《益普索：2023年生成式AI時代的醫患關系研究報告（英文版）（14頁）.pdf》由會員分享，可在線閱讀，更多相關《益普索：2023年生成式AI時代的醫患關系研究報告（英文版）（14頁）.pdf（14頁珍藏版）》請在三個皮匠報告上搜索。

1、Five reasons why the time to act for pharma is nowTHE DOCTOR-PATIENT RELATIONSHIP IN THE GENERATIVE AI ERAROBERTO CORTESE Senior Director,Syndicated Healthcare ResearchChances are,this isnt the first piece on Generative AI in healthcare that youve encountered.If youre feeling overwhelmed by the ubiq

2、uity of this topic in your feed,you are not alone.In this perspective article,we take stock of an under-discussed yet disruptive consequence of this technology in healthcare:its impact on a cornerstone of care,that is,the doctor-patient relationship.This is being reshaped by unprecedented access to

3、free,human-like,and seemingly authoritative healthcare advice.Introduction2 Ipsos 2023 All rights reserved.THE DOCTOR-PATIENT RELATIONSHIP IN THE GENERATIVE AI ERAIn Exploring the Changing AI Landscape,Ipsos highlighted how the breakthrough of“supercharged”Large Language Models(LLMs),such as OpenAIs

4、 ChatGPT,will have seismic implications on our lives,enhancing the capabilities of those who utilise it effectively.In the realm of healthcare,this translates to patients now having easy access to a wealth of medical advice never seen before,fundamentally transforming how they acquire,comprehend,and

5、 act upon medical information.This development,which will be further driven by the impending release of Googles Med-PaLM 2(designed to provide high-quality answers to medical questions),will have more far-reaching effects than the traditional Dr Google.With the medical profession already grappling w

6、ith immense shortages and escalating stress levels,AI-powered chatbots can streamline administrative activities and support patient services.However,the influx of patients equipped with advice from these LLMs,which may or may not be accurate or relevant,is likely to put an additional burden on physi

7、cians who will need to help patients address misconceptions in the information gathered while still making swift and effective clinical decisions in time-constrained consultations.A recent JAMA study even suggested that ChatGPT may outperform actual physicians in terms of empathy and advice,comparin

8、g real doctors and ChatGPTs responses to actual medical queries in an online forum.Pharmaceutical companies have a crucial role to play in mitigating risks and seizing opportunities in this significant paradigm shift.While doctors are far from becoming obsolete,patients are becoming more empowered a

9、nd possibly more confident and influential in health decisions than ever before,both within and outside of the consultation room.There is a window of opportunity for pharmaceutical companies.The medical community remains largely unprepared,while patients are using this technology not only to check s

10、ymptoms but also to seek information about treatments.To avoid being left behind,companies must proactively shape how patients and doctors navigate this transformation.This comes with significant challenges,but companies that get it right stand to gain huge rewards.Its not a question of if this chan

11、ge will occur,but when and that when is now.In the remainder of this article,we present five key reasons why immediate and proactive action on the part of pharma will ultimately benefit all stakeholders.Our recommendations are supported by the results of a recent online perceptual survey among 283 U

12、S HCPs(including cardiologists,dermatologists,endocrinologists,haematologists,neurologists,obstetrician-gynaecologists,oncologists,PCPs,and pharmacists),conducted by Medefield on behalf of Ipsos,and the multi-year knowledge gathered through Ipsoss Digital Doctor,the industrys longest-running tracker

13、 of digital HCP behaviours and attitudes,covering 20 markets.3FIVE REASONS WHY THE TIME TO ACT FOR PHARMA IS NOWThe online survey amongst US doctors regarding Generative AI and its impact on doctor-patient relationships reveals that 41%of the HCPs surveyed are aware that their patients have already

14、utilised technology,such as ChatGPT and Med-PaLM,for medical-related purposes in the past three months.While this figure includes patients who have used it sparingly,the actual prevalence of this phenomenon is likely underreported,given that patients may not always disclose their use of Generative A

15、I prior to their visit.1.For patients,the Generative AI era has already begunLooking ahead,we anticipate a significant surge in the number of patients engaging with LLM chatbots in the coming months.In part,this growth will be fuelled by the potential integration of these technologies into healthcar

16、e platforms like Babylon or patient services provided by medical centres.FIGURE 1:Level of Patient Usage of Generative AI(HCP perceptions)%surveyed HCPs stating frequency their patients have used Generative AI(e.g.,ChatGPT,Med-PaLM)in the last three monthsSource:Data collected online from 283 HCPs(O

