التنقل في علاقة الطبيب بالمريض والذكاء الاصطناعي – دراسة مختلطة لآراء الأطباء تجاه الذكاء الاصطناعي في الرعاية الأولية Navigating the doctor-patient-AI relationship – a mixed-methods study of physician attitudes toward artificial intelligence in primary care

المجلة: BMC Primary Care، المجلد: 25، العدد: 1
DOI: https://doi.org/10.1186/s12875-024-02282-y
PMID: https://pubmed.ncbi.nlm.nih.gov/38281026
تاريخ النشر: 2024-01-27

التنقل في علاقة الطبيب بالمريض والذكاء الاصطناعي – دراسة مختلطة لآراء الأطباء تجاه الذكاء الاصطناعي في الرعاية الأولية

ماثيو ر. ألين , صوفي ويب , عمار ماندفي , مارشال فريدن , مينغ تاي-سيل و جين كالنبرغ

الملخص

الخلفية: الذكاء الاصطناعي (AI) هو مجال يتقدم بسرعة ويبدأ في دخول ممارسة الطب. الرعاية الأولية هي حجر الزاوية في الطب وتتعامل مع تحديات مثل نقص الأطباء والإرهاق الذي يؤثر على رعاية المرضى. يتم تقديم الذكاء الاصطناعي وتطبيقه عبر الصحة الرقمية بشكل متزايد كحل محتمل. ومع ذلك، هناك نقص في الأبحاث التي تركز على مواقف أطباء الرعاية الأولية (PCP) تجاه الذكاء الاصطناعي. تدرس هذه الدراسة آراء PCP حول الذكاء الاصطناعي في الرعاية الأولية. نستكشف تأثيره المحتمل على مواضيع ذات صلة بالرعاية الأولية مثل علاقة الطبيب بالمريض وسير العمل السريري. من خلال القيام بذلك، نهدف إلى إبلاغ أصحاب المصلحة في الرعاية الأولية لتشجيع الاستخدام الناجح والعادل لأدوات الذكاء الاصطناعي المستقبلية. دراستنا هي الأولى من نوعها، حسب علمنا، التي تستكشف مواقف PCP باستخدام حالات استخدام محددة للذكاء الاصطناعي في الرعاية الأولية بدلاً من مناقشة الذكاء الاصطناعي في الطب بشكل عام. الطرق: من يونيو إلى أغسطس 2023، أجرينا استطلاعًا بين 47 طبيب رعاية أولية مرتبطين بنظام صحي أكاديمي كبير في جنوب كاليفورنيا. قام الاستطلاع بتحديد مواقف تجاه الذكاء الاصطناعي بشكل عام وكذلك فيما يتعلق بحالتين محددتين لاستخدام الذكاء الاصطناعي. بالإضافة إلى ذلك، أجرينا مقابلات مع 15 من المشاركين في الاستطلاع. النتائج: تشير نتائجنا إلى أن PCPs لديهم آراء إيجابية إلى حد كبير حول الذكاء الاصطناعي. ومع ذلك، غالبًا ما كانت المواقف تعتمد على سياق التبني. بينما كانت بعض المخاوف التي أبلغ عنها PCPs بشأن الذكاء الاصطناعي في الرعاية الأولية تركز على التكنولوجيا (الدقة، السلامة، التحيز)، كانت العديد منها تركز على عوامل تتعلق بالأشخاص والعمليات (سير العمل، العدالة، التعويض، علاقة الطبيب بالمريض). الخلاصة: تقدم دراستنا رؤى دقيقة حول مواقف PCP تجاه الذكاء الاصطناعي في الرعاية الأولية وتبرز الحاجة إلى توافق أصحاب المصلحة في الرعاية الأولية حول القضايا الرئيسية التي أثارها PCPs. قد تواجه مبادرات الذكاء الاصطناعي التي تفشل في معالجة كل من المخاوف التكنولوجية ومخاوف الأشخاص والعمليات التي أثارها PCPs صعوبة في تحقيق تأثير.

الكلمات الرئيسية: الذكاء الاصطناعي، المواقف، العمل الرقمي، الرعاية الأولية، التكنولوجيا، البحث النوعي، الصحة الرقمية

الخلفية

بينما تم مناقشة التأثير المحتمل للذكاء الاصطناعي في الطب لفترة طويلة، أصبحت التطبيقات الواقعية التي تواجه الأطباء للذكاء الاصطناعي حقيقة فقط مؤخرًا [1-3]. إدارة الأمراض المزمنة المدعومة بالذكاء الاصطناعي، دعم التشخيص، والأعمال الإدارية (مثل التوثيق، الفوترة، ورسائل المرضى) لديها إمكانات كبيرة لتحسين الطب وتخفيف بعض الأعباء عن الأطباء مما يسمح لهم بالتركيز على رعاية المرضى على مستوى الأطباء [4]. علاوة على ذلك، فإن استخدام الذكاء الاصطناعي السريري هو جزء من تحول أوسع في الطب نحو “الصحة الرقمية” حيث يتم إجراء العديد من جوانب الرعاية الطبية عن بُعد، بوساطة وسيط تكنولوجي مما يؤدي إلى تحسينات محتملة في الكفاءة والوصول [5، 6]. هذه التطورات من شأنها أن تحدث تأثيرًا كبيرًا في الرعاية الأولية، وهو مجال يواجه حاليًا معدلات عالية من إرهاق الأطباء، وتعويض غير كاف، ونقص متزايد في الأطباء [7-9]. ومع ذلك، هناك قلق من أنه إذا تم دمج الذكاء الاصطناعي بشكل سيء، فقد يؤدي ذلك إلى تفاقم “فجوة القيم المهنية وواقع ممارسة الرعاية الأولية” [10، 11]. على سبيل المثال، على الرغم من الدور الحاسم لعلاقة الطبيب بالمريض في الطب، لا يزال تأثير الذكاء الاصطناعي والصحة الرقمية على هذا المكون الأساسي من الرعاية الأولية غير مستكشف بشكل كافٍ [12-14].
على الرغم من وجود الكثير على المحك، إلا أن الأدبيات حول آراء PCP تجاه الذكاء الاصطناعي محدودة [15، 16]. لقد جرت معظم الأبحاث الموجودة في سياق نظري بحت تستكشف الذكاء الاصطناعي بشكل عام مع الأطباء الذين لم يكن لديهم خبرة في استخدام أنظمة مدعومة بالذكاء الاصطناعي. نقترح أن هناك حاجة لمزيد من مشاركة المستخدم النهائي مع الأطباء لمناقشة حالات استخدام محددة وملموسة للذكاء الاصطناعي السريري [17]. من خلال تسليط الضوء على حالات استخدام محددة للذكاء الاصطناعي، نأمل في استنباط مخاوف ومواقف جديدة قد تبقى مخفية عند مناقشة الذكاء الاصطناعي بشكل عام. إن الفشل في إشراك المستخدمين النهائيين في تصميم أدوات الصحة الرقمية المدعومة بالذكاء الاصطناعي يؤدي إلى دمج غير فعال أو غير ناجح لهذه الأدوات في سير العمل السريري مما يؤدي إلى زيادة إرهاق الأطباء وحتى ضرر المرضى [10، 18، 19].

الرعاية الأولية، التكنولوجيا والعدالة الصحية

تدرك دراستنا إمكانات التكنولوجيا في تفاقم أو تحسين الفجوات الموجودة في الرعاية الصحية [20-24]. أنظمة الذكاء الاصطناعي معرضة بشكل خاص لتفاقم العدالة الصحية بسبب عوامل مثل البيانات المتحيزة المحتملة التي تصبح متجذرة في أنظمة الذكاء الاصطناعي أو التوزيع غير المتكافئ لأدوات الذكاء الاصطناعي الجديدة المطورة [25]. تعتبر اعتبارات العدالة ضرورية بشكل خاص في سياق الرعاية الأولية. غالبًا ما يكون PCPs هم نقطة الاتصال الأولى للمرضى وهم مركزون في تقديم الرعاية الصحية للمجتمعات ذات الوصول المحدود بسبب عوامل جغرافية أو اقتصادية أو اجتماعية [26-30]. كما أن PCPs يشكلون أكبر مجموعة محتملة من مستخدمي الذكاء الاصطناعي بين المهنيين الصحيين [9]. على الرغم من الطبيعة الأساسية للرعاية الأولية، فإن هذا المجال
قد عانى لفترة طويلة من نقص في الاهتمام والموارد والاعتراف مقارنة بالتخصصات الطبية الأخرى [31-33]. وقد ساهم ذلك في نقص نسبي في تقدم الذكاء الاصطناعي وتنفيذه في الرعاية الأولية على الرغم من الحاجة الكبيرة والإمكانات [9، 22، 34-36]. وبناءً عليه، فإن العدالة هي اعتبار رئيسي للذكاء الاصطناعي في الرعاية الأولية.

الهدف

في سعيينا نحو تصميم يركز على المستخدم، استخدمنا نهجًا مختلطًا للتعمق في مواقف PCP بشأن التأثير التحويلي المحتمل للذكاء الاصطناعي والتحول الأوسع نحو الرقمنة في الرعاية الأولية. هدفنا الأول هو إبلاغ أصحاب المصلحة في الرعاية الأولية بمخاوف PCP بشأن الآثار السلبية المحتملة للذكاء الاصطناعي في الرعاية الأولية. تكشف نتائجنا عن عوامل محورية يمكن أن تسهل أو تعيق دمج أنظمة الذكاء الاصطناعي في الرعاية الأولية. هدفنا على المدى الطويل هو استخدام هذه النتائج لتطوير أدوات الذكاء الاصطناعي الموضحة في المخطوطة بهدف تحسين رعاية المرضى. على الرغم من أن هذه ليست أول دراسة استكشافية لمواقف PCP حول الذكاء الاصطناعي، إلا أنها، حسب علمنا، هي الدراسة الأولى التي تمتد إلى ما هو أبعد من المجال النظري، وتدمج حالات استخدام محددة ومدخلات من PCPs مع خبرة حقيقية في استخدام أدوات الذكاء الاصطناعي والرعاية الصحية الرقمية.

الطرق

مشاركة المشاركين مع الذكاء الاصطناعي والصحة الرقمية

يشارك المشاركون في دراستنا مع مركز طبي أكاديمي (AMC) يشارك بنشاط في تطوير أو اختبار تجريبي أو تنفيذ عدة تطبيقات للذكاء الاصطناعي ضمن نظام الرعاية الصحية الخاص به. هنا، نسلط الضوء على حالات استخدام محددة ذات صلة بالرعاية الأولية، مع تسليط الضوء على دورها المحوري في الدراسة. تقع هذه الحالات تحت فئة الصحة الرقمية الأوسع [37].

فحص الأمراض المعزز بالذكاء الاصطناعي: انقطاع النفس النومي الانسدادي

مع الاعتراف بالانتشار المتزايد لانقطاع النفس النومي الانسدادي (OSA) وحالته التي غالبًا ما تكون غير مكتشفة، حدد فريق البحث في مركزنا الطبي OSA كهدف مناسب لفحص الأمراض المعتمد على الذكاء الاصطناعي [38]. يبني هذا المبادرة على أبحاث سابقة تستخدم بيانات السجلات الصحية الإلكترونية (EHR) لتحديد الأفراد المعرضين لخطر مرتفع من OSA [39، 40]. تعتبر هذه الحالة نموذجًا للعديد من أنواع فحص الأمراض في الرعاية الأولية وتثير أسئلة مهمة مثل ما يجب القيام به مع نتائج الفحص الإيجابية من أداة الذكاء الاصطناعي التي تم تشغيلها على مجموعة من المرضى.

إدارة الأمراض الميسرة بالذكاء الاصطناعي: ارتفاع ضغط الدم

تستكشف مؤسستنا استراتيجية صحة رقمية لإدارة ارتفاع ضغط الدم، من خلال دمج قياسات ضغط الدم المنزلية ودعم اتخاذ القرار السريري المدعوم بالذكاء الاصطناعي من خلال سجل على مستوى المجموعة [41، 42].
النهج لديه القدرة على مساعدة أطباء الرعاية الأولية في تقديم رعاية دقيقة لارتفاع ضغط الدم بناءً على الخصائص الفريدة للمرضى، مع تزويدهم أيضًا بالأدوات والكفاءات للقيام بذلك على مستوى السكان. بالإضافة إلى ذلك، قدمت مؤسستنا خدمة صحة سكانية تتيح لأطباء الرعاية الأولية إحالة المرضى الذين يعانون من ارتفاع ضغط الدم إلى إدارة الأدوية الرقمية التي تتم عن بُعد بواسطة الممرضين والصيادلة. هذه الحالة تمثل نموذجًا لإدارة الأمراض المزمنة في الرعاية الأولية وتثير تساؤلات مثل: كيف يمكن للذكاء الاصطناعي تعزيز قدرات أطباء الرعاية الأولية أو تنسيق الرعاية بين أعضاء فريق الرعاية الأولية المختلفين؟

المهام الإدارية المدعومة بالذكاء الاصطناعي: رسائل المرضى

زيادة الضغط على صندوق الوارد، الذي تفاقم خلال جائحة كوفيد-19، يساهم في الإرهاق و”وقت البيجامة” في الرعاية الأولية [45]. للتخفيف من هذا التحدي، يقوم مركزنا الطبي حاليًا بتجربة استخدام نماذج اللغة الكبيرة مثل ChatGPT لصياغة ردود رسائل المرضى ضمن السجل الصحي الإلكتروني [45]. هذه الحالة تمثل نموذجًا لكيفية مساعدة الذكاء الاصطناعي في المهام الإدارية وتثير تساؤلات تتعلق بالتأثيرات المحتملة على العلاقة بين الطبيب والمريض.

استطلاع رقمي

كخطوة أولى في نهجنا المختلط، استخدمنا استبيانًا رقميًا (مرفق) تم تطويره خصيصًا لمشروعنا لقياس مواقف مقدمي الرعاية الأولية. نظرًا للطبيعة الجديدة لتركيز بحثنا وغياب استبيانات موثوقة مسبقًا، قاد أعضاء فريق البحث لدينا ذوي الخبرة في البحث النوعي إنشاء أداة الاستبيان، مما ضمن الحيادية، والصرامة المنهجية، والقدرة على التقاط الرؤى الدقيقة. كانت أداة الاستبيان لدينا وصفية في طبيعتها بشكل أساسي ووفرت مصدر بيانات قيم لفهم تكرار الردود وتقديم إطار للمقابلات اللاحقة. باستخدام مقاييس ليكرت، استكشف الاستبيان مستويات راحة المشاركين وإدراكاتهم حول الذكاء الاصطناعي في الرعاية الصحية. كما جمعنا بيانات ديموغرافية غير محددة الهوية لوضع وجهات النظر في سياقها. وقد التقط الاستبيان ردودًا من مجموعة متنوعة من أطباء الرعاية الأولية. ” )، موفرًا وجهات نظر من تخصصات الرعاية الأولية المختلفة وإعدادات الممارسة. شملت العينة أطباء من كلية الطب الداخلي في AMC ( )، كلية الطب العائلي في AMC ( AMC الطب الباطني السريري ) ، الطب العائلي السريري في AMC ( ).
تراوحت أعمار المستجيبين بين 25-34 عامًا )، من 35 إلى 44 عامًا ( )، من 45 إلى 54 عامًا ( ) ، سنة كان التنوع الجنسي واضحًا مع ذكر ) و أنثى ) المستجيبين. اختلفت سنوات الخبرة في الممارسة (المتوسط سنوات؛ )، بما في ذلك سنوات ( ) ، سنوات ( ) ، سنوات ( ) ،
سنوات سنوات سنوات 31-35 سنة هذا الطيف يضمن رؤى واسعة حول وجهات نظر PCP عبر مختلف الفئات السكانية ومراحل الحياة المهنية.

مقابلة شبه منظمة

في الاستطلاع، طُلب من المشاركين إبداء استعدادهم للمشاركة في مقابلة متابعة. باستخدام تنسيق مقابلة شبه هيكلية مع دليل مقابلة (ملحق) تم تطويره بشكل تكراري لمشروعنا بالتعاون مع خبراء البحث النوعي، قدمنا للمشاركين منصة مفتوحة وقابلة للتكيف لمشاركة وجهات نظرهم، والتي قمنا بعد ذلك بدراستها باستخدام التحليل الموضوعي. تم إجراء المقابلات في بيئة سرية عبر برنامج مؤتمرات عن بُعد (زووم). تم استخدام برنامج النسخ الآلي (أوتار.إيه.آي“) تم استخدامه لتوليد النسخ وتم تحليل بيانات المقابلات التي تم جمعها بشكل دقيق باستخدام برنامج كويركوس لتحليل البيانات النوعية. تضمنت هذه الطريقة عملية منهجية لترميز وتصنيف الردود لتحديد الأنماط المتكررة والرؤى والمواضيع الناشئة. من خلال التحسين التكراري، استخرجنا مواضيع ذات مغزى تعكس جوهر آراء مقدمي الرعاية الأولية بشأن الذكاء الاصطناعي في الرعاية الأولية. بعد المقابلات الأولية لدينا ( ) التي سلطت الضوء على مخاوف PCP بشأن زيادة عبء العمل والتوقعات، أضفنا سؤالاً للتعمق أكثر في كيفية تأثير الذكاء الاصطناعي على علاقة الطبيب بالمريض. كان العدد النهائي للمقابلات 15 .

