This study demonstrates patient acceptance of a virtual health coach while waiting for the physician to enter the primary care exam room.
ABSTRACT
Objectives: Virtual assistants formed through the union of artificial intelligence (AI) and natural language understanding (NLU) technologies (eg, Siri) are routine in our daily lives. The purpose of this study was to test patients’ acceptance of a virtual health coach (VHC) who emulates medical staff to activate the patient to engage in a conversation with their physician about healthy lifestyles.
Study Design: Descriptive survey study.
Methods: Eighty-nine primary care patients agreed to test “new technology” while waiting in the exam room for their physician. Patients used a handheld tablet computer to interact with an animated VHC who asked questions about common health concerns, including weight, smoking, drinking, stress, or medication adherence. During the brief intervention, the VHC advised the patient to have a conversation with the physician about the identified lifestyle concern.
Results: Patients were comfortable with the technology and said it positively impacted their feelings about the organization. They talked to the VHC, rated the technology as easy to use, and felt that they would use it again and that other patients would enjoy it. Many patients tried to talk to the VHC about the specific reason for their visit.
Conclusions: VHCs using AI and NLU technologies are acceptable and useable, and they positively impact patients during exam room “downtime.” The technology easily integrates into the clinic workflow and does not extend the length of the visit. Although patients were willing to talk with the VHC, in many cases, they confused the VHC’s line of questioning about general lifestyle concerns with the purpose of their visit.
The American Journal of Accountable Care. 2016;4(4):41-45
Patient “activation” is a recognized concept that describes a person’s knowledge, confidence, and skills to manage his or her own health and healthcare; thus, patients who are activated in their health and healthcare generally have better outcomes and lower costs.1 Yet, 75% of patients are not adequately informed or confident, nor do they have the necessary skills to manage their health. In particular, healthy lifestyle behaviors are problematic for most patients. Those with lower activation rarely participate in behaviors that require active self-management, such as exercise, eating a healthy diet, and managing stress.2
Interventions to increase patients’ activation in their health have focused on the development of skills and advice tailored to the patient’s level of activation. For example, training patients to ask their doctors questions and giving them support to do so has been shown to increase both activation levels and participation in healthcare.3,4 Observations of patient—physician encounters suggest only 10% of the primary care visit is devoted to lifestyle topics and, further, that these 2 minutes mostly entail the physician bringing up a lifestyle concern. Such limited and ineffective interactions have prompted calls for additional communication training for primary care providers to help them engage patients in lifestyle issues in a time-efficient and effective manner.5
A virtual health coach (VHC) offers a solution to the problem of patient activation and patient—physician discussions of lifestyle issues. Current technology enables the design of an animated VHC who can understand everyday language and respond to patients in a conversational manner. Patients are accepting of virtual agents and, interestingly, some disclose more to a computer interviewer than to its human counterpart.6-8 Standardized behavioral change and motivational interventions can be built into VHC interactions.9,10 The ability to standardize conversations allows challenging and time-consuming interactions to be brief, consistent in quality, and scalable. VHCs could ultimately cost less than real providers in specific contexts.11 Thus, they are an attractive alternative to training physicians to be proficient in effective lifestyle discussions by using motivational interviewing strategies, which has proven to be challenging at best.12,13
The purpose of the current study was to test the feasibility of a technology innovation in the primary care setting, building on an earlier unpublished study of a prototype. We wanted to determine patients’ acceptance of a VHC—with artificial intelligence (AI), natural language understanding (NLU), and animated capabilities—by engaging in a verbal interaction with the coach while waiting in the exam room for the doctor. The VHC is intended to emotionally connect with, encourage, and support patients in preparation for a conversation with their doctor about a lifestyle issue that the patient identifies as concerning.
METHODS
Setting and Procedures
Physicians and their medical assistants (MAs) from women’s health and adult medicine clinics participated in the study. MAs gave patients a handheld tablet and provided the instruction, “We are testing new technology and need your help.” There was a prominently displayed “start” button. The Northern California Kaiser Permanente Institutional Review Board allowed a waiver to a signed consent document.
The VHC oriented the patient to the purpose of the tablet app and provided instructions for the microphone. As the VHC “spoke,” the tablet’s screen simultaneously displayed the VHC and the text of the VHC’s words on the screen to allow patients to read the conversation. The patient was asked to select a general health concern (eg, weight, smoking, drinking, stress, taking medications, or none of the above) and respond to a series of questions, interspersed with empathic and supportive statements, all with the purpose of encouraging the patient to bring up the concerning health topic as soon as the doctor entered the room. The VHC used motivational questions to prime the patient for the conversation.14 At the end, there was a “learn more” button for reading materials relevant to the lifestyle topic the patient selected. Upon entry, the physician asked, “What did you discuss with the health coach?”
