The Predictive Analytics Journey

So what is the foundation for innovation and transformation?

Our leaders need to hold the mirror up and ask "is the information I am getting helping me make decisions and take effective action at the earliest point?" Then ask "can I get more from my data to drive better outcomes?".

The basic underpinning of all analytics is: Data - Insight - Action.

The astute and visionary leaders have realized that as organizations have become more complex, centralized, globalized and digitalized – they have also become data rich but insight poor and siloed. Some have lost sight of the basics in service delivery. There are so many data sources being captured and stored in multiple locations – a blizzard of data swirling around with untapped value. Data; one of the most valuable assets in any organization is being lost. It is not uncommon for healthcare trusts to have hundreds of systems operating that carry huge latent value.

Governance over efficiency, effectiveness and legitimacy is sub-optimal. In many cases mediocre at best. If you use a data driven approach and ‘turn some stones’ – opportunities will fall at your feet quite readily.

And this is where disrupters in public sector emerge. Our innovators are treating data as a critical strategic imperative – using it to shine a light on every bit of the business. They are ‘tin opening’ the organization’s culture and delivery using a modern analytics blueprint. The business is taking the lead; democratizing and industrializing every bit from the board to the front-line; creating a line-of-sight or golden thread. People, process and system is laid bare for scrutiny and improvement. Most of this scrutiny is in the hands of the ‘front-line’ – empowering people closest to decision and action. Applying these approaches shifts assurance, productivity, quality and overall outcomes and ultimately the culture for delivery. This is not about producing dashboards with KPIs – this is about providing actionable intelligence to drive business. Different thinking.

Once you have built your analytics platform and connected your data, innovators start to think about moving up the Business Intelligence Curve. The business intelligence curve shows the relationship between how more advanced analytical techniques yield a greater competitive advantage. Innovators are moving insight from raw data through descriptive, diagnostic, predictive to finally a prescriptive approach in a rapid roadmap approach. Traditional waterfall approaches don’t cut it – an agile approach delivers value back to the business in weeks; not years.

Moving up the curve and embarking on the predictive analytics journey

Once a preserve of bearded boffins in darkened rooms, predictive analytics technology is now readily accessible and affordable to the public sector. Technology has been moving so fast in this field; the limiting factor is the confidence and creativity to exploit what it offers. You can deploy predictive analytics through integrating Python/R or visual modeling software such as Big Squid/Data Robot into your analytics platform. You really can get started quickly if you have the vision, ambition and knowledge.

So how does it work in simple terms?

Probability is the underpinning logic behind predictive analytics; algorithms working the 'math' to generate the likelihood of an event occurring.

Once you have the prepped the 'lifeblood' data flowing in (the most important bit), you can build and train your predictive models on key outcomes (the target) you want to shift, for example; Vulnerability risk (any agency), length of stay (healthcare), offender risk (law enforcement), CSE risk (Local Authority), high risk of road flooding (Highways) and demand forecasts (any agency). The model will identify the key predictors that influence the target outcome from a true data driven approach. When run against your new data, the model generates a risk score or likelihood of the outcome occurring.

Evaluation is key; understand the accuracy of your models - and any key limitations. Trial them in the field, identify model improvements - aim for better accuracy. Remember, modeling is a journey.

What type of resource do you need to carry out predictive modeling? Some parts of private industry may require a ‘hoard’ of data scientists when extracting out a 1% model gain means a multi-million worldwide benefit, however; this is not the case for public sector. One data scientist, hired in or trained internally can make a significant impact on the organization. The critical success factors are; simplicity, connection with the operational business leads, building the models into your analytics platform, building into operational tasking.

Deploy your models and act

With the combination of the right technology and a data scientist, there should be a suite of 'rolls Royce' insight popping out at your fingertips. Build your analytics into work-flows and tasking processes.

And this is the key point: it provides an evidenced based source of insight that enables a professional to make better decisions more of the time. In the public sector, things are complex and require wisdom. People are good at this; algorithms are not. Apply a pragmatic and common sense approach to the results and a confidence will build, and good results will be generated. The world is changing. Automated and augmented risk management techniques should be embraced to help us all tackle the ethical dilemmas that modern day organizations face.

What are they doing in healthcare…

A growing number of predictive apps being used to support early intervention opportunities.

