Feeding the learning from clinical and non-clinical data back into existing processes is essential to fully realizing the benefits of digital technology. We have already highlighted numerous examples of how this continuous cycle of improvement and learning can take place throughout the report: using patient flow technology to identify where patients wait in the system and re-designing services; collecting data to understand where variation in care occurs and why, and using data to predict and target at-risk groups.
Once you are starting to use data in a systematic way, there are numerous ways in which this supports your on-going benchmarking and performance improvement. You can’t improve what you don’t measure.
— Ran Balicer Clalit Research Institute Israel
A core part of Intermountain’s digital strategy is system learning and improvement from the data they collect. Leaders at Intermountain choose the data they collect carefully, keeping in mind the estimate that each data item collected costs a dollar per patient. Data analysts are fundamental to this process. Intermountain’s Institute for Health Care Delivery Research employs 17 statisticians at Masters level or higher to analyze registry data and produce routine reports on care delivery performance. The intention is to make performance transparent to the clinical teams at an individual patient level and at a process level. According to Brent James, Intermountain has carried out three formal evaluations to understand if these data analysts could be replaced by a business intelligence system. However, they have always found that analysts are the preferable option given that they offer considerably more flexibility than a technological solution. Intermountain embed their analysts in clinical teams and believe they represent very good value for money, often costing less than a nurse but offering insight that could save their salary many times over. Intermountain have already taken out 10 percent of their costs in the last 3 years. They believe that these systems will enable them to reduce the cost of care by 50 percent — both through reductions in waste and non-value adding activity and improvements in clinical outcomes. [Source: BJ Interview].
The use of data can also help drive improved care pathways and ensure that patients receive optimal care. For example, Advocate Healthcare in Chicago estimate that they are saving US$200 million a year from an algorithm that offers recommendations to physicians and patients about what level of care someone should be discharged to (e.g. nursing home, their own home with nursing support, or a hospice). We are also starting to see significant investment in artificial intelligence, with the best known example being Watson. This is some distance from mainstream adoption but signals a direction of travel.
Watson is a supercomputer built by IBM, which is able to process and understand data in a novel way to answer complex questions put to it by the user. It is able to extract meaning from free text enabling it to store data from any written source. It has a wide range of applications across multiple industries, including healthcare. Watson is being trained by oncology experts at Memorial Sloane Kettering (MSK) so that it might be able to inform decision making in cancer care. The idea is that Watson will be able to analyze the patient’s medical record to identify key characteristics that might influence outcomes. It can then identify potential evidence-based treatment options, rank treatment options and present these to the user with supporting evidence from a wide range of sources. This can allow clinicians to match individual patient characteristics to the vast and complex research and knowledge base and provide tailored and evidencebased treatments. Definitive outcomes for its use in cancer care are awaited as Watson is still undergoing training and testing at MSK. However, results presented at the 2014 American Society of Clinical Oncology meeting demonstrate that Watson is able to choose the preferred treatment option with 89–100 percent precision, depending on cancer type.1
Gain patient consent for use of data beyond direct care: England’s scaled-back care.data program has highlighted the sensitivities around using patient data for reasons beyond direct care without adequate engagement and consultation. Organizations need to be entirely transparent about how they will use patient data and think carefully about how they will gain informed consent for data analysis, articulating clearly intended benefits of using data in this way.
The skills required to enable digital healthcare — big data, user experience, cybersecurity — are in limited supply and have not historically been core to healthcare. But accessing these capabilities — whether in-house or externally — will be the ultimate arbiter of progress.
— Liam Walsh KPMG in the US
1Epstein AS, Zauderer MG, Gucalp A, Seidman AD, Caroline A, Fu J, et al. (2014) Next steps for IBM Watson Oncology: Scalability to additional malignancies. Journal of Clinical Oncology.