This document provides an overview of business intelligence, analytics, and their application in healthcare. It defines descriptive, predictive, and prescriptive analytics and provides examples of each. Descriptive analytics describes past performance through metrics like averages and trends. Predictive analytics uses historical data to predict future probabilities and outcomes. Prescriptive analytics indicates optimal actions by incorporating predictive analytics, rules, and models. The document emphasizes how these analytical approaches can simplify data to derive meaningful insights for improving healthcare delivery and decisions.
1. The Discipline of Business Intelligence(Course 3)
Introduction
This section provides an overview of the
discipline of business intelligence, as applied to
the
business of healthcare service delivery for the
improvement of decisions. It includes the types
of analytics and examples, timeframe based
viewpoints that engaging analytics provides and
presents these viewpoints in relation to
decision-making power generated. This section
will also
discuss the concept of decision management,
decision types and analytic tools.
Analysis, Analytics and Business Intelligence
A good deal of discussion and some
confusion exists as the science of working
with data
evolves, especially in healthcare. When there
is talk of engaging business intelligence(BI),
often
a question arises,”What does this encompass?”
Various definitions are put forth and another
question arises, “How is business intelligence
different from business analytics? How does
this
differ from Clinical informatics being engaged
in many healthcare organizations?” And now
2. the
water is very muddy. This will require
filtering, clarity and foundation.
Analytics
The science of analytics-processes for separation
and/or manipulation of data(simultaneously
applying statics, mathematics, operations research
and computer programming) to develop
greater understanding is a more recent idea,
however, its roots are, by extension of the
above, a
branch of logic dealing with analysis.
Analytics also favors the visual display of
information to
communicate insights.
Analysis
At its core, all of this work is analysis,
that is, separation of a whole or complex
idea or topic into
its component parts to gain a greater
understanding. These techniques have been
applied in the
study of mathematics and logic since before
Aristotle.
Business Intelligence(BI)
BI does the same work of simplifying data
to amplify meaning in relation to internal,
structured
data and business processes to transform raw
data into meaningful and useful information for
business (operating versus market competitive)
purposes.
3. Clinical Information
This is the application of information science
and information technology to the delivery of
healthcare services. Clinical informatics is
concerned with information use in healthcare by
clinicians. It is also referred to as applied
clinical informatics and operational informatics.
Much
of the analytics work and output in
healthcare will be through collaborative
effortsofBI/Analytics
practitioners and Clinical Informationists.
AMIA(American Medical Informatics Associations)
considers informatics, when used for
healthcare delivery, to be essentially the same
regardless of the health professional group
involved-whether nurse, physician, pharmacist,
imaging technician or other health professionals.
Clinical informatics includes a wide range of
topics ranging from clinical decision support to
visual images(e.g. Radiological, pathological,
dermatological, ophthalmological, etc); from
clinical documentation to provider order entry
systems, from system design to system
implementation and adoption issues.*
You, the BI/A consultant, should be working
with physician and nursing informaticists
(assuming
your organization has these members on
board). They are heavily involved with
electronic
4. health records(EHR) and other suh system
implementations and these are experts in
clinical/operations workflows which are crucial to
understand in performing many analytic
activities.
Business Intelligence
Business intelligence(BI) is a set of
methodologies, processes, architectures and
technologies
that encompass those three types of analytics.
This section will walk through the continually
emerging discipline as applied to healthcare
service delivery.
What are the types of analytics?
What are the simplifications?
What are the meanings that can be derived?
What are the benefits of this work?
What is the emerging work that couples
analytics and decision management to drive
better
decision processes in the delivery of
healthcare?
Analytics
Now, one might ask:
Why analytics? Why spend the time?
The answer is:
“Scientific management is moving from a skill
that creates competitive advantage to an ante
that
gives a company the right to play the
game.”
5. -Ian & Elizabeth Stephenson, Ten Trends to
Watch in 2006,
McKinsey Quarterly, Jan. 2006
Now we can easily see these trends are
taking shape more quickly than was thought
when the
piece was written. These trends apply to all
industries, however the trend may be amplified
in
healthcare now based on the need to achieve
significant change in a short period of time.
Section 2
Types of Analytics
The three types of analytics presented in
this course are foundational, straightforward and
immediately related to organizational action in
service to customers. BI as a discipline
continues to evolve and other types of
analytics are beyond the scope of this course.
These types of analytics denote what happened,
what could happen, and what one should do
about it. They are not necessarily used in
locked in locked sequence. They all should
be used
as a continuing sequence of stages, tasks or
events that can occur in any direction. Data
limitations and unaccounted-for external forces,
among other issues, can distort or derail
6. output
produced by any of these.
Descriptive Analytics
Descriptive analytics provides an understanding
of past performance. This understanding is
unlocked from the historical data by
summarizing and comparing or mining for
patterns. The
patterns may point to the reasons for success
or failure. Patterns in data described through
the
presentation of data in the form of tables
and charts. Almost all management reporting
falls
under descriptive analytics. It is the
predominant analytic type in use, mostly
because it has
been used for many years and does not
require extensive computational power to yield
significant results. Summarizations and
comparisons are most often made with readily
available tools, such as Excel.
● Percentiles
● Frequency distributions and standard deviations
● Scatter plotting
● Trend lines``
Predictive Analytics
The practice of relating what you do know
to what you do not know provides better
7. information
that is valuable when a decision is needed.
