Artificial Intelligence and Telemedicine: Human Judgment in the Digital Era with Dr. Jan Niclas Strickling

The age of artificial intelligence is reshaping healthcare delivery worldwide, with telemedicine at the forefront of this transformation. In episode five of the DevelopmentAid Dialogues podcast, host Hisham Allam speaks with Dr. med. Jan Niclas Strickling, a German board-certified interventional cardiologist who has played a key role in advancing telemedicine through Germany’s certified centers and holds multiple certifications from the German Society of Cardiology.  

“AI is redefining what’s possible in telemedicine—but at every step, it’s collaboration, not competition,” Strickling said, unpacking how digital tools are changing patient care.” 

Across Germany and the EU, AI-driven triage systems, medical imaging analysis, and real-time language translation are making healthcare more accessible. “If AI takes the strain out of documentation—which is half my daily work—it frees me to focus on the patient,” Strickling explained. Wearables like the Apple Watch, CPAP machines, and glucose sensors generate continuous data streams that help identify patients’ needs remotely, especially in underserved areas. 

But he cautioned that technology alone isn’t enough. “AI can bridge gaps, but equity depends on broadband access, device availability, and whether AI models are trained on diverse populations.” Without representative data, AI risks missing or misdiagnosing patients from different demographic groups. 

Alongside opportunity, risks persist. Strickling described “alert fatigue” where oversensitive AI systems overwhelm clinicians with notifications, potentially obscuring urgent issues. The bigger danger is “automation bias”—over-relying on AI recommendations while sidelining clinical judgment. “The final decision must remain human,” he stressed. He recalled uploading his own ECG to ChatGPT, which wrongly diagnosed a life-threatening arrhythmia. “For patients, that can cause needless fear and erode trust in doctors.” 

Highlighting the promise of AI, Strickling described a heart failure project in Germany where wearable defibrillator vests and smart scales transmit continuous health information. AI analyzes daily blood pressure, weight, and body movement to preempt hospitalizations by advising medication adjustments. “The data flood makes sense only when paired with human judgment to determine who needs attention now.” 

Hybrid care models blending remote monitoring with targeted in-person visits are expanding, with virtual rounds led by nurses and specialists joining as needed. Yet, the human connection—empathy, understanding, and trust—remains irreplaceable 

As digital health advances Strickling calls for transparency, patient consent, and robust regulation. “We must disclose AI’s use and limits, monitor for biases, and ensure privacy through encryption and strict data controls.” The need for accountable human oversight is paramount. “Who bears responsibility for AI-driven errors? That must be a clinician.” 

Echoing the complex future, he said, “Experience and learning from mistakes remain at medicine’s core. AI assists but can’t replace the wisdom patients deserve.” 

Listen to the full episode on DevelopmentAid Dialogues. Stay informed. Stay engaged.

The podcast is sponsored by DevelopmentAid. Procurement notices, funding and grants to opportunities, lists of potential partners, insights into market trends, databases of development professionals, webinars, latest news, and much more. Stay informed and connected.

Subscribe and Stay Connected

Transcript

Season 3. Episode 5: Artificial Intelligence and Telemedicine: Human Judgment in the Digital Era with Jan Niclas Strickling 

This podcast is sponsored by developmentaid.org, your gateway to international tenders, grants, and the largest job database in the aid sector. Secure free registration and access funding opportunities. Calls for partners, exclusive insights, and weekly News Digest. Become a member today to achieve your sustainable development goals.  

Hisham Allam: Hello everyone and welcome to DevelopmentAid Dialogues. I'm your host, Hamma Allam. Today we are diving into an exciting and rapidly evolving topic, the impact of artificial intelligence on telemedicine. These digital health services are reshaping how care is delivered remotely. Offering new opportunities and raising important questions. 

Our discussion will explore both the promises and the challenges AI brings to telemedicine and how it might shape the future of healthcare for patients and providers alike. Our guest today is Dr. Jan Nicholas Strickling, a Jim Board certified interventional cardiologist. Dr. Strickling brings extensive expertise in heart failure and cardiac device therapy, as well as significant experience advancing telemedicine in Germany through one of its certified telemedicine centers. 

He holds multiple certifications from the German Society of Cardiology and is an active member of both the German and European societies of cardiology. Dr. Strickling, welcome to DevelopmentAid Dialogues. 

