Yes - I’ve always found those letters patronising (though I’ve never been described as ‘charming’ I have to say). Power to you and these campaigners - much better to write direct to the patient and copy in the GP. Who wants to read a letter about yourself written in the third person?
Interesting that Dr Rayner a renal physician put this forward. I was looked after by an amazing renal
Physician at the Churchill Hospital in Oxford for 27yrs till he retired. He always wrote the post clinic letter as if he was writing to me. With a bit at the end telling the GP off for something. There is a difference between physicians and surgeons. He used to say surgeons only like patients when they are out cold under a GA! All drs should try to make the letter helpful for the patient first.
This is an excellent summary of the webinar and the issues discussed. Thank you, Rory. The webinar will be available for viewing for free in a week or so. I hope you can advertise it.
Thank you! I have my own pile of letters from my consultants, but hadn’t realised how viscerally I dislike this way of communicating, until I read your post.
It is not only the fact that the patient is a “third person” and therefore not part of the communication. It is also that they become reduced to be a “clinical case” or if you will, an “object”, and that takes them out of the conversation about their treatment (and their own lives!).
Further, many chronic patients never get seen by their GPs, so there’s no point in the pretence of doctor-to-doctor update. Patient-engagement and patient-centricity should start with communication.
And a communication that is tailored to patient’s needs, not abusing specialised language beyond the reach of patients, which I would argue probably sometimes not even GPs might fully grasp. ;)
Thanks for this, long iverdue as you note. I have commented in the NHS change consultation askingbthat patients have the ability to comment/ annitate too. Not so much for physical health, which as you say can be technical, but certainly for mental health where details can be misremembered by the clinician too
It would be so good for a patient to receive a letter, (or an email if they prefer,) that explains in clear language what the condition is, the symptoms being experienced, the symptoms you may experience, what the treatment is and the potential timeframe for the treatment and ongoing care.
I have had to do my best to explain letters to service users, to my mum, and my mother in law, looking up phrases where needed. It would be nice to think that by the time my son might need to explain them to me, that he won't have to because I'll be able to understand them myself!
What a good idea! Like you Rory I’ve a number of long term conditions to deal with and having navigated the transition from one health trust to another ( moving house) which has been an absolute nightmare and is ongoing -anything to make life easier and more patient centred regarding NHS bureaucracy would be welcome
As a governor and committee chair for my NHS Foundation Trust, and with help from AI re Q&A, I have submitted the following proposal:
Q. Is it feasible for data recorded in a patient’s Electronic Patient Record to be passed through Ai so as to provide a concise, intelligible summary or synopsis for the patient of the nature of the intervention, the outcome and the next steps? If yes, how would this be best achieved?
A. Yes, it is feasible to use Ai to process data recorded in a patient’s Electronic Patient Record (EPR) to provide a concise, intelligible summary for the patient. Such a system could help patients better understand their medical interventions, outcomes, and next steps in their care. Here’s how this could be best achieved:
1. Data Collection and Structuring
Extract Relevant Data: The Ai system needs to extract relevant data from the EPR, such as the diagnosis, treatment interventions, outcomes, and recommended follow-ups. This involves identifying key sections within the EPR, like clinical notes, lab results, medications, and discharge summaries.
Natural Language Processing (NLP): NLP techniques can be used to parse and understand unstructured text within the EPR, allowing the Ai to identify important details and contextual information.
Data Structuring: Organise the extracted data into a structured format that is easier to process, categorise, and summarise.
2. Ai Model Development
Pre-trained Medical NLP Models: Utilise pre-trained models like BERT, GPT, or MedPaLM, which are already fine-tuned on medical literature and clinical notes. These models can be adapted to the specific task of summarising medical data.
Fine-tuning: Fine-tune the chosen model using a dataset of annotated EPRs, where each record is paired with a human-written summary. This will teach the model to generate concise summaries that are patient-friendly.
Incorporate Domain Knowledge: Use medical ontologies like SNOMED CT or ICD-10 to help the Ai system understand medical terminology and improve the accuracy of the summaries.