النتائج

الآراء العامة حول الذكاء الاصطناعي في الطب

أعربت الغالبية العظمى من المشاركين في الاستطلاع (76.6%) عن وجهة نظر متفائلة بشأن إمكانيات الذكاء الاصطناعي في الطب. اختلفت مستويات الراحة في دمج التقنيات المعتمدة على الذكاء الاصطناعي في الممارسة السريرية عبر مجالات مختلفة (الجدول 1).
بينما أفاد بعض الأطباء بأنهم يشعرون بالراحة في توصيل دور الأدوات المعتمدة على الذكاء الاصطناعي للمرضى (مريح جداً: مريح إلى حد ما: نسبة كبيرة لم تفعل (غير مريحة بعض الشيء: غير مريح للغاية: ). من المهم، من الأطباء الذين تم استطلاع آرائهم وصفوا نهجهم في التعلم عن الذكاء الاصطناعي في الطب بأنه “التعلم بشكل غير نشط من خلال مصادر الأخبار الشعبية أو المحادثات غير الرسمية” مع فقط “السعي بنشاط للحصول على التعليم من خلال منظمات معترف بها، أو دورات دراسية، أو محاضرات، أو مجلات مهنية، أو كتب.””

القلق بشأن الذكاء الاصطناعي في الرعاية الأولية

على الرغم من الإيجابية العامة التي تم قياسها في الاستطلاع، أعرب المشاركون في المقابلات عن العديد من المخاوف بشأن الذكاء الاصطناعي – خاصة عند مناقشة حالات استخدام الذكاء الاصطناعي المحددة. قمنا بتحليل بيانات المقابلات لدينا حول المخاوف.
الجدول 1 تمثل النسب المئوية في الجدول نسبة مقدمي الرعاية الأولية الذين أبلغوا عن مستويات متفاوتة من الراحة مع مشاركة الذكاء الاصطناعي في مجالات مختلفة كما تم الإبلاغ عنها من خلال الاستطلاع الرقمي.
نطاق مريح جداً مريح إلى حد ما محايد غير مريح إلى حد ما غير مريح للغاية
فحص الأمراض ٢٩.٨٪ ٤٦.٨٪ ٨.٥٪ ٨.٥٪ 6.4%
إدارة الأمراض المزمنة 25.5% ٤٦.٨٪ 12.8% ٨.٥٪ 6.4%
تشخيص الأمراض ٨.٥٪ 42.6% 10.6% ٢٣.٤٪ 14.9%
المهام الإدارية ٤٠.٤٪ 25.5% 12.8% 14.9% 6.4%
الشكل 1 مخاوف الأطباء العامين بشأن الذكاء الاصطناعي. الوصف: الموضوعات الناشئة من المقابلات مع الأطباء العامين بشأن الذكاء الاصطناعي مقسمة إلى مخاوف بشأن تقنية الذكاء الاصطناعي نفسها ومخاوف بشأن سياق وطريقة تنفيذ الذكاء الاصطناعي.
فيما يتعلق بالذكاء الاصطناعي، تم إجراء تحليل ترميز موضوعي وتحديد الموضوعات التالية التي تم تصنيفها على أنها مخاوف تتعلق بالتكنولوجيا أو بالأشخاص والعمليات (الشكل 1) [46].
تضمنت المخاوف التكنولوجية عوامل مثل التحيز الخوارزمي (مشارك واحد) أو الدقة والسلامة (7 مشاركين).
“الشيء الذي أشعر بالقلق بشأنه هو، كيف نقوم بتعليم الذكاء الاصطناعي هذه الأمور لأن بعض هذه التحيزات قد تتسرب.” [المشارك ج]
تتعلق مخاوفي حول الذكاء الاصطناعي في الطب بشكل أساسي بمساحة الدقة. وأداة أشعر أنها موثوقة. [المشارك ج]
كان القلق بشأن الصلاحية الخارجية وقدرة خوارزميات الذكاء الاصطناعي على تقدير الفروق الدقيقة للمرضى المحددين (8 مشاركين) اعتبارًا مهمًا للأطباء الممارسين.
لقد عرفت الكثير من مرضاي الآن لمدة 30 عامًا وأعرف الكثير عنهم. لا يمكن حقًا وضع ذلك في مجموعة بيانات يمكن للذكاء الاصطناعي الاستناد إليها. [المشارك F]
أفاد الأطباء الممارسون الذين تم مقابلتهم بآراء مختلفة حول ما إذا كانوا يعطون الأولوية للشفافية [47] في نماذج الذكاء الاصطناعي (3 مشاركين).
في هذه المرحلة، أريد أن أتمكن من الحصول على تفسير منطقي. [المشارك ل]
أو إذا لم يكن هذا ضروريًا بالنسبة لهم (10 مشاركين).
أنا بخير تمامًا إذا كانت تجربتي مع الوقت تتحسن أكثر فأكثر وأشعر، كما تعلم، أنها تعمل. لا تسألني كيف تعمل، لكنها تعمل. [المشارك أ]
أعتقد تقريبًا أن الأداة تضيف المزيد إلى عملية اتخاذ القرار إذا كانت تعمل خارج تلك العملية العقلانية القابلة للوصول البشري. [المشارك ب]
ومع ذلك، كانت العديد من المخاوف المبلغ عنها تركزت حول قضايا نظامية بدلاً من قضايا تكنولوجية. كانت إحدى المخاوف الشائعة (5 مشاركين) تتعلق بالآثار الطبية القانونية للعمل، أو عدم العمل، بناءً على توجيهات الذكاء الاصطناعي.
إذا كان النظام يقول، ‘مرحبًا، هذه الشخص لديه انقطاع تنفس أثناء النوم الشديد، وماذا لو تعرض لحادث سيارة غدًا وكان لدينا تلك البيانات اليوم؟ [المشارك O]
عند مناقشة إمكانية الكشف الخوارزمي عن انقطاع النفس أثناء النوم، أشار مقدمو الرعاية الأولية (5 مشاركين) إلى أنه دون تعزيز قدرة النظام على تشخيص وعلاج المزيد من المرضى الذين يعانون من انقطاع النفس أثناء النوم بشكل قاطع، فلن يكون أداة الذكاء الاصطناعي مفيدة.
عندما نقوم بتحميل نظامنا من هذا الجانب، نحتاج إلى أن تكون الموارد متاحة من الجانب الآخر. [المشارك E]
أعتقد أننا بحاجة إلى أن نكون مجهزين بشكل أفضل قبل أن نبدأ في إخبار الناس بهذا لأنه، كما تعلم، الأمر يشبه، مرحبًا، قد يكون لديك انقطاع النفس أثناء النوم. انتظر ستة أشهر لدراسة نومك… [المشارك ج]
بالإضافة إلى ذلك، أبلغ 10 مشاركين عن قلقهم من أن دمج الذكاء الاصطناعي في الرعاية الأولية قد يؤدي إلى زيادة عبء العمل وإرهاق الأطباء.
قلقي هو أنه مثل كل شيء آخر حاولنا القيام به لتحسين الأمور في الطب، فإنه في الواقع يجعل الأمور أصعب على الطبيب ويخلق المزيد من العمل لنا بدلاً من تقليل العمل. [المشارك د]
أبلغ المشاركون عن طرق متعددة يمكن أن يحدث بها ذلك بما في ذلك أدوات الذكاء الاصطناعي التي تفوض العمل للأطباء الذي يمكن أن يتعامل معه أعضاء الفريق الآخرين،
هل يجب أن يتعامل الأطباء مع هذا في المقام الأول؟ [المشارك ك]
حاجة للتحقق المستمر أو إعادة القيام بالعمل الذي أنجزه الذكاء الاصطناعي،
إنه مثل وجود طالب معي طوال الوقت، حيث يجب أن أتحقق من كل شيء مرتين. [المشارك د]
أو تركيز مفرط على الإنتاجية.
سنضيف مريضين إضافيين لكل جلسة لأنه الآن لدينا مساعدة هناك. لذا، للأسف، أحيانًا تُستخدم المزيد من المساعدة بطريقة سلبية. [المشارك هـ]
لقد رأيت دائمًا أن النظام يريد الإنتاجية، والطريقة التي يتم تعريف الإنتاجية بها تعتمد على عدد المرضى الذين تم رؤيتهم. [المشارك ل]
لم يشارك جميع أطباء الرعاية الأولية هذا القلق بشأن زيادة عبء العمل بسبب الذكاء الاصطناعي، حيث أعرب أحد المشاركين عن أن الكفاءة المتزايدة بسبب الذكاء الاصطناعي ستكون مرحبًا بها حتى لو كان ذلك يعني رؤية المزيد من المرضى.
إذا كنت أستطيع رؤية 30 مريضًا في يوم واحد، وأنهيت سجلاتي بحلول الساعة 6 مساءً، مبتسمًا، وأعود إلى المنزل لتناول العشاء، سأكون سعيدًا. [المشارك ز]
كما أبلغ الأطباء عن مخاوف بشأن كيفية تأثير الذكاء الاصطناعي على العلاقة بين الطبيب والمريض. أعرب بعضهم عن آمال إيجابية في أن يحسن الذكاء الاصطناعي العلاقة بين الطبيب والمريض (10 مشاركين) من خلال القيام بأشياء مثل تخفيف عبء الأطباء أو تحسين تفاعل المرضى.
ربما تكون أكثر تعاطفًا في اللقاء، لأنك لم تضطر إلى القيام بكل ذلك الحمل الذهني. [المشارك م]
لكن العديد – بعضهم أيضًا أعرب عن مشاعر إيجابية – كانوا قلقين من أن الذكاء الاصطناعي قد يضر بالعلاقة بين الطبيب والمريض (12 مشاركًا) بسبب عوامل مثل تشويه توقعات المرضى أو إعطاء الأولوية لاحتياجات المرضى على رفاهية الطبيب.
قد ينتهي الأمر بالمرضى إلى الشعور بأنه، كما تعلم، إذا كان بإمكان الذكاء الاصطناعي إخباري بذلك، فلماذا أزعجت نفسي بالقدوم إليك؟ [المشارك ل]
كلما دربنا المرضى على توقع الأشياء بسرعة وكفاءة، زادت التوقعات على الطبيب ليحقق ذلك بنفس الطريقة. [المشارك د]
كما أعرب أطباء الرعاية الأولية المشاركون عن أسفهم لعدم التركيز على رفاهية الأطباء عند تنفيذ تقنيات جديدة (9 مشاركين).
أشعر أنه بمجرد أن أكون فعالًا، يتم تقديم شيء جديد. [المشارك م]
النظام كله يتعلق برضا المرضى. هل هناك أي تركيز على رضا الأطباء؟ [المشارك هـ]
كانت نقطة رئيسية هي القلق من أن الطريقة الحالية التي يتم دفع الرعاية الصحية بها لا تشجع على طرق مبتكرة لتقديم الرعاية مثل أدوات الصحة الرقمية المدعومة بالذكاء الاصطناعي.
في جوهرها، نحن نقدم مجموعة من الرعاية المجانية، والتي، كما تعلم، ليست مستدامة. [المشارك ح]
لذا، يبدو أن أطباء الرعاية الأولية يرون انفصالًا بين ابتكار الرعاية والطريقة التي يُجبرون على الممارسة بسبب كيفية تعويض الرعاية.
يُطلب منهم القيام بالأمرين. لذا، هم حقًا هناك لإنتاجها، إنتاج هذه الوحدات القابلة للتعويض بينما يقومون أيضًا بممارسة الطب القائم على القيمة والطب القائم على السكان. [المشارك ي]
تم التعبير عن الحاجة إلى وقت مخصص للصحة الرقمية بالإضافة إلى نماذج تعويض بديلة بشكل متكرر (11 مشاركًا) كعامل رئيسي في اعتماد ونجاح أدوات الذكاء الاصطناعي.
أود أن يكون لدي وقت مخصص يوميًا أو على الأقل أسبوعيًا للمراجعة. خلاف ذلك، قد أراه فقط إذا رأيت المريض. [المشارك ف]
أعتقد أن هذا النظام يحتاج حقًا إلى إعادة التفكير في كيفية توظيفه للأطباء ومقدمي الخدمات. [المشارك ل]
تؤكد بيانات استطلاعنا هذه النتيجة بأن أطباء الرعاية الأولية غير متأكدين من مكان تناسب أدوات الصحة الرقمية في سير عملهم. عندما سُئلوا عن تفضيلاتهم بشأن متى يتلقون اتصالات من أداة ذكاء اصطناعي بشأن فحص مريض إيجابي لانقطاع النفس أثناء النوم، كانت الردود متنوعة على نطاق واسع مع 3 مستجيبين يستخدمون الاستجابة الحرة
للإشارة إلى أنهم لا يرغبون في أن يتم إبلاغهم. عندما سُئلوا عما إذا كانوا على علم بسجل إلكتروني موجود مسبقًا للمرضى الذين يعانون من ارتفاع ضغط الدم، أجاب ما يقرب من نصفهم (45.83%) بلا. بالنسبة لأولئك الذين كانوا على علم بالسجل، أبلغ أكثر من نصفهم أن استخدامهم للسجل كان “نادراً” ( ) أو “أبدًا” ( ) مع كون السبب الأكثر شيوعًا المبلغ عنه هو نقص الوقت.

