As a striking illustration of the
potential dangers of inadequate data modeling, insufficient granularity in
Medicare's current system of diagnosis-related groups (DRG's)
has caused revenue shifting over the past decade away from cash-starved urban
medical centers to smaller facilities. Many billions of dollars are involved.
This revenue shifting is due to the fact that the current 491 DRG categories do
not adequately capture the differences in severity of illnesses within a single
DRG, resulting in underpayment to some providers (i.e., those that have the
capability to handle very sick patients, often transferred from small
facilities) and overpayment to others (small facilities that keep the simpler
In fact, it is proposed that the number of DRG's be increased by almost one hundred percent, to a total of 900 groups. This reflects not an adjustment, not an iterative change, but a stunning complete overhaul of a defective DRG system. It is especially important to note that healthcare has not recently doubled in its complexity! (See "Panel may support doubling of DRG’s", Modern Healthcare, 31 Jan 2000.)
Such problems are complex, and it may be unfair to say the DRG situation might have been prevented. However, it is clear that correct modeling the medical world is not a process of "luck." The highest levels of medical informatics expertise must be deployed appropriately (in leadership roles) in organizational or national data initiatives that determine resource allocation, best practice determination, drug development, compensation, and other major healthcare decisions.
Yet, advertisements for healthcare and biomedical data project leadership often appear similar to the following:
"Data Architect: As a leader in data modeling and development for the company, this position entails supporting software development activities and internal database usage. Primary responsibilities logical and physical data modeling, data application development, supporting of R&D projects through the vending of data services (SQL development, data transformations, database design & development, etc.), and limited DBA functions. Candidates must have real experience in database design (1-3 yrs), advanced SQL (2+ yrs), implementing concurrent large distributed databases, dimensional data modeling, developing data-driven applications, and the ability to work on a team. Additionally, knowledge of Oracle 7+ (UNIX & NT), and Business Objects WebIntelligence Query tool, are highly desirable. Requires a BS in IS, or equivalent."
medicine, healthcare, or any biomedical field are considered optional or
unnecessary. As in the example above, often only a bachelor's degree in
"Information Systems" is called for. Unfortunately, optimal
healthcare data modeling requires a background in both biomedicine and
information science. Specific software or hardware package experience is far
less important, yet this is usually unrecognized in the job formulation and
hiring process. (It should also be noted that experience in management information
systems is not the same thing as experience in information science
and scientific computing).
The common blindness to the primacy of the medical modeling process over specific software and hardware skills in hospitals, industry and government is astonishing to many medical informaticists. They understand that medicine is of unparalleled informational complexity, and that mis-categorization or mismodeling of the healthcare environment can have significant, unexpected consequences on an individual, organizational or societal scale.
Such blindness usually stems from a common business and MIS belief that "medical data is like everybody else's data" and that "if it's data, we can do it." Instead of selecting individuals optimally, high-level cognitive abilities such as biomedical data modeling are sacrificed for fluency with the latest whiz-bang software platform. Interestingly, the business and MIS leaders who adhere to such beliefs and practices almost always have no clinical experience or training.