Medically Unexplained Symptoms - Digital tools and identification system
How can we identify MUS patients with medically unexplained symptoms (MUS) using primary and/or secondary patient information systems and assist in treating along a care pathway.

Ideas (Publish, Detailed Submission)

Innovation 'Elevator Pitch':
PredictX Machine Learning system has been successfully deployed to solve a similarly difficult problem of identifiying School Ready Children. This model is ideal for solving the problem of early identification of MUS.
Overview of Innovation:
The Complex Symptoms Service (CSS) will focus on the problem of MUS following on from the success of the BMUSS Phase 1 pilot. A system that automatically segments and identifies patients with high probability of MUS at an early stage will significantly increase the savings realised of the CSS as well as improve the well-being of patients. The complexity of MUS makes it difficult to identify manually especially at early stage.

Machine Learning (ML) has advanced to a stage that the challenging task of identifying Medically Unexplained Symptoms (MUS) automatically is now possible.

Our ML model can infer & predict results even with large variety of causal factors and disparate source data and where a single data point in isolation is a poor indicator. In the case of some long term conditions such as diabetes, a single source such as blood sugar levels can be a confirmatory indicator. MUS suffers from the fact that one episode whether via GP assessment or hospital admission does not indicate MUS on its own. Instead, it takes multiple events and episodes examined together in a holistic model that indicates the probability of MUS.

If a system could identify MUS with minimal historical data, early intervention avoids much more serious and costly problems later on. ML models such as the one PredictX developed for Essex County Council (see attachments) is ideal for working on a large range of data sets and inferring relationships not obvious to human observers.

An accurate MUS identification system must be able to meet several requirements.
  1. Robustness of the model. The model must adapt to differences in local data availability and preserve the accuracy by using proxy data if core data is missing.
  2. Leverage a wide range of data sources including social care and health
  3. Capability to read, understand, and classify free text information in any format including PDF as the source data management is often a challenge.
The PredictX Risk model deals with all of the issues. ML models are utilised not just for the forecasting models but also to facilitate understanding of free text utilising advances in Natural Language Processing (NLP).

The model has further advantage of having real-world analogues & explanability.  There could be a direct relationship from the features to the phenotypes greatly aiding the health professionals responsible for the interventions.

The model has effectively been deployed in a similar context at Essex CC.https://youtu.be/3J2S82vjb8U
Stage of Development:
Market ready and adopted - Fully proven, commercially deployable, market ready and already adopted in some areas (in a different region or sector)
WMAHSN priorities and themes addressed: 
Mental Health: recovery, crisis and prevention / Wellness and prevention of illness / Wealth creation / Digital health / Innovation and adoption / Person centred care
Benefit to NHS:
The direction of Travel for the NHS is the attention to early identification and treatment of disease and chronic conditions. As MUS is so difficult to identify at an early stage, patients can ricochet around the system for years before the correct diagnosis is made. MUS is a pervasive condition which has significant implications for cost and resource consumption in the system. MUS accounts for 11 % of the NHS budget for adults and comprises up to 50% of new hospital outpatient visits.

This figure is 11% of total NHS spend. The related cost to the economy in terms of sickness absences and restrictions to quality of life for people with MUS accounts for over £14 billion per annum to the UK economy.

The Complex Symptoms Service is  estimated to save in excess of £750,000 per year, based on data from local pilot services and NHS research studies. This estimate is based on current patient identification methods which logically will identify only a subset of patients at an early stage. A system that automatically identifies MUS patients at an early stage will increase the pool of patients where the interventions have greater impactability.

The value to patients who can benefit from early intervention will be measured in better quality of life, less days in hospital and improved outcomes.
 
Initial Review Rating
4.80 (2 ratings)
Benefit to WM population:
The West Midlands population is uniquely positioned to benefit from the System. Instead of introducing a new system before the delivery mechanism has been tested and rolled-out. The West Midlands have already mobilised dedicated services for intervening in MUS cases via the  Birmingham Medically Unexplained Symptoms Service (BMUSS), phase 1 and 2.

The service successfully implemented evidence–based practice for MUS, targeting mild, moderate and severe MUS based on a stepped care model. As the infrastructure for impacting patients with MUS have been set-up, tested and proven to be effective, the stage is set for enlarging the funnel and treating more patients with MUS at an early stage.
The system will be effective at identifying patients with MUS at an early stage. Identification alone will be ineffective without the capacity for impactful intervention. The West Midlands deployment will combine both the capacity for early identification with the capacity for intervention.

The close and strategic working relationship established over many years between PredictX and the Midlands and Lancashire CSU will help to ensure that projects are executed per agreed scope and delivered on time and on budget.
 
