i5 Coronavirus Health Risk Calculator

The specific problem we are solving is the inability to really know the levels of risk faced by individuals should they ultimately contract Covid-19. Knowing those risk levels can provide individuals, clinicians, hospital authorities and public authorities with critical data that will inform significantly decision making.

i5 Coronavirus Health Risk Calculator ("Calculator") establishes whether a person would fall into one of four health risk categories: Low, Medium, High and Very High. The Calculator is trained on the historical medical records provided by the NHS to selected and trusted organisations like i5 – where we have expertise in Artificial Intelligence (AI) as applied to health data.

If scaled globally, it can support the saving of hundreds of millions of lives by providing sufficient knowledge to individuals, clinicians, hospital authorities and public authorities to support informed decisions.

The specific problem we are solving is the inability to really know the levels of risk faced by individuals should they ultimately contract Covid-19. Knowing those risk levels can provide individuals, clinicians, hospital authorities and public authorities with critical data that will inform significantly decision making.

The scale of the problem is worldwide. Without sufficient and more easily available information of this sort, there will be many individuals who risk making wrong lifestyle and self-care choices, overburdened clinicians particularly those treating populations which are under-resourced from a financial and facilities perspective, management challenges for hospitals that need to plan in the short and medium term for beds and ICUs and governments making less than optimal judgements on escalation and de-escalation of social distancing rules.

With the prospect of Covid-19 wave extending further around the world, the partogen mutating, another wave following and re-occurrences happening each year, there is even greater importance attached to such provision of information.

The totals of those that could become infected by Covid-19 are potentially enormous. They range from between 33m and 75m in the USA to between 1m to 2.3m in India (Caution is necessary in respect of all calculations).

The Calculator establishes whether a person falls into one of four health risk categories: Low, Medium, High and Very High. It is trained on the historical medical records provided by the NHS to trusted AI organisations like i5. The initial records were of sufferers from influenza and earlier strains of Coronavirus. The results helped train the AI Neural Network that underlies the Calculator (see academic paper https://www.i5analytics.com/HealthRiskInPatientsWithCoronavirus.pdf).

The Calculator directly available on the internet (https://www.coronavirusrisk.org/risk-calculator/ ) is composed of a series of questions answerable through choosing from a dropdown list of ten medical condition categories. The more ‘industrial scale’ access that allows multi-million records to be analysed in real time (and is that used by the NHS) is available through an API on https://www.i5analytics.com/freehealthcare-ai . 

The best performing Neural Network had 255 inputs (features), with 244 diagnosis and 11 age band codes. This network was able to re-categorise 98.9% of patients in the training data set into four bands of outcomes: low, medium, severe and died. The accuracy of re-categorisation of new patients into the three bands was 80.1% with an error rate of 19.9%. Levels of prediction will be improved once new data on SARS-CoV-2 is available.

The on-line version of the Calculator ( https://www.coronavirusrisk.org/risk-calculator/ ) is capable of serving all those persons, throughout the world, with access to the internet (personal computer or mobile phone) and with a knowledge of their medical history. Thus, individuals can check their own risk levels or help family members establish theirs. Doctors, in hospitals or in the community, can establish the state of affairs of their patients. 

This individual use can impact in the following ways:

• Deciding the levels of quarantine needed to apply to individuals in order to minimise the risks of infection
• Supports clinicians, whether face to face with patients or remote, in deciding which pathways may be the most appropriate for patients  

The ‘industrial strength’ API version of the Calculator, available at (https://www.i5analytics.com/freehealthcare-ai ), can be applied to the following ends for the benefit of individuals:

• Clinicians and Hospital Management, for which the calculation for admitted patients with Covid-19 can help prioritising care and improve bed management
• For hospitals, helping establish the immediate or medium term demands for ICU facilities
• Public Health authorities and Governments in planning for disease control and levels of quarantine -- critical in making decisions escalation and de-escalation per locality

The specific problem we are solving is the inability to really know the levels of risk faced by individuals should they ultimately contract Covid-19. Knowing those risk levels can provide individuals, clinicians, hospital authorities and public authorities with critical data that will inform significantly decision making.

The Calculator addresses this challenge by establishing, through using AI to analyse medical records, whether a person would fall into one of four health risk categories: Low, Medium, High and Very High.