نقاش

الذكاء الاصطناعي كحد سيفين

تكشف نتائجنا عن الطبيعة المزدوجة للذكاء الاصطناعي في الرعاية الصحية، كاشفة عن إمكانيته في تخفيف أو تفاقم التحديات في الرعاية الأولية. بعض المخاوف التي حددناها بشأن اعتماد الذكاء الاصطناعي في الرعاية الصحية، بما في ذلك نقص الصلاحية الخارجية، وإمكانية التحيز، وقضايا السلامة، تم توثيقها جيدًا في الأدبيات [48-50]. توسع دراستنا في هذه المخاوف، مشددة على أنه بالنسبة للأطباء، قد تكون آليات الذكاء الاصطناعي نفسها في المرتبة الثانية مقارنة بتأثيرها المحتمل على حياتهم المهنية، ورفاهيتهم الشخصية، وعلاقاتهم مع المرضى [51]. نجادل بأن المخاوف مثل القلق بشأن زيادة عبء العمل تنبع من شعور أوسع بين أطباء الرعاية الأولية بأن التقدم في الرعاية الصحية غالبًا ما يعطي الأولوية للإنتاجية على رفاهية الطبيب أو يضع الاعتبارات المالية فوق العلاقات الإنسانية [5254]. وبناءً عليه، قد تكون بعض مخاوف أطباء الرعاية الأولية بشأن الذكاء الاصطناعي انعكاسًا لخيبة الأمل مع المشهد المتطور للطب بشكل عام. في هذا السياق، يُنظر إلى إدخال الذكاء الاصطناعي على أنه حالة أخرى حيث قد تُقمع مصالح الأطباء لصالح الكفاءة التنظيمية. هذه المخاوف ليست غير مبررة، حيث اقترحت الأدبيات السابقة استخدام التكنولوجيا لزيادة القدرة كواحدة من الحلول لمواكبة نقص الأطباء المتزايد [55، 56]. علاوة على ذلك، يُشعر بالفعل أن أحجام لوحات أطباء الرعاية الأولية مفرطة، وقد يكون الخوف من استخدام الذكاء الاصطناعي لتبرير إضافة المرضى مبررًا [57-59]. يجب تطبيق الصحة الرقمية والذكاء الاصطناعي في الرعاية الأولية بعناية لتجنب المزيد من إبعاد أطباء الرعاية الأولية عن قيمهم المهنية.
أعتقد أننا جميعًا نشعر بالقلق من أن المزيد من العمل هو ما تهدف إليه الأمور. [المشارك ف]
هذه ليست مخاوف أنانية حيث أن رفاهية الأطباء مرتبطة ارتباطًا وثيقًا بنتائج المرضى وتتوافق مع الهدف الرباعي للرعاية الصحية [60، 61]. هذه المخاوف بشأن رفاهية الأطباء ذات أهمية خاصة في سياق الرعاية الأولية – حجر الزاوية للرعاية الصحية الحيوية لتوفير الوصول إلى الفئات السكانية المحرومة التي يتم التقليل من قيمتها بشكل مزمن من قبل نظام الرعاية الصحية [62، 63]. يجب أن يكون أطباء الرعاية الأولية قادرين على المشاركة في الفوائد مثل الوقت أو التوفير في التكاليف الناتجة عن تنفيذ أنظمة الذكاء الاصطناعي. تشير نتائجنا إلى أن
الشكل 2 تطور العلاقة بين الطبيب والمريض. الوصف: كان لظهور الإنترنت تأثيرات كبيرة على العلاقة بين الطبيب والمريض. يشعر أطباء الرعاية الأولية بمزيج من القلق والتفاؤل بشأن كيفية تأثير الذكاء الاصطناعي على نفس الشيء
الشكل 3 الذكاء الاصطناعي كعضو في فريق الرعاية الصحية. الوصف: الذكاء الاصطناعي يصبح عضوًا في فريق الرعاية الأولية
إذا كانت كل الفوائد تذهب إلى المنظمة، فستظل شهية الأطباء للاعتماد منخفضة.
تسلط أعمالنا الضوء أيضًا على الدور المتطور لطبيب الرعاية الأولية [64]. كان يُنظر إليهم سابقًا كمصدر للحقيقة الطبية، يتعايش الأطباء الآن مع “دكتور جوجل”، وهو مستودع رقمي للمعلومات الصحية يمكّن المرضى من المشاركة بشكل استباقي في رعايتهم [65]. كان لهذا التغيير تأثيرات مختلطة على العلاقة بين الطبيب والمريض، ولكن يمكن أن تكون إيجابية إذا شارك الطرفان في التواصل المناسب واتخاذ القرار المشترك [66-68]. تعتمد هذه التأثيرات الإيجابية أيضًا على عوامل مثل قوة معرفة المرضى بمعلومات الصحة ووقت التواصل الكافي بين الطبيب والمريض [69].
في ظل هذا السياق، ستحدث عملية دمج الذكاء الاصطناعي في الرعاية الصحية ثورة مماثلة في ديناميكية العلاقة بين الطبيب والمريض.
يجب تكرار وتوسيع الكثير من الأبحاث التي أجريت لاستكشاف تأثير معلومات الصحة على الإنترنت على علاقة الطبيب بالمريض في سياق الذكاء الاصطناعي. على عكس المعلومات الثابتة على الإنترنت، فإن أنظمة الذكاء الاصطناعي التفاعلية مستعدة لتولي دور أكثر نشاطًا في تشكيل التفاعلات بين المرضى والأطباء. إن ضمان تأثير إيجابي للذكاء الاصطناعي على علاقة الطبيب بالمريض أمر ضروري للحفاظ على العقد الاجتماعي للطب مع المجتمع.
دور الذكاء الاصطناعي في تزويد المرضى بالمعلومات، سواء كانت دقيقة أو مضللة، يمكن أن يربك الديناميكية بين الطبيب والمريض. المرضى الذين يصلون بمعلومات مستندة إلى الذكاء الاصطناعي
يمكن أن تعقد المعلومات اتخاذ القرارات التعاونية. تشمل العواقب المحتملة الالتزام الصارم بنصائح الذكاء الاصطناعي دون النظر في التاريخ الطبي الفردي أو الصعوبات التي يواجهها الأطباء في محاولة التوفيق بين خبراتهم واقتراحات الذكاء الاصطناعي. لا يهدد هذا فقط بتآكل دور طبيب الرعاية الأولية، بل يمكن أن يعيد تشكيل العلاقة بين الطبيب والمريض إلى نموذج المستهلك والمزود.
أنا قلق بشأن الذكاء الاصطناعي، حيث يتم تعزيز المعلومات المضللة… إذا جاء مريض وقال، مرحبًا، كما تعلم، أخبرني WebMD.GPT أنني بحاجة إلى تصوير بالرنين المغناطيسي، فهناك شيء آخر قوي أجادل ضده. [المشارك ب]
جزء أساسي من ضمان تأثير إيجابي للذكاء الاصطناعي على علاقة الطبيب بالمريض هو تعزيز توقعات واقعية ومتوافقة بشأن الذكاء الاصطناعي من خلال التعليم للأطباء والمرضى قبل تنفيذ أدوات الذكاء الاصطناعي.
“بدلاً من مجرد الإفصاح عن الأمر ورؤية ما سيحدث، تريد التأكد من أن كل من سيتفاعل مع ذلك لديه توقعات دقيقة وقد تم توعيته بالدور الذي من المفترض أن يلعبه.” [المشارك ج]
لقد أبرزت الأدبيات السابقة المخاوف من أن المعرفة غير المتكافئة أو الخلفيات المختلفة في علاقة الطبيب بالمريض قد تؤدي إلى تفاقم عدم المساواة في الصحة. يمكن أن يشجع التنفيذ المدروس والعادل للذكاء الاصطناعي على زيادة مشاركة المرضى وفهمهم، مما يؤدي إلى تحسين التواصل بين الطبيب والمريض وزيادة العدالة.

مستقبل سير العمل في الرعاية الأولية

تدعو نتائجنا إلى إعادة النظر في الأسئلة الأساسية المتعلقة بسير العمل في الرعاية الأولية. إذا كان من المفترض أن يكون مقدمو الرعاية الأولية مشاركين نشطين في أشكال جديدة من تقديم الرعاية الصحية، بما في ذلك الصحة الرقمية المدعومة بالذكاء الاصطناعي، متى يُفترض بهم القيام بذلك العمل؟ إن هذا التحول نحو الصحة الرقمية يحدث بالفعل، وإن كان بطريقة غير مجدولة وغير مدفوعة الأجر.
معظم رعايتنا تُقدم عبر MyChart. دعونا نكون صادقين، هكذا يتم تقديمها. [المشارك ج]
عبء صندوق الوارد هو مشكلة معروفة، لكنه مجرد بداية للصراع بين عبء العمل الرقمي غير المتزامن والعبء الشخصي المتزامن في الرعاية الأولية. لقد فشلت أنظمة الصحة في تحديد الطرق المناسبة لتخصيص الوقت والموارد والمعايير لهذه الطرق الجديدة للتفاعل مع المرضى، وقد أدى ذلك بالفعل إلى تأثيرات كبيرة.
العواقب [77]. من المحتمل أن تصبح الأدوات الرقمية المدعومة بالذكاء الاصطناعي لإدارة الأمراض المزمنة وفحص الأمراض، المعززة بأنظمة مراقبة المرضى عن بُعد، أكثر شيوعًا [78، 79]. وبناءً عليه، يجب النظر في كيفية تخصيص وقت الأطباء بشكل أفضل لدعم الصحة الرقمية. النجاح في الصحة الرقمية يتجاوز مجرد حل مشكلة تراكم الرسائل الواردة أو إرهاق الإنذارات، بل يتعلق بتحقيق تحول أساسي في كيفية تفاعل الرعاية الأولية مع المرضى ورعايتهم [6]. نقترح أن تجربة جداول العمل الهجينة التي تجمع بين الحضور الشخصي والافتراضي يمكن أن تمكن الأطباء من تحقيق إمكانيات الصحة الرقمية [80].
اعتبار آخر هو: كيف يمكن دمج الذكاء الاصطناعي في الرعاية الأولية المتمحورة حول المريض والمعتمدة على الفريق؟ يتكون نهج صحة السكان في الرعاية الأولية من طبيب يعمل كـ “قائد الرعاية الصحية” المسؤول عن صحة مجموعة كاملة من المرضى، بغض النظر عن كيفية أو مكان أو من يقدم كل عنصر من عناصر الرعاية. غالبًا ما أشار المشاركون في المقابلات إلى أنه ليس من الضروري أن يكون الأطباء دائمًا نقطة الاتصال الرئيسية لتوصية الذكاء الاصطناعي. يمكن أن يساعد تحديد متى يمكن لأعضاء الفريق الآخرين مراجعة والتصرف بناءً على التوجيهات التي ينتجها الذكاء الاصطناعي مع الحفاظ على طبيب الرعاية الأولية في الصورة في التخفيف من المخاوف حول إمكانية أن يؤدي الذكاء الاصطناعي والصحة الرقمية إلى زيادة عبء العمل على الأطباء. يجب أن يتم ذلك بعناية بطريقة تعزز – بدلاً من تقويض – العلاقة الأساسية بين الطبيب والمريض.
بشكل عام، تنسيق الرعاية هو تحدٍ أساسي في الرعاية الأولية. يجب إجراء المزيد من الأبحاث حول استخدام الذكاء الاصطناعي كوسيلة لتسهيل متابعة المهام، والتفويض، ومكونات أخرى من تنسيق الرعاية (الشكل 3) [84]. تصميم سير العمل القياسي للرعاية الأولية يمثل تحديًا [85]. ومع ذلك، فإن تمكين مقدمي الرعاية الأولية من الوقت، وأعضاء الفريق، والحوافز لاستخدام أدوات رقمية مدعومة بالذكاء الاصطناعي تتعامل مع بعض جوانب الرعاية عن بُعد وبشكل غير متزامن يمكن أن يسهل زيارات العيادة الشخصية بشكل أكثر معنى وفعالية وتركيزًا [86-88]. كما أظهرت الدراسات أن قضاء المزيد من الوقت في طرق الرعاية الرقمية يحسن أيضًا مقاييس جودة الرعاية [89]. حاليًا، يشارك مقدمو الرعاية الأولية بالفعل في هذا النوع من العمل، لكنهم يفعلون ذلك على حساب الوقت مع مرضاهم أو عائلاتهم [75، 90]. يجب على المنظمات أن تكون حذرة من الاعتماد كثيرًا على الإيثار الطبي لإيجاد الوقت لاستخدام الأدوات الرقمية، مما يجبر الأطباء على الاختيار بين رفاهيتهم الشخصية ورفاهية مرضاهم [91-93].

اعتبارات لأصحاب المصلحة في الرعاية الأولية

كخطوة أولى، يحتاج المعنيون إلى ضمان تقييم أنظمة الذكاء الاصطناعي في الرعاية الأولية بدقة من حيث الدقة والسلامة والتحيز. ومع ذلك، يجب أن تُعطى الاهتمام اللاحق لدمج سير العمل وتأثيره على رفاهية الأطباء (الجدول 2) [86، 94]. لقد أظهرت الأدبيات السابقة أن مثل هذه العوامل غير التقنية
الجدول 2 التوصيات لأصحاب المصلحة في الرعاية الأولية
المعنيون توصية
الدافعين تنفيذ نماذج سداد مبتكرة لمقدمي الرعاية الأولية وأعضاء فريق الرعاية الأولية الآخرين المشاركين في الصحة الرقمية
أنظمة الرعاية الصحية جدولة الوقت وتحديد المعايير لمقدمي الرعاية الأولية للمشاركة في الصحة الرقمية
أنظمة الرعاية الصحية تزويد مقدمي الرعاية الأولية بأعضاء فريق إضافيين مثل الصيادلة أو منسقي المرضى الذين يمكنهم التفاعل رقميًا مع المرضى
أنظمة الرعاية الصحية تطوير وتوزيع مواد تعليمية حول الدور الصحيح للأدوات الجديدة للذكاء الاصطناعي للمرضى والأطباء قبل تنفيذ الأداة
الباحثون قم بإجراء تجارب عشوائية محكومة، تجارب عملية، أو أبحاث قائمة على الممارسة بين سير العمل التقليدي وسير العمل المعزز رقميًا لمقدمي الرعاية الأولية.
الباحثون تقييم أدوات الرقمية المدعومة بالذكاء الاصطناعي الجديدة في سياق سير عمل الأطباء بدلاً من بيئة معزولة
PCPs الدفاع بشكل فردي وجماعي عن أدوات الذكاء الاصطناعي التي تحسن جودة رعاية الأطباء، والرفاهية، وعلاقة الطبيب بالمريض
تعتبر ضرورية لتعزيز اعتماد التقنيات الجديدة [95]. يجب توخي الحذر لضمان عدم تسبب الذكاء الاصطناعي في “تقليل عمل الأطباء وزيادة العمل المكتبي” مما يؤدي إلى شعور مقدمي الرعاية الأولية بالاستغلال من قبل نظام الرعاية الصحية [10]. بدلاً من ذلك، يجب استخدام الذكاء الاصطناعي كفرصة لمعالجة تحديات الرعاية الأولية مثل المساعدة في تقديم رعاية أفضل مع منع إرهاق الأطباء [96].
بالإضافة إلى ذلك، فإن الطريقة التي يتم بها تعويض خدمات الرعاية الصحية لها تأثير كبير على سلوك وتخصيص الوقت لمقدمي الرعاية الأولية. تم القيام ببعض المحاولات، مثل فرض رسوم على المستخدمين مقابل رسائل المرضى وفوترة الدافعين مقابل الزيارات الإلكترونية (أي، الرد على رسائل المرضى)، لتعويض خدمات الصحة الرقمية في نموذج الدفع مقابل الخدمة. ومع ذلك، من المحتمل أن تكون الصحة الرقمية أكثر ملاءمة للرعاية الأولية القائمة على القيمة، التي من شأنها تحفيز وتوفير المرونة للأطباء للمشاركة في أدوات الصحة الرقمية غير المتزامنة على مستوى السكان. يجب على المنظمات التي ترغب في جني فوائد الذكاء الاصطناعي في الرعاية الأولية مواجهة هذا التحدي بشكل مباشر وأن تكون مستعدة لإعادة تصور كيفية تقديم الرعاية ودفع تكاليفها بدلاً من تعزيز الأنظمة والأساليب القديمة. هذا الأمر مهم بشكل خاص عند النظر في طبيعة تعويض الرعاية الصحية في الولايات المتحدة. هناك جهود جارية لتحديث أنظمة الدفع، ولكن الفشل في التقدم بسرعة وابتكار نماذج الدفع لدينا قد يؤدي إلى تخلف أنظمتنا عن الدول التي لا تعتمد على نموذج الدفع مقابل الخدمة من حيث الابتكار في الذكاء الاصطناعي والصحة الرقمية.

الذكاء الاصطناعي والعدالة

الناس الذين سيفهمون ذلك هم الأشخاص الذين يمكنهم دفع ثمن الحوسبة. لذا فإن أكبر مخاوفي هي
أننا سنستبعد أفقر الناس من الحصول على أفضل رعاية. – المشارك ج.
هذا الشعور الذي عبّر عنه العديد من المشاركين في الدراسة يبرز أهمية ضمان الوصول العادل إلى حلول الرعاية الصحية المدعومة بالذكاء الاصطناعي للمرضى من مختلف الخلفيات الاجتماعية والاقتصادية. يجب ألا تؤدي التحولات التي أحدثها الذكاء الاصطناعي عن غير قصد إلى تعزيز الفجوات القائمة، بل يجب أن تكون أداة لتخفيفها. على سبيل المثال، يجب ألا يخفف الذكاء الاصطناعي من إرهاق الأطباء ويحسن نتائج المرضى فقط في المراكز الطبية الأكاديمية الكبيرة التي تمتلك موارد كافية للمشاركة في تطوير وتنفيذ الذكاء الاصطناعي، بل يجب أن يكون متاحًا وذو تطبيق واسع عبر بيئات ومؤسسات الرعاية الصحية المتنوعة، مما يضمن الوصول العادل إلى فوائده للجميع. يجب أن تشمل ثورات الذكاء الاصطناعي في الرعاية الأولية أنظمة الصحة التي تقدم الدعم، والمراكز الصحية المؤهلة اتحاديًا، والمناطق الريفية، وغيرها من بيئات الممارسة التي قد تُستبعد.
“الواقع هو أن مرضانا يحتاجون إلى الكثير من الرعاية الصحية التي لا يمكننا تقديمها، لذا سيحتاجون إلى أدوات الذكاء الاصطناعي.” – مشارك .

