers were aware of. It aimed to provide an overview of the study, capture questions they may have about the study, explore any hesitations in their patients participating, and obtain their perspectives on the most important and challenging aspects of the study to convey to their patients. In contrast, the patient guide aimed to explore general concepts related to genetic testing such as familiarity and experience with genetic testing and information needed to decide whether to get genetic testing and then probed more specifically about whether they would feel any differently about getting genetic testing through a research study and what information would support decision-making.

Following the discussion pertaining to population genetic screening, both patients and providers participated in a demonstration of the Gia chatbot. Gia is a HIPAA-compliant clinical chatbot in which a virtual genetic counselor engages in a synchronous chat-like conversation with a patient. The chatbot included user interaction at various points including acknowledgement of understanding of the presented information and indication of the desire for more explanation. (Figure 1) Participants either navigated the chatbot demonstration themselves (if they were able to access the link provided) or the interviewer navigated the chatbot demonstration on behalf of the participant per their direction. Participants were encouraged to verbalize their thoughts and reactions as they reviewed the chatbot. Participants were asked an identical set of questions regarding ease of use, understanding, and adequacy of information for patient decision-making around receipt of genetic testing. These questions were a mix of closed and open-ended questions to follow participants’ review and spontaneous reactions during the chatbot demonstration.

Figure 1.

Gia Screen Examples

Qualitative interviews were conducted via Zoom to ensure a convenient and accessible platform for participants. Primary care patient interviewees joined the interview via a study-provided iPad from a private room at the health center just following their appointment to maintain confidentiality and create a comfortable environment for discussion. Primary care providers participated from their offices, joining the Zoom from their own device. The research coordinator, who was an employee of the same health system, assisted participants with connecting to the Zoom interview as needed, but was not present during the interview. To minimize bias and ensure the integrity of the data collection process, interviews were led by one of two public health researchers, who identify as female, from RTI International with each over a decade of experience conducting qualitative research (R.R.M and S.M.A). Interviews were up to 60 minutes long and were audio recorded and transcribed. Participants received a $50 gift card as a token of appreciation. All study materials and procedures were approved by the University of Florida’s Institutional Review Board (IRB). A waiver of written signed consent was granted by the IRB.

Data Analysis and Interpretation

A rapid qualitative analysis was undertaken by three analysts (S.M.A., R.R.M., O.K.). This methodology is commonly used in health services and evaluation research when there is a need to triangulate findings from multiple qualitative and quantitative data sources and quickly obtain insights to make timely decisions regarding strategy and practice.^{9, 10} Rapid qualitative analysis is ideal for studies in which a semistructured guide is used to obtain qualitative data on predetermined constructs of interest.¹¹ This methodology facilitated a deductive content analysis¹² in which analysts categorized data according to pre-determined interview domains (e.g., For patients: perceived risks and benefits of genetic testing; For providers: aspects most important to communicate before and after genetic testing; For both about GIA: Ease of navigation, sufficiency of information for making a decision, etc.).

The research team developed a summary template in which section headings represented key interview domains. The template was refined, and inter-analyst agreement was established on use of the template, via an iterative comparison and consensus process. Two analysts piloted each summary template (patients: S.M.A., O.K.; providers: S.M.A., R.M.) using the same interview transcripts and compared summaries to ensure that the template was ideally structured to capture the interview content relevant to each domain of interest and that the analysts reliably noted the same key points under each domain. Following this process, transcripts were divided among analysts, and the final template was used to create a summary of content related to the domains of interest for each interview. Bulleted summaries were transferred from MS Word documents (one per data collection episode) into an analytic matrix¹³ developed in MS Excel representing all data collection episodes. The rows in the matrix represented topics of interest (i.e., the same topics of interest that were captured by section headings in the summary template) and columns represented the participants. The analytic matrix was used to summarize across all interview participants to identify themes. Illustrative quotes were extracted from transcripts.

RESULTS

Participants

Of the 20 primary care patients who participated, 55% were female and 40% were between 55 and 74 years of age. Patients self-reported their race/ethnicity from the options listed in Table 1 along with “Other,” with 70% reporting White, 25% Black or African American, and 90% indicating not of Hispanic, Latino, or Spanish origin. Race or ethnicity data were not reported for three participants. Nearly half (45%) had an Associate’s degree or higher.

Table 1.

