Today’s blog post is the third in a series looking at health value and outcomes. In the first post, I set forth my belief that a single approach like activity-based costing (though vitally important) cannot in isolation drive transformation across a complex ecosystem like healthcare – a broader approach addressing the “critical 4” dimensions of knowledge, incentives, values, and transparency is required. Last week, I described how our lack of a more comprehensive approach to the critical 4 severely limits the utility of mathematical approaches to assessing and managing costs and outcomes; high variability in the “inputs” of the value equation produce high variability in the “outputs.” This week, I hope to broaden the view of the opportunities associated with applying advanced health analytics.
Today, health care is largely preoccupied with descriptive statistics – establishing and reporting quality measures, for example. Descriptive statistics are an important and necessary precursor to asking and answering more sophisticated types of business and scientific questions. But I would argue that descriptive statistics are much less useful in driving consensus on medical knowledge, incentives, values, and transparency. Why?
Knowledge. Descriptive statistics answer “what”, but offer little-to-no insight into “why” or “how.” For example, research showing that 57% of patients with a particular condition respond to a specific therapy is generally useful, but not if you are a newly diagnosed patient and don’t know if you are one of the 57%.
Incentives. Descriptive statistics are good at looking retrospectively at what has happened, but are poor in predicting what will happen in the future, particularly when the playing field is shifting under your feet. By definition, incentives are about influencing future behaviors in the light of both current state and most likely future scenarios.
Values. Descriptive statistics offer little in terms of understanding the causal relationships which people use as part of their belief systems. Many of the "hot topics" we find in our industry – especially those hyped by consumer media – simply fan the flames of under-informed personal values and flip-of-the-coin approaches to complex decision making.
Transparency. Because of the inherent limitations in descriptive statistics, people and institutions are often hesitant to be overly transparent out of concern that the information derived through descriptive statistics can be misinterpreted and mis-used (especially in the absence of more compelling, comprehensive contextually sensitive information). National efforts around quality measures and patient safety organizations have gone a long way towards managing these constraints, but they cannot remove the inherent limitations in a descriptive approach.
So what can more advanced approaches to health analytics offer? Simply, they allow us to draw conclusions -- specific, actionable conclusions on what, why, and how. Here are 9 examples:
- Patient Population Segmentation -- not simply the way that patient populations segment along traditional dimensions like age, gender, and geography, but how individuals and groups should actually be characterized in terms of issues like risk propensities, therapeutic response probabilities, optimal engagement models, likelihood of both short term and sustained behavioral change as well as other predictive dimensions.
- Redefining Medical Conditions -- balancing and actually optimizing for individual patients the multiplicity of factors that should be included in defining the boundaries of a medical condition: genetic traits, associated personal risk factors, care process standardization, safety management, reimbursement, partitioning epochs of care, etc.
- Redefining Medical Outcomes -- using multi-factorial analysis to optimize the selection of the "right" set of personalized outcome measures (i.e., powering the "numerator" of the formula I presented in the last post)
- Medical Indicator Selection -- assessing the relative strength and predictive value of an ever-growing inventory of indicators associated with medical outcomes, including the detection of previously unknown or under-utilized indicators. (e.g., the early probabilistic identification of those signals that will become like needles in a haystack of clinical data).
- Applied Quality Metrics -- using analytics to uncover the relationships between a comprehensive catalog of quality measures and their potential relevance to individual patients (i.e., their indicators, risk profiles, etc.).
- Care Process Refinement -- developing a true predictive understanding of the relationships between minute-to-minute, day-to-day, and month-to-month care process variations and alternatives, medical outcomes, quality, complications, safety and costs.
- Tradeoff Optimization -- using advanced analytics to aid in the detection, assessment and optimal selection (at the level of an individual patient) of the tradeoffs between complementary care approaches, conflicting care options, and risk management.
- Outcomes Risk Adjustment -- using more sophisticated forms of predictive analytics to adjust patient-specific expected medical outcomes, taking into account individual choice and preference.
- Redefining Reimbursement -- enabling organizations to predictively model the risk/reward ratio of the myriad of emerging reimbursement alternatives in terms of costs, outcomes, variability, profitability, and patient segmentation and risk factors.
There are of course many more opportunities for health analytics -- I'm just touching the surface. But by deepening the sophistication of the questions we ask and the answers we derive, health analytics can achieve what purely descriptive statistics cannot -- namely, furthering the development of our knowledge, incentives, values, and transparency across the health ecosystem (including patients).