3. Patient-Centric Summarisation
Simplified Language: Ensure that the Ai system can translate medical jargon into layman’s terms. This is crucial for making the summary understandable to patients who may not have a medical background.
Personalisation: Tailor the summary to the specific patient, taking into account their medical history, literacy level, and preferences. For example, some patients might prefer a more detailed explanation, while others might want only the essentials.
Contextual Relevance: Ensure the Ai focuses on the most relevant information for the patient, like what they need to do next (e.g., take medication, attend a follow-up appointment) and what to expect in terms of outcomes.
4. Validation and Safety
Clinical Review: Before deploying the Ai-generated summaries, have them reviewed by healthcare professionals to ensure accuracy and appropriateness. This helps avoid potential misunderstandings or errors.
Patient Feedback Loop: Incorporate a feedback mechanism where patients can ask follow-up questions or clarify doubts about the summary. This helps refine the Ai’s performance over time.
5. Implementation
Integration with EPR Systems: Implement the Ai within existing EPR systems, ensuring it can access the data in real-time. The Ai should be an embedded tool that clinicians can use to generate summaries during or after a patient’s visit.
User Interface Design: Create a user-friendly interface where patients can access their summaries, either through a patient portal or a mobile app. The design should focus on clarity and ease of use.
6. Privacy and Security
Data Encryption: Ensure that all patient data is encrypted and that access is controlled to protect patient privacy.
Compliance with Regulations: The system must comply with healthcare regulations like GDPR to ensure that patient data is handled securely and ethically.
7. Continuous Learning and Improvement
Monitor and Update: Continuously monitor the Ai’s performance and update it with new data and patient feedback. Regularly retraining the model ensures it stays accurate and effective.
Collaborative Learning: Collaborate with medical professionals to periodically review and refine the summaries, keeping the Ai system aligned with evolving medical practices and patient needs.
Feasibility Considerations:
.Technical Feasibility: Current NLP technologies are advanced enough to handle this task, especially with fine-tuning. However, the quality of the output heavily depends on the quality and structure of the EPR data.
Operational Feasibility: Integrating Ai into EPR systems is technically complex but achievable. It requires collaboration between healthcare providers, IT specialists, and Ai developers.
Ethical and Legal Feasibility: Strict adherence to privacy laws and ethical guidelines is essential. The system must be designed to avoid bias and ensure fairness in how information is presented to all patients.
In summary, while challenges exist, particularly around data privacy, accuracy, and integration, using Ai to provide patient-friendly summaries from EPRs is feasible and can significantly enhance patient understanding and engagement in their healthcare.
Sharing patient-friendly summaries of treatments and outcomes should not simply be viewed as a process, but as an integral element of healthcare. With the right programming, it seems that Ai is in a perfect place to do this with both accuracy and programmed compassion, especially if the output is checked by a clinician before being given to the patient.
Perhaps now is a good time to address this issue? It may be that, with a written (or in time, voice) summary, the clinicians themselves may learn a little more about the implications for the patient of their medical and surgical interventions?
One thing. The digital dictation software providers to the NHS (Dragon, Bighand and Scribenet) all have models that will voice transcribe directly, rather than having to use someone to type up the dictation. The thing is of course, this is more expensive for Trust’s in terms of licensing, so as with most things, it usually comes down to money.
Excellent idea!! As an exGP, have to admit I didn’t understand always the hospital letters and would have welcomed being copied in to letter to patient instead…a few do that and then add a note to GP (usually an unrealistic task…)
The title of your blog is so appropriate - you are always "on" in a TV/reporter sense, even if not necessarily in a PD sense. All your work and reports regarding many aspects of the NHS are so valuable and helpful. I am very grateful for your hard work.
Yes - I’ve always found those letters patronising (though I’ve never been described as ‘charming’ I have to say). Power to you and these campaigners - much better to write direct to the patient and copy in the GP. Who wants to read a letter about yourself written in the third person?
Interesting that Dr Rayner a renal physician put this forward. I was looked after by an amazing renal
Physician at the Churchill Hospital in Oxford for 27yrs till he retired. He always wrote the post clinic letter as if he was writing to me. With a bit at the end telling the GP off for something. There is a difference between physicians and surgeons. He used to say surgeons only like patients when they are out cold under a GA! All drs should try to make the letter helpful for the patient first.