Wrightington Wigan and Leigh (UK NHS) are looking at how analytics is driving new ways to manage and improve A&E performance. They have developed an app with predictive capabilities that looks to forecast and support better resource and risk management decisions. They scrum down and review results every two hours to improve the service delivered to the public.

Morecombe Bay (UK NHS) have developed a stroke prevention app to support the reduction in this life changing problem. It is unique in drawing both acute and primary care data together and putting this in the hands of our healthcare professionals. This pioneering is supporting the development of an innovative and transformative culture.

In Australia, Aginic (Qlik partner) have produced deployed a Length of Stay predictive app. When the patient presents, the initial screening information is presented, and the algorithm calculates the predictive length of stay to enable a better care and customer package to be worked through. It also allows an evidenced based way to better manage bed management further upstream in the patient flow process.

Other examples include the risk scoring for chronic illness, readmissions into the emergency department, appointment ‘no-shows’, patient satisfaction and patient self-harm.

What are they doing in law enforcement…

Avon and Somerset Police have deployed predictive analytics and visualization in an industrial fashion to solve some very real business problems. At scale, understanding risk, the escalation of risk and profiling vulnerable people every day is just not possible using traditional techniques. Also, live time research on people within a command and control context is time consuming with every minute potentially compromising a more effective response. General time to research subjects takes a significant amount of time across the organization. Early intervention is mission critical when aiming to prevent harm.

The vulnerable persons risk and profiling app developed provides quick and simple subject profiles and overviews to allow the understanding of risk and early intervention opportunities. The app has been built in to role profiles, tasking and accountability processes.

Analytic Insights:

• 600,000 people risk scored every day
• Prioritisation of risk
• Vulnerable people with escalating risk can be identified quickly
• People and cohort profiles at a force, local policing area and beat level within seconds

And in local Government…

A good example of how predictive analytics is being applied is within a Troubled Families Program, based in the UK City of Bristol. They have brought together the most complete picture of social issues families are facing that the city has ever seen.

The Troubled Families Program is designed to help families who struggle with factors such as debt, homelessness, mental health issues, domestic violence, poor parenting, illness or substance misuse. The partners include the council, police, public health, and the social care service.

The process of bringing together over 30 datasets securely (very important), and in a timely manor, has proved pivotal to the success of Bristol's Troubled Families Program. The data was derived from individual events, grouped by person, before being built into aggregate family and household entities. Considerable care was taken around entity matching between datasets - recognizing the quality of the dataset will allow more advanced analytical approaches to be used. It was not easy. There were challenges in sharing and reconciling datasets, renegotiation with existing partners, along with engaging new ones was key issues to overcome.

A significant part of this process has been the use of predictive analytics to understand individual risk for members in the scheme. This has ranged from the vulnerability of domestic abuse, young people being at risk of street conflict, and vulnerability of anti-social behavior and child sexual abuse.

Are we sliding towards an integrated services approach?

The application of true predictive analytics in the public sector has, and still remains fairly limited, but there is a huge appetite building to change this picture. There are many leaders in the public sector who are great visionaries. They also invest in empowering their staff to push new technological boundaries within the limitations of their information technology structure. People are beginning to 'get' the potential of this technology. There is a real drive by leaders to 'make it happen'.

Better public outcomes are not the sole preserve of one agency, but are nearly always a combination of all of them. Let’s take crime; is it the police who are responsible for reducing crime – no, it is a combination of health, probation services, the council and many others. Often the first early trigger signs (predictors) are from agencies such as health e.g. a vulnerable adult has just started taking drugs again, or is drinking more alcohol etc – often the pathways to crime.

Having predictive technology working across agencies will enable targeted prevention activity at the earliest opportunity – moving from hindsight to foresight.

All agencies are trying to tackle the issue of reducing resources, whilst meeting increased demand. There is a strong investment and vision to develop integrated services further, and the role of big data and predictive analytics is pivotal to the success and future of public services.

Wrapping it up…

Treatment rather than prevention will grind services down to a point of surrender. To create a sustained and successful public sector; we must reduce demand, risk and harm – at the earliest point possible. Predictive technology, when applied with confidence and creativity can transform an organization. Our leaders can adopt established blueprints for success and start the journey now.