Patient Participant Characteristics

	N=20	%
Age (n)
18–24	1	5
25–34	3	15
35–44	3	15
45–54	3	15
55–64	5	25
65–74	3	15
75–84	1	5
Sex (n)
Female	11	55
Male	9	45
Race (n)
White	14	70
Black or African American	5	25
American Indian and Alaska Native	0	0
Asian	0	0
Native Hawaiian and Other Pacific Islander	0	0
Missing	1	5
Ethnicity (n)
Not of Hispanic, Latino, or Spanish origin	18	90
Missing	2	10
Education (n)
Less than high school	1	5
High school graduate	1	5
Some college, no degree	9	45
Associate’s degree	3	15
Bachelor’s degree	3	15
Graduate or professional degree	3	15
Annual Household Income (n)
Less than 25,000	5	25
$25,000–34,000	4	20
$35,000–49,999	1	5
$75,000–99,999	5	25
$100,000–149,999	3	15
$150,000 to 199,999	1	5
$200,000 or more	1	5

Providers did not self-report demographic information. All nine providers interviewed provided some primary care services to UF Health Jacksonville patients. Most worked in primary care clinics, including two who worked at OB-GYN clinics, considered primary care within the healthcare system.

Attitudes Toward Population-level Genomic Screening

Patients’ Attitudes Toward, Understanding of, and Experience With Genetic Testing

All patients demonstrated a basic understanding of the term “genetic testing.” Many described that genetic testing could tell a person about their health risks and that it reflects traits that are passed through families. Most patients had not personally undergone genetic testing; a few knew someone who had.

Most said that they would consider getting genetic testing with some reporting if it would help others in their family. Patients described two main potential benefits: to inform their or their family member’s healthcare (e.g., health decisions, learn about a condition that may run in the family, ability to tell children if they are at a higher risk) and contribution to medical knowledge.

“ I know that like, on my, on my dad's side that like, there was heart attacks and strokes on that side, but I don't know if it was because he smoked or not. So that's always been in the back of my head. I would love to find out if I have that gene. Because I think if I found out, you know, there could be a way that I can modify it.”

– Patient, male, age 65–74 years, White [P5]

“You know, what, actually, maybe only because I be having so many things these days going on with me. And my dad can say that, like, my aunt dealt with this or someone down the family dealt with this. And it kind of makes me curious, because I know that some things can kind of come from down in the family, and you just don't know. So yeah, I would like to know.”

- Patient, female, age 35–44, Black or African American [P15]

“If I'm at risk for, you know, the familial high cholesterol, then, you know, I'll take preventative measures for that. Instead of waiting until I'm, you know, older for lipid screening, I would do it now. And, you know, pretty much that just kind of keeping an eye on things if I need to”

- Patient, female, age 25–34, White [P14]

P4: There's just so much potential right now. If it could help the, the rest of the medical community to address diseases or pandemics or anything like that, yes, I would be I would consider it beneficial and I would the, the benefits would outweigh the risks.”

- Patient, female, age 75–84, White [P4]

Some patients expressed hesitations toward getting genetic testing. Patients reported concerns over privacy and negative impacts of the results (e.g., potential for employment discrimination, inability to get insurance), anxiety about the results and anticipating a positive result, lack of perceived benefit to testing for older patients for whom “not a lot can be done,” and invasiveness of testing (e.g., fear of blood draws). One participant said they would not consider testing unless they were displaying risk factors or symptoms.

“Come on call me crazy. I think that maybe somewhere down the road that insurance companies might require it before they determine coverage?”

- Patient, female, age 25–34, White [P11]

“Yeah I work in IT. So I've had to have HIPAA training myself. So I know that there's a lot of laws around protecting people's privacy in terms of medical information. But being in it, I also know that people with the right access can easily abuse data”

- Patient, female, age 55–64, race & ethnicity not self-reported [P17]

“Well, at the same time, as I would be, like, excited I’d also be, like, a little nervous. Because once you find out like, exactly, you know... that that could be scary…So it’s kind of like a double-edged sword.”

- Patient, male, age 65–74, White [P5]

“I don’t like pain. So if I can be painful, that would be one thing that might scare me a little bit.”

– Patient, female, age 35–44, Black or African American [P15]

When patients were specifically asked about genetic testing through a research study, their perspectives were similar, with some noting that they did not consider genetic testing in a research study to be any different than as part of clinical care, although more concerns were raised about privacy and confidentiality of research data than about clinical data. Most patients reported that they would consider having genetic testing through a research study.