This is an excellent summary of the webinar and the issues discussed. Thank you, Rory. The webinar will be available for viewing for free in a week or so. I hope you can advertise it.
Thank you! I have my own pile of letters from my consultants, but hadn’t realised how viscerally I dislike this way of communicating, until I read your post.
It is not only the fact that the patient is a “third person” and therefore not part of the communication. It is also that they become reduced to be a “clinical case” or if you will, an “object”, and that takes them out of the conversation about their treatment (and their own lives!).
Further, many chronic patients never get seen by their GPs, so there’s no point in the pretence of doctor-to-doctor update. Patient-engagement and patient-centricity should start with communication.
And a communication that is tailored to patient’s needs, not abusing specialised language beyond the reach of patients, which I would argue probably sometimes not even GPs might fully grasp. ;)
Thanks for this, long iverdue as you note. I have commented in the NHS change consultation askingbthat patients have the ability to comment/ annitate too. Not so much for physical health, which as you say can be technical, but certainly for mental health where details can be misremembered by the clinician too
Thanks Rory.
What a huge difference a sensible and small change could make
It would be so good for a patient to receive a letter, (or an email if they prefer,) that explains in clear language what the condition is, the symptoms being experienced, the symptoms you may experience, what the treatment is and the potential timeframe for the treatment and ongoing care.
I have had to do my best to explain letters to service users, to my mum, and my mother in law, looking up phrases where needed. It would be nice to think that by the time my son might need to explain them to me, that he won't have to because I'll be able to understand them myself!
Some pilots have been done recently in public libraries around helping people decipher health information and use NHS app and similar. https://www.cilip.org.uk/page/health-and-digital
A fascinating piece Rory. Letters to the patient rather than the GP such a simple patient focused solution.
Mind you, here in Wales id like a first step to be actually able to access my own patient record online!
Damn! I really thought it was only me that was described in this way after each consult, so thanks for piddling on my bonfire...
What a good idea! Like you Rory I’ve a number of long term conditions to deal with and having navigated the transition from one health trust to another ( moving house) which has been an absolute nightmare and is ongoing -anything to make life easier and more patient centred regarding NHS bureaucracy would be welcome
Great idea to write to the patient. Time for change.
I received an email direct to my NHS app, hours after a recent hospital consultation. Fabulous!
As a governor and committee chair for my NHS Foundation Trust, and with help from AI re Q&A, I have submitted the following proposal:
Q. Is it feasible for data recorded in a patient’s Electronic Patient Record to be passed through Ai so as to provide a concise, intelligible summary or synopsis for the patient of the nature of the intervention, the outcome and the next steps? If yes, how would this be best achieved?
A. Yes, it is feasible to use Ai to process data recorded in a patient’s Electronic Patient Record (EPR) to provide a concise, intelligible summary for the patient. Such a system could help patients better understand their medical interventions, outcomes, and next steps in their care. Here’s how this could be best achieved:
1. Data Collection and Structuring
Extract Relevant Data: The Ai system needs to extract relevant data from the EPR, such as the diagnosis, treatment interventions, outcomes, and recommended follow-ups. This involves identifying key sections within the EPR, like clinical notes, lab results, medications, and discharge summaries.
Natural Language Processing (NLP): NLP techniques can be used to parse and understand unstructured text within the EPR, allowing the Ai to identify important details and contextual information.
Data Structuring: Organise the extracted data into a structured format that is easier to process, categorise, and summarise.
2. Ai Model Development
Pre-trained Medical NLP Models: Utilise pre-trained models like BERT, GPT, or MedPaLM, which are already fine-tuned on medical literature and clinical notes. These models can be adapted to the specific task of summarising medical data.
Fine-tuning: Fine-tune the chosen model using a dataset of annotated EPRs, where each record is paired with a human-written summary. This will teach the model to generate concise summaries that are patient-friendly.