Jan Nicholas Strickling: Hi, good morning everyone. Thanks for having me at this exciting podcast on this exciting artificial intelligence and telemedicine topic. 

Hisham Allam: Let's begin by exploring how AI is currently shaping telemedicine. From your experience, what are the most impactful changes you have seen so far? 

Jan Nicholas Strickling: So let me share some of the most interesting changes of the use of AI in telemedicine with you. In my opinion, triage systems like symptom checkers, risk score, and automatic calculation of risk scores help to direct the patient to the right and correct level of care even before they visit the clinician. 

This will shorten the waiting time and avoid misleading of the patient to the wrong specialist. One additional huge area is the documentation. So, AI has the opportunity to scribe, capture and structure tele visit notes, freeing my time as a clinician to focus more on the patient. Even remote monitoring. 

So, we are using an Apple watch and WHOOP, we have some input data of CPAP machines, what is used by a patient with chronic obstructive pulmonary disease or glucose sensors, this NFC chips you use in diabetes. And we can use all of this data to see the right patient at the right time. One huge topic where we see and where we use AI is diagnostics and imaging. 

AI supports telemedicine and dermatology in ophthalmology where we can scan the background of the eye, what is a fundoscopy with an AI driven tool, and only send the results and the potential findings to the ophthalmologist to discuss with the patient and AI assists in analyzing medical images like X-ray, MRI scans, CT scans, accelerating the diagnostic process. 

Improving the diagnostic accuracy at the end, assisting the clinician. And one additional topic is language. So real time translation. There was the involvement of the new Apple speakers what can offer real time translation. And this leads to a like patient friendly instruction that increases the patient education and adherence at the end. 

So, the use of AI in telemedicine, creates accessibility in rural and underserved regions for sure. 

Hisham Allam: Do you see that artificial intelligence as a threat to the traditional practice of telemedicine, or rather as a powerful push that will accelerate its adoption and evolution? 

Jan Nicholas Strickling: So in my opinion, it is mostly an accelerator, but if governance is strong, because it is mandatory to build up trust and quality, because when systems and AI systems and models are deployed without oversight, explainability, or equitable assess for the patient itself, this can create fear to use this AI model. So, AI and telemedicine will make me as a doctor perform better in assisting and to concentrate on things that are very crucial. So, it is more like a filter to exclude unnecessary visits and triage the patient along with their risks, symptoms, and findings at the end. 

Hisham Allam: Don't you feel that ordinary people that are using or relying a lot on Gemini or Chat GPT to get some diagnosis or reading analysis or x-rays. Do you think that this is happening so rapidly the reliance of people are increasing on this even without getting any consultation from doctors? 

Jan Nicholas Strickling: Of course I see that in a regular basis. So, I see patient in my teleconsultation or in op in my OPD visits. They bring their prompt from Chat GPT and want to be sure if it's correct or not. So this is something I see definitely and it's increasing. So it is very easy because there are free software available, Gemini and Chat GPT, where you can upload your symptoms, where you can upload even your medical reports or x-rays and you get a prompt what can be true. But in most cases it should be like checked by a human clinician at the end. 

Hisham Allam: But for you, doctor how are you using AI in your specialty?  

Jan Nicholas Strickling: Let me explain in an easy way. So there is a patient bringing me his prompt and, I need to check if this prompt matches with the symptoms with my experience. 

So, it is a lot about experience of the clinician to get to the right diagnose. So for me, there are several areas where I use AI. One area is, for example - documentation. So there are softwares available. They listen to my consultation. So while I'm doing the consultation, my medical report is like in real time is written by AI, so it reduces my time for documentation, what makes like 50% of my daily business so I can concentrate more on what is needed for the patient. So what is face-to-face. And not only like typing all the  symptoms and diagnosis while I'm sitting face-to-face to the patient, not facing the patient, just facing my computer to drop in my documentation. 

So, this is something where I use AI. In like MRI CT or x-ray cases, it is very easy and there are several studies available, where AI assisted clinicians are compared to clinicians without the use of AI and the AI assist clinician, they perform much better. And it's not about like cheating. 

Yeah. So at the end it is. More accuracy, more safety, and a higher probability to get to the right diagnose for the patient. So it's not cheating. So it is accelerating and assisting the doctor. So it's more like a collaboration and not a completion at the end. 