نقاط القوة، والقيود، والاتجاهات للبحوث المستقبلية

على عكس الكثير من الأدبيات الموجودة، جميع المشاركين في دراستنا لديهم خبرة فعلية مع الصحة الرقمية وبعض المشاركين في دراستنا لديهم خبرة فعلية مع الذكاء الاصطناعي الطبي. نظرًا لأن هذه التحولات التكنولوجية بدأت للتو، فإن الأطباء الذين لديهم وجهات نظر مستندة إلى خبرة فعلية نادرة، مما يجعل نتائجنا أكثر قيمة. ومع ذلك، فإن أحد القيود هو أن مصطلحات مثل “الذكاء الاصطناعي” أو “الصحة الرقمية” لها تعريفات متطورة وقد لا تعني الشيء نفسه لأفراد مختلفين. بينما حاولنا التخفيف من هذا التأثير من خلال تأصيل مناقشاتنا في حالات استخدام ملموسة وأمثلة، قد تكون الأفكار المسبقة المختلفة للمشاركين حول الذكاء الاصطناعي قد أثرت على ردود المشاركين. علاوة على ذلك، قد يكون التركيز على حالات استخدام معينة دون غيرها قد أثر على آراء مقدمي الرعاية الأولية المبلغ عنها حول الذكاء الاصطناعي بشكل عام. يجب أن تسعى الجهود المستقبلية إلى تقييم تصورات الذكاء الاصطناعي في الرعاية الأولية بشكل أكثر شمولاً لضمان عدم تأثر المواقف المبلغ عنها لمقدمي الرعاية الأولية بشكل مفرط بأي حالة استخدام معينة. بالإضافة إلى ذلك، بينما تم تطوير استبياننا وأدلة المقابلات بطريقة صارمة بالتعاون مع خبراء في الأساليب النوعية، يجب أن تحاول الأبحاث المستقبلية تطوير أدوات نوعية موثوقة لتقييم مواقف مقدمي الرعاية الأولية تجاه الذكاء الاصطناعي في الرعاية الأولية.
ركز بحثنا فقط على مواقف أطباء الطب الباطني وطب الأسرة تجاه الذكاء الاصطناعي والصحة الرقمية. يحتاج البحث المستقبلي إلى تضمين أعضاء آخرين من فريق الرعاية الأولية بما في ذلك أطباء الأطفال في الرعاية الأولية والممرضات.
ومقدمي الرعاية المتقدمين مثل ممارسي التمريض ومساعدي الأطباء. بالإضافة إلى ذلك، هناك حاجة إلى مزيد من البحث حول كيفية تجربة المرضى – وخاصة أولئك الذين قد يكونون مهمشين – للتحول نحو الذكاء الاصطناعي والصحة الرقمية في الرعاية الأولية. أخيرًا، يجب أن يتوسع البحث المستقبلي أيضًا ليشمل تطبيقات الذكاء الاصطناعي الأخرى ذات الصلة بالرعاية الأولية.
حجم عيّنتنا الصغيرة نسبيًا يحد من إمكانية تعميم نتائجنا على مجموعات سكانية أكبر. بالإضافة إلى ذلك، جميع المستجيبين لدينا من نفس المنظمة. على الرغم من أننا حاولنا تقييم مواقف الأطباء الأكاديميين وغير الأكاديميين، إلا أن هذه التجانس التنظيمي قد يحد من إمكانية التعميم بشكل أكبر. علاوة على ذلك، يجب الاعتراف بإمكانية وجود تحيز في الاستجابات الذاتية المبلغ عنها.
في ضوء هذه القيود، نؤكد على الحاجة إلى جهود بحثية مستقبلية تستخدم الأساليب الكمية لاستكشاف الأسئلة المتعلقة بالذكاء الاصطناعي في الرعاية الأولية، ولتضمين عينات أكبر وأكثر تنوعًا من مختلف بيئات الرعاية الصحية. يمكن أن يوفر دمج منظمات متعددة، خاصة تلك التي لا تتمتع بتمويل جيد مثل مراكز الرعاية الأكاديمية في البيئات الحضرية، منظورًا أوسع حول اعتماد الذكاء الاصطناعي والصحة الرقمية في الرعاية الأولية.

الخاتمة

كانت هذه الدراسة الأولى التي تحقق في مواقف مقدمي الرعاية الأولية تجاه الذكاء الاصطناعي في الرعاية الأولية مع التركيز على حالات استخدام محددة للذكاء الاصطناعي. اختلفت المواقف المبلغ عنها، لكن ردود مقدمي الرعاية الأولية أظهرت تفاؤلاً عاماً بشأن الذكاء الاصطناعي في الرعاية الأولية مع وجود بعض المخاوف. بينما ركزت بعض المخاوف على عوامل تكنولوجية مثل دقة الخوارزميات والسلامة والتحيز، ركزت أخرى على عوامل تتعلق بالأشخاص والعمليات مثل التأثيرات على سير عمل الأطباء، والعدالة، والتعويض، وعلاقة الطبيب بالمريض. تشير هذه النتائج إلى أن مبادرات الذكاء الاصطناعي التي تفشل في معالجة كل من المخاوف التكنولوجية ومخاوف الأشخاص والعمليات التي أثارها مقدمو الرعاية الأولية قد تواجه صعوبة في تحقيق تأثير. يجب على المعنيين بالرعاية الأولية استخدام هذه النتائج لإبلاغ تطوير وتنفيذ الذكاء الاصطناعي في الرعاية الأولية.

الاختصارات

الذكاء الاصطناعي
طبيب الرعاية الأولية PCP
مركز AMC الطبي الأكاديمي
انقطاع النفس النومي الانسدادي (OSA)
سجل الصحة الإلكتروني (EHR)

معلومات إضافية

تحتوي النسخة الإلكترونية على مواد إضافية متاحة علىhttps://doi. org/10.1186/s12875-024-02282-y.
المادة التكميلية 1: أسئلة من الاستطلاع الرقمي
المادة التكميلية 2: أسئلة مستمدة من المقابلة شبه المنظمة

شكر وتقدير

غير قابل للتطبيق.

مساهمات المؤلفين

قام MRA وAM وGK بوضع التصميم الأولي للدراسة. قدمت MTS مساهمات جوهرية في تصميم الدراسة وتحسين أداة الاستطلاع. قام MRA وSW بتنفيذ الدراسة وصياغة النسخة الأولية من المخطوطة. قام MRA بترميز نصوص المقابلات وSW تحقق من رموز المقابلات. ساعد MF وAM وMTS وGK في مراجعة المخطوطة وتحليل البيانات وقدموا قيمة فكرية للمخطوطة. جميع المؤلفين يوافقون على النسخة النهائية من المخطوطة.

تمويل

تم دعم هذا العمل من قبل منحة الذكاء الاصطناعي/تعلم الآلة من المجلس الأمريكي لطب الأسرة وزمالة البحث في كلية الطب بجامعة كاليفورنيا في سان دييغو. لم يكن للجهات الممولة أي دور في أي من عناصر الدراسة.

توفر البيانات

ستُدرج أسئلة الاستطلاع الرقمي وأسس أسئلة دليل المقابلة كملحق للمخطوطة. ستُتاح بيانات المقابلات والاستطلاعات الكاملة عند الطلب المعقول.

الإعلانات

تمت الموافقة على طرق الدراسة والبروتوكولات من قبل لجنة تنسيق وتحسين الجودة والبحث والتقييم في جامعة كاليفورنيا سان دييغو (مشروع ACQUIRE رقم 751) التي تعمل تحت سلطة مجلس المراجعة المؤسسية في جامعة كاليفورنيا سان دييغو. تم الحصول على موافقة مستنيرة من جميع المشاركين. تم تنفيذ جميع الطرق وفقًا للإرشادات واللوائح ذات الصلة أو إعلان هلسنكي.
غير قابل للتطبيق.

المصالح المتنافسة

يعلن المؤلفون عدم وجود مصالح متنافسة.
تاريخ الاستلام: 5 أكتوبر 2023 / تاريخ القبول: 19 يناير 2024
نُشر على الإنترنت: 27 يناير 2024