Provider Attitudes Toward a Population Screening Pilot Study

When probed, no providers reported aspects of a population screening pilot study that would make them hesitant to recommend to their patients, and some made explicitly positive comments about the study being a “good idea” and providing a potential benefit to patients.

“And so, and it’s the way our country is going and so to have an additional resource that we can use to have early identification and have early intervention, I think is something that that could change the way we do medicine. It’s something that I’m excited about. I’m excited to see kind of how it how it works.”

– Provider [P526]

“Regarding the testing, you know, that I am probably biased because I’ve been a researcher and I’m always about getting data to patients and to me. So I’m generally pro - get as much information as we can.”

– Provider [P16]

An overall facilitator of the study described by providers is that most patients are familiar with basic concepts of genetics (e.g., through television) and will likely be interested in genetic testing even without a family history of one of the Tier 1 conditions. Some stated that most patients want to know as much as they can about their health risks and want access to this type of testing.

“I think people will be pretty open to it. Again, know, this is my perspective, but just even learning from my patients, coming to me to ask about genetic testing, I've had a number of patients who want to get tested, particularly with the family history. I don't see lots of pushback. Particularly with the breast and the ovarian. Right? Everybody knows about it. Pretty much if you know the risk, you get concerned about it.”

- Provider [P616]

Providers reported on the potential barriers to engaging patients about the study. One example given was patients’ inherent distrust of the healthcare system among some socioeconomic or cultural groups, such as the Black community distrusting research because of historical abuses, mixed levels of trust in the Coronavirus vaccine, and patient concerns about the safety of their data.

“…or not trusting that “I'm not going to let the government know” they people don't want to even people cover their license plates on their cars, let alone they don't want to let people know their genes and things like that. So I imagine that that's going to one group of people might not be willing to give up their information, even if it's to help them.”

– Provider [P13]

Additionally, providers reported that young, healthy patients and those with no known family history may not be as interested or see a need to participate. Providers recommended that the “major selling point” to communicate is that early identification can lead to improved health outcomes and longer life. And further, providers should communicate the risk of these conditions in the general population so that they will understand that despite a family history, they could be at risk.

Providers also noted that some patients are not a good fit for this type of testing, because knowing their risk in the absence of symptoms or active illness will cause anxiety. Providers recommended that they could have a role in talking to patients about the study and helping them determine appropriateness of participation.

Providers also expressed concerns about potential gaps in communication or lack of clear processes that could result in patients not receiving appropriate follow-up about their results.

Informed Decision-Making Needs

Patient Perspectives

Patients wanted information that affirmed the types of benefits that made them interested in testing (e.g., utility of results to themselves or family) and disclosed potential harms that gave them hesitations (e.g., pain, being denied insurance coverage). Patients also desired information about the process, such as where and how testing happens (e.g., blood draw), timeline from testing to results, what would be tested, what the results would tell them, and information about data security protocols.

“Probably just the process itself is what I would probably be more interested in. Maybe how soon I may get results back, um, stuff like that.”

- Patient, age 45–54, White [P10]

“mostly what I was being tested for and what I can do with it. I mean, it's kind of pointless to have this information if I don't know what to do with it. Kind of be the biggest thing.”

- Patient, female, age 25–34, White [P14]

When patients were specifically asked about genetic testing through a research study, their perspectives were similar, although they would like information about time commitment and compensation and assurance that they would get their results and assurance of the data privacy and would only participate if it was a reputable organization.

“I'm a grandma, I do things when my grandkids everyday, take them to school, and I watch one of my grandbabies. So, all of that is a factor for me. So, if it's something where it's just gonna take maybe an hour or so, then I'm good enough. You know, but if there's something that's gonna be take a long time now, I will be interested, but it wouldn't work for me at this time.”

- Patient, female, age 55–64, race & ethnicity not self-reported [P17]

“I would think someone like this organization, UF Health with that cachet that I would feel much more sure. If it was someone did not have the confidence in that, then I think the, they like I said the I have a lot of confidence in the UF, the health organization.