The target population comprises individuals throughout the world (with direct or indirect access to the internet), clinicians, hospitals and government authorities.

The innovative approach has developed out of existing i5 methodology within i5 Diagnosis Stratification (i5 DST) whereby a Web API accepts patient’s medical history coded in ICD-10 and JSON format. This information is used by i5 AI inference engine and analysed to provide a probability indicator of the patient having an undiagnosed Long Term Condition or may develop the condition in the next 12 months. 

The following example academic papers explain the science underlying i5 DST:

https://www.researchgate.net/publication/326735666_Heart_Failure_Screening_Programmes_for_Population_Health_Management_Supported_by_Artificial_Neural_Networks

https://www.researchgate.net/publication/325335156_Artificial_Neural_Networks_for_Population_Health_Management_to_support_Atrial_Fibrillation_Screening_Programmes

The Calculator, with (like i5 DST) AI Neural Network  algorithms trained on years of NHS data, establishes whether a person would fall into one of four health risk categories in the event the person contracts Covid-19: Low, Medium, High and Very High.

As to comparison with other solutions: There are tests becoming available to establish whether a person has already contracted Covid-19. This might have manifested itself in one way or another or remained asymptomatic for some people (ie carrying the disease but having no symptoms). Decisions based on the tests can be made accordingly. Even without them, clinicians might have enough awareness of the threat to patients (particularly the elderly) to give precautionary advice. However, the mass of the populations will not be automatically within any surveillance system. The Calculator changes all that as only the medical records (centralised in the UK) are required.  

Also innovative is the existence of alternative distribution systems: 1. through an on-line device and 2. through an API.

The trained AI algorithms that underpin the Calculator is described in the academic paper available at:

https://www.i5analytics.com/HealthRiskInPatientsWithCoronavirus.pdf

Training has involved:

WP1 - Collect/categorise data

•             Three year medical records with over 200 fields including 21 diagnosis and 21 procedures and diagnosis present prior to Covid-19

•             Data on related lengths of stay, ICU attendance and needs eg ventilators

•             Categorise into balanced cohorts (Low, Medium, High and Very High) for optimal separation between Sensitivity and Specificity

WP2 - Principal Component Analysis (PCA)

•             PCA carried out on diagnosis codes to ensure only conditions used are pertinent for prediction

WP3 -- Data split

•             Create separate data sets for: AI training, monitoring of overfitting, performance testing

•             Balancing sets to contain patients with various outcomes and avoid over-representation

WP4 -- Coding and Training

•             Short term -- risk level prediction; Medium/long term -- Healthcare Planning 

•             Train various AI models and topologies to evaluate optimum combinations 

WP5 -- Test beds

Using the Technology

On line Calculator (https://www.coronavirusrisk.org/risk-calculator/ )

After selecting age, choice is of pre-existing conditions. Pressing an “Assess My Risk” button will then show one of four categories: Low, Medium, High and Very High.

API (https://www.i5analytics.com/free-healthcare-ai)

The i5 DST AI-Web API accepts medical history coded in ICD-10 and JSON format. This information is used by i5 AI inference engine and analysed to probability indicator of the patient having a specific Covid-1 risk level. The i5 server has the capacity to analyse millions of records in real time.

The Calculator is a less common technology. An academic paper on the methodology underlying it is available at: https://www.i5analytics.com/HealthRiskInPatientsWithCoronavirus.pdf 

The paper is co-written Professor Dr Harald Braun (Professor of Health Economics and Artificial Intelligence at University of Chester; Director i5) Prof David Patterson (Emeritus Professor of Cardiovascular Medicine at University College London) Dr Aoife Molloy (Imperial College London and specialist in Infectious Diseases) Keith Davies (BSc Economics, Attorney-at-Law, Managing Director i5).

The recently created on-line Calculator system,at the date of this submission, has already been used over 200,000 times. A key organisation within the NHS, the key Commissioning Support Unit in London, NEL (North East London) CSU, has used the API system to evaluate the risk situation of 4,100,000 Londoners (after de-selecting potential low risk individuals)

The Calculator has been included in FutureNHS – a Collaboration Platform which is a digital cloud-based system which supports people from across the health and social care sector. FutureNHS is helping to deliver the NHS Long Term Plan. The Calculator has been used by top executives and experts within the NHS and drawn considerable praise from within and outside the UK by users and during international conferences.

