This week I attended the National Quality Summit in Dallas, Texas. The event, sponsored by the National Association for Healthcare Quality, was chaired by Dr. Drew Harris, a nationally known expert on population health, and co-chaired by NAHQ’s own Len Parisi and Nancy Terwood.

Disclosure: I’m a member of the NAHQ board of directors, and part of our attendance in Dallas included our quarterly board meetings. However, my summary of the summit does not constitute a NAHQ-endorsed communication. My observations, below, are solely my own.

The summit included roughly 250 in-person and more than 400 virtual attendees. The event highlighted current trends in population-health management presented, partially, as case studies in how health quality professionals can help move the needle on the health of patients at a population level.

The “population” distinction is significant. Much of healthcare today focuses on the treatment of individual patients. Although some initiatives, like Pay for Value, roll up results at an aggregate level, the truth is, we don’t manage the health of cohorts very well. Even the vaunted Patient Centered Medical Home model is, at heart, a treatment paradigm for individual human beings. But treatment at a population level requires different incentives, different skills and (potentially) a different political climate.

Some core learnings from the summit:

• “No margin, no mission” migrates to “no outcomes, no money.”
• Hospitals might decline as incentives against inpatient care mount, but health systems will endure. Those systems must be adaptive, not reactive, if they are to thrive.
• Population health features a dual focus: from the “patient out” (looking at individual humans within a system) to the “population in” (looking at the community as a whole to remediate socioeconomic problems contributing to poor health outcomes.
• The community aspect requires a reliably delivered, broad set of preventative interventions for prevalent but inadequately addressed health risks. Advanced preventative care mixes care coordination across the continuum with effective chronic disease management and personalized prevention services. However, “personalized prevention” is wildly under-employed.
• Programs must be relevant to the targeted population. It’s foolish to develop a program and then find a population to channel toward it.
• Good pop-health programs use a portfolio of 30-35 robust, evidence-based interventions. No one-size, one-program approaches work. Analytics should rely on statistical process capability techniques to assess ongoing effectiveness (before a formal evaluation study) and make better use of geomapping capabilities.
• Population targeting presents an interest set of challenges —
  • What’s the aim of the targeting initiative?
  • Why obsess over 1-5% of costs if other populations with lower costs can yield better clinical results? [A CMS study suggested better cost savings with low-to-medium risk targeting (aversion) instead of high-risk (management) activities.]
  • Can we target rising-risk patients?
  • Can we move beyond claims data to identify a population?
• An optimal mix for program targeting might include moderate-to-high risk patients based on diagnosis, physician opinion and health-risk assessments. A typical candidate pool might be roughly one-fifth of the 65+ population and with good outreach, perhaps 40-50% can be enrolled.
• The system perversely incentivizes wasteful care precisely because some stakeholders profit from it. But when average costs for a median family of four consume 40 percent of disposable income — there’s a problem. The fee-for-service model creates a structural barrier to the care of populations. And despite their unpopularity, narrow networks consolidate care teams and health information much more effectively than wide networks.
• Over-reliance on primary care physicians, instead of comprehensive care teams spanning stakeholders, isn’t helping advance the cause. PCPs can be an entry point, but they can’t do it alone. Clinical integration (including coordinated care, evidence-based practices, waste reduction and network management) is essential, as are the infrastructure pieces like health IT, info exchanges, predictive modeling, population segmentation and a stronger emphasis on behavioral-health and wellness services.
• Could an Uber-like model disrupt healthcare at a consumer level?
• We have a lot of data, but not much capability to transform data into actionable information. If you can’t measure it, you can’t improve it.
• We tend to focus on high-cost/high-risk patients, but perhaps we should invert the pyramid. It’s easier and better in the long run to focus on wellness and routine gaps in care than to let people progress through rising risk until they hit a “you’re at the edge of the cliff” threshold after which — when it’s too late — the care team rallies.
• Partnership with health plans is essential (integrated wellness, behavioral health services, patient engagement) but the messaging has to be deliberate. In one case study, telephonic care management engagement jumped from 15 to 80 percent when the caller-ID string changed from the insurer to the provider.
• The move from “volume” to “value” will be hugely disruptive.
• Clinical interventions take time — from 3 months to 5 years — to generate value. Need to level-set expectations accordingly and avoid premature evaluation and trending.
• Pharmacists are one of the least appreciated professionals in the industry. PharmD’s make a real difference when integrated into the care team. Coordinating pharmacists, dietitians and SNFs as part of overall home-health outreach may be worth exploring.
• Decision science — using aggregate data to inform strategy — is under-valued in health analytics. Metric-based decision making can be a barrier to effective population health. Solo dashboards — data viz w/o substantive analysis — are rarely helpful and sometimes harmful. You can’t do “analytics” without storytelling.
• Problems in most health-analytics projects: Not asking the right questions from the outset, disregarding qualitative information, not reviewing published research, not acting on the results of the analysis.
• Let some analysis follow from “microtrends” (the small forces behind big changes). Census-block analysis is more helpful than it appears.
• Near-real-time surveillance is key, and can be built by individual institutions with adequate resources. BRFSS is broad, NHANES is detailed — but both lag. By the time you know your population, you’ve lost your actionability window, unless you do your own surveillance using claims and lab data with incremental census data. Knowing something useful about your community helps shape the tree of questions during individual patient encounters — i.e., you’re treating your patient as being a member of the population they represent instead of an automaton disconnected from their community.

Because of the board meeting, we were not able to attend the second-day sessions. But the first day certainly left me with many valuable take-aways.