This is a white paper combining data from the Autoderm integration into the myGP app that has 3 million subscribers, and cost data from published studies.
Mass population screenings for skin diseases using artificial intelligence (AI) could significantly enhance early detection and accessibility to dermatological care, particularly in light of the global shortage of dermatologists. It is estimated that there are only 150 000 board certified dermatologists in the world. Ranging from 1 dermatologist per 30 000 citizens in the US, 1 dermatologist per 150 000 citizens in the UK, 1 dermatologist per 150 000 citizens in Indonesia and 1 dermatologist per 7 million citizens in Ethiopia.
Successful implementation depends on the accessibility to AI on a large scale. Autoderm has partnered with the myGP app in the UK, which has over 3 million subscribers, and corresponds to roughly 5% of the UK population. The myGP app offers their users:
The shortage of dermatologists globally presents a significant challenge for effective skin disease diagnosis and management. In this context, mass population screenings using AI could be a viable solution.
AI technologies, particularly those utilizing machine learning and computer vision; close to 200 peer review studies have now demonstrated strong potential in identifying skin cancers and other skin conditions at early stages. Early detection is crucial for effective treatment, and AI can enhance diagnostic accuracy, potentially matching or exceeding that of human dermatologists in some cases.
AI-driven tools can provide rapid assessments of skin conditions using images uploaded by users. For example, Autoderm can analyze a skin ailment taken with a smartphone within a second or, 90 000 skin ailments per day, offering insights into potential skin issues instantaneously. This immediacy can help alleviate the burden on dermatology services by allowing patients to receive preliminary evaluations without needing an immediate in-person consultation.
An early version of our AI was studied clinically and the results published in Nature. The AI model served as a decision support tool (DST) for non-specialist doctors, improving their ability to interpret skin conditions. This is particularly beneficial in primary care settings, where a significant number of skin-related consultations occur. It showed that Autoderm could help streamline referrals to dermatologists by sorting cases based on urgency and complexity.
There are about 2500 skin diseases documented in the literature. Autoderm has been developed to recognize the most common skin diseases and updated with a diverse set of new skin diseases of any skin color as the AI models are being trained, not reducing accuracy. This versatility makes AI a valuable asset in mass screening initiatives on any demography, as it can address a wide array of dermatological issues.
Screening tests can lead to false positives, resulting in unnecessary anxiety and invasive procedures. Conversely, false negatives may lead to missed diagnoses of skin cancer.
While mass screenings may not be effective in reducing mortality, they can improve awareness and encourage individuals to perform self-examinations, which is crucial for early detection.
At Autoderm, our AI models undergo rigorous clinical validation to ensure accuracy across diverse populations, skin types and user testing. Many existing AI models have been trained on limited datasets and the “wrong” kind of images, like clinical images taken with a high resolution camera, not a smartphone camera, and may not adequately represent all demographics.
AI as we know it today is not 100% perfect. For example, when we use any of the large language models (LLM), like OpenAI, we can see that it is about 80% correct and human interpretation is needed, but it is still very good at achieving the desired end result and sometimes it helps improve the end result. Clear user instructions and expectations should be presented to the end user. Personal data is not collected, only the image is used for analysis, it can be chosen to be discarded or be collected for further use to train future AI models.
It has also been made clear that AI should complement, not replace, the expertise of healthcare professionals, to ensure that definitive diagnosis and treatment plans involve human oversight. However, it is very good at triaging a patient to the right level of healthcare at the right time.
During our partnership with myGP, as of May 2024 over 370 000 API calls have been done. Since we are a Software as a Medical Device (SaMD) we do Product Market Surveillance (PMS) reports every 6 months.
You can read the full report here.
Screening for diseases in healthcare is a valuable tool that, when applied appropriately, can significantly improve health outcomes, reduce healthcare costs, and enhance public health. However, it is crucial to balance the benefits with the potential risks and limitations. Effective screening programs should be evidence-based, targeting populations where the benefits outweigh the harms, and should be accompanied by clear guidelines and patient education to ensure informed decision-making.
It is estimated that 20% of GP visits are skin related. Autoderm did roughly 18 000 screenings in May 2024 and from the data, about 50% could have been triaged with further information retrieval for self care with or without over the counter medication from a pharmacy.
NHS documentation shows that average cost to the NHS for a 9-minute GP face-to-face consultation was estimated to be £42 in 2021/22. That would equal a £4 536 000 (9000*12) saving to the NHS, if 50% of the screening were triaged for self care.
This is a complicated task to pinpoint, there are many studies that have tried to calculate the number, however because of the different stages of melanoma, there are different costs associated. According to our online research, the cost to treat a case of melanoma in the UK is estimated to be around £2 600 per case, these are figures from 2013 by the Department of Health.
