This Congress was an assembly of seven summits. These included: Care Coordination and Transitions, Connected Health: RPM and Telehealth, Bundled Payment, ACO Population Health Management, Merit-Bases Incentive Payment System (MIPS), Health Plan Chronic Disease Management, Provider Strategies on Community Health and Social Determinants of Health.

Topics presented were best of breed in each area. Common themes emerged from the 3 day group of talks. These included data science, CMS priorities, culture change from the top down in large health care organizations, operations and workflow, models of care financial incentives, care transitions, population management and risk stratification.

To summarize the take home message, it is obvious to say that health care as an industry and a profession is being transformed at a rate never seen before, ever. The scope of the transformation covers all corners, none is too small, nor is any too big to control or prevent change. The drivers of change are assumed to be high cost and fragmentation, but something more fundamental seems to be in play. This Congress is unusual in the scope of the organizations present. This is an ideal forum to see the commonalities of the perception of change among the industries and professions effected. The energy the groups show in responding to change is refreshing. The fundamental driver is not unique to health care, but exists in the world economy as a whole whereby mobile devices are democratizing businesses, and data science is migrating into health care as it has in other industries. Old intractable problems see new solutions.

Foremost are the discussions through all topics of data science and the relevance to design, operations, and evaluation of programs. Artificial Intelligence and machine learning are credited with having an application for everything. These data science tools are generally relevant for big data. This will place them under the wing of plans, population in general, and data from large integrated provider networks. Clarify Health Solutions is a remarkable company that analyzes care paths and programs for health care systems using Medicare data available to anyone. The presentation by their CEO Jean Drouin and their VP Julie Cohen gave insight into what AI could and could not do. AI works best when there is a defined objective and lots of data such as speech recognition, and image representation. He claims AI is not good for open ended problems. The best of AI works in the moment to deliver solutions real time. AI to cause change management comes from solutions that people will actually use; it is best at pattern recognition, not episode management. The patterns come from ability to contextualize, and from its ability to personalize. This company is in the sweet spot of care transformation.

In population health presentations, AI and machine learning came up frequently. For example, Laurie Evans of Blue Cross Blue Shield of Tennessee gave a masterful talk about risk stratifying patients, where those deemed at high risk received call center attention from nurses whose list of calls were prioritized by severity of risk. This is more predictive modelling than it is AI; this highlights how the use of the AI term is presently quite loose with the difference between predictive modelling and data science not clear. Predictions start with data, data science starts with modelling which is completed before the unknown subject data is analyzed. This is an important difference.

Representing the top of the population scale, Richard Wild, head of Atlanta Medicare Division, gave an introductory talk on the MACRA/MIPS program titled "Explore CMS Priorities in the Transition to Value Based Care". These programs are highly data driven, with the analysis of these programs realized as benchmarks for ACOs, PCMH, and Episode of Care technology. However Dr. Wild is clearly very sensitive to the problem of care fragmentation by noting it is time for Marcus Welby to reappear, a doctor willing to do many things to overcome fragmentation of inpatient, outpatient, and long term care, by someone who knows the patient. One wonders why fragmentation of patient care is so obvious at the largest population level. To consider this issue from the point of view of data and analytics, one needs to know how data is generated, and how the analytics are produced for users of the data. The scale factor helps here. For N of 1, the patient and the physician, the Marcus Welby, data is important, but not meaningful as a population metric because standards and programs that depend on benchmarks come from groups. Not being important as a population surrogate, the wise decisions that are needed to coordinate care at the patient level, cannot generalize to population level accountable benchmarks. The policy of Medicare in MIPS recognizes the potential of small scale provider level incentives in the MIPS scoring program to lower costs. Here is an opportunity for data science to compare and contrast these small scale incentives to population incentives of the accountable programs on population level costs.

Samuel Le Church Family Physician of Synergy Health presented "Get Ahead of the Game and Be Prepared for the Cost Measure." This measure as it is phased in to account for 30% of the MIPS score by 2020, applies to individual providers, but the cost measure is derived from all patient level expenditures. The metric will drive collaboration, but the mechanism to do this collaboration is not in place yet. This is an open issue for MIPS scoring, whereby there is an opportunity to correlate codes from all providers to make this metric a collective function. MIPS as a provider centered metric offers an opportunity to serve a multilevel function ultimately rolled into population level metrics, although this is not a development envisioned by CMS. Lee Pearce of the Alabama Quality Assurance Foundation elaborated on provider adoption of the MIPS program. The most difficulty is for small practices who are provided adoption support by this organization. Here is an issue that seems to exist for all providers, in that incentives to make wise decisions for clinical care in a cost effective manner are small single provider scale, yet these incentives are designed for effect at provider group scale. MIPS brings out the need to coordinate scales for care starting with the smallest scale. For small practices this scale is vulnerable because of resource needs that are hard to meet. Even for large groups with sufficient resources for qualified information systems, the locus for wise decisions is still the individual provider. Despite the frustration even on the part of provider consultants there appears to be an opportunity for MIPS to be the leader in cost containment that serves as a foundation for population scale accountable cost control.

Optimizing high performing large integrated provider networks was described by Wright Lassiter III of the Henry Ford Hospital system . This was an excellent description of leadership driving culture change. The high level objectives of culture change are coordination, value and measuring outputs. The barriers to culture change include turf issues, resistance to change, hierarchy, beurocracy, and lack of bias for action, trust, openness, teamwork, can do attitude and customer focus. Neil Gomes of Thomas Jefferson University spoke about The Jefferson hospital system, highlighting multimodality technology, system integration, and information transparency. It is good to know how well large networks can function effectively.

Any discussion of scale from N of 1 to CMS's millions of beneficiaries must always account for the wise decisions of single providers. The ecological fallacy comes to mind here, whereby if you know the tree you can know the forest, but if you only know the forest you do not know the tree. For the purposes of health care cost reduction "knowing the tree" is essential. In the alternative payment model world, this is exemplified by the differences of episode of care technology vs. accountable care technology. A presentation by Emily Walker of Anthem, "Align Payors and Providers to Develop Mutually Beneficial Episode of Care Models" exemplified the distinction. Though not all would agree, by far and away the most important factor in cost reduction is physician engagement. In observing the evolution of the joint replacement bundle for 7 years, Ms. Walker noted the time and continuous attention needed for engaging orthopedists in the bundle from the inception of the program. Clearly this was a process of invention only possible with a high level of engagement on the part of providers. On the other hand accountable care organizations are caught in the process of receiving packaged benchmarks derived from outside their organization with no connection or reporting of data that derive the benchmarks. The focus of bundles needs to exist in the context of broader population measures, but there is no way to do this as yet. The problem may be the administrative and information management intensity of interacting with providers who wisely cling to the high dimensionality of direct patient care.

To cap the meeting was a presentation by Jean Drouin of Clarify Health whose topic was "The Precision Workflow Revolution ~ What if Amazon, FedEX, and JPMorgan Combined to Deliver Delightful Care"? The ERISA plans are the most likely force to drive health care reform. Dr. Drouin gets it; he is at the sweet spot of all the themes of this very important meeting.