– Patient, female, 75–84, White [P4]

Provider Perceptions of Patient Information Needs About Population Genomic Screening

Although providers were asked about information and decision-making needs of patients in the context of population screening offered through a research study, some of their responses applied to population screening more broadly. Many providers expressed that it is important to convey the potential benefits of testing to patients, specifically that early identification can improve morbidity and mortality. Additionally, providers deemed that it was important to convey to patients that participation is voluntary, and they aren’t being “targeted” (i.e., it is being offered widely). Providers also expressed the importance of telling patients the population risk for the conditions screened, how their results could have implications for their family, or how family members having these conditions affects their own risk. Further, providers want their patients to know that the genetic testing being offered doesn’t test for all of the potential genetic causes of hereditary breast, ovarian, and colon cancer, and that screening could impact insurability (e.g., life insurance).

“I think that that, you know, if they do have a positive, there's that potential impact on like, life insurance, health insurance, that could I mean, I know that there's the non-discrimination act, but still, there is some barriers to getting, like life insurance. So I think patients understanding that as long as the potential like, because of it being a genetic component, that it impacts offspring, and family members, so knowing that information could impact you know, their family. And then also, like, knowing that the genetic testing doesn't test for all of the major, or all of the potential genetic causes of hereditary breast and ovarian and colon cancer.”

– Provider [P712]

“That this is actually applied to, that you're not targeting a group of people who already feel marginalized. That this is something that’s been offered to, you know, everyone. It’s really a benefit for all people/populations. Not just they’re like a guinea pig for it.

- Provider [P609]

Providers reported that despite patients having some basic familiarity with genetics, there may be concepts more difficult to explain to patients including condition heritability and impact of test results beyond the patient’s immediate health risk (e.g., implications for other family members and insurability). Providers also acknowledged the varied literacy of their patient population and the fact that there will be a subset of the population who need a basic education in what genetic testing is and what it can tell you. Some providers added that it is a challenge to address the needs of those with lowest literacy without making it “boring” for those who already understand. Additionally, providers reported that patients may find this type of testing somewhat “novel” and it can “sound scary.”

Providers shared suggestions to help address challenges including keeping the information that is conveyed to patients as simple as possible (e.g., providing patients with a “1-pager” that is separate from the consent that provides an overview of precision medicine, for example, how long these studies have been in use, what kinds of populations have participated, and how many lives have been saved because of this type of testing), an explanation of what the process entails (e.g., how the sample will be taken), and incorporating stories, analogies, and humor to engage patients.

“Let's give us some bullet points. You know, patients are going to really trust them a whole lot more, and if nothing more than just to get them to the gate, and then you can have the two arms with genetic counselors or the chat bots to kind of give some more details about what is Lynch syndrome etcetera, etcetera, probabilities, they can kind of handle that.

- Provider [P0614]

To address the barrier of potential distrust, providers suggested building on doctor-patient trust and rapport when introducing the study and including doctors of color, and being transparent about the study data in terms of where it is held, how it is used, and whether it is identifiable.

Providers also wanted their patients to be informed after the testing. Some providers suggested that the study team provide primary care providers with information about topics patients might ask about, such as next steps after testing and how long they will take, information about the accuracy of the test, and best practice recommendations for people who screen positive. Providers suggested that patients may need to be notified of their results in multiple ways to ensure that they don’t miss an electronic notification (e.g., call, send a letter, provide via patient portal). Providers also stressed the importance of timely follow-up with patients about their results and clear information on how to interpret.

“Because in our regular referral process, there are cracks, because nothing is perfect. I would hate for a patient of the study to fall through one of those cracks and not the advice of an appointment or not the advised of the lack of an appointment, etc. Whereas if whoever is doing the research project that they can facilitate the happening of that referral, they can be calling whoever needs to be called for this., for that to happen, then that will probably be easy.”

– Provider [P613]

“I would definitely like to get the results of the testing to show me okay, this person does have BRCA1, BRCA2 mutation. But from a follow up standpoint, I definitely would want the genetic counselor and the [study] team to communicate with the patient results and to have a follow up.”

– Provider [P616]

“So just some kind of confirmation thing, so that it's not like, Okay, I've gone on the app or I've talked to the counselor I've done this thing, but then it kind of I get busy with life and things and I just kind of blow off and never really follow up on the results.”

– Provider [P0526]

Acceptability and Effectiveness of Chatbot as a Decision-making Tool

Patients

Almost all patients said that the chatbot was easy or very easy to use. Patients reported that it was good that there were only one or two responses to select from and that the information flowed in “one string.” Two patients indicated that some assistance would be desired when using the chatbot because they were not adept with technology and considerations of vision impairments.