17、b/Gyns,n=89;General/Family Doctors,n=53;Dermatologists,n=38;Cardiologists,n=31;Endocrinologists,n=25;Neurologists,n=25;Oncologists,n=10;Pharmacists,n=8;Haematologists,n=4)in the US between May June 2023 by Medefield on behalf of Ipsos.Medefield14%3%2%2%20%59%NeverRarelyOccasionally/sometimesOftenAlw

18、aysDont know/NA4 Ipsos 2023 All rights reserved.THE DOCTOR-PATIENT RELATIONSHIP IN THE GENERATIVE AI ERAThe same group of HCPs indicated that Generative AI is far more than a mere symptom checker,a role traditionally filled by“Dr Google”.While symptom checking and self-diagnosis top the list of stat

19、ed uses of Generative AI by patients,some of the HCPs with patients using this technology report a more comprehensive and profound use,such as gathering information about treatment availability(28%)and their side effects(26%),efficacy(14%),and costs(11%),as well as understanding tests or procedures(

20、26%).2.Beyond Dr Google:Patients are accessing human-like digital advisorsSource:Data collected online from 283 HCPs(Ob/Gyns,n=89;General/Family Doctors,n=53;Dermatologists,n=38;Cardiologists,n=31;Endocrinologists,n=25;Neurologists,n=25;Oncologists,n=10;Pharmacists,n=8;Haematologists,n=4)in the US b

21、etween May June 2023 by Medefield on behalf of Ipsos.Base of this analysis are HCPs citing their patients have used Generative AI rarely/occasionally/often/always(n=115)MedefieldFIGURE 2:Objectives of Patients Use of Generative AI(HCP perceptions)%surveyed HCPs citing objectives of Generative AI use

22、 among their patientsFor self-diagnosisTo get general medical adviceTo check symptomsTo check available treatmentsTo get information about possible side effectsTo get information about tests or proceduresTo get information about treatment efficacyTo get information about treatment costsTo find out w

23、hich HCPs to speak toTo get guidance about lifestyle modificationsTo find mental health supportTo find practical supportOtherI dont know35%35%34%28%26%26%14%11%10%10%6%6%4%12%CONDITIONTREATMENT&MANAGEMENTSUPPORTThe use of Generative AI by patients should not come as a surprise.As noted in the aforem

24、entioned JAMA article,ChatGPT scored 9.8 times higher than physicians in the prevalence of empathetic responses and 3.6 times higher in the prevalence of high-quality responses.What is more surprising and potentially riskier is its use in retrieving information about treatments.The Conversation pose

25、d the same cancer-related questions to ChatGPT and“Dr Google”.When asked about the side effects of a specific cancer drug “Does pembrolizumab cause fever and should I go to the hospital?”ChatGPT provided five different responses each time it was queried.While all responses recommended seeking profes

26、sional medical advice,not all conveyed the urgency or clearly defined how potentially serious this side effect was.One response even stated that fever was not a common side effect but did not explicitly say it could occur.This risk may be further amplified,considering that ChatGPT is trained on data

27、 only up to 2021.The free and human-like insights provided by chatbots may also have deeper implications.Patients may struggle to distinguish truth from fabricated facts(hallucinations are an intrinsic limitation of these models)and,satisfied with the advice received,may delay or entirely avoid cont

28、acting their healthcare professional(HCP).The patients themselves may fuel this misinformation loop as they inadvertently feed incorrect or misplaced information into the algorithm,which will later be shared with other patients.Digital Doctor 2023 an ongoing Ipsos study among 3,248 HCPs recently dis

29、covered that,despite the evident benefits of patient empowerment,most HCPs surveyed expressed concern that connected health devices and tools could potentially lead patients to misinterpret data or self-diagnose without physician supervision.These developments have clear implications for the pharmac

30、eutical industry,individual health,and the broader healthcare system.6 Ipsos 2023 All rights reserved.THE DOCTOR-PATIENT RELATIONSHIP IN THE GENERATIVE AI ERAAsked how they expect all of this to impact doctor-patient interactions,the HCPs who participated in the US survey demonstrate a clear cry for

31、 assistance.According to their responses,the use of Generative AI by patients risks exacerbating their existing challenges in multiple ways:Requiring more time for doctors to address any misconceptions created by Generative AI according to 42%of surveyed HCPs Requiring them to learn new skills accor

32、ding to 30%of surveyed HCPs Eroding their authority and trust according to 27%of surveyed HCPs3.Beleaguered HCPs fear their existing challenges will increaseSource:Data collected online from 283 HCPs(Ob/Gyns,n=89;General/Family Doctors,n=53;Dermatologists,n=38;Cardiologists,n=31;Endocrinologists,n=2

