JUL 11, 2013 12:26pm ET

Extending an All-Payer Claims Database (APCD) to Support Population Health


Six months ago, working on a big state HIX solution, I first heard the term APCD (All-Payers Claims Database). Perhaps I was behind in knowing that acronym, but since then I’ve heard it so many times that I wanted to get to the bottom of the hoopla.  I soon realized that like so many health care concepts, APCD means different things to different people with some core similarities. More importantly, I realized that the current concept of an APCD should be extended if maximum value will be derived, more like APCD+ rather than APCD.

What do I mean by APCD+?

A baseline, vanilla variant of APCD is not much more than another data warehouse -- granted it has the potential to be a single source of truth, but another warehouse nonetheless.  The objective is collecting data from insurers in the state using a mechanism (Master Member Index, a variant of Master Patient Index) to correlate and store the information in either a raw or rolled up cumulative format. This process by itself likely won’t help reduce costs, improve quality or deal with and abuse, the most commonly spoken benefits of APCD.

What can really perform the desired tricks is the quality and focus of the analytics one runs against the APCD and the type of data applied to those analytics.  For example, even the base analytics that can be performed against pure play financial data (which is pretty much what payers have) is limited. Before you protest, let me qualify that, yes, the 837s contain medical codes (ICDs, CPTs), but that provides only a pseudo-clinical perspective. In other words, these claims records are at best a rough surrogate for what is going on in a real clinical setting.  Not really equivalent to medical charts, are they? Unfortunately, to realize the promise of the APCD and to start moving towards a platform that enables population health, this is necessary but wholly insufficient.

That gave me food for thought to propose an enhanced concept that I call APCD+.  (Nobody can ever blame me for being a great name-coiner.) The concept defines three generations for APCD.

Generation 1 encompasses the core building blocks of APCD, e.g., Master Member Index, Intake mechanism from varied payers systems, an extensible data model to store extracted information under broad categories and some basic financial analytics along with dash boarding.  This stage primarily establishes a member-centric single source of truth and allows for broad categorization of focus areas for further evaluation.  It also answers basic questions, such as what are the biggest spend areas for my population?  How is my network in terms of efficiency and adequacy?  How are the baseline prescribing trends among my providers?  And which cohort of my population is costing me most?

Generation 2 takes a big step forward and involves intake and integration of “real” clinical data. This data could come from state HIEs, regional RHIOs, IDNs, provider groups or even outpatient and long-term care facilities. Now we are talking about an interesting data set that could start diving meaningful insights into managing population health. This stage allows one to answer in much more detail the questions posed in generation 1. For example, instead of simply identifying which cohort of your patient population is costing you the most, you can explore why.  Perhaps  it is not only the chronic type 2 diabetic population that is most expensive to maintain, it is a subset of 20% type 2 diabetics who are also asthmatic and obese, who are costing me 60% of the total spend. Answering the “why” is critical if efficient action is going to be taken that can meaningfully impact cost while maintaining quality.

In classic problem solving, we basically identified the problem area in generation 1 while in generation 2 we dig into the core components of the problem.

So, are we done?

Not yet. We have identified a problem area and causation, but we still are not 100% sure that the high cost is really a problem and not simply a manifestation of the disease condition. If one has a certain condition, one has to spend a certain amount to cure it. No two ways about it, even if it is a high spend.

So the next logical step is to compare your costs to what others are spending with a similar cohort. That is where Gen 3 comes into picture, wherein Real World Evidence (RWE) or population data is brought into the APCD in the form of benchmarks, serving as a barometer of where one is missing the mark. While the cost of care for your diabetic patients with asthma and obesity as comorbid might be a big ticket item, if your cost is lower than most other networks/states, then you can’t do much about it. It must be a causation of the disease condition. On the other hand, you might be spending only a fraction of your overall spend on managing your population’s pneumonia symptoms, but if it’s costing you double on a per capita basis compared to other networks, it is definitely an area of consideration.

Once one identifies the problem areas, analyzes their core components and compares it to the rest to identify the gap, what next?  That brings us to the last and final step of the APCD+ concept, which I consider part of Gen 3 because it’s intertwined with benchmarking. It is the incorporation of best practices and Therapeutic Pathways into the APCD system. That can allow one to figure out how to solve the issue that has been identified hitherto.

So this is my modified concept of APCD:  first, build the baseline financial centric data set, superimpose it with clinical data, compare it to real world evidence (RWE) and, finally, leverage best practices to affect change in the population.

Sounds simple. What do you think?


Rajiv Sabharwal serves as Director, Deloitte LLP. He can be reached at rsabharwal@deloitte.com



Comments (2)
APCDs are playing an increasingly large role in Insurance Exchange creation and administration, health planning activities, transparency, and delivery system reform.

The functions you describe--aggregation/basic analysis (Gen1), complex analysis (Gen2), and benchmarking (Gen3) are all available now through APCDs.

Adding EMR data will undoubtedly add richness (and leading APCDs are already working in that direction, see civhc.org to learn more about Colorado's efforts) but the already-existing power of APCDs are being vastly underused. Leading states, such as Massachusetts, are using their APCDs to fulfill multiple functions: support of HIX, consumer education, setting policy goals, serving as the basis for health planning efforts, benchmarking providers, etc.

Learn more from http://www.linkedin.com/groups/All-Payer-Claims-Database-Group-3362505, www.apcdcouncil.org, or freedmanhealthcare.com.

John Freedman MD MBA
Posted by John F | Thursday, July 11 2013 at 8:22PM ET
I've just decided to create a blog, which I have been wanting to do for a while. Thanks for this post, it's really useful!
Paradise Valley Blinds
Posted by zeeshan t | Sunday, April 06 2014 at 3:29PM ET
Add Your Comments:
Not Registered?
You must be registered to post a comment. Click here to register.
Already registered? Log in here
Please note you must now log in with your email address and password.

Blog Archive for Rajiv Sabharwal

HIMSS13 - postscript
What’s Ahead for Insurance Exchanges?
Will An Aspirin A Day Keep The Doctor Away -- For Me?
SCOTUS and Beyond…
Health Care: A Not Too Distant Future

More from Rajiv Sabharwal »

Blog Index »

loading time...

Stay Connected


HDM Clinical Visionary John Showalter has seen the future of predictive analytics, and it starts right now.

Already a subscriber? Log in here
Please note you must now log in with your email address and password.