Patients reported that the information was not difficult to understand, and many patients said the information provided in the chatbot was sufficient to decide about participating.

Some patients desired additional information including logistics (e.g., cost, exact procedure, frequency, timing), conditions tested, more information about factors related to the diseases being tested (e.g., cholesterol), privacy and confidentiality, and more explanation of familial disease and heredity to demonstrate why testing is useful.

“Again, just how, like the information just seem to flow and how just didn't seem to just go from one thing to the other… Like as you were thinking about it, like, how long is this gonna take? It's gonna take two to three weeks. All right, the second you still want to proceed in learning more information? Yes, no, it kept giving you options like that as well. So it wasn't like forcing you to know like a bunch of automated information. It was like give me a chance to kind of know what you wanted.”

– Patient, male, age 18–24, White [P1]

“Actually, it was straight to the point, there wasn't a lot of song and dance stuff going on? Just simple. comments and questions.”

– Patient, female, age 65–74, White [P13]

“If you said, Yes, I like to have this testing, then maybe put some way, put another little clip on doing about what's your privacy information, where it's going and all that stuff?”

- Patient, female, 25–34, White [P11]

“I don't know if you can pinpoint things, it'd be nice to know. Exactly. An example of specifics, you know, is it looking at? What do you mean by cardiac diseases? What diseases? That kind of thing, aside from just cancers and cardiac diseases, leaves it very broad. And maybe there's some things that some people don't associate as, as that disease or, you know, that can be better. Maybe if there was more specific, they might be more apt to saying, Oh, well, yeah. You know, grandpa had that.”

– Patient, female, age 25–34, White [P8]

Providers

Providers reported that the chatbot was easy to navigate and self-paced for older patients to use. Providers said that it was generally enjoyable, response options were user-friendly, and the routing options met differing patient needs for information. Providers liked the images used to break up the text and found that the tool was good at parceling the information and flow.

“I think from an idea standpoint, it sounds great. Particularly because I think we have two of the big ones, even the lynch syndrome. So I think targeting these specific conditions, making history. And again, the fact that a chatbot initially is having an informative visit, I think is definitely sounds interesting.”

–Provider [P0616]

“… from my perspective, it's very easy and even from a patient's perspective, you know, I, my grandparents, my grandfather's 91, and I think he could kind of get through this thing without too much issues and stuff.

- Provider [P0526]

Providers shared that the chatbot generally used plain, personable language, and content was written at a level that most patients can understand, including patients with low literacy. Providers appreciated that when a patient clicks to learn more, the information answers questions rather than creating new ones. Further, providers described that concepts were explained well (e.g., what is meant by “blood relatives” and actionability of the tests).

Providers reported that the chatbot mostly provided the right information for participants to make an informed choice. They commented that the chatbot mentioned important key concepts like difference between ancestry testing and genetic testing, possible need to inform family, and that negative results don’t mean the patient will never have cancer or heart disease. Providers discussed that some older patients may be reluctant to participate because of internet connectivity and generally not being “up on technology” as it related to the use of a chatbot. As such, providers recommended that patients be given clear explanations of what to expect, to give them confidence to navigate the system.

Providers offered suggestions for improving the accessibility of the chatbot for patients including incorporation of an audio feature for patients with low literacy or vision impairments, more visuals, making navigation buttons more prominent, and a feature that enables patients to type in questions.

“I'll be I'll be honest with you. I have had a couple of patients who have told me that they can read but their literacy is low and they're slow readers and so if it if it was able to read it as they're reading it, if there was some kind of audio overlay I think it would help some people kind of navigate it a little easier.”

– Provider [P0526]

DISCUSSION

Our study sought to understand perspectives of patients and primary care providers who would be directly impacted by a population genomic screening research study. Participants were receptive to the implementation of population genomic screening and the use of a chatbot to provide pretest education. Patients held similar views about genetic testing regardless of whether it was part of clinical care or offered as a research study. Participants provided feedback around information that should be provided to patients when communicating population screening opportunities (Figure 2).

Figure 2.