References

  1. Strohm L, Hehakaya C, Ranschaert ER, Boon WPC, Moors EHM. Implementation of artificial intelligence (AI) applications in radiology: hindering and facilitating factors. Eur Radiol. 2020;30(10):5525-32.
  2. Park CJ, Yi PH, Siegel EL. Medical Student perspectives on the impact of Artificial Intelligence on the practice of Medicine. Curr Probl Diagn Radiol. 2021;50(5):614-9.
  3. Briganti G, Le Moine O. Artificial Intelligence in Medicine: Today and Tomorrow. Front Med [Internet]. 2020 [cited 2023 Aug 6];7. Available from: https:// www.frontiersin.org/articles/https://doi.org/10.3389/fmed.2020.00027.
  4. Seneviratne MG, Shah NH, Chu L. Bridging the implementation gap of machine learning in healthcare. BMJ Innov. 2020;6(2):45-7.
  5. Zimlichman E, Nicklin W, Aggarwal R, Bates D, Health Care. 2030: The Coming Transformation. NEJM Catal Innov Care Deliv [Internet]. (March 1, 2021). Available from: https://catalyst.nejm.org/doi/full/https://doi.org/10.1056/ CAT.20.0569.
  6. Pagliari C. Digital health and primary care: past, pandemic and prospects. J Glob Health. 2021;11:01005.
  7. Rabatin J, Williams E, Baier Manwell L, Schwartz MD, Brown RL, Linzer M. Predictors and outcomes of Burnout in Primary Care Physicians. J Prim Care Community Health. 2016;7(1):41-3.
  8. Petterson SM, Liaw WR, Tran C, Bazemore AW. Estimating the Residency Expansion required to avoid projected primary care physician shortages by 2035. Ann Fam Med. 2015;13(2):107-14.
  9. Lin SA, Clinician’s, Guide to Artificial Intelligence (AI). Why and how primary care should lead the Health Care AI revolution. J Am Board Fam Med. 2022;35(1):175-84.
  10. Agarwal SD, Pabo E, Rozenblum R, Sherritt KM. Professional Dissonance and Burnout in Primary Care: a qualitative study. JAMA Intern Med. 2020;180(3):395.
  11. Amano A, Brown-Johnson CG, Winget M, Sinha A, Shah S, Sinsky CA, et al. Perspectives on the Intersection of Electronic Health Records and Health Care Team Communication, function, and well-being. JAMA Netw Open. 2023;6(5):e2313178.
  12. Kearley KE, Freeman GK, Heath A. An exploration of the value of the personal doctor-patient relationship in general practice. Br J Gen Pract J R Coll Gen Pract. 2001;51(470):712-8.
  13. Aminololama-Shakeri S, López JE. The Doctor-Patient Relationship with Artificial Intelligence. Am J Roentgenol. 2019;212(2):308-10.
  14. Nagy M, Sisk B. How will Artificial Intelligence Affect Patient-Clinician relationships? AMA J Ethics. 2020;22(5):E395-400.
  15. Blease C, Kaptchuk TJ, Bernstein MH, Mandl KD, Halamka JD, DesRoches CM. Artificial Intelligence and the future of primary care: exploratory qualitative study of UK General practitioners’ views. J Med Internet Res. 2019;21(3):e12802.
  16. Buck C, Doctor E, Hennrich J, Jöhnk J, Eymann T. General practitioners’ attitudes toward Artificial intelligence-enabled systems: interview study. J Med Internet Res. 2022;24(1):e28916.
  17. Kueper JK, Terry AL, Zwarenstein M, Lizotte DJ. Artificial Intelligence and Primary Care Research: a scoping review. Ann Fam Med. 2020;18(3):250-8.
  18. Marwaha JS, Landman AB, Brat GA, Dunn T, Gordon WJ. Deploying digital health tools within large, complex health systems: key considerations for adoption and implementation. Npj Digit Med. 2022;5(1):1-7.
  19. Downing NL, Bates DW, Longhurst CA. Physician burnout in the Electronic Health Record Era: are we ignoring the Real cause? Ann Intern Med. 2018;169(1):50.
  20. López L, Green AR, Tan-McGrory A, King RS, Betancourt JR. Bridging the Digital divide in Health Care: the role of Health Information Technology in addressing racial and ethnic disparities. Jt Comm J Qual Patient Saf. 2011;37(10):437-45.
  21. Latulippe K, Hamel C, Giroux D. Social Health Inequalities and eHealth: A literature review with qualitative synthesis of theoretical and empirical studies. J Med Internet Res. 2017;19(4):e136.
  22. Weiss D, Rydland HT, Øversveen E, Jensen MR, Solhaug S, Krokstad S. G Virgili editor 2018 Innovative technologies and social inequalities in health: a scoping review of the literature. PLOS ONE 134 e0195447.
  23. Ramsetty A, Adams C. Impact of the digital divide in the age of COVID-19.J Am Med Inform Assoc. 2020;27(7):1147-8.
  24. Yao R, Zhang W, Evans R, Cao G, Rui T, Shen L. Inequities in Health Care services caused by the Adoption of Digital Health Technologies: scoping review. J Med Internet Res. 2022;24(3):e34144.
  25. Joyce K, Smith-Doerr L, Alegria S, Bell S, Cruz T, Hoffman SG, et al. Toward a sociology of Artificial Intelligence: a call for Research on inequalities and Structural Change. Socius Sociol Res Dyn World. 2021;7:237802312199958.
  26. Salhi RA, Dupati A, Burkhardt JC. Interest in serving the Underserved: role of race, gender, and Medical Specialty Plans. Health Equity. 2022;6(1):933-41.
  27. Jetty A, Hyppolite J, Eden AR, Taylor MK, Jabbarpour Y. Underrepresented Minority Family Physicians more likely to care for vulnerable populations. J Am Board Fam Med. 2022;35(2):223-4.
  28. Starfield B, Shi L, Macinko J. Contribution of primary care to health systems and health. Milbank Q. 2005;83(3):457-502.
  29. Blumenthal D, Mort E, Edwards J. The efficacy of primary care for vulnerable population groups. Health Serv Res. 1995;30(1 Pt 2):253-73.
  30. Marrast LM, Zallman L, Woolhandler S, Bor DH, McCormick D. Minority Physicians’ role in the Care of Underserved patients: diversifying the physician workforce may be key in addressing Health disparities. JAMA Intern Med. 2014;174(2):289.
  31. Lambert SI, Madi M, Sopka S, Lenes A, Stange H, Buszello CP, et al. An integrative review on the acceptance of artificial intelligence among healthcare professionals in hospitals. Npj Digit Med. 2023;6(1):1-14.
  32. Stange KC, Ferrer RL. The Paradox of Primary Care. Ann Fam Med. 2009;7(4):293-9.
  33. Shi L. The impact of primary care: a focused review. Scientifica. 2012;2012:1-22.
  34. Liaw W, Kakadiaris IA. Primary care Artificial Intelligence: a branch hiding in Plain Sight. Ann Fam Med. 2020;18(3):194-5.
  35. Liaw W, Kakadiaris IA. Artificial Intelligence and Family Medicine: better together. Fam Med. 2020;52(1):8-10.
  36. Lin SY, Mahoney MR, Sinsky CA. Ten ways Artificial Intelligence Will Transform Primary Care. J Gen Intern Med. 2019;34(8):1626-30.
  37. Fatehi F, Samadbeik M, Kazemi A. What is Digital Health? Review of definitions. Stud Health Technol Inform. 2020;275:67-71.
  38. Motamedi KK, McClary AC, Amedee RG. Obstructive sleep apnea: a growing problem. Ochsner J. 2009;9(3):149-53.
  39. Ramesh J, Keeran N, Sagahyroon A, Aloul F. Towards validating the effectiveness of obstructive sleep apnea classification from Electronic Health Records Using Machine Learning. Healthc Basel Switz. 2021;9(11):1450.
  40. Maniaci A, Riela PM, lannella G, Lechien JR, La Mantia I, De Vincentiis M, et al. Machine learning identification of obstructive sleep apnea severity through the patient clinical features: a retrospective study. Life Basel Switz. 2023;13(3):702.
  41. Molina-Ortiz El, Vega AC, Calman NS. Patient registries in primary care: essential element for Quality Improvement: P ATIENT R EGISTRIES IN P RIMARY C ARE. Mt Sinai. J Med J Transl Pers Med. 2012;79(4):475-80.
  42. Hu Y, Huerta J, Cordella N, Mishuris RG, Paschalidis IC. Personalized hypertension treatment recommendations by a data-driven model. BMC Med Inform Decis Mak. 2023;23(1):44.
  43. Fisher NDL, Fera LE, Dunning JR, Desai S, Matta L, Liquori V, et al. Development of an entirely remote, non-physician led hypertension management program. Clin Cardiol. 2019;42(2):285-91.
  44. Visco V, Izzo C, Mancusi C, Rispoli A, Tedeschi M, Virtuoso N, et al. Artificial Intelligence in Hypertension Management: an Ace up your sleeve. J Cardiovasc Dev Dis. 2023;10(2):74.
  45. Holmgren AJ, Downing NL, Tang M, Sharp C, Longhurst C, Huckman RS. Assessing the impact of the COVID-19 pandemic on clinician ambulatory electronic health record use. J Am Med Inform Assoc. 2022;29(3):453-60.
  46. University NCS, Payton F, Pare G, Montréal HEC, Le Rouge C, Saint L et al. University,. Health Care IT: Process, People, Patients and Interdisciplinary Considerations. J Assoc Inf Syst. 2011;12(2):I-XIII.
  47. the Precise4Q consortium, Amann J, Blasimme A, Vayena E, Frey D, Madai VI. Explainability for artificial intelligence in healthcare: a multidisciplinary perspective. BMC Med Inform Decis Mak. 2020;20(1):310.
  48. Challen R, Denny J, Pitt M, Gompels L, Edwards T, Tsaneva-Atanasova K. Artificial intelligence, bias and clinical safety. BMJ Qual Saf. 2019;28(3):231-7.
  49. Wong A, Otles E, Donnelly JP, Krumm A, McCullough J, DeTroyer-Cooley O, et al. External validation of a widely implemented proprietary Sepsis prediction model in hospitalized patients. JAMA Intern Med. 2021;181(8):1065.
  50. Yu KH, Kohane IS. Framing the challenges of artificial intelligence in medicine. BMJ Qual Saf. 2019;28(3):238-41.
  51. Terry AL, Kueper JK, Beleno R, Brown JB, Cejic S, Dang J, et al. Is primary health care ready for artificial intelligence? What do primary health care stakeholders say? BMC Med Inform Decis Mak. 2022;22(1):237.
  52. Phillips RL, Bazemore AW, Newton WP. PURSUING PRACTICAL PROFESSIONALISM: FORM FOLLOWS FUNCTION. Ann Fam Med. 2019;17(5):472-5.
  53. Relman AS. The New Medical-Industrial Complex. N Engl J Med. 1980;303(17):963-70.
  54. Tai-Seale M, Baxter S, Millen M, Cheung M, Zisook S, Çelebi J et al. Association of physician burnout with perceived EHR work stress and potentially actionable factors. J Am Med Inform Assoc. 2023;ocad136.
  55. Bodenheimer TS, Smith MD. Primary care: proposed solutions to the physician shortage without training more Physicians. Health Aff (Millwood). 2013;32(11):1881-6.
  56. Green LV, Savin S, Lu Y. Primary care physician shortages could be eliminated through use of teams, nonphysicians, and Electronic Communication. Health Aff (Millwood). 2013;32(1):11-9.
  57. Raffoul M, Moore M, Kamerow D, Bazemore A. A primary care panel size of 2500 is neither Accurate nor reasonable. J Am Board Fam Med JABFM. 2016;29(4):496-9.
  58. Harrington C. Considerations for patient panel size. Del J Public Health. 2022;8(5):154-7.
  59. Salisbury C, Murphy M, Duncan P. The impact of Digital-First Consultations on workload in General Practice: modeling study. J Med Internet Res. 2020;22(6):e18203.
  60. Halbesleben JRB, Rathert C. Linking physician burnout and patient outcomes: exploring the dyadic relationship between physicians and patients. Health Care Manage Rev. 2008;33(1):29-39.
  61. Sikka R, Morath JM, Leape L. The Quadruple Aim: care, health, cost and meaning in work. BMJ Qual Saf. 2015;24(10):608-10.
  62. Streeter RA, Snyder JE, Kepley H, Stahl AL, Li T, Washko MM. The geographic alignment of primary care Health Professional Shortage Areas
    with markers for social determinants of health. Shah TI, editor. PLOS ONE. 2020;15(4):e0231443.
  63. Liu J, Jason. Health Professional Shortage and Health Status and Health Care Access. J Health Care Poor Underserved. 2007;18(3):590-8.
  64. Ahuja AS. The impact of artificial intelligence in medicine on the future role of the physician. PeerJ. 2019;7:e7702.
  65. Van Riel N, Auwerx K, Debbaut P, Van Hees S, Schoenmakers B. The effect of Dr Google on doctor-patient encounters in primary care: a quantitative, observational, cross-sectional study. BJGP Open. 2017;1(2):bjgpopen17X100833.
  66. Murray E, Lo B, Pollack L, Donelan K, Catania J, White M, et al. The impact of Health Information on the internet on the physician-patient relationship: patient perceptions. Arch Intern Med. 2003;163(14):1727.
  67. Tan SSL, Goonawardene N. Internet Health Information seeking and the patient-physician relationship: a systematic review. J Med Internet Res. 2017;19(1):e9.
  68. Liaw W, Kakadiaris IA, Yang Z. Is Artificial Intelligence the Key to Reclaiming relationships in Primary Care? Am Fam Physician. 2021;104(6):558-9.
  69. Luo A, Qin L, Yuan Y, Yang Z, Liu F, Huang P, et al. The Effect of Online Health Information seeking on Physician-Patient relationships: systematic review. J Med Internet Res. 2022;24(2):e23354.
  70. Cruess SR. Professionalism and Medicine’s Social Contract with Society. Clin Orthop. 2006;449:170-6.
  71. Resnicow K, Catley D, Goggin K, Hawley S, Williams GC. Shared decision making in Health Care: theoretical perspectives for why it works and for whom. Med Decis Mak Int J Soc Med Decis Mak. 2022;42(6):755-64.
  72. Verlinde E, De Laender N, De Maesschalck S, Deveugele M, Willems S. The social gradient in doctor-patient communication. Int J Equity Health. 2012;11(1):12.
  73. Willems S, De Maesschalck S, Deveugele M, Derese A, De Maeseneer J. Socioeconomic status of the patient and doctor-patient communication: does it make a difference? Patient Educ Couns. 2005;56(2):139-46.
  74. Arndt BG, Beasley JW, Watkinson MD, Temte JL, Tuan WJ, Sinsky CA, et al. Tethered to the EHR: Primary Care Physician Workload Assessment using EHR Event Log Data and Time-Motion observations. Ann Fam Med. 2017;15(5):419-26.
  75. Tai-Seale M, Olson CW, Li J, Chan AS, Morikawa C, Durbin M, et al. Electronic Health Record logs Indicate that Physicians Split Time evenly between seeing patients and Desktop Medicine. Health Aff (Millwood). 2017;36(4):655-62.
  76. Murphy DR, Meyer AND, Russo E, Sittig DF, Wei L, Singh H. The Burden of Inbox Notifications in Commercial Electronic Health Records. JAMA Intern Med. 2016;176(4):559.
  77. Tai-Seale M, Dillon EC, Yang Y, Nordgren R, Steinberg RL, Nauenberg T, et al. Physicians’ well-being linked to In-Basket messages generated by Algorithms in Electronic Health Records. Health Aff (Millwood). 2019;38(7):1073-8.
  78. Muller AE, Berg RC, Jardim PSJ, Johansen TB, Ormstad SS. Can remote patient monitoring be the New Standard in Primary Care of Chronic Diseases, Post-COVID-19? Telemed E-Health. 2022;28(7):942-69.
  79. Willis VC, Thomas Craig KJ, Jabbarpour Y, Scheufele EL, Arriaga YE, Ajinkya M, et al. Digital Health Interventions To Enhance Prevention in Primary Care: scoping review. JMIR Med Inform. 2022;10(1):e33518.
  80. Nochomovitz M, Sharma R. Is it time for a New Medical Specialty? The Medical Virtualist. JAMA. 2018;319(5):437.
  81. Goldberg DG, Beeson T, Kuzel AJ, Love LE, Carver MC. Team-Based Care: a critical element of Primary Care Practice Transformation. Popul Health Manag. 2013;16(3):150-6.
  82. American College of Physicians, Smith CD, Balatbat C, National Academy of Medicine, Corbridge S. University of Illinois at Chicago, Implementing Optimal Team-Based Care to Reduce Clinician Burnout. NAM Perspect [Internet]. 2018 Sep 17 [cited 2023 Aug 1];8(9). Available from: https://nam.edu/ implementing-optimal-team-based-care-to-reduce-clinician-burnout.
  83. Loeb DF, Bayliss EA, Candrian C, deGruy FV, Binswanger IA. Primary care providers’ experiences caring for complex patients in primary care: a qualitative study. BMC Fam Pract. 2016;17(1):34.
  84. Wagner EH, Sandhu N, Coleman K, Phillips KE, Sugarman JR. Improving care coordination in primary care. Med Care. 2014;52(11 Suppl 4):33-8.
  85. Holman GT, Beasley JW, Karsh BT, Stone JA, Smith PD, Wetterneck TB. The myth of standardized workflow in primary care. J Am Med Inform Assoc JAMIA. 2016;23(1):29-37.
  86. Committee on Implementing High-Quality Primary Care, Board on Health Care Services, Health and Medicine Division, National Academies of Sciences, Engineering, and Medicine. Implementing High-Quality Primary Care: Rebuilding the Foundation of Health Care [Internet]. McCauley L, Phillips RL, Meisnere M, Robinson SK, editors. Washington, D.C.: National Academies Press. ; 2021 [cited 2023 Jul 27]. Available from: https://www.nap.edu/ catalog/25983.
  87. Tai-Seale M, McGuire TG, Zhang W. Time Allocation in Primary Care Office visits: Time Allocation in Primary Care. Health Serv Res. 2007;42(5):1871-94.
  88. Hilty DM, Torous J, Parish MB, Chan SR, Xiong G, Scher L, et al. A literature review comparing clinicians’ approaches and skills to In-Person, Synchronous, and Asynchronous Care: moving toward competencies to ensure Quality Care. Telemed E-Health. 2021;27(4):356-73.
  89. Rotenstein LS, Holmgren AJ, Healey MJ, Horn DM, Ting DY, Lipsitz S, et al. Association between Electronic Health Record Time and Quality of Care Metrics in Primary Care. JAMA Netw Open. 2022;5(10):e2237086.
  90. Saag HS, Shah K, Jones SA, Testa PA, Horwitz LI. Pajama Time: Working after Work in the Electronic Health Record. J Gen Intern Med. 2019;34(9):1695-6.
  91. Pellegrino ED, Altruism. Self-interest, and Medical Ethics. JAMA J Am Med Assoc. 1987;258(14):1939.
  92. Sajjad M, Qayyum S, Iltaf S, Khan RA. The best interest of patients, not selfinterest: how clinicians understand altruism. BMC Med Educ. 2021;21(1):477.
  93. Jones R. Declining altruism in medicine. BMJ. 2002;324(7338):624-5.
  94. Yin J, Ngiam KY, Teo HH. Role of Artificial Intelligence Applications in Real-Life Clinical Practice: systematic review. J Med Internet Res. 2021;23(4):e25759.
  95. Choudhury A, Asan O, Medow JE. Clinicians’ perceptions of an Artificial Intelligence-based blood utilization calculator: qualitative exploratory study. JMIR Hum Factors. 2022;9(4):e38411.
  96. Thomas Craig KJ, Willis VC, Gruen D, Rhee K, Jackson GP. The burden of the digital environment: a systematic review on organization-directed workplace interventions to mitigate physician burnout. J Am Med Inform Assoc JAMIA. 2021;28(5):985-97.
  97. Gosden T, Forland F, Kristiansen I, Sutton M, Leese B, Giuffrida A et al. Capitation, salary, fee-for-service and mixed systems of payment: effects on the behaviour of primary care physicians. Cochrane Effective Practice and Organisation of Care Group, editor. Cochrane Database Syst Rev [Internet]. 2000 Jul 24 [cited 2023 Jul 31];2011(10). https://doi.org/10.1002/14651858.CD002215.
  98. Diaz N. More health systems charging for MyChart messages. Becker’s Health IT [Internet]. 2022 Aug 28 [cited 2023 Aug 17]; Available from: https:// www.beckershospitalreview.com/ehrs/more-health-systems-charging-for-mychart-messages.html.
  99. Porter ME, Pabo EA, Lee TH. Redesigning primary care: a Strategic Vision to improve Value by Organizing around patients’ needs. Health Aff (Millwood). 2013;32(3):516-25.
  100. AMA Future of Health issue brief: Commercial Payer Coverage for Digital Medicine Codes. American Medical Association [Internet]. 2023 Sep 18 [cited 2023 Sep 26]; Available from: https://www.ama-assn.org/practice-management/ digital/ama-future-health-issue-brief-commercial-payer-coverage-digital.
  101. Richardson S, Lawrence K, Schoenthaler AM, Mann D. A framework for digital health equity. Npj Digit Med. 2022;5(1):119.
  102. Frimpong JA, Jackson BE, Stewart LM, Singh KP, Rivers PA, Bae S. Health information technology capacity at federally qualified health centers: a mechanism for improving quality of care. BMC Health Serv Res. 2013;13(1):35.
  103. Guo J, Li B. The application of Medical Artificial Intelligence Technology in Rural areas of developing countries. Health Equity. 2018;2(1):174-81.
  104. Davlyatov G, Borkowski N, Feldman S, Qu H, Burke D, Bronstein J, et al. Health Information Technology Adoption and Clinical Performance in Federally Qualified Health Centers. J Healthc Qual. 2020;42(5):287-93.
  105. Iyanna S, Kaur P, Ractham P, Talwar S, Najmul Islam AKM. Digital transformation of healthcare sector. What is impeding adoption and continued usage of technology-driven innovations by end-users? J Bus Res. 2022;153:150-61.
  106. Richardson JP, Smith C, Curtis S, Watson S, Zhu X, Barry B, et al. Patient apprehensions about the use of artificial intelligence in healthcare. Npj Digit Med. 2021;4(1):140.

ملاحظة الناشر

تظل Springer Nature محايدة فيما يتعلق بالمطالبات القضائية في الخرائط المنشورة والانتماءات المؤسسية.

  1. *المراسلة:
    ماثيو ر. ألين
    Mattallen876@gmail.com; maa007@health.ucsd.edu
    قسم طب الأسرة، جامعة كاليفورنيا سان دييغو، لا
    جولا، كاليفورنيا 92093، الولايات المتحدة الأمريكية
    قسم المعلوماتية الحيوية، جامعة كاليفورنيا سان دييغو، لا
    جولا، كاليفورنيا 92093، الولايات المتحدة الأمريكية

Journal: BMC Primary Care, Volume: 25, Issue: 1
DOI: https://doi.org/10.1186/s12875-024-02282-y
PMID: https://pubmed.ncbi.nlm.nih.gov/38281026
Publication Date: 2024-01-27

Navigating the doctor-patient-AI relationship a mixed-methods study of physician attitudes toward artificial intelligence in primary care

Matthew R. Allen , Sophie Webb , Ammar Mandvi , Marshall Frieden , Ming Tai-Seale and Gene Kallenberg

Abstract

Background Artificial intelligence (AI) is a rapidly advancing field that is beginning to enter the practice of medicine. Primary care is a cornerstone of medicine and deals with challenges such as physician shortage and burnout which impact patient care. AI and its application via digital health is increasingly presented as a possible solution. However, there is a scarcity of research focusing on primary care physician (PCP) attitudes toward AI. This study examines PCP views on AI in primary care. We explore its potential impact on topics pertinent to primary care such as the doctorpatient relationship and clinical workflow. By doing so, we aim to inform primary care stakeholders to encourage successful, equitable uptake of future AI tools. Our study is the first to our knowledge to explore PCP attitudes using specific primary care AI use cases rather than discussing AI in medicine in general terms. Methods From June to August 2023, we conducted a survey among 47 primary care physicians affiliated with a large academic health system in Southern California. The survey quantified attitudes toward AI in general as well as concerning two specific AI use cases. Additionally, we conducted interviews with 15 survey respondents. Results Our findings suggest that PCPs have largely positive views of AI. However, attitudes often hinged on the context of adoption. While some concerns reported by PCPs regarding AI in primary care focused on technology (accuracy, safety, bias), many focused on people-and-process factors (workflow, equity, reimbursement, doctorpatient relationship). Conclusion Our study offers nuanced insights into PCP attitudes towards AI in primary care and highlights the need for primary care stakeholder alignment on key issues raised by PCPs. AI initiatives that fail to address both the technological and people-and-process concerns raised by PCPs may struggle to make an impact.

