Empowering clinicians to better solve problems – The AEGLE Approach

Author: John Chang (Director/Consultant at Croydon Health Services NHS Trust)


The global population is improving in health, leading to better standards of living. As the population aged, chronic diseases such as type 2 diabetes becomes more prevalent, leading to increased morbidity, as well as health care costs, adding burden to its healthcare cost.

Type 2 diabetes is thus a scourge of developed countries and an increasing issue in developing countries. At present, there are over 195 million people affected worldwide, and this figure is expected to rise to over 330 million by 2030. Healthcare, therefore, needs to be aware of this tsunami of patients heading towards health needs, both developing a strategy for management as well as empowering the patient to better care for themselves. Health regulators also need to improve research and gain a better understanding of the disease through research into causation as well as complications; as it is only by improving a better understanding of the disease, complications and medical therapy, can we then look forward to improving the management of such a chronic disease, and better improve the quality of life for the patient as well as reducing the disease burden non-the wider health economy.

Mission statement:

To deliver high-performance computing infrastructure to the masses such that harnessing Big Data analysis readily available for all interested stakeholders, thereby increasing the capabilities of interested stakeholders to help tackle and defeat this scourge.

As part of a consortium of partners dedicated towards developing an IT solution to aid healthcare, we aim to develop an infrastructure that allows for the fostering of a community spirit and collaboration of interested players in a secure web-based environment which has analytical functions for processing data to better gain an understanding of the disease process, leading to potential solutions to the issues raised.

What is AEGLE?

AEGLE is a Big Data Analytics platform designed to handle Big Data datasets, in an encrypted environment, allowing analytics and sharing of data and results with authorised end users. It is unique in having both a dataset management function, as well as analytics package that users can utilise (the AEGLE tools library) if they do not want to use their own analytics programmes.

How does this apply to Type 2 Diabetes?

Our team has clinical partners (in this case diabetologists) as well as statisticians and IT software developers. We are all pulling together to utilise predefined datasets that have already been collected as part of routine care at a number of sites, and in whom we have obtained ethical and data custodian approval for use in research in developing software programme for helping to analyse, visualise and then apply the results of these findings to support improving the care of patients with type 2 diabetes.

At present we have 3 datasets, covering over 25,000 patients from which we are developing the analytical packages. To date, we have developed modelling for such common complications as visual impairment, renal impairment, vascular impairment and ultimately death in terms of survival curves and statistical analytics of relevance for the common known issues for variables known to have linkage to complications to type 2 diabetes (e.g. lipid profiles, HBA1C levels, ethnicity to name but a few).

However, by enhancing the use of heatmaps, then it is possible to see unexpected associations that could presage an unknown causation, thus allowing hypothesis to be generated and enable other avenues to be explored for managing intervention in the future. This is of relevance for researchers both within the academic field, as well as by pharma industry to hope to capture novel methods of addressing the issues within type 2 diabetes.

An illustrative heat map is shown below:

Heat map of relationships for type 2 diabetes

(In this case, RRT is illustrative of renal impairment; blindness is illustrative for visual impairment)

It is not just using heat maps on their own: AEGLE also has the functional ability to analyse the data into population pyramids: this has the advantage for both health care providers, as well as researchers keen to identify at-risk population, and affected cases so that the planning of intervention and trial design into the feasibility of clinical  trials can be undertaken quickly and efficiently: this helps reduce costs for trial set up, as well as aid healthcare providers in predicting future health needs. This is best illustrated by the example seen below:

Population pyramid for type 2 diabetes and limb amputations in relation to HbA1c levels

The high and low markers are for high and low HB A1C levels whilst the high_amps and low_amps are for amputations. The age cohorts (x axis) are at 10-year intervals (y axis are number of patients)

For example, a clinical trial aiming to achieve the reduction in amputation effects would target the 60 to 80 age band as best return for effects of the intervention. For health planners aiming to manage complications, then they would need to plan ahead for a cohort of amputees coming on stream at around 60 years of age onwards.

What about the patient in all of this?

The beauty of AEGLE is that the analytics also includes an interactive predictive modelling. This enables patient data to be analysed on a personalised level, and to predict complications arising in set time frames from diagnosis. By including adjustable parameters, that leads to the recalculation of risk prediction, the clinician can utilise this during consultation with the patient and see the effects of changes in the survival curve. An example of the personalised prediction modelling is shown below.

Please note blindness is used for simplification purposes – the data is for visual impairment, inclusive of blindness as well as progressive diabetic retinopathy. Note there are three curves drawn – this reflects the effects on risk of visual impairment with high, average and low HDL lipid levels.

Predictive modelling of visual impairment

Furthermore, the development of a meta-analysis of pharmacological agents available, we can then refine the various combinations of both old and new anti-glycaemic agents available to identify the combination which best suits each individual patients – development of personalised healthcare, leading to reductions in hypoglycemic episodes as well as be vigilant for the complications that may be seen with the newer drug therapies.

Is this all?

I am excited that these are just illustrative examples of the potential of AEGLE. The challenges facing the project in dealing with type 2 diabetes is that single large datasets are not readily accessible. However, by using filters, then it may be possible to analyse national datasets but filter out subsets for type 2 diabetes, or for that matter any other disease entity that one is interested in and then analysing that data.

As this is a Web based infrastructure, being ‘sharable’ to authorised end users, ultra-rare disease registry on an international level could lead to collaboration and collation of large datasets of rare illnesses at an international level, never seen before, that may create the touch paper to enable health care management and health care research to fully cross national borders and lead to improved understanding of diseases and lead pharma as well as academics to study and define better treatments for citizens in the future –  if we are already seeing examples of this in such concepts as ‘Patient knows best’ or the international Vermont – Oxford database for neonatology, then why cannot  we  foster this to truly take healthcare into a global powerhouse for our children?