Melanoma treatment costs are highly dependent on the stage at diagnosis, with Stage IV cases costing over 25 times more than early Stage I cases. Costs are particularly high in the first year after diagnosis for advanced stage disease. Promoting primary prevention and early detection of melanoma could lead to substantial cost savings.
So in summary, the available evidence indicates the cost to treat a single case of malignant melanoma in the UK is approximately £2 600 on average, although the total costs can vary depending on the severity and treatment required.
Patients diagnosed with early stage melanoma (Stage I or II) have much lower treatment costs compared to those diagnosed at later stages. One study found the cost of managing a Stage IV case (€122 985) was over 25 times more expensive than a Stage IA case (€4 269).
Another study estimated individual patient costs ranging from €1 689 for Stage I to €88 268 for Stage IV. The largest cost differences were between Stage IA and IB-IIA, and between Stages III and IV.
In the first year after diagnosis, hospitalization costs are 8-16 times higher for advanced stage melanoma compared to early stage. The costs of outpatient services and inpatient drugs also decrease gradually as the stage at diagnosis is earlier.
The higher expenditure associated with more advanced stages is mainly driven by increased inpatient drug use and more intensive treatment required.
Overall, the average direct costs related to melanoma are highest in the first year after diagnosis (€2 903) and then decrease over time
Now that we know the costs, we can calculate the yearly cost savings for the NHS detecting 79 melanomas per month using Autoderm. We need to consider the cost of treating each melanoma and the number of screenings performed.
We will be conservative in our calculations, because we do not know what stages the melanomas were at. We will assume the average cost to be £2 600 per case. The total cost of treating the 79 melanomas detected in May would be:
The cost of performing 18 000 screenings can be estimated at £18 000 assuming £1 per screening.
£205 400*12 (treatment cost) - £18 000*12 (screening cost) = £2 248 800
The yearly cost savings for the NHS due to the detection of 79 malignant melanomas using AI in dermatology screening would be approximately £2 248 800.
Total cost savings considering 50% are triaged to self care is £4 536 000.
£2 248 800 + £4 536 000 = £6 784 800.
If we had screened the whole UK population (100%) it would theoretically be a £135 696 000 savings to the NHS.
AI mass screenings show great potential when there are not enough dermatologists to go around. More data needs to be collected and more research done over time. But 2024 could be a good year to start following this data on a macro level.
This is a white paper combining data from the Autoderm integration into the myGP app that has 3 million subscribers, and cost data from published studies.
Mass population screenings for skin diseases using artificial intelligence (AI) could significantly enhance early detection and accessibility to dermatological care, particularly in light of the global shortage of dermatologists. It is estimated that there are only 150 000 board certified dermatologists in the world. Ranging from 1 dermatologist per 30 000 citizens in the US, 1 dermatologist per 150 000 citizens in the UK, 1 dermatologist per 150 000 citizens in Indonesia and 1 dermatologist per 7 million citizens in Ethiopia.
Successful implementation depends on the accessibility to AI on a large scale. Autoderm has partnered with the myGP app in the UK, which has over 3 million subscribers, and corresponds to roughly 5% of the UK population. The myGP app offers their users:
The shortage of dermatologists globally presents a significant challenge for effective skin disease diagnosis and management. In this context, mass population screenings using AI could be a viable solution.
AI technologies, particularly those utilizing machine learning and computer vision; close to 200 peer review studies have now demonstrated strong potential in identifying skin cancers and other skin conditions at early stages. Early detection is crucial for effective treatment, and AI can enhance diagnostic accuracy, potentially matching or exceeding that of human dermatologists in some cases.
AI-driven tools can provide rapid assessments of skin conditions using images uploaded by users. For example, Autoderm can analyze a skin ailment taken with a smartphone within a second or, 90 000 skin ailments per day, offering insights into potential skin issues instantaneously. This immediacy can help alleviate the burden on dermatology services by allowing patients to receive preliminary evaluations without needing an immediate in-person consultation.
An early version of our AI was studied clinically and the results published in Nature. The AI model served as a decision support tool (DST) for non-specialist doctors, improving their ability to interpret skin conditions. This is particularly beneficial in primary care settings, where a significant number of skin-related consultations occur. It showed that Autoderm could help streamline referrals to dermatologists by sorting cases based on urgency and complexity.
There are about 2500 skin diseases documented in the literature. Autoderm has been developed to recognize the most common skin diseases and updated with a diverse set of new skin diseases of any skin color as the AI models are being trained, not reducing accuracy. This versatility makes AI a valuable asset in mass screening initiatives on any demography, as it can address a wide array of dermatological issues.
Screening tests can lead to false positives, resulting in unnecessary anxiety and invasive procedures. Conversely, false negatives may lead to missed diagnoses of skin cancer.