Hisham Allam: That's clear. Yeah. How can AI help telemedicine overcome some of its current limitations or, and barriers to more widespread acceptance? 

Jan Nicholas Strickling: So there are several areas where AI can help telemedicine to overcome these limitations we are fighting against. So as I told you already, where I use AI is documentation and when documentation is 50% of my daily business, and AI can take this time and can concentrate more on things that are needed. 

So, and there are several, reliable examination and studies about burnout in clinicians just because of this documentation. So if AI can use and can be used to reduce this time, it's a gift. Yeah. And we can perform even in a higher quality, to support my decision as a clinician. So when I enter a new medication, for example, and I already know about the current medication the patient is taking, and then get directly for example, a drug interaction alert due to AI. This helps me to reduce my miss rates and, there is a continuity in AI. It is available 24/7 and it can summarize and stitch together fragmented records for at the end, coherent tele visit and regarding costs like automatic scheduling, eligibility, and preauthorized workflows they shrink together my administrative workflow and in a very cost-effective way. 

Hisham Allam: Other scenarios where AI might unintentionally disrupt or challenge existing telemedicine workflows negatively. 

Jan Nicholas Strickling: Yeah, so of course there are still a lot of challenges we are facing at the moment. For example, when I open up my laptop in the morning and I got like 1000 red alerts on my laptop and my AI software, it is an over sensitive model. 

And there, there is a risk for the clinician to get alert fatigue and miss some very important alerts because these models need to be trained and they need to learn. So, we need to create models that are not over sensitive and overwhelming the doctor with any potential findings. One challenge we are facing is data drift. 

Data drift is when the model is trained on one setting, but it is used in another. And this causes silent performance drops. And I spoke about this documentation. So it is a software, it's a microphone, what is listening to the combination I have, so it should be a tool what is very, very stick to the effect. 

And some of the medications, for example, they sound very similar. And when this tool hallucinates or misattributes some statements of the patient or for me, this can lead to an error at the end. So, it needs to be rechecked before I finalize my medical report. 

Hisham Allam: Building on what you just said, in your view, could an over line on AI and telemedicine risk diminishing the value of personal clinical judgment? 

Jan Nicholas Strickling: Yeah, of course. So, this risk exists and it has a name. What is automation bias. And automation bias is real, and occurs when I as a clinician or even the patient over relies on the recommendation from AI. Potentially overlooking clinical judgment and missing errors. And it requires the opportunity for me to override this prompt or this finding when I disagree. 

And because the final decision to find the clinical judgment is on me as a human doctor. And AI can assist. Data analysis, pattern recognition, and information retrieval. But at the end, for me as a doctor, the critical thinking, the final decision making and the patient doctor interaction is a human one. 

So, as I told you, it is a collaboration. 

Hisham Allam: Before we move on, can you share specific situations where AI has proven especially valuable in remote healthcare delivery? 

Jan Nicholas Strickling: Mm-hmm. Yes. There are a lot of different scenarios where we effectively and efficiently use AI in telemedicine. 

To prevent complications and it is a good indicator for chronic disease management. Let me share a project we started in Germany, what is in heart failure monitoring. So, we attached so-called variable defibrillator vest. It's a device what is worn by the patient for 24/ 7. In patient with a severe heart failure, this defibrillator vest has the opportunity to check the ECG 24/ 7 and has the opportunity to deliver a shock when it's needed, when there is a live threatening arrhythmia because this patient with a severe heart failure, they are in on under risk for sudden cardiac death. So, we provided this vest. In addition to that, we provided BP monitor and the scale, and the patient was advised to send the results daily automatically. 

But he needs to do the blood pressure monitoring and he needs to step on the scale for sure. So, this vest was not only able to deliver a shock when it's needed. So, has the opportunity to see the heart rate over the day when it has the opportunity to take an ECG. 

It has the opportunity to see the sleeping position of the patient, the body movement, and the steps taking per day. So it was easy for us to adjust the medication. And to avoid any hospitalization because in heart failure patient, we know that every hospitalization will worse the outcome of the patient at the end. 

So, it was a huge amount of data, but with daily BP, blood pressure, daily weight, any like gain in weight, any risk for the congestion decompensation of too much fluid in your body, what is the risk in heart failure patient. We could avoid this hospitalization earlier and we can adjust the medication. 

