Actuarial Science, Accountable Care Organizations, and Workflow

Posted on by

//platform.twitter.com/widgets.js

Today’s Actuarial Challenge tweetchat is a welcome opportunity to ponder the future of actuarial science, health IT, and accountable care organizations, in a single blog post.

For background, see:

How will Accountable Care Organization (ACO) IT look in 10 years? How will Actuarial Science fit into that infrastructure? How can workflow technology get us there?

I am not an actuary, but I did get Accountancy and Industrial Engineering degrees (on the way to med school!). In doing so I studied fundamental Actuarial Science concepts: economic risk, random variables, time value of money, and modeling and optimizing stochastic processes (stock and trade for actuaries). Eventually I designed and deployed health IT software. But I’ve kept an interest in financial security systems. From Wikipedia:

“A financial security system finances unknown future obligations. Such a system involves an arrangement between a provider, who agrees to pay the future obligations, often in return for payments from a person or institution who wish to avoid undesirable economic consequences of uncertain future obligations.[1] Financial security systems include insurance products as well as retirement plans and warranties.[2]”

Here are my ten-year “What Will ACO IT” Look Like” predictions:

1. ACO enterprise SW will be ‘process-aware’ (workflow engines executing declarative process models). It will be essential for turning actuarial insight into automated ACO workflows, in almost real-time.

2. Stochastic simulation will literally be built into ACO enterprise software. Simulation is already built into many Business Process Management (BPM) workflow platforms. Actuarial simulation and workflow simulation will increasingly complement and even merge.

3. Virtual ACO enterprises will be built across workflow interoperable healthcare subsystem organizations. For extended discussion of task, workflow, and pragmatic interoperability, see my five-part series

4. ACOs will know exactly how much each service line (chronic Dx, procedure, etc) costs. Comprehensive ACO workflow IT platforms will seamlessly drive sophisticated event-driven activity-based cost management systems. Other industries know exactly what their smartphones and vehicles cost. Healthcare needs to do so too.

5. Virtual ACO enterprises will systematically optimize ROI on collections of targeted workflows. If each of predictions 1-4 become true (workflow tech infrastructure, embedded stochastic simulation, pragmatic workflow interoperability, and virtual ACOs), ACO will become truly intelligent learning healthcare systems.

Relative to intelligent learning systems, I should mention another of my degrees, an MS in Artificial Intelligence. Artificial intelligence, machine learning, workflow technology, business process management, and data pipeline management systems are increasingly leveraging each other’s strengths, and in some cases, even merging. While process-aware workflow technologies will increasingly form virtual ACO IT infrastructure, these workflows will be highly “tunable.” The additional data made possibly by workflow technology about what happened when will increasingly feed into stochastic models, which, in turn, will be essential for systematic improvement of workflows both driven by, and generating the data. This is the “intelligent learning” to which I am specifically referring.

I also have a whole bunch of questions! For example, what, exactly, does “stability” mean? (Probability that premiums will be sufficient to claims cash flows?) What is the current state-of-the-art for actuarial simulation? Are “state models” (as in, Markov models of disease progression) routinely used in ACO actuarial calculations? And so on.

For now, I’ll just close with some thoughts on the intersection among actuarial science, accountable care, and my favorite topics: healthcare workflow and workflow technology!

What is the connection between workflow, workflow technology, actuarial science, and accountable care?

I’ve taken entire courses in workflow. I’ve looked at hundreds of definition of workflow. The following is what I eventually “settled” upon.

“Workflow⁰ is a series¹ of steps², consuming resources³, achieving goals⁴.”

⁰ process
¹ thru graph connecting process states (not necessarily deterministic)
² steps/tasks/activities/experiences/events/etc
³ costs
⁴ benefits

Workflow technology is any technology that represents workflow as a model, explicitly (declaratively) or implicitly (neural network weights), and operates on the model/representation to automatically execute workflows or automatically support human execution of workflows. Academic workflow researchers call these “process-aware information systems.” The best know PAIS are BPM systems. However process-aware workflow tech is rapidly appearing in IT systems, such as Customer Relationship Management systems (CRM) and data and language “pipeline” platforms not typically referred to as BPM systems.

If one modifies my definition of workflow, though within my subscripted limits, to…

Process is a series of events, consuming expected resources, achieving expected benefits

…you arrive at a stochastic process closely resembling actuarial science’s generalized individual model (page 35 in Fundamental Concepts of Actuarial Science).

