I resisted signing up to Netflix until very recently because I was struggling to find the time to watch the programmes that was filling up my Sky box. I signed up because of my kids (that’s my story and I’m sticking to it).
Apart from the interesting and original content, I have long been an admirer of how Netflix maximise the use of data to identify what their customers would like to see and how they would be like to be engaged with.
They analyses what users watch, for how long, when they pause, which devices they watch it on, what day and time, and a number of other metrics. This allows Netflix to identify new app features that enhances the viewing experience, encourages users to watch for longer, and it feeds the sophisticated recommendation algorithms that provides users with a steady stream of content to choose from. They regularly experiment (and analyses the associated user action data) to figure out how to customise images and trailers to show to users (what actors or other elements to highlight, which words to use to describe content, etc). At a strategic level, it uses information on the success of similar genres, popularity of directors and actors to make investment decisions on new content.
Having seen Netflix in action over the last couple of days, a fascinating study exploring how data can help identify tailored interventions to overcome implementation barriers caught my attention today. The study was based on a 2 year programme aimed at implementing professional services (i.e. screening services, medication management services, etc) across community pharmacies in Australia.
Change facilitation was used as a key implementation strategy and pharmacies received tailored interventions. Barriers (i.e. an inability to plan for change, lack of internal supporters for the change) and facilitation strategies (i.e. empower stakeholders to develop objectives and solve problems) were identified. Facilitators also recorded whether barriers were resolved following implementation of the chosen strategies. All of the data was then coded in a structured manner by the research team. Machine learning techniques were used to identify a predictive ‘score’ for how successful an intervention would be for a particular barrier.
A Microsoft Excel based tool was provided to the change facilitators to record barriers, interventions, and their effectiveness. This in itself provide more structure to the implementations. The data collected and the predictions calculated would inform future implementations be more successful by selecting more effective facilitation strategies.
Thinking of wider (and more ambitious applications), what if we can capture different interventions used within a health setting to manage various symptoms and diagnoses? Could we use existing outcome measures (supplemented by surveys or Patient Reported Outcome Measures) to minimise extra work required or even automate? What if we could then extend this to capture interventions chosen and implemented by the users themselves (think Google, user support groups and non-medical intervention that do not involve healthcare professionals)? This could give some really interesting insights into what works and what doesn’t, and help inform design of services.
One of the challenges currently is that healthcare data is siloed. Interventions in hospitals may have an impact on the user in the community (i.e. increase or reduction in GP appointments or visits to the physiotherapist). With the increasing importance of Integrated Care Systems (and some of the technology and data systems that will need to follow to facilitate truly joined up user care pathways), this may (hopefully) change and there may be opportunities to follow each user through their journey across different organisations and care settings.
From another perspective, some of this data is being brought together already. I recently shared my thoughts about how Apple are bringing together data from medical records, user generated sources and sensors, and how they have strengths in 3 key areas (data being one) which could help them build the best wellbeing services.
As the required data becomes available in one place and we can start extracting (near) real time insights, experiment safely and suggest tailored interventions that we know works in that particular scenario, we may truly see health services with 33 million different versions (as Joris Evers, Director of Global Communications at Netflix once described the Netflix service).
Very happy to hear your comments below or feel free to email me to share ideas – firstname.lastname@example.org