We would expect our solution to have a positive impact on the problem across all the categories aimed at - giving strong societal and humanitarian benefits.

For each of the major targeted ‘constituencies’ there are clear linkages between Activities, Outputs and Longer-term Outcomes - as follows: 

In the cases of an individuals using the on-line app, the activity of self-interrogation leads directly to categorisation as output; that categorisation can help the  individuals in becoming more sensitive about the dangers they face in being in the High and Very High risk categories and adjusting their lifestyles accordingly (eg Social Distancing, treating more seriously the risk of Long Term Conditions that predominate in these two categories...). The longer-term outcomes from those adjustments, multiplied across a population, can be enormous in terms of health, longevity, healthcare costs and family cohesion.

Clinicians checking a patient’s records with the Calculator can be informed of a risk situation and/or be reinforced in their opinion on the optimum treatment. The longer-term outcome is ensuring proper care along a patient's pathway.

Hospital management, using the API version for their local population’s health records, can be better prepared, given the Calculator results for their localities, in terms of beds, length of stay, equipment of ICUs....Again, patient care and financial outcomes benefit in the longer-term. 

Government authorities can use the results from the API version to support decision making eg deciding on selective measures for populations in different localities in the face of a pandemic and being able to ramp up/ramp down measures accordingly – with significant longer-term effects countrywide.

A key example of the last category is the recent use of the Calculator in London by the NHS where the medical records of 4.1m people have been subject to analysis to identify for shielding those within the High Risk category. This analysis has been able to extend significantly the lists originally put forward by general practitioners.

Current number of people served (after six weeks): circa 4,250,000

Serving in one year: On the basis of just the countries noted above and the implementation by government authorities: 300,000,000

Serving in five years: 2 billion

We have developed a battery of other AI Neural Network tools which have a potentially global effect. In particular, there is i5 Diagnosis Stratification (i5 DST) which identifies individuals with, as yet, an undiagnosed Long Term Condition.

As with the Calculator, it uses WHO ICD coding and is therefore applicable in over 120 countries. i5 DST covers a number of the key Long Term Conditions (LTCs) and is firmly within the most forward thinking area of health care: Prevention. 

Actually realizing the potential of i5's tools on the scale envisaged is a major challenge for an SME like i5. Our preferred strategy is self funded growth but that is unrealistic given the ambition. We would seriously consider an investment from a major multinational in the IT and/or health field. That might, additionally, give our tools the 'footprint' for distribution it needs.

The barriers for i5 are principally a function of its size -being the smallest of SMEs. From this flows the following barriers:

- Availability of finance beyond that which is obtained through revenues and grants - which finance can fund recruitment of highly skilled data experts and marketing expertise

- The limited distribution network available to the organisation for what we believe are services that can be applied in over 120 countries around the world (the number of states that are members of the World Health Organisation and therefore use the ICD coding system)

- Linked with the above are limited marketing/publicity opportunities

It should be emphasised here that there are no technical barriers to advancement; our USP is being able to create unique tools for the health sector and, in doing so, being thought leaders not followers - with an ability to always be ahead of the game. What helps underlie that advantage is that i5 has the privilege of being trusted with access, each year, to the medical records of 56 million people in the UK. These are provided to us, under the strictest terms, by the NHS.

We plan to address these barriers as follows:

- Accessing finance 

We would be keen to bring in appropriate third party funding. By this we mean equity investment by a corporation that has complementary roles in the healthcare field. Alternatively, we would be interested in funds that specialised in the health technology field. 

That funding would enable us to invest in more data analysts and code writers as well as marketing expertise.

- Developing Distribution

Our target territories specifically the circa 120 members of the World Health Organisation that are users of the ICD coding system (integral to the algorithms for all our services). It is possible that a multinational healthcare corporation might have the appropriate 'footprint' already that we could benefit from. 

- Publicity opportunities 

These would include addressing international conferences, giving interviews and getting articles published. These functions would be undertaken by Professor Dr Harald Braun and Keith Davies.