During my student days, we spent a lot of time estimating parameters and distributions, and then predicting behaviors of these stochastic processes. Sometimes we did so analytically with complicated equations (Markov Models). Sometimes we fell back on computer simulation (Monte Carlo).

A quick review of actuarial science literature indicate many of these same techniques are used today. I found questions about them on actuarial science exams and interesting papers by actuarial science researchers. I’ve appended links to some examples at the end of this post.

By the way, I believe my five predictions are incredibly relevant to a very topical topic: MACRA’s “virtual groups.” So I’ll close with this quote (stretches in italic due to me).

“Virtual Groups

“…the MACRA Proposed Rule will likely put pressure on solo practices and small group practices, while favoring large groups.

Fortunately, the MACRA legislation offers a possible “salvation” for solo practitioners in that the rule allows for the formation of “virtual” groups. This would presumably enable smaller practices to band together (virtually) and to function as a larger group, spreading risk and potentially taking advantage of APMs and other benefits the MACRA legislation offers larger practice groups.

Unfortunately, CMS has decided that virtual groups, while mandated by law, were too complicated to set up. Consequently, it is proposing to delay the implementation of virtual groups until 2018.

What is especially ironic about this is that CMS states that virtual groups will be delayed due to the difficulty of establishing an efficient and effective “technical infrastructure” by the beginning of the 2017 performance period. Yet (of course) providers and software vendors are granted no such relief, even though they too will have “technical infrastructure” needs that will have to be enabled in their EHRs in that same restrictive timeframe.

The net result is that without the relief of virtual groups, the majority of small and solo practitioners may be even more unlikely to meet the MIPS standards during 2017, and are more unlikely to avoid penalties being assessed in 2019.”

What’s my point? Well, my predictions are ten years out, and therefore not likely to benefit small medical practices next year. However, I do think the concepts I invoke — virtual ACOs, pragmatic workflow interoperability, true costs, intelligent learning systems — are highly relevant in the long run. Therefore, even when fighting short-term fires, we need to keep our eye on long term goals, and paths to those goals.

Relevant to the #ActuarialChallenge, virtual intelligent learning ACOs will require actuarial science knowledge and experience to be successful. But, I also firmly believe, actuaries must leverage workflow technology to achieve the kind automatic, transparent, flexible, and systematically improvable workflow necessary to merge actuarial science secret sauce directly into ACO IT infrastructure.


//platform.twitter.com/widgets.js

window.twttr=function(t,e,r){var n,i=t.getElementsByTagName(e)[0],w=window.twttr||{};return t.getElementById(r)?w:(n=t.createElement(e),n.id=r,n.src=”https://platform.twitter.com/widgets.js”,i.parentNode.insertBefore(n,i),w._e=[],w.ready=function(t){w._e.push(t)},w)}(document,”script”,”twitter-wjs”)@wareFLO On Periscope!

firetalk-button2

P.S. Here are some links and resources I consulted while writing this blog post.

Fundamental Concepts of Actuarial Science (fantastic introduction to AS if you hate math)

Introduction to Actuarial Science (edX course: videos, but also PDFs, ends with Monte Carlo simulation of life insurance)

Health Insurance: Basic Actuarial Models (Amazon, heavier going, but Fundamentals monograph and Intro edX course are good prereqs)

Society of Actuaries 2014 Exam MLC Models for Life Contingencies (includes questions about multiple state models)

Actuarial Calculations Using a Markov Model

Critical Review of Stochastic Simulation Literature and Applications for Health Actuaries

Stochastic Process (Wikipedia)

Constructing Probabilistic Process Models based on Hidden Markov Models for Resource Allocation (process mining to estimate state model transition probabilities)

Mathematical modelling of social phenomena

An actuarial multi-state modeling of long term care insurance

Estimation of disease-specific costs in a dataset of health insurance claims and its validation using simulation data

How Predictive Modeling Is Helping Employers Gain Control of Health Care Costs

Stochastic Modeling in Health Insurance

The ACO Conundrum: Safety-Net Hospitals in the Era of Accountable Care(Lean Six Sigma to ID and eliminate inefficient processes)

CMS ices reinsurers out of an ACO program

“The group practices were unable to track and manage the fluctuations in risk, and many went out of business. In recent years, organizations have tried to develop better cost tracking systems”

Multiple State Models

Actuarial Uses of Health Service Locators

Estimation of disease-specific costs in a dataset of health insurance claims and its validation using simulation data

HOW PREDICTIVE MODELING IS HELPING EMPLOYERS GAIN CONTROL OF HEALTH CARE COSTS

Transforms the Insurance Industry with Cloud Modeling Platform

Leave a comment