Data mining

Health care costs are rising these days.
In fact, it seems everything is more
expensive, impacting your bottom line. But there are actions you can take to
cut those costs and maintain the plan’s
maximum benefit to your employees.

Take data mining, for instance. Done
effectively, says Tim Sullivan, director of
medical informatics and performance metrics for Care Choices in Farmington Hills,
this process can help companies implement fact-based solutions that can become
cost-effective bottom-line management
tools. That includes cutting health care
costs and suggesting more appropriate
medical treatments.

Smart Business spoke to Sullivan about
data mining and the numerous benefits it
can provide.

What exactly is data mining and how is it
useful to a company?

Data mining is the process of analyzing
data and summarizing it into useful information to identify patterns or relationships. It gives an employer and/or health
plan the ability to summarize data in order
to create predictive models that can assist
an employer or plan sponsor in implementing fact-based solutions that, when
implemented, become cost-effective bottom-line management tools.

Data mining allows a user to analyze data
from many different dimensions and angles,
and identify correlations and patterns among
dozens of data fields. It results in useful information that can be used to cut health care
costs, suggest more appropriate medical
treatments, predict medical outcomes, and
assist the plan administrator in designing
more effective benefits among many users.

For example, an employer may want to
use data mining to respond to human
resource/benefit questions like how much
emergency room visits cost the plan. By
evaluating data, the employer may find that
the benefit design in place encourages ER
visits through lower co-pays than physician office visits or the urgent care center.
In this case, data mining can be used to
restructure the benefit to direct high-cost
care to a lower-cost setting.

How long has data mining been around?

Data mining has been around since the
1980s but has become more popular in the
1990s and into the present due to several
factors — mainly, the ability of personal
computers to better handle large amounts
of data. The creation of more advanced
software applications such as SAS has also
aided in the popularity surge.

Increases in health care costs and complexity in health care as a result of
advances in medicine, pharmacy and technology that were experienced in the late
1990s and early 2000s forced employers to
begin understanding how their health care
dollars were being spent, how they could
maximize their budgets, and how they
could improve employees’ productivity
and design benefits that would maximize
those benefit dollars.

Finally, health plans are providing
more integrated care management programs such as disease management and
health coaching than ever before. Data
mining tools are utilized by many health
plans to identify at-risk patients and
intervene with care management programs prior to the member becoming a
high-cost one.

Is there one better way of conducting data
mining over another?

The basics of data mining are that you
have a database and a tool to query the
database and a way to present the data in a
useful format. The database is where you
collect and store data from various sources
(health plans, pharmacy benefit managers,
workers’ compensation carriers, disease
management firms, wellness programs).
There are many tools that can either build
the database, query the database, present
the data, or do all of these tasks, such as
SAS, SPSS, ACCESS or Excel.

Data mining can be accomplished by
either looking at all of the data or by taking
a subset of it. One can take a random sample of the population to create a smaller
subset of the entire population in which
the results can then be extrapolated to the
overall population. The random-sampling
methodology can be beneficial if your
entire population is so large that it would
impact the analysis run-time negatively.

Can a company conduct data mining itself, or
should it rely on an outside party? What are
the advantages/disadvantages to this?

Skills required for data mining and
obtaining useful information from the
mountains of data available can be compared to the same skills required of a detective. To conduct data mining, you need a
database that has significant and detailed
historical data as well as a tool to query the
database. Also, it’s necessary to understand the relationship of the data, to have
the ability to understand and manipulate it,
and to have some knowledge of statistical
methodologies if you intend to compile statistical analyses. Most businesses don’t
have in-house expertise to do this.

TIM SULLIVAN is director of medical informatics and performance metrics for Care Choices. Reach him at (248) 848-2140.