Keywords Artificial intelligence, Attitudes, Digital work, Primary care, Technology, Qualitative research, Digital health

Background

While the potential impact of AI in medicine has been long discussed, real-life, clinician-facing applications of AI have only recently become a reality [1-3]. AI-assisted chronic disease management, diagnostic support, and administrative work (such as documentation, billing, and patient messaging) have significant potential to improve medicine and to take some burden off physicians allowing them to focus on physician-level patient care [4]. Further, use of clinical AI is part of a broader shift in medicine toward “digital health” where many aspects of medical care are conducted remotely, mediated by a technological intermediary leading to potential improvements in efficiency and access [5, 6]. These developments stand to make a substantial impact in primary care, a field that is currently grappling with high rates of physician burnout, inadequate compensation, and a growing shortage of physicians [7-9]. However, there is concern that if AI is poorly integrated it could exacerbate the “disconnect between professional values and the realities of primary care practice” [10,11]. For example, despite the crucial role of the doctor-patient relationship in medicine, the impact of AI and digital health on this essential component of primary care remains underexplored [12-14].
Despite so much on the line, there is limited literature on PCP views toward AI [15, 16]. Much of the research that does exist has taken place in a purely theoretical context exploring AI in general terms with physicians that did not have experience using AI-powered systems. We propose that more end-user engagement with clinicians discussing tangible, specific use cases of clinical AI is needed [17]. By highlighting specific AI use cases, we hope to elicit new concerns and attitudes that would remain hidden when discussing AI in general. Failure to engage end users in the design of AI-powered digital health tools leads to inefficient or unsuccessful integration of these tools into clinical workflow leading to added clinician burnout and even patient harm [10, 18, 19].

Primary care, technology and health equity

Our study recognizes the potential of technology to exacerbate or ameliorate existing inequalities in healthcare [20-24]. AI systems are particularly at risk of worsening health equity due to factors like potentially biased data becoming engrained in AI systems or unequal distribution of newly developed AI tools [25]. Equity considerations are especially vital in the context of primary care. PCPs are often the first point of contact for patients and are central in providing healthcare to communities with limited access due to geographical, economic, or social factors [26-30]. PCPs also make up the largest potential group of AI end users among health professionals [9]. Despite the foundational nature of primary care, this field
has long endured a lack of attention, resources, and recognition compared to other medical specialties [31-33]. This has contributed to a comparative lack of AI progress and implementation in primary care in spite of huge need and potential [9, 22, 34-36]. Accordingly, equity is a key consideration for AI in primary care.

Objective

In our pursuit of user-centric design, we employed a mixed-methods approach to delve into PCP attitudes regarding the potential transformative influence of AI and the broader shift towards digitalization in primary care. Our initial aim is to inform primary care stakeholders of PCP apprehensions regarding potential adverse effects of AI in primary care. Our findings reveal pivotal factors that can either facilitate or hinder the integration of AI systems in primary care. Our long-term aim is to use these findings to develop the AI tools outlined in the manuscript with the goal of improving patient care. While this is not the first exploratory investigation of PCP attitudes about AI, to our knowledge it is the first study that extends beyond the theoretical realm, weaving in specific AI use cases and input from PCPs with real-world experience using primary care AI and digital health tools.

Methods

Participant engagement with AI and digital health

Our study participants are affiliated with an academic medical center (AMC) actively engaged in the development, pilot testing, or implementation of several AI applications within its healthcare system. Here, we spotlight specific use cases relevant to primary care, highlighting their pivotal role in the study. These use cases, characterized by their remote and asynchronous elements, also fall under the broader category of digital health [37].

Al-enhanced disease screening: obstructive sleep apnea

Acknowledging the growing prevalence of Obstructive Sleep Apnea (OSA) and its often-undetected status, a research team at our AMC identified OSA as a suitable target for AI-based disease screening [38]. This initiative builds upon prior research employing electronic health record (EHR) data to identify individuals at high risk of OSA [39, 40]. This use case is an archetype for multiple types of disease screening in primary care and raises important questions such as what to do with positive screening results from an AI tool run on a patient panel.

Al-facilitated disease management: hypertension

Our institution is exploring a digital health strategy for hypertension management, integrating home blood pressure measurements and AI-powered clinical decision support through a panel-level registry [41, 42]. This
approach has potential to help primary care physicians give precise hypertension care based on unique patient characteristics while also giving them tools and efficiencies to do so at a population level [42-44]. Additionally, our institution has introduced a population health service enabling PCPs to refer patients with hypertension to digital medication management facilitated remotely by nurses and pharmacists. This use case is an archetype for chronic disease management in primary care and raises questions such as: how can AI augment PCP abilities or coordinate care between different primary care team members?

AI-facilitated administrative tasks: patient messaging

Inbox overload, which was exacerbated during the Covid-19 pandemic, contributes to burnout and “pajama time” in primary care [45]. To mitigate this challenge, our AMC is currently piloting the utilization of Large Language Models like ChatGPT for drafting patient message responses within the EHR [45]. This use case is an archetype for AI assisting with administrative tasks and raises questions including potential impacts on the doctorpatient relationship.

Digital survey

As the first step in our mixed-method approach, we employed a digital survey (appendix) specifically developed for our project to quantify PCP attitudes. Given the novel nature of our research focus and the absence of pre-existing validated questionnaires, members of our research team with qualitative research expertise lead the creation of the survey instrument which ensured impartiality, methodological rigor, and ability to capture nuanced insight. Our survey instrument was primarily descriptive in nature and served as a valuable data source for understanding the frequency of responses and providing a framework for the subsequent interviews. Using Likert scales, the survey explored participants’ comfort levels and perceptions about AI in healthcare. We also gathered deidentified demographic data to contextualize perspectives. The survey captured responses from a diverse group of primary care physicians ( ), providing perspectives from different primary care specialties and practice settings. The sample included physicians from AMC Faculty Internal Medicine ( ), AMC Faculty Family Medicine ( ), AMC Clinical Internal Medicine ( ), AMC Clinical Family Medicine ( ).
The respondents’ ages encompassed a wide range: 25-34 years old ( ), 35-44 years old ( ), 45-54 years old ( ), years old ( ). Gender diversity was evident with male ( ) and female ( ) respondents. Years of experience in practice varied (mean years; ), including years ( ), years ( ), years ( ),
years years years , 31-35 years ( ). This spectrum ensures broad insights into PCP perspectives across different demographics and career stages.

Semi-structured interview

In the survey, respondents were asked if they would be willing to engage in a follow-up interview. Employing a semi-structured interview format with an interview guide (appendix) iteratively developed for our project in conjunction with qualitative research experts, we provided participants with an open and adaptable platform to share their perspectives which we then scrutinized using thematic analysis. Interviews were conducted in a confidential environment via remote teleconferencing software (Zoom). Automated transcription software (Otter.AI) was used for transcription generation and collected interview data underwent rigorous thematic analysis using Quirkos qualitative analysis software. This method involved a systematic process of coding and categorizing responses to identify recurring patterns, insights, and emerging themes. Through iterative refinement, we extracted meaningful themes that captured the essence of PCP’s views regarding AI in primary care. Following our initial interviews ( ) that highlighted PCP’s concerns about increased workload and expectations, we added a question to delve deeper into how AI could affect the doctor-patient relationship. Our final number of interviews was 15 .

Results

General perceptions of a.i. in medicine

The majority of survey respondents (76.6%) held an optimistic perspective regarding the potential of AI in medicine. Comfort levels in integrating AI-based technologies into clinical practice varied across different domains (Table 1).
While some physicians reported feeling comfortable communicating the role of AI-based tools to patients (very comfortable: , somewhat comfortable: ) a sizeable percentage did not (somewhat uncomfortable: , very uncomfortable: ). Importantly, of surveyed physicians described their approach to learning about AI in medicine as “passively learning via popular news sources or casual conversation” with only “actively seeking education through established organizations, coursework, lectures, professional journals, or books.”

Concerns about AI in primary care

Despite the general positivity quantified in the survey, interview participants expressed numerous concerns about AI-especially when discussing specific AI use cases. We subjected our interview data on concerns
Table 1 Table percentages represent the proportion of PCPs reporting varying levels of comfort with AI involvement in different domains as reported via the digital survey
Domain Very Comfortable Somewhat Comfortable Neutral Somewhat Uncomfortable Very Uncomfortable
Disease Screening 29.8% 46.8% 8.5% 8.5% 6.4%
Chronic Disease Management 25.5% 46.8% 12.8% 8.5% 6.4%
Disease Diagnosis 8.5% 42.6% 10.6% 23.4% 14.9%
Administrative Tasks 40.4% 25.5% 12.8% 14.9% 6.4%
Fig. 1 PCP concerns about Al. Description: Emergent themes from interviews with primary care physicians regarding AI divided into concerns about AI technology itself and concerns about the context and manner of AI implementation
regarding AI to a thematic coding analysis and identified the following themes which have been categorized as concerns regarding technology or people-and-processes (Fig. 1) [46].
Technological concerns included factors such as algorithmic bias (1 participant) or accuracy and safety (7 participants).
The thing I’m apprehensive about is, how are we teaching AI these things because some of those biases could leak in. [Participant C]
My concerns around AI in medicine have most to do with the space of accuracy. And a tool that I feel is reliable. [Participant G]
Concern about external validity and the ability of AI algorithms to appreciate the nuances of specific patients (8 participants) was an important consideration for PCPs.
I’ve known a lot of my patients now for 30 years and know a lot about them. That can’t really be put into a data set that AI can draw upon. [Participant F]
Interviewed PCPs reported differing opinions on whether they prioritized explainability [47] in AI models (3 participants),
At this point, I want to be able to get a logical explanation. [Participant L]
or if this was not imperative for them (10 participants).
I’m perfectly okay if my own experience with time goes better and better and I feel like, you know, it works. Don’t ask me how it works, but it works. [Participant A]
I almost think that the tool adds more to the deci-sion-making process if it’s operating outside of that human accessible reasoning process. [Participant B]
However, many reported concerns centered around systemic issues rather than technological ones. One common concern (5 participants) was that of the medicolegal implications of acting, or failing to act, on AI guidance.
If the system is saying, ‘Hey, this person has severe sleep apnea, and what if they get in a car accident tomorrow and we had that data today? [Participant O]
When discussing the potential of algorithmic detection of OSA, PCPs (5 participants) pointed out that without augmenting the system’s ability to definitively diagnose and treat more patients with OSA, the AI tool would not be helpful.
When we are loading our system from this side, we need to have the resources on the other side. [Participant E]
I think we need to be better equipped before we start telling people this because, you know, it’s like, Hey, you might have sleep apnea. Wait six months for your sleep study… [Participant C]
Additionally, 10 participants reported concern that the integration of AI into primary care could potentially lead to increased workload and physician burnout.
My concern is that like everything else that we have tried to do to make things better in medicine is that it actually makes things harder on the physician and creates more work for us instead of less work. [Participant D]
Interviewees reported multiple ways in which this could happen including AI tools delegating work to physicians that could potentially be handled by other team members,
Is it really the physicians that should deal with this in the first place? [Participant K]
a need to constantly verify or redo work done by AI,
It’s like having a student with me all the time, where I’ve got to just double check everything. [Participant D]
or an excessive focus on productivity.
We are going to add two extra patients per session because now we have help there. So unfortunately, sometimes more help is used in a negative way. [Participant E]
I’ve always seen that the system wants productivity, and the way productivity is defined is based on the number of patients seen. [Participant L]
Not all PCPs shared this concern of increased workload due to AI, with one participant expressing that increased efficiency due to AI would be welcome even if it meant seeing more patients.
If I could see 30 patients in a day, and actually close out my charts by 6pm, smiling, and get home for dinner, I’d be happy. [Participant G]
Physicians also reported concerns on how AI might impact the doctor-patient relationship. Some expressed positive hopes for AI to improve the doctor-patient relationship (10 participants) by doing things such as alleviating clinician burden or improving patient engagement.
Maybe you are actually then more compassionate in an encounter, because you haven’t had to do all of that mental lifting. [Participant M]
But many-some that had also expressed positive senti-ment-worried that AI could harm the doctor-patient relationship ( 12 participants) by factors such as warping patient expectations or prioritizing patient needs over physician well-being.
Patients may end up feeling that, you know, if the AI can tell me that then why did I bother to come to you? [Participant L]
The more we sort of train patients to expect things quickly and efficiently, the more expectations are on the doctor to then produce in the same way. [Participant D]
Participating PCPs also lamented a lack of focus on phy-sician-wellbeing when implementing new technologies (9 participants).
I feel like right when I get efficient, something new gets introduced. [Participant M]
The system is all about the patient’s satisfaction. Is there any of that focus on physician satisfaction? [Participant E]
One key point was the concern that the current way healthcare is paid for does not encourage innovative ways of care delivery such as AI-powered digital health tools.
In essence, we’re providing a bunch of free care, which, you know, is not sustainable. [Participant H]
Thus, it appears that PCPs see a disconnect between care innovation and the way they are forced to practice due to how care is reimbursed.
They’re told to do both things. So they’re really there to crank it out, crank out these RVUs while also doing value based medicine and population based medicine. [Participant J]
Dedicated time for digital health as well as alternative reimbursement models were frequently voiced (11 participants) as a key determiner of the uptake and success of AI tools.
I’d like dedicated time daily or at least weekly to review. Otherwise, I might only see it if I see the patient. [Participant F]
I think that this system really needs to rethink how it employs physicians and providers. [Participant L]
Data from our survey corroborates this finding that PCPs are unsure of where digital health tools fit into their workflow. When asked about preferences about when to receive communication from an AI tool regarding a patient screening positive for OSA, responses varied widely with 3 respondents using the free response
to indicate that they would not like to be notified. When asked if they were aware of a pre-existing EHR registry of patients with hypertension, nearly half (45.83%) responded no. For those that were aware of the registry, more than half reported that their usage of the registry was “infrequently” ( ) or “never” ( ) with the most common reported reason being a lack of time.

Discussion

Al as a double-edged sword

Our findings reveal the dual nature of AI in healthcare, uncovering its potential to alleviate or exacerbate challenges in primary care. Some of our identified concerns about AI adoption in healthcare, including lack of external validity, potential for bias, and safety issues, have been well-documented in the literature [48-50]. Our study expands upon these concerns highlighting that, for clinicians, the mechanics of AI itself may take a back seat to its potential impact on their professional lives, personal well-being, and their relationships with patients [51]. We argue that concerns such as apprehension about increased workload stem from a broader sentiment among PCPs that advancements in healthcare often prioritize productivity over physician well-being or put financial considerations over human relationships [5254]. Accordingly, some PCP concerns about AI may be a reflection of a disillusionment with the evolving landscape of medicine in general. In this context, the introduction of AI is perceived as yet another instance where physician interests may be subjugated to organizational efficiency. These concerns are not unfounded with previous literature proposing using technology to add capacity as one of the solutions to keep up with the increasing physician shortage [55, 56]. Further, PCP panel sizes are already felt to be excessive and a fear regarding AI being used to justify the addition of patients may be rational [57-59]. Digital health and AI in primary care must be applied thoughtfully to avoid further ostracizing PCPs from their professional values.
I think we all worry that more work is what things are aimed at. [Participant F]
These are not selfish concerns as the well-being of physicians is intrinsically linked to patient outcomes and is aligned with the Quadruple Aim of healthcare [60, 61]. This concern for physician well-being is especially pertinent in the context of primary care-a cornerstone of healthcare critical for providing access to underserved populations that is chronically undervalued by the healthcare system [62, 63]. PCPs should be able to share in benefits such as time or cost-savings produced by the implementation of AI systems. Our findings suggest that
Fig. 2 The evolution of the doctor-patient relationship. Description: The advent of the internet had significant impacts on the doctor-patient relationship. Primary care physicians have a mix of concern and optimism about how AI may do the same
Fig. 3 Al as a member of the healthcare team. Description: Al becoming a member of the primary care team
if all benefit goes to the organization, physician appetite for uptake will remain low.
Our work also highlights the evolving role of the primary care physician [64]. Once commonly viewed as the source of medical truth, physicians now coexist with “Dr. Google,” a digital repository of health information that empowers patients to engage proactively in their own care [65]. This change has had mixed effects on the doc-tor-patient relationship but can be positive if both parties engage in proper communication and shared-decision making [66-68]. These positive effects also hinge on factors such as strong patient health information literacy and adequate doctor-patient communication time [69].
Against this backdrop, The impending integration of AI into healthcare will similarly revolutionize the doctorpatient dynamic (Fig. 2).
Much of the same research that was done to explore the impact of internet health information on the doctorpatient relationship needs to be repeated and expanded upon in the context of AI. Unlike static online information, interactive AI systems are poised to assume a more active role in shaping the interactions between patients and physicians. Ensuring a positive impact of AI on the doctor-patient relationship is essential to maintaining medicine’s social contract with society [70].
AI’s role in providing patients with information, be it accurate or misleading, could confound the physicianpatient dynamic. Patients arriving with AI-informed
information could complicate collaborative decisionmaking [71]. Possible consequences include rigid adherence to AI-driven advice without considering individual medical history or difficulties for physicians attempting to reconcile their expertise with AI suggestions. This not only risks eroding the PCP’s role but also could reshape the doctor-patient relationship into a consumer-provider model.
I’m worried about AI, introducing a dynamic where misinformation is enhanced… if a patient comes in and they’re like, hey, like, you know, WebMD.GPT told me that I need an MRI then there’s another powerful thing that I’m arguing against. [Participant B]
A key part of ensuring a positive impact of AI on the doc-tor-patient relationship is promoting realistic and aligned expectations regarding AI via education for doctors and patients before implementation of AI tools.
Instead of just letting the cat out of the bag and seeing what happens, you want to make sure that everyone that is going to be interacting with it has accurate expectations and has been educated on what role this is supposed to play. [Participant J]
Previous literature has highlighted concerns that unequal knowledge or differing backgrounds in the doctor-patient relationship could exacerbate health inequity [72, 73]. Thoughtful and equitable implementation of AI could encourage increased patient engagement and understanding leading to more effective doctor-patient communication and increased equity.