Incorporate Domain Knowledge: Use medical ontologies like SNOMED CT or ICD-10 to help the Ai system understand medical terminology and improve the accuracy of the summaries.
3. Patient-Centric Summarisation
Simplified Language: Ensure that the Ai system can translate medical jargon into layman’s terms. This is crucial for making the summary understandable to patients who may not have a medical background.
Personalisation: Tailor the summary to the specific patient, taking into account their medical history, literacy level, and preferences. For example, some patients might prefer a more detailed explanation, while others might want only the essentials.
Contextual Relevance: Ensure the Ai focuses on the most relevant information for the patient, like what they need to do next (e.g., take medication, attend a follow-up appointment) and what to expect in terms of outcomes.
4. Validation and Safety
Clinical Review: Before deploying the Ai-generated summaries, have them reviewed by healthcare professionals to ensure accuracy and appropriateness. This helps avoid potential misunderstandings or errors.
Patient Feedback Loop: Incorporate a feedback mechanism where patients can ask follow-up questions or clarify doubts about the summary. This helps refine the Ai’s performance over time.
5. Implementation
Integration with EPR Systems: Implement the Ai within existing EPR systems, ensuring it can access the data in real-time. The Ai should be an embedded tool that clinicians can use to generate summaries during or after a patient’s visit.
User Interface Design: Create a user-friendly interface where patients can access their summaries, either through a patient portal or a mobile app. The design should focus on clarity and ease of use.
6. Privacy and Security
Data Encryption: Ensure that all patient data is encrypted and that access is controlled to protect patient privacy.
Compliance with Regulations: The system must comply with healthcare regulations like GDPR to ensure that patient data is handled securely and ethically.
7. Continuous Learning and Improvement
Monitor and Update: Continuously monitor the Ai’s performance and update it with new data and patient feedback. Regularly retraining the model ensures it stays accurate and effective.
Collaborative Learning: Collaborate with medical professionals to periodically review and refine the summaries, keeping the Ai system aligned with evolving medical practices and patient needs.
Feasibility Considerations:
.Technical Feasibility: Current NLP technologies are advanced enough to handle this task, especially with fine-tuning. However, the quality of the output heavily depends on the quality and structure of the EPR data.
Operational Feasibility: Integrating Ai into EPR systems is technically complex but achievable. It requires collaboration between healthcare providers, IT specialists, and Ai developers.
Ethical and Legal Feasibility: Strict adherence to privacy laws and ethical guidelines is essential. The system must be designed to avoid bias and ensure fairness in how information is presented to all patients.
In summary, while challenges exist, particularly around data privacy, accuracy, and integration, using Ai to provide patient-friendly summaries from EPRs is feasible and can significantly enhance patient understanding and engagement in their healthcare.
Sharing patient-friendly summaries of treatments and outcomes should not simply be viewed as a process, but as an integral element of healthcare. With the right programming, it seems that Ai is in a perfect place to do this with both accuracy and programmed compassion, especially if the output is checked by a clinician before being given to the patient.
Perhaps now is a good time to address this issue? It may be that, with a written (or in time, voice) summary, the clinicians themselves may learn a little more about the implications for the patient of their medical and surgical interventions?
Thank you, this is a very helpful post.
One thing. The digital dictation software providers to the NHS (Dragon, Bighand and Scribenet) all have models that will voice transcribe directly, rather than having to use someone to type up the dictation. The thing is of course, this is more expensive for Trust’s in terms of licensing, so as with most things, it usually comes down to money.
Excellent idea!! As an exGP, have to admit I didn’t understand always the hospital letters and would have welcomed being copied in to letter to patient instead…a few do that and then add a note to GP (usually an unrealistic task…)
Hope you succeed in this important quest..
Ann Robinson
https://annrobinson.contently.com/
The title of your blog is so appropriate - you are always "on" in a TV/reporter sense, even if not necessarily in a PD sense. All your work and reports regarding many aspects of the NHS are so valuable and helpful. I am very grateful for your hard work.
Gillian this means a lot to me, thank you so much
I couldn’t agree more. And sometimes the patient is not even sent a copy, as has happened to me recently - twice.