Patient and Provider Recommendations for Communicating Population Screening Opportunities

When considering genetic testing offered through a research study, some patients reported concerns related to their privacy, confidentiality of their health data, and how their screening results may negatively impact health or life insurance coverage. Similar concerns about privacy arose in interview studies conducted with Vietnamese American primary care patients about population genomic screening.¹⁴ These studies demonstrate that patients need information to address privacy and insurability concerns and education about existing protections like the Genetic Information Nondiscrimination Act (GINA). Additional education for providers may also be beneficial, as providers in our study anticipated such questions and concerns from their patients. Similarly, in a mixed-methods study of interviews and surveys with primary care physicians involved in a pilot population genetic testing program, providers voiced concerns about privacy and confidentiality of their patients’ information and were unfamiliar with GINA.¹⁵

Our study underscored the importance of engaging with providers as stakeholders prior to the implementation of a population genomic screening pilot. Additionally, not reported in this paper, we asked providers what questions or concerns they had for the investigators leading the pilot population-based genomic testing study that was being planned to take place in their healthcare system. While these questions were specific to the planned randomized controlled trial study, thus not reported here, providers had questions related to the study design, and the study’s impact on their clinic and workflow.

Providers in our study were not averse to their patients taking part in a pilot study offering screening for Tier 1 conditions, citing the benefits of testing. However, they also openly discussed potential implementation barriers within their healthcare system and made suggestions for how to best equip themselves with the knowledge and tools they need to support their patients during both the decision-making process and post-testing follow-up. Interviews conducted with primary care providers of Latino patients at a community-based health center who received positive screening results revealed that despite the benefits of testing, there were also challenges related to the follow-up of patients who tested positive. This included the coordination of care but also inability for patients to receive care based on their financial status.¹⁶ Providers in our study anticipated such barriers. Use of the chatbot technology was not viewed as a barrier to patients or providers, but some participants suggested modifications to enhance accessibility and engagement (e.g., addition of an audio component, visuals) and the inclusion of additional information important for decision-making. Schmidlen and colleagues¹⁷ conducted a qualitative assessment of Gia as it relates to acceptability, usability, and understanding of study consent, and use for the purposes of post-results follow-up. These participants were actively involved in the MyCode Community Health Initiative and thus had more familiarity with genetic research. Authors reported that their participants supported the use of a chatbot to consent for genetic research and concluded that future research is needed to explore acceptability with diverse patient populations and seek perceptions of healthcare providers. Our study results confirm acceptability of Gia among providers and patients of a diverse healthcare system fills this need. Although other assessments have been conducted with Gia, our study is the first to seek acceptability of its integration in population-level genomic screening. Our findings illustrate that a chatbot may facilitate informed decision-making by patients and make such screening feasible.

This study was not without limitations. Interview concepts were simplified to be accessible to a low-literacy population, which may limit generalizability of the findings. For example, interviewers explained to patients at the beginning of the interview that they would be asked about a research study offering genetic testing for three health conditions to patients who are not known to be at risk for having them. The context of testing not being driven by risk factors was not reiterated; instead, questions referred broadly to “genetic testing for certain cancers or heart disease.” It is unclear whether patient responses reflect an understanding of the distinction between getting genetic testing with and without known risk factors or family history. Additionally, it is unclear whether any patients interviewed had a personal or family history for any Tier 1 conditions, as this was not an explicit screening criterion. So, we cannot say for certain whether this affected their responses.

At time of this formative research, a pilot study was planned for the primary care division of the health system. The protocol was described to primary care provider participants. Although this may have led to a more relevant, richer discussion, a different protocol/approach to the pilot study may have yielded different provider opinions.

Participant perspectives may not be representative of the broader provider and primary care patient population as convenience sampling was used to recruit participants within a single healthcare system. Further, we acknowledge there may have been some interview selection biased introduced by the provider weighing in on which patients to offer the interview.

Our study assessed receptivity to population-level genomic screening and acceptability of a chatbot from primary care patients who had limited personal experience with genetic testing. Our study can inform future studies/use of a chatbot based on the perspectives received, including what adaptations could be made generally (benefits of genetic testing to self and family members), for application in population level screening or for site-specific uses where the information may be variable (what conditions are tested, how the test is done, etc.). Additionally, providers gave insights into what information they might need to support patients and the implementation of a population-level genomic screening study at their health center, which may lead to healthcare provider endorsement and successful execution.