For example, to increase the dosage of a diuretic, and therefore we could help patient with a severe heart failure. What is a good example with a huge amount of data input, but at the end, AI helps us to get it sorted and to see the patient who need an interaction, a human interaction now. 

Hisham Allam: Turning to the future, how might AI driven telemedicine change the roles and responsibilities of healthcare providers in the near future? 

Jan Nicholas Strickling: For me as a clinician in my tailor or OPD consultations. The first minutes are about the current symptoms and the clinical background of the patients, so when it could be possible that AI provides the complete medical background, the current medication, I can definitely better focus and have more time for the current problem of the patient and bring it in line with the patient history. 

In addition to that, AI can do my administrative work and documentation, I have more time for face-to-face and for the physical examination of the patient. 

Hisham Allam: That's something for our listeners to think about. Can AI serve as a breach to expand telemedicine access globally, or might it create inequities due to technological divides? 

Jan Nicholas Strickling: These are both possible in my opinion. So let's start with AI serving as a bridge. For example, this low-cost triage, the language support and real-time translation. This will help to bring medical treatment to underserved areas. But at the, on the other side, there are inequities for sure like limited broadband access device gaps, and one crucial thing in my opinion is we don't know what the models the AI models are trained on and more maybe they are trained on non-representative data. I read a lot about this training of AI models and some suggestions are AI models are mostly trained on white college aged males. 

So, this is not my daily business patient. For example in skin cancer diagnosis, there are a lot of apps available where you can scan your molds by your own and you don't need to see any physician and you get a very accurate result. But on the one hand, there are pay for tech without any reimbursement from insurance. 

And also these models are struggling in darker skin colors. So, there are a lot of steps to go. We are just at the beginning, but we need to know what the models are trained on, and we need to know that there are potential gaps and these models are potentially struggling and different population groups. 

Hisham Allam: This question came to my mind now. Do you expect that AI could change the way that medicine is studied at universities and colleges that you don't need to study this or to remember this as long as you can ask AI. Or to find this information just by one click. 

Jan Nicholas Strickling: So, it is so easy to get a lot of information due to just entering something into Chat GPT. 

What is this disease? What, what are the symptoms common for? It's that easy. When I studied medicine, this was not available. I need to read the book. And there is. One fear inside of me. What is residency? When I was a young doctor you need to learn by mistakes. So sure, doing mistakes and medicine is something we definitely want to avoid, but making mistakes will lead to do that never, ever again. 

So, this is something what is deep in your brain. And it's not only mistakes done by you, your colleagues, but this is something what makes you learn even better and when you are not allowed to make mistakes, what is at the end, patient safety for sure. But when you're not allowed to make mistakes and you have this easy access to all this medical information, this definitely will change the medical education, but I'm not quite sure, is it something bad or something positive? So, it is something in between. Because on the one hand side, it is positive because it's easy to reach and everyone has a smartphone in his pocket and can lead to a more accurate diagnosis, but you need to create experience to make the right decision. Because in some situations you don't have the opportunity to check the Chat GPT and I don't want to sit in front of the patient and use Chat GPT and type in his symptoms. 

This looks like, as I told you already, like cheating. Yeah. And this will reduce the trust the patient has in me as a clinician, as an experienced clinician. 

Hisham Allam: But Chat GPT also make mistakes. 

Jan Nicholas Strickling: Yeah, of course, of course. Let me tell you one story about Chat GPT. I just recently added an ECG, my own ECG to Chat GPT and asked Chat GPT, what is the ECG showing and Chat GPT answers. Okay. Your heart rate is 200, no, it was 75. And it is probably life-threatening arrhythmia. What is an ventricular tachycardia. And I gave him another chance. So, I give him another chance, another attempt. Are you really sure I have a life-threatening arrhythmia?  And Chat GPT answers, okay I cannot do the final clinical judgment, but I can help you to interpret this ECG. And again, it says most probably it is a life threatening arrhythmia. you need to see an ER if you're suffering of chest pain, palpitations, fast heart rate or dizziness. And I asked, I don't, I answered, I don't have this symptoms at the moment. And the answer of Chat GPT was, okay you are not suffering of the symptoms, but this can change suddenly better see and ER.. 