The future of primary care workflow

Our findings call for a reconsideration of fundamental questions regarding primary care workflow. If PCPs are going to be active participants in new forms of healthcare delivery, including AI-powered digital health, when are they supposed to do that work? This shift toward digital health is already occurring albeit in an unscheduled and uncompensated way [45, 74, 75].
“Most of our care is delivered in MyChart. Like let’s just be honest, that’s how it’s getting delivered. [Participant G]
Inbox burden is a well-known problem, but it is only the beginning of asynchronous digital versus synchronous in-person workload conflict in primary care [76]. The failure of health systems to identify proper ways of allocating time, resources, and standards to these new ways of interacting with patients has already had substantial
consequences [77]. AI-powered digital tools for chronic disease management and disease screening augmented by remote patient monitoring systems will likely become increasingly common [78, 79]. Accordingly, consideration needs to be given to how best to allocate physician time to support digital health. Succeeding in digital health is more than solving inbox overload or alarm fatigue but rather realizing a fundamental shift in how primary care interacts with and takes care of patients [6]. We propose that experimenting with hybrid in-person and virtual work schedules could empower physicians to actualize the potential of digital health [80].
Another consideration is: how can AI be integrated into patient-centered, team-based primary care [81]? A population health approach to primary care consists of a physician acting as a “healthcare quarterback” who is responsible for the health of an entire patient panel, regardless of how, where, or by whom each component of care is delivered [81, 82]. Our interview participants frequently indicated that physicians need not always be the primary point of contact for an AI recommendation. Identifying when other team members can review and act upon AI-produced guidance while maintaining the PCP in the loop could mitigate concerns around the possibility of AI and digital health creating more work for physicians [82, 83]. This must be done carefully in a manner that enhances-rather than erodes-the core doctorpatient relationship [11, 12].
In a more general sense, care coordination is a core challenge of primary care. More research should be done on using AI as a facilitator of task follow-up, delegation, and other components of care coordination (Fig. 3) [84]. Designing standardized primary care workflow is challenging [85]. However, enabling PCPs with time, team members, and incentives to use AI-powered digital tools that handle some aspects of care remotely and asynchronously could facilitate more meaningful, effective, and focused in-person clinic visits [86-88]. More time spent on digital avenues of care has also been shown to improve quality of care metrics [89]. Currently, PCPs already engage in this sort of work but they do it at the expense of time with their patients or their families [75, 90]. Organizations should be wary of relying too much on physician altruism to find time to use digital tools, forcing physicians to choose between their personal wellbeing and that of their patients [91-93].

Considerations for primary care stakeholders

As a first step, stakeholders need to ensure that primary care AI systems are rigorously evaluated from an accuracy, safety, and bias standpoint. However, subsequent attention must be given to workflow integration and impact on physician well-being (Table 2) [86, 94]. Previous literature has shown that such non-technical factors
Table 2 Recommendations for primary care stakeholders
Stakeholder Recommendation
Payors Implement innovative reimbursement models for PCPs and other primary care team members engaging in digital health
Healthcare Systems Schedule time and establish standards for PCPs to engage in digital health
Healthcare Systems Provide PCPs with additional team members such as pharmacists or patient coordinators who can engage digitally with patients
Healthcare Systems Develop and disseminate educational materials on the proper role of new AI tools to patients and physicians before tool implementation
Researchers Run RCTs, pragmatic trials, or practice-based research between traditional and digitally enhanced PCP workflows
Researchers Evaluate new AI-powered digital tools in the context of physician workflow instead of an isolated environment
PCPs Advocate individually and collectively for Al tools that improve physician care quality, wellbeing, and the doctor-patient relationship
are essential to promote uptake of new technologies [95]. Caution must be taken to ensure that AI does not result in “less doctor work and more office work” leading to PCPs who feel exploited by the healthcare system [10]. Rather, AI should be used as an opportunity to address primary care challenges such as helping deliver better care while preventing physician burnout [96].
Additionally, the way in which healthcare is compensated has a substantial impact on the behavior and time-allocation of PCPs [97]. Some attempts, such as charging a user fee for patient messages and billing payers for e-visits (i.e., responding to patient messages), have been made to reimburse digital health services in a fee-for-service model [98]. However, digital health is likely more well suited for value-based primary care that would incentivize and provide flexibility for physicians to engage in asynchronous, population-level digital health tools [99]. Organizations wishing to reap the benefits of AI in primary care must tackle this challenge head on and be willing to reimagine how care is delivered and paid for rather than further ingraining legacy systems and approaches. This is especially pertinent in considering the nature of healthcare reimbursement in the United States. Efforts at payment modernization are underway, [100] but failure to quickly advance and innovate our payment models could lead to our systems lagging behind non-fee-for-service nations in terms of AI and digital health innovation.

Al and equity

The people who will get it are the people who can pay for the compute. And so that’s my biggest fear is
that we will leave out the poorest people from getting the best care. – Participant G.
This sentiment expressed by multiple study participants underscores the gravity of ensuring equitable access to AI-powered healthcare solutions for patients across the spectrum of socioeconomic backgrounds. The transformation brought by AI should not inadvertently reinforce existing disparities, but rather serve as a tool to alleviate them [101]. For example, AI must not alleviate physician burnout and improve patient outcomes only at large AMCs that have adequate resources to engage in AI development and implementation but should be broadly accessible and applicable across diverse healthcare settings and institutions, ensuring equitable access to its benefits for all. Safety-net health systems, federally qualified health centers, rural areas, and other practice settings that could be left out need to be included in the primary care AI revolution [102-104].
The reality is that our patients have much more healthcare that they need delivered than we can ever deliver, so they’re going to need AI tools. – Participant .

Strengths, limitations, and directions for future research

In contrast to much of the existing literature, all our study participants have had actual experience with digital health and some of our study participants have actual experience with medical AI. Because these technological shifts are just beginning, clinicians with perspectives informed by actual experience are rare, making our findings more valuable. However, a limitation is that terms such as “AI” or “digital health” have evolving definitions and may not mean the same thing to different individuals. While we tried to ameliorate this effect by grounding our discussions in tangible use-cases and examples, participants differing preconceived ideas of AI may have affected participant responses. Further, focusing on certain use-cases over others may have influenced reported PCP views on AI in general. Future efforts should more comprehensively evaluate perceptions of AI in primary care to ensure that reported PCP attitudes are not overly influenced by any particular use-case. Additionally, while our survey and interview guides were developed in a rigorous manner in collaboration with qualitative methods experts, future research should attempt to develop validated qualitative tools for assessing PCP attitudes toward AI in primary care.
Our research focused only on internal medicine and family medicine physician attitudes toward AI and digital health. Future research needs to include other primary care team members including primary care pediatricians, nurses,
and advanced practice providers such as nurse practitioners and physician assistants. In addition, greater research is needed on how patients-especially those that may be marginalized-are experiencing a shift toward AI and digital health in primary care [105, 106]. Finally, future research should also expand beyond our selected AI use cases to incorporate other AI applications pertinent to primary care.
Our relatively small sample size limits the generalizability of our findings to larger populations. Additionally, all our respondents originated from the same organization. While we attempted to assess attitudes of both academic and non-academic physicians, this organizational homogeneity might further limit generalizability. Moreover, the potential for response bias in self-reported responses should be acknowledged.
In light of these limitations, we emphasize the need for future research endeavors to employ quantitative methods to explore questions regarding AI in primary care and to incorporate larger and more diverse samples from various healthcare settings. Incorporating multiple organizationsespecially those that are not well-funded AMCs in urban environments-can provide a broader perspective on the adoption of AI and digital health in primary care.

Conclusion

This study was the first to investigate PCP attitudes toward AI in primary care focusing on specific AI use cases. Reported attitudes varied, but PCP responses showed general optimism around AI in primary care tempered by certain concerns. While some concerns focused on technological factors like algorithmic accuracy, safety, and bias, others focused on people-and-process factors such as effects on physician workflow, equity, reimbursement, and the doctor-patient relationship. These findings suggest that AI initiatives that fail to address both the technological and people-and-process concerns raised by PCPs may struggle to make an impact. Primary care stakeholders should use these findings to inform development and implementation of AI in primary care.

Abbreviations

Al Artificial Intelligence
PCP Primary Care Physician
AMC Academic Medical Center
OSA Obstructive Sleep Apnea
EHR Electronic Health Record

Supplementary Information

The online version contains supplementary material available at https://doi. org/10.1186/s12875-024-02282-y.
Supplementary Material 1: Question stems from the digital survey
Supplementary Material 2: Questions stems from the semi-structured interview

Acknowledgements

Not applicable.

Author contributions

MRA, AM, and GK devised the initial study design. MTS provided substantive contributions to study design and survey instrument refinement. MRA and SW carried out the study and drafted the initial version of the manuscript. MRA coded the interview transcripts and SW verified the interview codes. MF, AM, MTS, and GK aided in manuscript revision, data analysis, and contributed intellectual value to the manuscript. All authors approve of the final version of the manuscript.

Funding

This work was supported by the American Board of Family Medicine Artificial Intelligence/Machine Learning grant and the UCSD School of Medicine Research Fellowship. The funders played no role in any elements of the study.

Data availability

The digital survey questions and interview guide question stems will be included as an appendix to the manuscript. Complete interview and survey data will be made available upon reasonable request.

Declarations

Study methods and protocols were approved by the Aligning and Coordinating QUality Improvement, Research, and Evaluation Committee of the University of California San Diego (ACQUIRE Project #751) acting under the authority of the Institutional Review Board of the University of California San Diego. Informed consent was obtained from all subjects. All methods were carried out in accordance with relevant guidelines and regulations or declaration of Helsinki.
Not applicable.

Competing interests

The authors declare no competing interests.
Received: 5 October 2023 / Accepted: 19 January 2024
Published online: 27 January 2024