What is known about this topic:

The Centers for Disease Control and Prevention (CDC) has classified a set of Tier 1 genes that provide vital information on risks for hereditary cancer syndromes and familial hypercholesterolemia, thus suitable for genomic testing in primary care. Population-based genomic screening will be more widely offered to patients.

What this paper adds to the topic:

This paper provides insights into primary care patient and provider views on population genomic screening at a diverse medical center where a study was being planned. Informed decision-making needs expressed by interviewees and receptivity to chatbot technology holds promise for future implementation of population-based genomic screening studies in primary care settings.

Appendix: Formative Research Interview Guides

For Primary Care Providers [Healthcare system name redacted]

Welcome & Introduction

Thank you very much for agreeing to participate in this interview that will explore what you think about research into population screening among primary care patients who are not known to be at increased risk for a hereditary condition. As you know, most patients undergo genetic testing when there is a known family history. Population screening would mean offering genetic testing for a condition to all patients, regardless of family history.

Today, the research team hopes to learn from you what thoughts and questions you have about your primary care patients participating in a study that offers genetic testing for three conditions that were designated as Tier 1 tests (ready for population screening) by the Center for Disease Control (CDC), the American College of Medical Genetics and Genomics and the Institute of Medicine. Those conditions are: inherited breast and ovarian cancer, Lynch syndrome (familial colon cancer) and familial hypercholesterolemia. The information we gather from this interview will help us to design a population screening study that facilitates the opportunity for primary care patients to make an informed choice about whether to accept or decline the testing that is offered.

Description of Research Study Involving Genetic Testing for Risk Assessment

First let me describe the population screening study that is being planned for the [Healthcare system name redacted] primary care clinic patients in partnership with [partner organization name and funder redacted]). We will offer primary care patients the opportunity to participate in a randomized controlled trial that compares the offer of genetic testing presented by a genetic counselor to that of a chatbot which is called the Genetic Information Assistant (GIA). Both the genetic counselor and GIA will describe the three conditions that are tested for and how they can be managed to reduce negative health outcomes for those who test positive for an increased risk. In comparing the effectiveness of the study arms, we will assess the proportion of participants that made an informed choice whether to undergo testing. That is, how many had sufficient understanding of essential information to select an option that was in keeping with their values.

Discussion of Genomic Sequencing Pilot Study

1. If your primary care patients are offered the chance to participate in a study like the one just described, what questions would you have for the study investigators about the study? [At this stage, the intent of me asking this is to help the study team understand physicians’ concerns so that we can address concerns and refine the study. So even if I don’t answer your questions, they will be conveyed to the study team, and [name redacted] is also on site to answer any questions you might have right after the interview.]

2. Is there anything that would make you hesitant to recommend the study to your patients?

3. What aspects of the study do you think are important to convey to patients making their decision whether to participate?

4. What do you anticipate being most challenging to convey to primary care patients about this type of testing?

   • Probe/explanation: It may be counterintuitive to patient expectations as we are offering testing to those who are NOT known to be at increased risk, when genetic testing has traditionally targeted to those who are at increased risk.

5. Do you have specific advice for how the research team presents key information via genetic counseling or via GIA the chatbot to primary care patients?

6. We expect that about 3% of those who accept testing through either study arm will test positive and require follow-up care. As a primary care provider, what would you want to know about the study to feel confident in supporting the 3% who test positive with access to follow up care?

7. What concerns do you have about facilitating referrals for the 3% of participants who will test positive for a pathogenic variant in one of the genes tested? How can the research team best support primary care providers?

8. What advice would you give to the research team about study facilitators and potential barriers… and strategies for overcoming them?

9. Is there anything more that you want to contribute to help the research team design and execute best practices in trial work?

GIA Demo

Thank you for addressing these questions. We would like to take a few minutes to navigate GIA, the chatbot, that we are validating for the study. We’d like you to think about how your patients would experience GIA, including those of lower literacy. By the way, GIA will ask you some questions about your health and family history; feel free to give “fake” responses.

Can you please click on the link in the chat and share your screen with me so I can navigate the chat with you? Please ensure there is no patient information shown anywhere on your screen. [Link redacted]

I will give you a few moments to try out the chat. While navigating GIA, please feel free to think out loud or react to what you’re seeing. Then I’ll ask you some questions.

Post-Demo Follow-Up Questions

1. How easy is GIA to navigate? Easy, somewhat easy, somewhat hard, hard

   1. What did you like best about the navigation? What did you find least satisfactory about the navigation?
   2. Do you predict that your patients will find GIA enjoyable to navigate?
   3. Are you concerned that some patients will find GIA annoying or difficult to navigate?

