Reporter, European Medical Journal
At the ‘Big Data in Use’ conference held in London earlier this year, Nondas Sourlas, Director of Healthcare Analytics at Bupa, spoke about how the UK’s largest private medical insurer is using Big Data to improve healthcare.
For the past 5 years, Sourlas explained, Bupa has had its own healthcare analytics team which began with just five people. “We are now about 50 people and it is not all analysts. There are clinicians in the team, specialists in reporting, data processing, as well as the core analysts,” he said. Since its beginning, the team has utilised the large amounts of data generated by the company as both a global payer of healthcare costs and from its involvement in healthcare provisions. This data has then been used to design and to improve the services that Bupa offers to its millions of customers.
Nearly 4 years ago, the team launched a programme that uses predictive analytics to identify which individuals require support at different points in their care. The programme can be used to predict which patients are at a high risk of being admitted to hospital within the next 12 months. Once a patient is identified, they are then contacted via telephone and informed about available treatment options or procedures that might be offered once they visit a consultant. According to Sourlas, the outcomes of the programme have been successful, including an annual reduction of >1,500 outpatient visits and >100 surgical procedures. As of 2 years ago, 79% of patients reported being extremely or very satisfied with the programme. No data was provided during the talk for how many patients this involved, or on what aspects of the programme the patient satisfaction referred to.
Another way in which the analytics team has used the healthcare data generated by Bupa is to carry out consultant profiling and referral management. Sourlas said that previous research has shown how many general practitioners feel they cannot make an informed decision when referring a patient to a consultant, due to a lack of available data. To address this problem, the analytics team have begun to look at data that Bupa has on consultants to assess their affordability in combination with the clinical outcomes measured as a result of the care they have previously given. “With the information we have, we try to profile the consultants within specialities,” Sourlas said, “[our] clinicians are looking at clinical measures based on clinical guidelines that exist, and it is also based on affordability where we look at the cost of treatment over the whole [care] pathway.” Predictive models are also used to estimate costs, as well as to make risk adjustments that alter the mix of patient cases a consultant receives according to previous performance, such as the number of severe cases dealt with.
Sourlas also spoke about the challenges faced when measuring the results of these predictive services. “The way we are evaluating whether our predictions are right and whether we are actually providing a good service is based on healthcare outcomes,” he said. However, data on healthcare outcomes is sparse because some hospitals and consultants are not currently required to disclose these outcomes, he explained. “They will be required starting in 2017, at least for the private sector,” he added. This refers to the establishment of the Private Healthcare Information Network (PHIN), set up by the UK government, requiring private healthcare providers and consultants to make public various healthcare outcomes. This includes mortality rates, infection rates, and the number of patients transferred to an NHS hospital from a private hospital. The government has said that the PHIN will be fully operational by 2017.
Bupa does not have access to complete medical records that cover entire patient histories since UK citizens can move between accessing private medical care or by going through the NHS. A key source of data instead comes from the insurance invoices for services that healthcare providers bill Bupa for. “They need to provide diagnostic codes, codes that say you were diagnosed with X-Y-Z condition, and then the procedure code which is what [the healthcare provider] actually did; that might be a test, it might be surgery, or a consultation. Those standardised codes are the information we decipher. The clinicians in the team try to make some clinical sense out of what it is that the person had and whether the treatment was appropriate and billed at the appropriate price based on what we were contracted for”.
At the end of the talk, Sourlas was asked by a member of the audience, who said they worked for the NHS, how he thought the public sector could make better use of data. Sourlas replied: “I guess the simple answer would be: make some of that data available. The NHS over the last 2 years seems to have gone the opposite direction. Bupa was providing some services to the NHS from an analytics perspective and we were getting access to NHS data, like secondary care data – Secondary Use Services (SUS) datasets from hospitals and providers.” Sourlas explained that access to that information has become more difficult and requirements for doing so have become more stringent. “I think if we are not able to get all the data together from different payers to try and paint the right picture, then we are missing something,” he said.