Full time: 4

Contractors: 6

CORE TEAM

Prof Dr Harald Braun 

Harald’s is one of the leading experts in the UK on Artificial Intelligence as applied to Healthcare. He has provided solutions for NHS finance, commissioning, contracting, service transformation, risk stratification and patient pathways. Harald is the Professor for Health Economics and Artificial Intelligence at University of Chester.    

Specific deliverables include commissioning opportunities evaluation, the identification of patients not being optimally treated for LTCs and at risk, forecasting sufferers from LTCs, invoice validation, A&E forecasting, Urgent Care modelling, End of Life service analysis and risk stratification.

Keith Davies 

Keith has high level experience of business and governmental relations in the UK and across Europe, the Middle East, China, Australasia, the USA and Canada. Keith headed up European wide insurance groups and led the £1.36 billion Turnaround Programme at the NHS. He was interim COO at Allied Irish Banks before becoming co-founder and CEO of i5. Keith is on the Advisory Board of the government’s Healthcare UK.

Jeanetta McLean 

Jeanetta McLean initially provided data analysis for the insurance industry - following which she has worked extensively on NHS projects.

A recent view about this i5 team - from a senior NHS executive:

“……….more importantly, i5 health is our primary partner in NHS England for delivering complex data dashboards powered by their own AI………We spoke…….last week of getting more sophisticated business intelligence into the exec suite, during and post Covid.. That's exactly what i5 have done for us in NHS England”.

NHS Arden & GEM CSU - Business Intelligence partners; our algorithms are integrated into the CSU's servers.

NHS NEL CSU - Business Intelligence partners; The CSU has used the Calculator to analyse the medical records of 4.1 million Londoners.

NHS England - The relationship between i5 and the managing body of the NHS is best reflected a few days ago in a message of one of the top execs at NHS England to the chair of a national body: “……….more importantly, i5 Health is our primary partner in NHS England for delivering complex data dashboards powered by their own AI………We spoke…….last week of getting more sophisticated business intelligence into the exec suite, during and post Covid. That's exactly what i5 have done for us in NHS England”.

IBM Watson Health - i5 is in discussion with IBM on placing a number of i5 products - including the Calculator - on the IBM Cloud system.

Cerner - i5 is a partner in the Cerner i.s.h.med Member Program and is currently investigating with Cerner the potential application of i5 DST in Latin America.

i5 is a specialist health technology company providing Healthcare data analytics using Artificial Intelligence, Machine Learning and Neural Networks. i5 principal services are:

• Identifying, through preliminary diagnosis, of individuals who, if not targeted for screening in a timely manner, might suffer unnecessarily and have to undergo expensive secondary care treatment (i5 Diagnosis Stratification – i5 DST).
• Interrogating health data, provide evidence and candidates for out-of-hospital initiatives and thereby support clinicians in referring patients into other, more appropriate and often less 7 costly services (i5 Commissioning Opportunity – i5 COP and i5 Targeted Social Prescribing – i5 TSP).
• Calculating the healthcare needs of individuals and populations, support transformational changes to better align services to patients‟ needs and reduce costs. (i5 Commissioning Opportunity – i5 COP and Targeted Social Prescribing – i5 TSP).
• Interrogating secondary care data at either population or individual levels, predict outcomes like treatment waiting times, hospital lengths of stay (LoS), complications and readmissions (i5 Outcomes Prediction – i5 OP). 

After targeted marketing and with the support of the UK’s Department for International Trade and the Department of Health, i5 is developing business opportunities with organisations on four continents.  Prospective partnerships between i5 and major multinationals are in progress.  The contracts/partnerships referred above include: 

• Saudi Arabia  - Ministry of Health and their Business Intelligence partner, Lean Business Services
• UAE and North Africa - Mubadala owned organisations
• Latin America - Cerner
• UK - NHS Arden & GEM CSU, NHS NEL CSU, IBM Watson and Cerner 

Our prospective funding is as follows:

Grants

We have just won an award from a UK government health innovation fund and a European Union fund to support the further development of one of our key tools and are shortlisted by a different  UK government fund for the development of another. We have also now been invited, given our particular capabilities, onto a UK government funding framework open for the next six months.