Patients completed surveys disseminated by the authors and brief interviews at the end of their visit. Physicians were surveyed via e-mail.
Techonology
The VHC (an AI-based cognitive software solution) was created by a contracted vendor, and it spoke English and Spanish. It was built on the Alme platform, which combines a healthcare domain model and NLU with an avatar to drive interactive conversations. Samsung tablets were connected to a secure, corporate wireless network.
Outcomes, Measurements, and Analysis
Data were captured via the tablet app, including verbal interactions, the lifestyle concern selected, and the mode of communication (talking, typing, and tapping an icon). Surveys used a 1-to-7 Likert scale to measure aspects of the experience, and patients also provided verbal feedback. Means and standard deviations were calculated for survey responses, and the user experience data were also described.
RESULTS
Eighty-nine patient—VHC conversations were captured. One patient refused to participate, and 63 patients (12 male, 48 female) were available for surveys and interviews. Thirty-four patients were from women’s health clinics and 29 were from adult medicine clinics. The average age of males was 56 years (range = 33-82 years), and the average age of females was 41 years (range = 18-84 years). Eight physicians participated.
Patients talked out loud to the VHC in 82% of the interactions and typed responses 6.6% of the time. When an icon was available, patients “tapped” on it 74% of the time versus voicing their lifestyle concern audibly. Most patients rated the technology as easy to use, felt other patients would enjoy using it, and were willing to use it again, while the modal response was “neutral” about their own enjoyment. Many reported the positive impact of the technology on their feelings toward the healthcare organization. Patients’ survey data are presented in Table 1.
Patients said they felt encouraged by the VHC to talk with their doctor, and 51% of them reported they engaged in a conversation with their physician about the health topic they selected on the tablet. Twenty-six percent reported they had not talked with the physician about this topic previously. “None of the above” was the most frequently (66.3%) selected lifestyle concern. Of that 66.3%, 24 patients specifically stated, “I’m healthy,” 26 vocalized a specific reason for their visit, and 9 of the verbal responses of patients were unintelligible. Weight was the most frequent concern (18%), followed by stress (7%), medication adherence (6%), drinking (2%), and smoking (1%).
The verbal comments regarding feedback on the VHC included it was “fun,” “interesting,” “very cool,” “impressive,” as well as “It was great to have something to do while waiting” and “I love this about this organization.” A number of other comments were similar to “It would be great if it related to the reason I’m here.” Comments specific to the VHC included, “weird movements,” “robotic voice,” and 1 patient stated, “I didn’t like it.”
Sixty-seven percent of the patient—VHC conversations were completed, meaning patients reached the final screen. There were Internet stability issues in some cases, and technology malfunctions in others; for some, the physician entered the room before the patient–VHC exchange was completed.
There was a significant negative correlation between age and comfort with using “touch screens” (r = —0.51; P <.00003; R² = 0.26). No relationship was detected between the variables of age and comfort using voice recognition technology. Physicians reported minimal impact of the technology on the duration of the visit, and they were, on average, neutral about the usefulness of the VHC. Table 2 displays the physician ratings.
DISCUSSION
The purpose of this feasibility study was 2-fold: first, to determine patients’ acceptance of a VHC while waiting for their physician, and second, to ascertain whether patients would engage in a conversation with their physician about a lifestyle concern identified through the virtual interaction. We believe this is the first report of using AI/NLU technologies with patients in the primary care setting and consider the findings worthy of further investigation.
Patient Experience
Patient responses to the technology exceeded expectations. Regardless of age, patients were willing to use a VHC while in the exam room and to speak aloud to it. As in earlier reports, patients were positive about the technology.7,15 Patients also felt encouraged by the VHC to engage in a conversation with their physician. Interestingly, although patients think others will enjoy the VHC, their own enjoyment ratings were “neutral.”
Patient responses regarding the intent of the VHC were perplexing. A sizable number selected “none of the above” regarding lifestyle concerns, which contradicts current healthy lifestyle investigations.16 Many patients neglected the instructions about “overall health” and “not your specific reason” for today’s visit and told the VHC about a specific problem. Nevertheless, half of the patients reported they had a conversation with their physician about the topic they discussed with the VHC, and about half of this group said they had not discussed this topic with their physician previously.