While mass screenings may not be effective in reducing mortality, they can improve awareness and encourage individuals to perform self-examinations, which is crucial for early detection.
At Autoderm, our AI models undergo rigorous clinical validation to ensure accuracy across diverse populations, skin types and user testing. Many existing AI models have been trained on limited datasets and the “wrong” kind of images, like clinical images taken with a high resolution camera, not a smartphone camera, and may not adequately represent all demographics.
AI as we know it today is not 100% perfect. For example, when we use any of the large language models (LLM), like OpenAI, we can see that it is about 80% correct and human interpretation is needed, but it is still very good at achieving the desired end result and sometimes it helps improve the end result. Clear user instructions and expectations should be presented to the end user. Personal data is not collected, only the image is used for analysis, it can be chosen to be discarded or be collected for further use to train future AI models.
It has also been made clear that AI should complement, not replace, the expertise of healthcare professionals, to ensure that definitive diagnosis and treatment plans involve human oversight. However, it is very good at triaging a patient to the right level of healthcare at the right time.
During our partnership with myGP, as of May 2024 over 370 000 API calls have been done. Since we are a Software as a Medical Device (SaMD) we do Product Market Surveillance (PMS) reports every 6 months.
You can read the full report here.
Screening for diseases in healthcare is a valuable tool that, when applied appropriately, can significantly improve health outcomes, reduce healthcare costs, and enhance public health. However, it is crucial to balance the benefits with the potential risks and limitations. Effective screening programs should be evidence-based, targeting populations where the benefits outweigh the harms, and should be accompanied by clear guidelines and patient education to ensure informed decision-making.
It is estimated that 20% of GP visits are skin related. Autoderm did roughly 18 000 screenings in May 2024 and from the data, about 50% could have been triaged with further information retrieval for self care with or without over the counter medication from a pharmacy.
NHS documentation shows that average cost to the NHS for a 9-minute GP face-to-face consultation was estimated to be £42 in 2021/22. That would equal a £4 536 000 (9000*12) saving to the NHS, if 50% of the screening were triaged for self care.
This is a complicated task to pinpoint, there are many studies that have tried to calculate the number, however because of the different stages of melanoma, there are different costs associated. According to our online research, the cost to treat a case of melanoma in the UK is estimated to be around £2 600 per case, these are figures from 2013 by the Department of Health.
Melanoma treatment costs are highly dependent on the stage at diagnosis, with Stage IV cases costing over 25 times more than early Stage I cases. Costs are particularly high in the first year after diagnosis for advanced stage disease. Promoting primary prevention and early detection of melanoma could lead to substantial cost savings.
So in summary, the available evidence indicates the cost to treat a single case of malignant melanoma in the UK is approximately £2 600 on average, although the total costs can vary depending on the severity and treatment required.
Patients diagnosed with early stage melanoma (Stage I or II) have much lower treatment costs compared to those diagnosed at later stages. One study found the cost of managing a Stage IV case (€122 985) was over 25 times more expensive than a Stage IA case (€4 269).
Another study estimated individual patient costs ranging from €1 689 for Stage I to €88 268 for Stage IV. The largest cost differences were between Stage IA and IB-IIA, and between Stages III and IV.
In the first year after diagnosis, hospitalization costs are 8-16 times higher for advanced stage melanoma compared to early stage. The costs of outpatient services and inpatient drugs also decrease gradually as the stage at diagnosis is earlier.
The higher expenditure associated with more advanced stages is mainly driven by increased inpatient drug use and more intensive treatment required.
Overall, the average direct costs related to melanoma are highest in the first year after diagnosis (€2 903) and then decrease over time
Now that we know the costs, we can calculate the yearly cost savings for the NHS detecting 79 melanomas per month using Autoderm. We need to consider the cost of treating each melanoma and the number of screenings performed.
We will be conservative in our calculations, because we do not know what stages the melanomas were at. We will assume the average cost to be £2 600 per case. The total cost of treating the 79 melanomas detected in May would be:
The cost of performing 18 000 screenings can be estimated at £18 000 assuming £1 per screening.
£205 400*12 (treatment cost) - £18 000*12 (screening cost) = £2 248 800
The yearly cost savings for the NHS due to the detection of 79 malignant melanomas using AI in dermatology screening would be approximately £2 248 800.
Total cost savings considering 50% are triaged to self care is £4 536 000.
£2 248 800 + £4 536 000 = £6 784 800.
If we had screened the whole UK population (100%) it would theoretically be a £135 696 000 savings to the NHS.
AI mass screenings show great potential when there are not enough dermatologists to go around. More data needs to be collected and more research done over time. But 2024 could be a good year to start following this data on a macro level.