And then I reviewed the secret. I answered, okay, I am a cardiologist and I do think I have a normal sinus rhythm, normal ECG, without any risk for this life threatening arrhythmia, and then the answer was, oh, okay this changes the frame of discussion a lot. I was thinking, okay imagine a patient I saw in the morning in my OPD or in an emergency room and we took an ECG and everything was fine, and she takes a picture of this and uploads this picture to Chat GPT and she is telling you have a life-threatening arrhythmia? This creates a lot of fear, anxiety in the patient, and the trust of the patient into my knowledge is like zero, and it is so hard to convince the patient that the Chat GPT was wrong when he's now again at the ER and brought there by an ambulance. So, imagine all this struggle and this anxiety what is created. So it is something what needs to learn what needs to be trained. We are just at the beginning and but we need to tell our patients that is something what can assist but will not give the correct final clinical judgment at the end. 

Hisham Allam: This is a very impressive example. Doctor, let's look at the broader healthcare system. How can providers and policy makers ensure that AI integration in telemedicine remains ethical and equitable. 

Jan Nicholas Strickling: We need to offer transparency and, at the end, a high quality output to create this wide acceptance. 

We need to disclose AI use and its limits and then obtain meaningful content of the patient. So, the patient needs to accept that we use AI in his case. And we know about this biases I told you about this automation bias. We need to monitor these biases, and we need to monitor our performance and by race, ethnicity, language, sex and we need to address these gaps. 

Because independent validation is crucial. I mean, we need, for example, like external site testing and continuous post-deployment monitoring. And one major and crucial question is who will be held responsible in an AI driven era? So, the final clinical decision and judgment should be on a human doctor because at the moment, you need someone who is reliable for this potentially AI assisted error, and so at the moment, it should be a human clinician. 

Hisham Allam: Do you foresee AI enabling new hybrid models of care that blend remote and in-person services more effectively?  

Jan Nicholas Strickling: Yes, exactly. There are a lot of hybrid models you can imagine. For example, this triage system with a risk stratified schedule, for example, a low-risk patient, they stay mostly remote with an periodic in-person checkup. 

And when there is a high-risk patient with an, when we come back to this example with a heart failure patient, there is a gain in weight. We see the steps are like significantly lower than the days before. And we see that his body position, for example, you need to sleep with an upright, upper body what is a sign that there could be a fluid congestion. So, we need to proactively outreach to this patient to target face-to-face visits or medical treatment adjustment. And another example could be team-based virtual rounds on the ICU, for example. So, the round is led by a nurse and only when there are some findings the specialist is escalated. Yeah. So, these hybrid models, they work as a triage filter to reduce the unnecessary work for the doctor and to concentrate on tasks where I am really needed. So, but we need to keep in mind that the fundamental aspect of healthcare is the human connection between the doctor and the patient. 

So the ability to provide emotional support, empathy, and understanding is essentially impossible for now to replicate by non-human. So, if you think about opening up your laptop, doing a teleconsultation with an like an AI driven avatar compared to an human doctor, and they're telling you exactly the same. I think you will trust the human doctor for now more than the AI avatar. 

Hisham Allam: As we come to a close, considering potential risks, what if guards should be put in place during AI integration to protect patient safety and privacy? 

Jan Nicholas Strickling: So, we need to answer the question again. Who will be held responsible in an AI driven error? 

And therefore, I pledge for final decision making should be done by human clinician. And when we speak about privacy and security, we need end to end encryption and least privilege assess. And whenever possible, data processing should be on device. Because we need to know on which servers the data is sent to and we need to establish a robust regulatory framework for AI driven diagnosis to remain crucial for the widespread and an effective integration of AI in telemedicine, because I don't want to send my sensitive patient data. To anyone unknown, this increases the risk of unauthorized assesses. For example, so imagine you want to go for a new healthcare insurance, and some years ago you uploaded some medical reports and you get rejected by any diagnosed, what was just written in, in the subtext. 

So, this risk exists and we need to do a regulatory framework to avoid this. 

Hisham Allam: Thank you, Dr. Jan Nicholas Strickling for sharing your expert insights and encouraging outlook on the role of artificial intelligence, and telemedicine. Your prospective sheds light on both the exciting benefits and the careful consideration needed as we embrace this new frontier in healthcare. 

To our listeners, if you find this conversation enlightening, please follow DevelopmentAid Dialogues for more in-depth discussion on that matter. Don't hesitate to share this episode within your network. I'm your host, Hisham Allam signing off. Until next time, stay informed and stay engaged. Goodbye.