Hello everyone and welcome back.
In my health care career,
I've had the opportunity to work in a variety of sectors and domains.
These areas include clinical research,
population health care improvement,
hospital profiling, quality improvement,
and review of service utilization.
In all of these areas I was asked to perform analysis that involve
questions about specific groups being at higher risk for some type of outcome.
For example, in clinical research what treatments lead to higher risk of complications.
With population health, what subpopulations are at
higher risk to experience some type of expensive or low quality medical outcome?
Concerning hospitals that might have higher mortality rates,
did they really have sicker patients making the risk of mortality higher?
For utilization, do patients that costs
the health plan the most have the most diseases that put
them at risk or of having future costs or
will most of them get healthy and not require interventions?
All of these questions have different analytical requirements, but in general,
these all requires some concepts from
risk evaluation and all of them require health care data.
Thus I'm excited to share with you what I know about this area.
I think that you also have the opportunity to get involved with these types of projects.
I have numerous objectives for this module.
In the first three lessons,
I will introduce you to the topic of risk stratification.
This is a common analytical tasks within healthcare analytics.
Thus, it is important that you have a good understanding of
the key concepts associated with this methodology.
I will cover a few topics.
First, I'll review the steps to stratify populations of
patients or members by risk of experiencing events of interests.
Second, although we have covered groupers in the past lesson,
it is very important to understand how groupers are used for risk stratification.
Third, we have already discussed model evaluation and accuracy.
However, I will say more on the context of accuracy
for predictive models in the domain of risk stratification.
Finally, as with all data analytic projects,
data issues can lead to important pitfalls,
it's important to think about data quality or missing data
that can have important impacts when performing risk stratification.
After that, I will continue the discussion about
medicare claims data and describe an altered and de-identified version of these data,
the CMS DE-SynPUF samples.
These claims data started as real medicare claims but were then altered to create
a de-identified sample of data that can be shared with researchers and analysts.
Although termed synthetic data,
these samples are close enough to the original samples for
researchers to use them with caution to learn about the medicare population.
My objective is to explain the data and provide
some data dictionaries so that you can perform some simple analysis,
this will allow you to apply groupers to real data.
It will also expose data weaknesses and
inconsistencies that come when analyzing claims data.
Finally, in the last lesson,
I want to offer some tips and reminders
about understanding and interpreting healthcare data.