Enterprise Information Management

Beside the obvious split between Healthcare and Life Science, the industry is further divided into minor sub categories, such as payers (primarily insurers), providers (hospitals, physicians, clinics etc) and PBM (Pharmacy Benefit Managers). On the life science side, the industry can be further broken down into pharma (drug manufacturers), medical device manufacturers, CRO (organizations that conduct clinical trials) and BioTech (organizations that will ultimately become pharmas provided their research turns successful).

Each of these categories are controlled thru different regulations and hence have drastically different information management requirements. For example the pharmas and biotechs are primarily controlled thru CFR 21 Part 11, a mandate that dictates many diverse issues including electronic signature, electronic record keeping, data management etc. The counterpart for CFR21Part11 on healthcare side is called HIPAA, which in turn controls data storage and delivery issues associated with clinical data.

The basic idea of this pre-amble was to impress upon the reader the complexity associated with data management (sorry, information management) in the industry. Afterall one is talking about data that could make the difference between life and death. There is no other industry (maybe with possible exception of food industry) where a small inconsistency in data stoareg and retrieval could lead to such catastrophes as is possible in HCLS. Imagine not retrieving a patient's allergy data and giving him/her pencillin by mistake. Worse yet, imagine not catching a particular toxin/side effect during clinical trial and mass-producing and circulating the drug. HCLS data management is serious business, literally and figuratively.

Today, I will delve a bit further in some specific HCLS scenarios to give you an idea about where ILM comes in handy before leaving the rest for future blogs, elaborating specific scenarios.

The ILM requirements for healthcare and lifesciences side of the industry are quite different especially from the perspective of business intelligence. Each of them attempt at synchronizing data from large number of underlying sources but in healthcare, the data sources are usually quite structured and are usually intra-organization whereas for life sciences, the underlying data sources are highly diversified and include quite a large portion of external public-domain information sources.

The healthcare organizations (at least on payer side) have largely grown inorganically, a heck of a lot of acquisitions. This has lead to situations where there are drastically different systems performing same basic task across multiple sub-organizations. It is also not easy at all to bring these organizations on a single platform as they each could indivdually cover huge number of members and usually have built very exclusive business rules specific to their business. For example an organization that serviced individual market may have drastically different claim adjudication rules compared to a payer who primarily services employer-based market. Once the orgs merge, it is not easy to merge the systems. The best the payers hope for (the holy grail) is an enterprise wide data access layer (EDL) that is smart enough to access disparate underlying sources and consolidate the retrieved data based on some predefined set of externalized rules. So basically, the EDL is working against a pre-defined data model (payer-specific attributes) and extracting data from disparate yet similar systems. Afterall, all claims adjudication systems require basic attributes such patient name, physician name, disease code, amount etc. On the provider side, the situation is very similar though there one works more against clinical data instead of financial data, as is the case for payers. I will discuss the interoperability, ETL, DW, and BI requirements for providers in detail in my next blog.

On the Life Sciences side, the situation is quite different. There the underlying data sources are not similar at all. Infact the sources may not be structured sources at all. How many of us know of scientists who mainbtain all their data on structured databases? Not many. Imagine some of the most crucial toxicity data for a compound being maintained by a research scientist in an excel spreadsheet (I am assuming even the theoritical scientists do not use tissue papers and napkins any more. Any bets?). Also the scientific community leverages a tremendous amount of research already performed and available in public domain. Obviously one would not want to redefine the double-helix definition of DNA. So now the data storage (more of data retrieval) problem becomes quite acute. We are not only talking about disparate systems, but we are talking about unstructured data, we are talking about public domain info (which is not bound to follow your organization's rules for data storage and retrieval. How I wish that every body had just listened to me, way back in 1995). So here the need is for more of a semantic integration, i.e., use some pre-defined finger-print to search the required data across inter and intra-organization domains, consolidate it against some pre-defined template and produce not '1 of 150Million' but preferably '1 of 20' result set.

I guess, this has gotten too long. Well I will shut up now (how does one do that when one is writing). Next time around I will go into details of using DW-ETL-BI techniques specific to an interoperability platform to setup a Health Information Exchange (the current rage in the provider market).