33、5;Neurologists,n=25;Oncologists,n=10;Pharmacists,n=8;Haematologists,n=4)in the US between May June 2023 by Medefield on behalf of Ipsos.*denotes option added during fieldwork and asked to a base of n=72 HCPs.MedefieldFIGURE 3:Expected Impact of Generative AI on Doctor-Patient Interactions%surveyed H

34、CPs citing expected impact of Generative AI on doctor-patient interactions in the futureIt will require more time for doctors to address any objections/misconceptionsNone of these I dont knowNegative expectationsNeutral expectationsPositive expectationsIt will require doctors to be more understandin

35、g/empatheticIt will make visits more efficientIt will improve the quality of our medical adviceIt will help with some staffs activities/shortages*I fear it may take over some functions of my role*It will increase patient influence on treatment decisionsIt will make patients more informed and empower

36、edIt can help doctors make better decisionsIt will hurt physicians authority and trustIt will require doctors to develop new skillsIt will create misunderstanding and false expectations42%33%30%27%22%19%16%15%14%12%10%6%13%1%7A more positive impact of Generative AI on doctor-patient interactions is

37、possible.Potential benefits,like improved decision-making,better patient empowerment,and increased efficiency,must come to the forefront.However,for this to happen,HCPs call for help must be listened to by pharma companies and by healthcare and political systems in general.8 Ipsos 2023 All rights re

38、served.THE DOCTOR-PATIENT RELATIONSHIP IN THE GENERATIVE AI ERAThese findings present an ideal opportunity for pharma companies to assist HCPs in building confidence with AI technology.By doing so,HCPs are more likely to view the benefits of Generative AI used by patients and see themselves as facil

39、itators of improved doctor-patient interactions,rather than as victims of technological advancement.4.HCPs see the benefits of AI but they need guidance hereOur 2023 Digital Doctor study revealed a promising trend:HCPs are warming up to AI.Specifically,the survey found that 40%of participating HCPs

40、in the 20 markets surveyed are excited about AIs role in healthcare,and a similar proportion believe it can automate repetitive tasks and enhance diagnostic efficiency and accuracy.However,only 35%feel they know a lot about it.The more recent US HCP survey,from which we have been drawing insights so

41、 far,reinforces this finding:while some of the HCPs are gaining personal experience with AI across various applications,most have not used it extensively in the past three months.Source:Data collected online from 283 HCPs(Ob/Gyns,n=89;General/Family Doctors,n=53;Dermatologists,n=38;Cardiologists,n=3

42、1;Endocrinologists,n=25;Neurologists,n=25;Oncologists,n=10;Pharmacists,n=8;Haematologists,n=4)in the US between May June 2023 by Medefield on behalf of Ipsos.MedefieldFIGURE 4:HCP Usage of Generative AI in Last 3 Months%surveyed HCPs citing usage and application of Generative AI in the last three mo

43、nthsI havent used AI in the last 3 monthsTo explore medical literatureTo explore treatmentsTo explore guidelinesTo do administrative tasks(e.g.data entry)To inform/support a diagnosisTo write emails/collaborate with my co-workersTo explore clinical trialsTo write medical contentTo inform/support tre

44、atment initiationTo support patient visits/follow-upsTo inform/support patient monitoringTo provide practical/emotional support to patientsI dont know64%10%10%9%7%6%5%4%4%3%3%2%2%3%95.Pharma has a unique window of opportunity to shape the impact of AIWith Generative AI already a reality for some pat

45、ients and HCPs,pharma companies must act swiftly to capitalise on the opportunities this structural shift presents,while mitigating the associated risks.Pharma companies can shape the impact of this technology on doctor-patient interactions by enhancing healthcare advice at two critical junctures:Ac

46、curacy:Make sure information about conditions and treatments is clear,concise,up-to-date,and supported by“real-life”keywords and phrases.This requires understanding patients needs and practices when seeking advice from LLMs.Consistency:Disseminate the same information across a wide range of credible

47、 sources likely to be included in the AIs training data.This requires an understanding of how AI training works.Education:As Ipsos Conversations with AI paper recently noted,a poorly crafted question can lead to a distorted perception of reality and misguided actions,while a well-crafted question ca

48、n lead to accurate and insightful answers.Therefore,it is crucial not only to inform patients about the risks and limitations of these tools but also to provide them with a repository of hints and example prompts to maximise the quality of advice received.Partnering with patient associations and HCP

49、s will be crucial here.Anticipate and provide training and resources to handle patient objections,such as FAQs to ChatGPT with common misconceptions and answers to handle patient conversations.Providing training on empathy and how to deal with possible conflicts,ultimately considering that engaged p

50、atients are key to better personalised care.Support HCPs in building familiarity and confidence with AI,showcasing use cases and involving them in initiatives that help them understand how to benefit from this technology.Upstream:Ensure patients access more valuable and trustworthy online informatio