Sale of Services

Selling the above services directly to UK national and international users eg government health authorities, hospital groups…. or through business partners such as those noted above

Raising investment capital

As an SME, to really take advantage of the world wide opportunities open to i5 services, we require external funding.

We are applying to Solve because our ambition, to provide the world with a means to calculate the levels of risk faced by individuals, faces significant barriers we believe Solve can help us overcome. As an SME, the types of assistance we seek from Solve are:

- Accessing finance to fund a project that has a significant pro bono element

 We do not discount monetising the Calculator at some point but do not see it as an immediate priority given the urgency of the Coronavirus situation. A specific grant for the solution is probably the key alternative. However, a philanthropic corporation identified by/through Solve that might consider being also an equity investor/partner given the financial benefits other i5 AI Neural Network tools could engender.

- Identifying distribution network opportunities 

Targeted for distribution would be specifically the 120 members of the World Health Organisation that are users of the ICD coding system (integral to our algorithms). 

- Providing us with publicity opportunities 

These would include addressing international conferences on the pandemic, giving interviews and getting articles published. 

- Mentoring

General mentoring from Solve as we go forward would be appreciated - particularly from those that have had a similar journey.   

Actually realising the potential of the Calculator - as well as other AI tools of i5 - on the scales envisaged is a major challenge for an SME like i5. Self funded growth is unrealistic given our international ambitions. We would seriously consider partnership with major organisations or multinationals in the IT and/or health field that would provide input/assistance principally in the following categories:

- Finance 

- Joint technical development

- Distribution on a substantial level

- Marketing - including publicity opportunities

Inter-American Development Bank

It is the IDB Group's priority commitment on COVID-19 in respect of the Latin American and Caribbean countries that initially chimes with us. We are currently in discussion with Cerner in Latin America on both COVID-19 and the identification of as yet undiagnosed sufferers from Long Term Conditions (LTCs). Funding in respect of COVID-19 is, of course, of great interest to i5 as is the prospect of some sort of synergy between the three organisations on medical conditions challenges in the region.

Alfred P. Sloan Foundation

The Foundation's Digital Technology commitment and its interest specifically in collecting, manipulating, storing, analysing and transmitting data fits well with what i5 is focused on: the development of Artificial Intelligence Neural Networks for healthcare. We think there are mutual benefits that can be achieved by some form of partnering.

The Patrick J. McGovern Foundation      

Please see our application below for the 'AI for Humanity Prize'.            Besides that which we have set out there, we believe i5 could learn a great deal from the The Patrick J. McGovern Foundation and even find together further ways of enabling artificial intelligence for social good.         

ELIGIBILITY

We believe we are eligible for consideration for The Patrick J. McGovern Foundation AI for Humanity Prize because fundamental to our activities, in this Coronavirus as well as other fields, is a desire to use our relatively rare skill sets for the improvement of the lives of individuals - particularly in the more economically and socially challenged parts of the world. 

Central to those skill sets are are a a deep understanding and long experience of creating highly advanced solutions which combine healthcare data and Artificial Intelligence (AI) to provide preliminary diagnosis, outcome prediction, service recommendations and health economy planning for social good. i5 AI systems are trained on several billions of healthcare records collected over many years and spanning across many settings by the one of the world’s largest and most comprehensive healthcare systems – the National Health Service (NHS) in the United Kingdom. i5 AI systems offer the opportunity to share the knowledge and experience extracted from these vast amounts of data with health economies around the world - particularly those most challenged.

It is the NHS which has very recently provided a strong endorsement of our AI Neural Network services. Mr Noel Gordon is the chair of NHS Digital (provider of health data) and of Healthcare UK (government body promoting UK health exports) as well as being a member of the Board of NHS England. Mr Gordon wrote to the head of a national health care organisation as follows:

“……….i5 is our primary partner in NHS England for delivering complex data dashboards powered by their own AI………We spoke…….last week of getting more sophisticated business intelligence into the exec suite, during and post Covid.. That's exactly what i5 have done for us in NHS England” 

We cite below evidence of our abilities in training AI Neural Networks in the healthcare environment and our philosophy in two specific areas: Diagnosis Stratification and the Coronavirus risk assessment.

i5 Diagnosis Stratification (i5 DST)

i5 DST is an algorithm-based Artificial Intelligence clinical evaluation tool that can identify undiagnosed patients with a high likelihood of having a Long Term Condition (LTC). DST facilitates proactive and preventative patient management that can reduce the threat, both health and financial, to individuals and their families. We have ensured that the system is made available free to health systems that would not, for size and funding reasons, otherwise be able to use it.