References

  1. Strohm L, Hehakaya C, Ranschaert ER, Boon WPC, Moors EHM. Implementation of artificial intelligence (AI) applications in radiology: hindering and facilitating factors. Eur Radiol. 2020;30(10):5525-32.
  2. Park CJ, Yi PH, Siegel EL. Medical Student perspectives on the impact of Artificial Intelligence on the practice of Medicine. Curr Probl Diagn Radiol. 2021;50(5):614-9.
  3. Briganti G, Le Moine O. Artificial Intelligence in Medicine: Today and Tomorrow. Front Med [Internet]. 2020 [cited 2023 Aug 6];7. Available from: https:// www.frontiersin.org/articles/https://doi.org/10.3389/fmed.2020.00027.
  4. Seneviratne MG, Shah NH, Chu L. Bridging the implementation gap of machine learning in healthcare. BMJ Innov. 2020;6(2):45-7.
  5. Zimlichman E, Nicklin W, Aggarwal R, Bates D, Health Care. 2030: The Coming Transformation. NEJM Catal Innov Care Deliv [Internet]. (March 1, 2021). Available from: https://catalyst.nejm.org/doi/full/https://doi.org/10.1056/ CAT.20.0569.
  6. Pagliari C. Digital health and primary care: past, pandemic and prospects. J Glob Health. 2021;11:01005.
  7. Rabatin J, Williams E, Baier Manwell L, Schwartz MD, Brown RL, Linzer M. Predictors and outcomes of Burnout in Primary Care Physicians. J Prim Care Community Health. 2016;7(1):41-3.
  8. Petterson SM, Liaw WR, Tran C, Bazemore AW. Estimating the Residency Expansion required to avoid projected primary care physician shortages by 2035. Ann Fam Med. 2015;13(2):107-14.
  9. Lin SA, Clinician’s, Guide to Artificial Intelligence (AI). Why and how primary care should lead the Health Care AI revolution. J Am Board Fam Med. 2022;35(1):175-84.
  10. Agarwal SD, Pabo E, Rozenblum R, Sherritt KM. Professional Dissonance and Burnout in Primary Care: a qualitative study. JAMA Intern Med. 2020;180(3):395.
  11. Amano A, Brown-Johnson CG, Winget M, Sinha A, Shah S, Sinsky CA, et al. Perspectives on the Intersection of Electronic Health Records and Health Care Team Communication, function, and well-being. JAMA Netw Open. 2023;6(5):e2313178.
  12. Kearley KE, Freeman GK, Heath A. An exploration of the value of the personal doctor-patient relationship in general practice. Br J Gen Pract J R Coll Gen Pract. 2001;51(470):712-8.
  13. Aminololama-Shakeri S, López JE. The Doctor-Patient Relationship with Artificial Intelligence. Am J Roentgenol. 2019;212(2):308-10.
  14. Nagy M, Sisk B. How will Artificial Intelligence Affect Patient-Clinician relationships? AMA J Ethics. 2020;22(5):E395-400.
  15. Blease C, Kaptchuk TJ, Bernstein MH, Mandl KD, Halamka JD, DesRoches CM. Artificial Intelligence and the future of primary care: exploratory qualitative study of UK General practitioners’ views. J Med Internet Res. 2019;21(3):e12802.
  16. Buck C, Doctor E, Hennrich J, Jöhnk J, Eymann T. General practitioners’ attitudes toward Artificial intelligence-enabled systems: interview study. J Med Internet Res. 2022;24(1):e28916.
  17. Kueper JK, Terry AL, Zwarenstein M, Lizotte DJ. Artificial Intelligence and Primary Care Research: a scoping review. Ann Fam Med. 2020;18(3):250-8.
  18. Marwaha JS, Landman AB, Brat GA, Dunn T, Gordon WJ. Deploying digital health tools within large, complex health systems: key considerations for adoption and implementation. Npj Digit Med. 2022;5(1):1-7.
  19. Downing NL, Bates DW, Longhurst CA. Physician burnout in the Electronic Health Record Era: are we ignoring the Real cause? Ann Intern Med. 2018;169(1):50.
  20. López L, Green AR, Tan-McGrory A, King RS, Betancourt JR. Bridging the Digital divide in Health Care: the role of Health Information Technology in addressing racial and ethnic disparities. Jt Comm J Qual Patient Saf. 2011;37(10):437-45.
  21. Latulippe K, Hamel C, Giroux D. Social Health Inequalities and eHealth: A literature review with qualitative synthesis of theoretical and empirical studies. J Med Internet Res. 2017;19(4):e136.
  22. Weiss D, Rydland HT, Øversveen E, Jensen MR, Solhaug S, Krokstad S. G Virgili editor 2018 Innovative technologies and social inequalities in health: a scoping review of the literature. PLOS ONE 134 e0195447.
  23. Ramsetty A, Adams C. Impact of the digital divide in the age of COVID-19.J Am Med Inform Assoc. 2020;27(7):1147-8.
  24. Yao R, Zhang W, Evans R, Cao G, Rui T, Shen L. Inequities in Health Care services caused by the Adoption of Digital Health Technologies: scoping review. J Med Internet Res. 2022;24(3):e34144.
  25. Joyce K, Smith-Doerr L, Alegria S, Bell S, Cruz T, Hoffman SG, et al. Toward a sociology of Artificial Intelligence: a call for Research on inequalities and Structural Change. Socius Sociol Res Dyn World. 2021;7:237802312199958.
  26. Salhi RA, Dupati A, Burkhardt JC. Interest in serving the Underserved: role of race, gender, and Medical Specialty Plans. Health Equity. 2022;6(1):933-41.
  27. Jetty A, Hyppolite J, Eden AR, Taylor MK, Jabbarpour Y. Underrepresented Minority Family Physicians more likely to care for vulnerable populations. J Am Board Fam Med. 2022;35(2):223-4.
  28. Starfield B, Shi L, Macinko J. Contribution of primary care to health systems and health. Milbank Q. 2005;83(3):457-502.
  29. Blumenthal D, Mort E, Edwards J. The efficacy of primary care for vulnerable population groups. Health Serv Res. 1995;30(1 Pt 2):253-73.
  30. Marrast LM, Zallman L, Woolhandler S, Bor DH, McCormick D. Minority Physicians’ role in the Care of Underserved patients: diversifying the physician workforce may be key in addressing Health disparities. JAMA Intern Med. 2014;174(2):289.
  31. Lambert SI, Madi M, Sopka S, Lenes A, Stange H, Buszello CP, et al. An integrative review on the acceptance of artificial intelligence among healthcare professionals in hospitals. Npj Digit Med. 2023;6(1):1-14.
  32. Stange KC, Ferrer RL. The Paradox of Primary Care. Ann Fam Med. 2009;7(4):293-9.
  33. Shi L. The impact of primary care: a focused review. Scientifica. 2012;2012:1-22.
  34. Liaw W, Kakadiaris IA. Primary care Artificial Intelligence: a branch hiding in Plain Sight. Ann Fam Med. 2020;18(3):194-5.
  35. Liaw W, Kakadiaris IA. Artificial Intelligence and Family Medicine: better together. Fam Med. 2020;52(1):8-10.
  36. Lin SY, Mahoney MR, Sinsky CA. Ten ways Artificial Intelligence Will Transform Primary Care. J Gen Intern Med. 2019;34(8):1626-30.
  37. Fatehi F, Samadbeik M, Kazemi A. What is Digital Health? Review of definitions. Stud Health Technol Inform. 2020;275:67-71.
  38. Motamedi KK, McClary AC, Amedee RG. Obstructive sleep apnea: a growing problem. Ochsner J. 2009;9(3):149-53.
  39. Ramesh J, Keeran N, Sagahyroon A, Aloul F. Towards validating the effectiveness of obstructive sleep apnea classification from Electronic Health Records Using Machine Learning. Healthc Basel Switz. 2021;9(11):1450.
  40. Maniaci A, Riela PM, lannella G, Lechien JR, La Mantia I, De Vincentiis M, et al. Machine learning identification of obstructive sleep apnea severity through the patient clinical features: a retrospective study. Life Basel Switz. 2023;13(3):702.
  41. Molina-Ortiz El, Vega AC, Calman NS. Patient registries in primary care: essential element for Quality Improvement: P ATIENT R EGISTRIES IN P RIMARY C ARE. Mt Sinai. J Med J Transl Pers Med. 2012;79(4):475-80.
  42. Hu Y, Huerta J, Cordella N, Mishuris RG, Paschalidis IC. Personalized hypertension treatment recommendations by a data-driven model. BMC Med Inform Decis Mak. 2023;23(1):44.
  43. Fisher NDL, Fera LE, Dunning JR, Desai S, Matta L, Liquori V, et al. Development of an entirely remote, non-physician led hypertension management program. Clin Cardiol. 2019;42(2):285-91.
  44. Visco V, Izzo C, Mancusi C, Rispoli A, Tedeschi M, Virtuoso N, et al. Artificial Intelligence in Hypertension Management: an Ace up your sleeve. J Cardiovasc Dev Dis. 2023;10(2):74.
  45. Holmgren AJ, Downing NL, Tang M, Sharp C, Longhurst C, Huckman RS. Assessing the impact of the COVID-19 pandemic on clinician ambulatory electronic health record use. J Am Med Inform Assoc. 2022;29(3):453-60.
  46. University NCS, Payton F, Pare G, Montréal HEC, Le Rouge C, Saint L et al. University,. Health Care IT: Process, People, Patients and Interdisciplinary Considerations. J Assoc Inf Syst. 2011;12(2):I-XIII.
  47. the Precise4Q consortium, Amann J, Blasimme A, Vayena E, Frey D, Madai VI. Explainability for artificial intelligence in healthcare: a multidisciplinary perspective. BMC Med Inform Decis Mak. 2020;20(1):310.
  48. Challen R, Denny J, Pitt M, Gompels L, Edwards T, Tsaneva-Atanasova K. Artificial intelligence, bias and clinical safety. BMJ Qual Saf. 2019;28(3):231-7.
  49. Wong A, Otles E, Donnelly JP, Krumm A, McCullough J, DeTroyer-Cooley O, et al. External validation of a widely implemented proprietary Sepsis prediction model in hospitalized patients. JAMA Intern Med. 2021;181(8):1065.
  50. Yu KH, Kohane IS. Framing the challenges of artificial intelligence in medicine. BMJ Qual Saf. 2019;28(3):238-41.
  51. Terry AL, Kueper JK, Beleno R, Brown JB, Cejic S, Dang J, et al. Is primary health care ready for artificial intelligence? What do primary health care stakeholders say? BMC Med Inform Decis Mak. 2022;22(1):237.
  52. Phillips RL, Bazemore AW, Newton WP. PURSUING PRACTICAL PROFESSIONALISM: FORM FOLLOWS FUNCTION. Ann Fam Med. 2019;17(5):472-5.
  53. Relman AS. The New Medical-Industrial Complex. N Engl J Med. 1980;303(17):963-70.
  54. Tai-Seale M, Baxter S, Millen M, Cheung M, Zisook S, Çelebi J et al. Association of physician burnout with perceived EHR work stress and potentially actionable factors. J Am Med Inform Assoc. 2023;ocad136.
  55. Bodenheimer TS, Smith MD. Primary care: proposed solutions to the physician shortage without training more Physicians. Health Aff (Millwood). 2013;32(11):1881-6.
  56. Green LV, Savin S, Lu Y. Primary care physician shortages could be eliminated through use of teams, nonphysicians, and Electronic Communication. Health Aff (Millwood). 2013;32(1):11-9.
  57. Raffoul M, Moore M, Kamerow D, Bazemore A. A primary care panel size of 2500 is neither Accurate nor reasonable. J Am Board Fam Med JABFM. 2016;29(4):496-9.
  58. Harrington C. Considerations for patient panel size. Del J Public Health. 2022;8(5):154-7.
  59. Salisbury C, Murphy M, Duncan P. The impact of Digital-First Consultations on workload in General Practice: modeling study. J Med Internet Res. 2020;22(6):e18203.
  60. Halbesleben JRB, Rathert C. Linking physician burnout and patient outcomes: exploring the dyadic relationship between physicians and patients. Health Care Manage Rev. 2008;33(1):29-39.
  61. Sikka R, Morath JM, Leape L. The Quadruple Aim: care, health, cost and meaning in work. BMJ Qual Saf. 2015;24(10):608-10.
  62. Streeter RA, Snyder JE, Kepley H, Stahl AL, Li T, Washko MM. The geographic alignment of primary care Health Professional Shortage Areas
    with markers for social determinants of health. Shah TI, editor. PLOS ONE. 2020;15(4):e0231443.
  63. Liu J, Jason. Health Professional Shortage and Health Status and Health Care Access. J Health Care Poor Underserved. 2007;18(3):590-8.
  64. Ahuja AS. The impact of artificial intelligence in medicine on the future role of the physician. PeerJ. 2019;7:e7702.
  65. Van Riel N, Auwerx K, Debbaut P, Van Hees S, Schoenmakers B. The effect of Dr Google on doctor-patient encounters in primary care: a quantitative, observational, cross-sectional study. BJGP Open. 2017;1(2):bjgpopen17X100833.
  66. Murray E, Lo B, Pollack L, Donelan K, Catania J, White M, et al. The impact of Health Information on the internet on the physician-patient relationship: patient perceptions. Arch Intern Med. 2003;163(14):1727.
  67. Tan SSL, Goonawardene N. Internet Health Information seeking and the patient-physician relationship: a systematic review. J Med Internet Res. 2017;19(1):e9.
  68. Liaw W, Kakadiaris IA, Yang Z. Is Artificial Intelligence the Key to Reclaiming relationships in Primary Care? Am Fam Physician. 2021;104(6):558-9.
  69. Luo A, Qin L, Yuan Y, Yang Z, Liu F, Huang P, et al. The Effect of Online Health Information seeking on Physician-Patient relationships: systematic review. J Med Internet Res. 2022;24(2):e23354.
  70. Cruess SR. Professionalism and Medicine’s Social Contract with Society. Clin Orthop. 2006;449:170-6.
  71. Resnicow K, Catley D, Goggin K, Hawley S, Williams GC. Shared decision making in Health Care: theoretical perspectives for why it works and for whom. Med Decis Mak Int J Soc Med Decis Mak. 2022;42(6):755-64.
  72. Verlinde E, De Laender N, De Maesschalck S, Deveugele M, Willems S. The social gradient in doctor-patient communication. Int J Equity Health. 2012;11(1):12.
  73. Willems S, De Maesschalck S, Deveugele M, Derese A, De Maeseneer J. Socioeconomic status of the patient and doctor-patient communication: does it make a difference? Patient Educ Couns. 2005;56(2):139-46.
  74. Arndt BG, Beasley JW, Watkinson MD, Temte JL, Tuan WJ, Sinsky CA, et al. Tethered to the EHR: Primary Care Physician Workload Assessment using EHR Event Log Data and Time-Motion observations. Ann Fam Med. 2017;15(5):419-26.
  75. Tai-Seale M, Olson CW, Li J, Chan AS, Morikawa C, Durbin M, et al. Electronic Health Record logs Indicate that Physicians Split Time evenly between seeing patients and Desktop Medicine. Health Aff (Millwood). 2017;36(4):655-62.
  76. Murphy DR, Meyer AND, Russo E, Sittig DF, Wei L, Singh H. The Burden of Inbox Notifications in Commercial Electronic Health Records. JAMA Intern Med. 2016;176(4):559.
  77. Tai-Seale M, Dillon EC, Yang Y, Nordgren R, Steinberg RL, Nauenberg T, et al. Physicians’ well-being linked to In-Basket messages generated by Algorithms in Electronic Health Records. Health Aff (Millwood). 2019;38(7):1073-8.
  78. Muller AE, Berg RC, Jardim PSJ, Johansen TB, Ormstad SS. Can remote patient monitoring be the New Standard in Primary Care of Chronic Diseases, Post-COVID-19? Telemed E-Health. 2022;28(7):942-69.
  79. Willis VC, Thomas Craig KJ, Jabbarpour Y, Scheufele EL, Arriaga YE, Ajinkya M, et al. Digital Health Interventions To Enhance Prevention in Primary Care: scoping review. JMIR Med Inform. 2022;10(1):e33518.
  80. Nochomovitz M, Sharma R. Is it time for a New Medical Specialty? The Medical Virtualist. JAMA. 2018;319(5):437.
  81. Goldberg DG, Beeson T, Kuzel AJ, Love LE, Carver MC. Team-Based Care: a critical element of Primary Care Practice Transformation. Popul Health Manag. 2013;16(3):150-6.
  82. American College of Physicians, Smith CD, Balatbat C, National Academy of Medicine, Corbridge S. University of Illinois at Chicago, Implementing Optimal Team-Based Care to Reduce Clinician Burnout. NAM Perspect [Internet]. 2018 Sep 17 [cited 2023 Aug 1];8(9). Available from: https://nam.edu/ implementing-optimal-team-based-care-to-reduce-clinician-burnout.
  83. Loeb DF, Bayliss EA, Candrian C, deGruy FV, Binswanger IA. Primary care providers’ experiences caring for complex patients in primary care: a qualitative study. BMC Fam Pract. 2016;17(1):34.
  84. Wagner EH, Sandhu N, Coleman K, Phillips KE, Sugarman JR. Improving care coordination in primary care. Med Care. 2014;52(11 Suppl 4):33-8.
  85. Holman GT, Beasley JW, Karsh BT, Stone JA, Smith PD, Wetterneck TB. The myth of standardized workflow in primary care. J Am Med Inform Assoc JAMIA. 2016;23(1):29-37.
  86. Committee on Implementing High-Quality Primary Care, Board on Health Care Services, Health and Medicine Division, National Academies of Sciences, Engineering, and Medicine. Implementing High-Quality Primary Care: Rebuilding the Foundation of Health Care [Internet]. McCauley L, Phillips RL, Meisnere M, Robinson SK, editors. Washington, D.C.: National Academies Press. ; 2021 [cited 2023 Jul 27]. Available from: https://www.nap.edu/ catalog/25983.
  87. Tai-Seale M, McGuire TG, Zhang W. Time Allocation in Primary Care Office visits: Time Allocation in Primary Care. Health Serv Res. 2007;42(5):1871-94.
  88. Hilty DM, Torous J, Parish MB, Chan SR, Xiong G, Scher L, et al. A literature review comparing clinicians’ approaches and skills to In-Person, Synchronous, and Asynchronous Care: moving toward competencies to ensure Quality Care. Telemed E-Health. 2021;27(4):356-73.
  89. Rotenstein LS, Holmgren AJ, Healey MJ, Horn DM, Ting DY, Lipsitz S, et al. Association between Electronic Health Record Time and Quality of Care Metrics in Primary Care. JAMA Netw Open. 2022;5(10):e2237086.
  90. Saag HS, Shah K, Jones SA, Testa PA, Horwitz LI. Pajama Time: Working after Work in the Electronic Health Record. J Gen Intern Med. 2019;34(9):1695-6.
  91. Pellegrino ED, Altruism. Self-interest, and Medical Ethics. JAMA J Am Med Assoc. 1987;258(14):1939.
  92. Sajjad M, Qayyum S, Iltaf S, Khan RA. The best interest of patients, not selfinterest: how clinicians understand altruism. BMC Med Educ. 2021;21(1):477.
  93. Jones R. Declining altruism in medicine. BMJ. 2002;324(7338):624-5.
  94. Yin J, Ngiam KY, Teo HH. Role of Artificial Intelligence Applications in Real-Life Clinical Practice: systematic review. J Med Internet Res. 2021;23(4):e25759.
  95. Choudhury A, Asan O, Medow JE. Clinicians’ perceptions of an Artificial Intelligence-based blood utilization calculator: qualitative exploratory study. JMIR Hum Factors. 2022;9(4):e38411.
  96. Thomas Craig KJ, Willis VC, Gruen D, Rhee K, Jackson GP. The burden of the digital environment: a systematic review on organization-directed workplace interventions to mitigate physician burnout. J Am Med Inform Assoc JAMIA. 2021;28(5):985-97.
  97. Gosden T, Forland F, Kristiansen I, Sutton M, Leese B, Giuffrida A et al. Capitation, salary, fee-for-service and mixed systems of payment: effects on the behaviour of primary care physicians. Cochrane Effective Practice and Organisation of Care Group, editor. Cochrane Database Syst Rev [Internet]. 2000 Jul 24 [cited 2023 Jul 31];2011(10). https://doi.org/10.1002/14651858.CD002215.
  98. Diaz N. More health systems charging for MyChart messages. Becker’s Health IT [Internet]. 2022 Aug 28 [cited 2023 Aug 17]; Available from: https:// www.beckershospitalreview.com/ehrs/more-health-systems-charging-for-mychart-messages.html.
  99. Porter ME, Pabo EA, Lee TH. Redesigning primary care: a Strategic Vision to improve Value by Organizing around patients’ needs. Health Aff (Millwood). 2013;32(3):516-25.
  100. AMA Future of Health issue brief: Commercial Payer Coverage for Digital Medicine Codes. American Medical Association [Internet]. 2023 Sep 18 [cited 2023 Sep 26]; Available from: https://www.ama-assn.org/practice-management/ digital/ama-future-health-issue-brief-commercial-payer-coverage-digital.
  101. Richardson S, Lawrence K, Schoenthaler AM, Mann D. A framework for digital health equity. Npj Digit Med. 2022;5(1):119.
  102. Frimpong JA, Jackson BE, Stewart LM, Singh KP, Rivers PA, Bae S. Health information technology capacity at federally qualified health centers: a mechanism for improving quality of care. BMC Health Serv Res. 2013;13(1):35.
  103. Guo J, Li B. The application of Medical Artificial Intelligence Technology in Rural areas of developing countries. Health Equity. 2018;2(1):174-81.
  104. Davlyatov G, Borkowski N, Feldman S, Qu H, Burke D, Bronstein J, et al. Health Information Technology Adoption and Clinical Performance in Federally Qualified Health Centers. J Healthc Qual. 2020;42(5):287-93.
  105. Iyanna S, Kaur P, Ractham P, Talwar S, Najmul Islam AKM. Digital transformation of healthcare sector. What is impeding adoption and continued usage of technology-driven innovations by end-users? J Bus Res. 2022;153:150-61.
  106. Richardson JP, Smith C, Curtis S, Watson S, Zhu X, Barry B, et al. Patient apprehensions about the use of artificial intelligence in healthcare. Npj Digit Med. 2021;4(1):140.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

  1. *Correspondence:
    Matthew R. Allen
    Mattallen876@gmail.com; maa007@health.ucsd.edu
    Department of Family Medicine, University of California San Diego, La
    Jolla, CA 92093, USA
    Division of Biomedical Informatics, University of California San Diego, La
    Jolla, CA 92093, USA