Editor’s note: This article is by Dr. Shahram Ebadollahi,
senior manager of Healthcare Systems and Analytics Research at IBM Research.
The more longitudinal medical data records
that are becoming available should mean that healthcare providers – from nurses
and public health officials, to specialists – have more insight into helping
solve their patients’ problems in the here and now. But now, the challenge is how
to elegantly analyze all that data and derive insights from it to help those
providers deliver better care to their patients.
My team at IBM Research developed the foundational analytics
for a healthcare solution, now called IBM Patient Care Insights.
These analytics can take into account all patient characteristics, such as treatments,
procedures, outcomes, costs, etc. – basically everything about a set of
patients that could be observed and captured over time (even the unstructured
information, such as a doctor’s notes on a chart).
The data, in a sense, captures the collective memory of the care delivery system and embedded in it
are insights about all the procedures and outcomes for all the patients. The
analytics that can help us extract that insight promises to lead to better, more-efficient,
and lower cost patient care.
 |
| How IBM Care Insights derives insight from population data to better personalize decision making. |
Medical data: analyzing, visualizing, predicting
So, how does Care Insights make
sense of years of data, from multiple sources, about thousands of people? All
to give healthcare providers a way to identify treatment and early intervention
options.
Care
Insights’ suite
of tools use innovative algorithms rooted in machine learning, data mining and
information retrieval techniques to look for patient similarity to derive tailored insights regarding a custom
course of actions that are delivered through easy-to-understand visuals.
The Patient Similarity
analytics tool finds all patients who display similar clinical
characteristic to the patient of interest. The resulting individualized insight
includes suggestions on how to manage care delivery to the patient, but perhaps
more-importantly predicts health issues that could arise in the future (because
patients with similar characteristics had experienced such health issues).
It can then match patients to specific physicians or
specialists who can potentially provide a better outcome; understand and analyze
utilization patterns (utilization of resources in the care delivery network) of
patients, and identify abnormal utilizations (over utilization or under
utilization), which could be an indicator of a potentially poor outcome or
unnecessarily high cost.
The similarity analytics suite of tools can also predict the
patient’s potential future adverse outcomes and conditions. Therefore, it can
identify opportunities for early intervention.
 |
| Visualizing the evolution patterns of patients with similar attributes to the patient of interest. |
What about Watson?
Watson can provide tailored and to-the-point answers with
supporting evidence to questions based on the corpus of knowledge it is
connected to. IBM
Care Insights complements
this knowledge-driven evidence, obtained from medical knowledge sources, with
data-driven insights derived from the large patient population medical records
discussed here.
Labels: analytics, electronic health records, healthcare