Physician Experiences
Physicians were “neutral” about the VHC. However, although several expressed verbal concern about increasing the encounter time, this outcome was not reported. Physicians were more positive in the earlier prototype test. One said, “It makes me have a conversation relevant to the patient. I want to discuss smoking, but this lets me know the patient is worried about stress. It saves time.” In the current study, little time was spent orienting the physicians about the purpose of the VHC, and this may, in part, be responsible for their lower satisfaction ratings. In addition, the desired outcome of the VHC interaction was for the patient and physician to have a discussion about a lifestyle concern, but this did not happen consistently.
Technology
The technology can improve. The voice recognition technology did not always capture the verbalizations of the patient, and the secure wireless connection was unstable. Some patients described the voice of the VHC as “robotic” and the movements as “unnatural” and “jerky.”
Based on greater success of an earlier prototype, we speculate that the appearance, movement, and/or voice of the VHC could be distractive to patients. In our unpublished study of 11 patients that displayed a posturized image of a coach on the tablet, patients followed instructions, identified a lifestyle concern, and discussed it with the physician in every case. In the current study, the VHC was an animated, semi-realistic virtual agent (similar to those in a video game) with empathic gestures and expressions. Elements, or the combination of these features, potentially diverted patients’ attention from the instructions, and, although the multi-featured VHC might be acceptable in a video game, the healthcare setting is inherently different and patients have ingrained expectations
Future Investigations
The application of behavioral and psychological strategies through technology (ie, behavioral intervention technologies) has tremendous promise for healthcare.17 We see the VHC as the inaugural attempt to integrate motivational and patient activation strategies with AI/NLU technologies. Future VHC iterations could include voice enhancement, modified movements, and altered appearance.18 Strengthening the instructions about the purpose of the interaction will be important, as well, to ensure the desired conversation takes place. Additional advancements could leverage data from electronic health records to personalize the exchange, and the VHC’s questions could address specific visits, such as birth control considerations for women’s health and alcohol screening in adult medicine.
CONCLUSIONS
The introduction of a VHC in the healthcare setting could be considered “disruptive technology,” which is defined as “a technology that shakes up an industry, or a groundbreaking product that creates a completely new industry.”19 According to Clayton Christensen, it often lacks refinement, does not perform well because it is new, and appeals to a limited audience. We think this describes our VHC concept: it needs refinement, better performance, and broader appeal, and it has untapped potential to improve quality, efficiency, health outcomes, cost, and patient satisfaction.
Acknowledgments
The authors would like to thank their colleagues from the iFund, NextIT, and The Permanente Medical Group administration, as well as physicians and medical assistants who made this study possible.
Author Affiliations: Kaiser Permanente (SDP, DER), Roseville, CA.
Source of Funding: This study was funded by the Kaiser Permanente Innovation Fund for Technology and The Permanente Medical Group, North Valley.
Author Disclosures: The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.
Authorship Information: Concept and design (SDP); acquisition of data (SDP, DER); analysis and interpretation of data (SDP, DER); drafting of the manuscript (SDP); critical revision of the manuscript for important intellectual content (SDP, DER); statistical analysis (SDP, DER); provision of study materials or patients (SDP); obtaining funding (SDP); administrative, technical, or logistic support (DER); and supervision (SDP).
Send Correspondence to: Dannielle E. Richardson, Kaiser Permanente, 1600 Eureka Rd, Roseville, CA 95661. E-mail: Dannielle.e.richardson@kp.org.
REFERENCES
1. Greene J, Hibbard JH, Sacks R, Overton V, Parrotta CD. When patient activation levels change, health outcomes and costs change, too. Health Aff (Millwood). 2015;34(3):431-437. doi: 10.1377/hlthaff.2014.0452.
2. Hibbard JH, Tusler M. Assessing activation stage and employing a “next steps” approach to supporting patient self-management. J Ambulatory Care Manage. 2007;30(1):2-8.
3. Deen D, Lu WH, Rothstein D, Santana L, Gold MR. Asking questions: the effect of a brief intervention in community health centers on patient activation. Patient Educ Couns. 2011:84(2):257-260. doi: 10.1016/j.pec.2010.07.026.
4. Post DM, Cegala DJ, Miser WF. The other half of the whole: teaching patients to communicate with physicians. Fam Med. 2002;34(5):344-352.
5. Ory MG, Peck BM, Browning C, Forjuoh SN. Lifestyle discussions during doctor-older patient interactions: the role of time in the medical encounter. MedGenMed. 2007;9(4):48.