51、nDownstream:Equip HCPs to handle AI-powered information effectively.BA10 Ipsos 2023 All rights reserved.THE DOCTOR-PATIENT RELATIONSHIP IN THE GENERATIVE AI ERAPHARMACEUTICAL COMPANIES,HAVING LONG HARNESSED AI TO BOLSTER PERFORMANCE,ARE NOW ON THE CUSP OF A DEFINING MOMENT11Ipsos will continue monit

52、oring this evolving landscape across several therapeutic areas and countries through a new syndicated study and thought leadership research to support our clients in this space.Our aim is to provide a greater understanding of how the world is changing.Learning how people decide upon and interact wit

53、h Generative AI,as patients and as healthcare professionals,will be our focus.THE DOCTOR-PATIENT RELATIONSHIP IN THE GENERATIVE AI ERALooking ForwardGenerative AI is revolutionising the medical advice landscape,redefining the doctor-patient dynamic in ways previously unimaginable.This transformative

54、 technology can empower patients to seize greater control of their health while supporting HCPs in crucial tasks,including providing insights to improve diagnosis or better communicate with patients.However,it also ushers in a new set of formidable challenges,which demand careful consideration.Patie

55、nts,now equipped with an unprecedented degree of advice,potentially inaccurate,are placing unprepared HCPs under considerable stress.Furthermore,easy access to illusory,confident and authoritative content engendered by AI may inadvertently lead patients to postpone or circumvent necessary interactio

56、ns with HCPs,or even to feed uncontrolled information to the algorithm,with huge consequences for patients as well as for regulators and industry players(think,for example,about the spread of confidential data or unchecked adverse events).Pharmaceutical companies,having long harnessed AI to bolster

57、their performance,from drug discovery to go-to-market strategies,are now on the cusp of a defining moment.They possess a historic opportunity to elevate health outcomes by guaranteeing that correct AI-driven health advice is found,discussed and implemented by HCPs and patients.Given the swift pace o

58、f technological evolution and the inherent risk of misinformation,the need for proactive measures is urgent.This is,however,a high-risk and high-reward field for pharma,with considerable legal and operational challenges.Those who take the lead stand to reap the most significant benefits.12 Ipsos 202

59、3 All rights reserved.About the Research The US HCP survey was a short perceptual survey conducted between May 29th and June 5th,2023,among 238 HCPs in the US.Participating HCPs covered the following specialities:cardiologists,dermatologists,endocrinologists,haematologists,neurologists,obstetrician-

60、gynaecologists,oncologists,PCPs,and pharmacists.Data referred to in this article were collected online.Sample sizes are provided alongside the relevant charts.The study was conducted by Medefield on behalf of Ipsos.Data copyright owned by Medefield.Digital Doctor,running since 2015,provides a compre

61、hensive overview of digital attitudes and behaviour among doctors worldwide.For Digital Doctor 2023,Ipsos conducted an online survey among 3,427 physicians across 20 markets using Sermo(fieldwork partner)from October 6th,2022,to February 16th,2023.Respondents included primary care physicians,paediat

62、ricians,oncologists and neurologists,recruited according to specific screening criteria.For detailed sampling information,please visit Ipsos Digital Doctor 2023.Data copyright owned by Ipsos.1.Timpone,R.and Guidi,M.(2023)Exploring the Changing AI Landscape From Analytical AI to Generative AI.Ipsos V

63、iews.https:/ 2.Legg,J.,Bangia A.and Timpone,R.(2023)Conversations with AI.How generative AI and qualitative research will benefit each other.Ipsos Views.https:/ al.l(2023)Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum.JAMA

64、 Internal Medicine https:/pubmed.ncbi.nlm.nih.gov/37115527/4.Mesk,B.(2023)Whats Next For ChatGPT In Healthcare?The Future of Digital Health.https:/ Chatbots Can Diagnose Medical Conditions at Home.How Good Are They?Scientific American https:/shorturl.at/cR0596.Hopkins,A.M.(2023)We asked ChatGPT and

65、Dr Google the same questions about cancer.Heres what they said.The Conversation https:/ C.and Margraff H.(2023)Chat GPT and AI in healthcare:Just what the doctor ordered?Simon Kucher https:/www.simon- Are Using Chatbots in an Unexpected Way.The New York Times.https:/ Roberto Cortese Senior Director,

66、Syndicated Healthcare Research Roberto.C References13About Ipsos Healthcare Service LineWe are a global insight,analytics and advisory partner to the healthcare sector.Our multi-disciplinary teams deliver integrated services and proprietary real-world evidence across the product lifecycle.This enables our clients to act with clarity,certainty and