Current and planned activity: 
PredictX works with the NHS at several levels currently. We work directly with Provider Trusts such as UCLH supplying them with advanced analytic solutions. We also work with CCGs and STPs focusing on advanced analytics that need to integrate social care and health information. We also have a strategic partnership with the West Midlands CSU working to serve their CCG members as well as investing in research and development efforts to apply Artificial Intelligence and Machine Learning techniques to address challenges the sector faces. This partnership has recently been codified with the launch of the Midlands and Lancashire Innovation Partnership - a partnership of the CSU, private sector and academia working together to deliver advanced, AI-enabled solutions.
What is the intellectual property status of your innovation?:
The IPR of our innovation is held 100% by PredictX. We are happy to co-invent new IPR and contribute to the sector.
Return on Investment (£ Value): 
high
Return on Investment (Timescale): 
6-12 mon
Ease of scalability: 
Simple
Read more
Hide details
Keesup Choe 25/05/2018 - 15:43 Publish Login or Register to post comments
0
0
Votes
-99999
Innovation 'Elevator Pitch':
UK developed patient focused Population Health solution pulls cross sector data (i.e. GP, Acute, Mental health, etc) and performs analytics and intelligent cohort finding that can identify patients with reported unexplained symptoms.  
Overview of Innovation:
ArtemusICS supports a health system to analyse population data and identify specific gaps, health inequalities and unwarranted variations in its population’s health. ArtemusiCS secure web accessed hosted system provides a wide range of tools, dashboards, selection and filtration options to identify and create patient cohorts. ArtemusICS can then further identify costs, utilisation and gaps in care.

Collecting data from Acute, GP, Community, Mental, Social and Ambulance, ArtemusICS provides commissioners and providers with a tool to identify cohorts of patients, and understand the integrated care needs of a population right down to an individual. It contains analytical tools to allow for LTC risk, EOL risk, Frailty, Social isolation, etc to be collated and presented in one picture. It also allows for evaluation of interventions (i.e outcomes) and identification of RoI.

It has a unique attributes based architecture, where all data items are linked to one or more “attributes”, based on their coding (i.e. Snomed, Read, ICD10, etc). Inherently future proof and expandable, it allows us to utilise Machine Learning and AI techniques to drill through large data sets to find correlations and trends in specified data types, such as the symptoms presented in the MUS cohorts.

ArtemusICS operates on three distinct levels:
  • Overview Dashboards: provide whole-population analytics with facilities to filter down to key patient cohorts. A range of Dashboards provides commissioners and clinicians pertinent population count, event counts, service and usage counts and costs; with the facility to then drill through to cohort and patient specific views.
  • Cohort Manager enables the user to review sub-lists of patients, caseload management groups or individual patients against a range of detailed views including conditions, events, measures, timelines and management lists. Venn diagrams offering cohort comparisons to enable care planners and clinicians to pin-point a specific cohort with a specific set of attributes for intervention and care management.
  • Patient view provides a range of views specific to patient conditions, medications, diagnosis, events and measures, supporting the clinical user to improve diagnosis and plan care.
In addition, our remote monitoring technology can also collect surveys such as PHQ,GAD7, HADs etc
Stage of Development:
Market ready and adopted - Fully proven, commercially deployable, market ready and already adopted in some areas (in a different region or sector)
WMAHSN priorities and themes addressed: 
Mental Health: recovery, crisis and prevention / Long term conditions: a whole system, person-centred approach / Advanced diagnostics, genomics and precision medicine / Wellness and prevention of illness / Wealth creation / Clinical trials and evidence / Digital health / Innovation and adoption / Patient and medicines safety / Person centred care
Benefit to NHS:
ArtemusICS currently identifies all patients with a mental health condition, including those with unexplained medical symptoms; this achieved by the use of Read, SNOMED and ICD10 codes.   Through the development of artificial intelligence and the use of inferential statistics, we would be able to provide rich case finding and provide greater insight in the nature of the specific problem associated with the patients.
Repeat admissions, vague multiple comorbidities and results of screening are all available in the ArtemusICS database and could inform the AI development. Using the ArtemusICS impactability management module, commissioners will be able to assess adherence to care pathways and also monitor the impact of the given interventions.
As patients with medically unexplained symptoms are often seen in multiple care and social settings, ArtemusICS would create and provide a clear picture of all the cross sector activities associated with the patients in order to develop bespoke interventions.
This affords both primary (and potentially) secondary care clinicians the opportunity to introduce patients to a specific MUS care pathway that will deal with the specific requirements of MUS in a timely manner that will prevent the condition becoming chronic.  Effective stratification of patients could aid in a reduction in need for acute secondary care outpatients, reducing unnecessary healthcare costs and freeing up resources.
Initial Review Rating
4.20 (1 ratings)
Benefit to WM population:
As for the health system above
Current and planned activity: 
Docobo are continuously adding additional functionality to its solutions, usually to meet the requests from our NHS customers or our own staff
The uniqueness of ArtemusICS is that Docobo fully owns and controls all of the product. It is not a 3rd party product that we sell, like many others, so we can easily add functionality and are not at the mercy of the overseas priorities of a 3rd party product
What is the intellectual property status of your innovation?:
Fully owned by Docobo
Return on Investment (£ Value): 
high
Return on Investment (Timescale): 
0-6 mon
Ease of scalability: 
3
Read more
Hide details
Adrian 23/05/2018 - 18:42 Publish Login or Register to post comments
6
2
Votes
-99999

Recent Activities

Machine Learning system for Early Automatic Identification of MUS has been moved from Initial Submission to Publish 20 weeks 6 days ago
JR voted on ArtemusICS 21 weeks 6 days ago
Sarah Appleby voted on ArtemusICS 21 weeks 6 days ago
ArtemusICS has been moved from Initial Submission to Publish 24 weeks 5 days ago

Active Campaigns

Top Contributors