6. Servan-Schreiber D. Artificial intelligence and psychiatry. J Nerv Ment Dis. 1986;174(4):191-202.
7. Bickmore T, Pfeifer L, Paasch-Orlow M. Using computer agents to explain medical documents to patients with low health literacy. Patient Education and Counseling. 2009;75(3):315-320. doi: 10.1016/j.pec.2009.02.007.
8. Bickmore T, Bukhari L, Vardoulakis LP, Paasche-Orlow M, Shanahan C. Hospital buddy: a persistent emotional support companion agent for hospital patients. In: Nakano Y, Neff M, Paiva A, Walker M. (eds). Intelligent Virtual Agents: 12th International Conference, IVA 2012, Santa Cruz, CA, USA, September, 12-14, 2012. Proceedings. Springer-Verlag Berlin Heidelberg; 2012: 492-495.
9. Schulman D, Bickmore T, Sidner CL. An intelligent conversational agent for promoting long-term health behavior change using motivational interviewing. In: AI and Health Communication—Papers from the AAAI 2011 Spring Symposium [SS-11-01]. Menlo Park, California: The AAAI Press; 2011.
10. Lisetti CL, Yasavur U, De Leon C, Amini R, Visser U, Rishe N. Building an on-demand avatar-based health intervention for behavior change. Presented at: Twenty-Fifth International FLAIRS Conference; May 23-25, 2012; Marco Island, FL.
11. Gratch J, Lucas GM, King AA, Morency LP. It’s only a computer: the impact of human-agent interaction in clinical interviews. In: Proceedings of the 2014 International Conference on Autonomous Agents and Multi-Agent Systems. International Foundation for Autonomous Agents and Multiagent Systems; 2014:85-92.
12. Hettema J, Steele J, Miller WR. Motivational interviewing. Annu Rev Clin Psychol. 2005;1:91-111. doi: 10.1146/annurev.clinpsy.1.102803.143833.
13. Kaner E, McGovern R. Training practitioners in primary care to deliver lifestyle advice. BMJ. 2013;346:f1763. doi: 10.1136/bmj.f1763.
14. Miller WR, Rollnick S. Motivational Interviewing: Preparing People for Change. 2nd edition. New York, NY: The Guilford Press; 2002.
15. Bickmore T, Picard R. Establishing and maintaining long-term human-computer relationships. ACM Trans Comput-Hum Interact. 2005;12(2):293-327. doi: 10.1145/1067860.1067867.
16. Loprinzi PD, Branscum A, Hanks J, Smit E. Healthy lifestyle characteristics and their joint association with cardiovascular disease biomarkers in US adults. Mayo Clin Proc. 2016;91(4):432-442. doi: 10.1016/j.mayocp.2016.01.009.
17. Mohr DC, Burns MN, Schueller SM, Clarke G, Klinkman M. Behavioral intervention technologies: evidence review and recommendations for future research in mental health. Gen Hosp Psychiatry. 2013;35(4):332-338. doi: 10.1016/j.genhosppsych.2013.03.008.
18. Ring L, Utami D, Bickmore T. The right agent for the job? In: Bickmore T, Marsella S, Sidner C (eds). Intelligent Virtual Agents: 14th International Conference, IVA 2014, Boston, MA, USA, August 27-29, 2014, Proceedings. Springer International Publishing: 2014: 374-384.
19. Christensen CM. The Innovator’s Dilemma: When New Technologies Cause Great Firms to Fail. Boston, MA: Harvard Business Review Press; 1997.
How English- and Spanish-Preferring Patients With Cancer Decide on Emergency Care
November 13th 2024Care delivery innovations to help patients with cancer avoid emergency department visits are underused. The authors interviewed English- and Spanish-preferring patients at 2 diverse health systems to understand why.
Read More
Geographic Variations and Facility Determinants of Acute Care Utilization and Spending for ACSCs
November 12th 2024Emergency department (ED) visits and hospitalizations for ambulatory care–sensitive conditions (ACSCs) among Medicaid patients constitute almost 40% of all ED visits and hospitalizations, with lower rates observed in areas with greater proximity to urgent care facilities and density of rural health clinics.
Read More
Pervasiveness and Clinical Staff Perceptions of HPV Vaccination Feedback
November 11th 2024This article used regression analyses to quantify how clinical staff perceive provider feedback to improve human papillomavirus (HPV) vaccination rates and determine the prevalence of such feedback.
Read More