2. Is the information presented at a level that your patients can understand?

3. Does it include the information your patients will need to understand to make an informed choice about whether they will undergo genetic testing?

4. What information, if any, did you find lacking?

5. Now that you have seen GIA, what arm of the study do you predict will lead to more informed choices whether to undergo genetic testing? Why?

6. (If not covered as part of #1 and its follow-up questions) Overall, what did you like best about GIA?

7. (If not covered as part of #1 and its follow-up questions) Overall, what did you like least about GIA?

8. Do you have any suggestions for us about GIA that we have not asked about?

Closing

Thank you for participating. Your input was extremely valuable, and we really appreciate your time.

For Primary Care Patients [Healthcare system name redacted]

Welcome & Introduction

Thank you very much for agreeing to participate in this discussion. We’ll talk about what you think about research that offers genetic testing for three health conditions to patients who are not known to be at risk for having them.

We want to know what thoughts and questions you might have about participating in research studies where genetic testing is offered that could help your doctor understand if you are at risk of developing specific types of cancer or heart disease. Information about genetic testing can be explained and offered to healthy people by a medical professional or through the help of technology. Later in our discussion, you’ll get to try an app called GIA that explains and offers genetic testing to patients.

To be clear: you are not being offered genetic testing as part of this discussion today. The information we gather as part of this interview will help us to design future studies that involve this type of genetic testing.

Discussion of Genetic Testing

1. What does the term “genetic testing” mean to you?

   Explain genetic testing: One way to describe genetic testing is a way to learn whether a person has a change in a gene (instruction for how the body works) that could lead to an illness. Sometimes genetic testing can help prevent an illness from starting, or find it early and treat it.

2. Have you or someone you know ever had a genetic test? If so, what kind? (Probe for what their thoughts were about whether it was useful to the recipient)

3. Would you consider having a genetic test?

4. Probe about when they would consider having genetic testing (positives).

   1. Can you think of a reason that knowing genetic information about you may be helpful to you? If yes, how?
5. Probe about why they might not consider having genetic testing (hesitations/questions/concerns).

   1. What might make you hesitate to have a genetic test?
   2. What concerns, if any, do you have about genetic testing?
   3. What would you want to know about the genetic test before making your decision?
   4. What questions, if any, do you have about genetic testing?
6. If you were offered an opportunity to participate in a research study that would offer you the choice whether to have genetic testing for cancer and heart disease, what would you want to know to decide whether to join the study?

   1. Is there anything that would make you MORE likely to participate in a research study that offered genetic testing?
   2. Is there anything that would make you LESS likely to participate in a research study that offered genetic testing?

7. Do you have any questions or comments about being offered genetic testing that we haven’t talked about already?

GIA Demo

Thank you for answering these questions. Now, we’d like to show you an example of how patients could be offered genetic testing. This is the app I mentioned earlier, called Genetic Information Assistant (or GIA, for short).

Here is a link to click on. I am putting it in the chat. Once you click the link, I will ask you to share your screen with me so I can see you clicking through GIA. I can walk you through how to do that. [Link redacted]

While going through the GIA app, feel free to tell me what you think about it/ any reactions you have. I’ll ask you to stop your review once we get to a certain spot, then I’ll ask you some questions.

Post-Demo Follow-up Questions

1. How easy is GIA to use? Easy, somewhat easy, somewhat hard, hard

2. What did you like best about how GIA works?

3. What did you like least about how GIA works?

4. Did you find GIA fun to use? Why or why not?

5. Did you find GIA annoying or difficult to use? If yes, in what ways?

6. Is any of the information confusing or hard to understand?

7. Is the information in GIA enough to help you make a decision about whether to have genetic testing?

   1. What information did you want, if any, that was missing from GIA?

8. (If not covered as part of #2 above) What did you like best about GIA?

9. (If not covered as part of #3 above) What did you like least about GIA?

10. Do you have any suggestions about ways to make GIA better that we have not asked about?

11. Is there anything more you want to add about using GIA?

Closing

Thank you for participating. Your input was extremely valuable, and we really appreciate your time.

Data Availability

De-identified transcripts from the interviews may be made available upon request. Requests should be made via email to corresponding author.