The following example academic papers explain the science underlying i5 DST:

https://www.researchgate.net/publication/326735666_Heart_Failure_Screening_Programmes_for_Population_Health_Management_Supported_by_Artificial_Neural_Networks

https://www.researchgate.net/publication/325335156_Artificial_Neural_Networks_for_Population_Health_Management_to_support_Atrial_Fibrillation_Screening_Programmes

Coronavirus Health Risk Calculator ("Calculator")

As set out with more detail within the submission to MIT 2020 Global Challenges (Health Security & Pandemics category), the specific problem we are seeking to solve is the inability to really know the levels of risk faced by individuals should they ultimately contract Covid-19. Knowing those risk levels can provide individuals, clinicians, hospital authorities and public authorities with critical data that will inform significantly life-saving decision making.

The AI science underlying the Calculator is set out in the following academic paper:

https://www.researchgate.net/publication/340405063_Predicting_Health_Risk_in_Patients_with_Coronavirus_or_Influenza_using_Artificial_Intelligence

The on-line version of the Calculator (www.coronavirusrisk.org ), available free around the world to individuals, is already translated into 11 languages in the hope of touching some of the 2.7 billion people speaking them. So far it has utilised over 200,000 times. The API version of the Calculator has been used by the NHS, also free of charge, to analyse the health records of 4.1 million citizens of London to identify individuals most at risk should they be infected with the disease.

The scale of the problem is worldwide. Without sufficient and easily available information of this sort, there will be, for example, many individuals who risk making wrong lifestyle and self-care choices, clinicians particularly those treating populations which are under-resourced  from a financial and facilities perspective, hospitals that need to plan in the short and medium term for beds and ICUs and governments that must make judgements on escalation and de-escalation of social distancing rules.

With the prospect of Covid-19 wave extending further around the world, the underlying Coronavirus mutating, another wave following and re-occurrences happening each year, there is even greater attached to such provision of information.

The totals of those that could become infected by Covid-19 alone are potentially enormous. They range from between 33m and 75m in the USA to between 1m to 2.3m in India - though caution is necessary in respect of all calculations.

USING THE PRIZE

It is our intention to use the AI for Humanity Prize to advance our Coronavirus solution in the following ways:

Further Upgrading the Calculator 

The AI Neural Network science underlying the Calculator was based on the analysis of medical records of patients who had suffered from earlier strains of Coronavirus or from Influenza - a close relation. The novel Coronavirus, SARS-CoV-2, is the pathogen responsible for the infectious respiratory disease COVID-19. As the novel Coronavirus keeps spreading around the world, it will probably keep changing  but it is difficult for scientists to predict how those virus changes might affect what happens. 

The priority for i5 is to investigate the historical medical health records of the sufferers from COVID-19 made available to it by the NHS. There could be well over 300,000 such patients by July 2020. The analysis will inform the upgrading of the AI Neural Network algorithms of the Calculator. However, as the novel Coronavirus, SARS-CoV-2, the pathogen responsible for COVID-19. keeps spreading around the world, it will almost certainly keep changing unpredictably. 

In the light of this, the likelihood is of a continuing need to upgrade the Calculator on a regular basis - as more and more patient data related to the changing pathogen becomes available. Right now, we can benefit from a one-off grants from the UK government grant and the European Union.  

We will use part of the AI for Humanity Prize to fund the upgrading of the Calculator from September 2020 onwards. We estimate our current upgrading exercise is costing between $100,000 to $125,000 and expect a similar range for future annual upgrading.

Improving Accessibility to the Service

We intend that the AI for Humanity Prize will be used to improve access to the service by:

• Making the www.coronavirusrisk.org website more user friendly
• Developing a mobile phone app 

Marketing and Distribution

We shall allocate part of the award for marketing and distribution of the Calculator to priority areas in the world. This exercise will entail writing and distributing articles for popular as well as for learned publications and for addressing international conferences. There shall be a targeted campaign to influence governments in particular.