Here, probabilities are applied to historical data
in
combination with rules, algorithms and possibly
external data to determine the probable
outcome of an event or likelihood of a
situation outcurring.
Meaning can be ascribe by:
● Indicating the probability of occurrence
● Pointing out possible actions to address the
patterns
● Indicating possible consequence of action or
inaction
In sum, one can turn uncertainty (what one
does not know) into usable probability and
act based
on the likelihood of occurrence.
Predictive analytics are presented as numbers,
scores, or percentages, depending on the use
of the information.
Prescriptive Analytics
Prescriptive analytics is a very new and still
emerging area that uses hybrid data (which
might
be any of the following: historical, real -time,
internal or external, structured or unstructured
and/or others) along with business rules and
mathematical models to indicate what should be
done. The models may or may not include
a predictive component, an thus, prescriptive
analytics is not necessarily predictive.
8. Decision paths can be presented or decision
implications can be illustrated using a what
if
analysis. As with predictive analytics, output
can be continuously generated and updated to
take advantage of new and better information
as it becomes available.
Descriptive Analytics Examples
● Count of healthcare users-inpatient admissions
● Cost per adjusted discharge
● Children and adolescents with non-urgent
health issues in the ED during primary
physician office hours.
● Average Length of Stay(LOS)
● Days of average LOS above three
● Median days in AR
● Door-to-administration of TPA time measure.
To describe the situation, use all the
following:
○ Range of minutes
○ Mean minutes
○ Median minutes
○ Standard deviation in number of minutes
● Number of cased miscoded
● Number of miscoded cases per shift
● Days cash on hand versus target
● Labor compensation ratio
● Inpatient census per hour per unit
9. ● 30-day readmits
● HCAHPS score
Predictive Examples
● Volume demand for services: ED/inpatient
● Labor hours needed next shift
● Respiratory therapist FTEs needed next March
● Probability of ED being over capacity
tomorrow
● Number of medication pumps needed per
month
● Billings versus collections over the next 30
days
● Unit expenses per month for the next
quarter’s flexible budget cycle
Prescriptive Analytics Examples
● One less RN needed next shift
● Order 200 surgical kit packs #2328 for
use over the next six months with delivery
scheduled
● Always review any case with code 12345A
for coding errors
● Accounts payable schedule
● Services needed per specific admit type
XXX to avoid 30-day readmit
● Inpatient transport to imaging on Mondays
first shift at 3 FTEs
● Increase flex budget for 3 North unit by
3.5%
10. These would be the result of calculations
performed using the methodologies described
rather
than ruleof thumb or estimations or guesses
after looking at descriptive information, which
are
used to make these operational decisions in
most cases currently.
Types of Analytics-Simplifications Provided
These types of analytics provide simplifications
that are easy to understand and use for
business decision-making. We’ll first discuss
descriptive analytics.
Descriptive Analytics
Descriptive analytics is generated by querying,
reporting and online analytical
processing(OLAP) tools and techniques that can
help answer the questions about:
● What happened(through general reporting)
● Why is this happening( by showing patterns
or trends)
Data is presented in terms of:
● Quantities
● Frequences
● Locations
● Classifications
● Combinations
11. ● Comparisons
● Trends
Descriptive Analytics Example
The example shown here indicates averages of
what happened over a specific time period.
This chart invites the consultant to investigates
why volumes on days/times such as Thursday
1300-1500 hours are below all other days of
the week that are otherwise quite uniformly
higher
in volume during this time frame. Why is
variability on all days in this timeframe
except
Thursday so limited when all days of the
week after 0900 and outside this timeframe
widely
variable? Further, it invites analysis to
determine the reasons for target(s) achieved or
missed.
For example, perhaps Thursday productivity is
below the daily averages.
Predictive Analytics
Predictive analytics refers to the skills,
technologies, applications and practices for
exploration
that answers questions:
● What if the trend continues?
● What will happen next? (prediction)
12. These can be generated continuously and/or
iteratively, as in performing what if scenarios.
Predictive analytics specifically applies inferential
statistics, probability testing confidence
intervals in the analysis.
Prescriptive Analytics
Iterative exploration on individual situations
throughout a course of care, or of larger
groups over
time, answers the following questions:
● What actions are needed?
● What is the best that can happen?
For example, “Turning Hospital Data into
Dollars,” in the February 2010 hfm magazine
(see
resource section), provides examples to answer
13. these questions. One related to charge coding
demonstrated the use of predictive modeling to
indicate specific action to the undertaken, or
not.
The case “demonstrates the scale and
comprehensiveness of predictive analytics. The
model
detected, with greater than 90% probability, 18
accounts as missing a charge for implantable
joint device. In this example, the model
analyzed all similar procedures for the previous
12
months and identified, with 93.6% probability,
accounts that, if four conditions were
performed in
conjunction, should have the artificial joint
charge on the bill.” Review action was
called for and
taken in all cases. Others were left alone.
Prescriptive analytics iswhat should be done,
as has been started. This may be based on
predictive forecast with optimization rules
applied to the prediction resulting in one or
more
courses of action prescribed. Predictions or
inferential statistics need not necessarily be
involved, however.
For example, to address central line infections,
a set of known information, such as
healthcare
user initial condition, clinical care rules, a nd
14. factors changing over time, such as healthcare
condition, along with possible corollary effects
based on time from the line’s placement, can
be
built into a rules engine to provide alerts
on current and upcoming central line care
and
recommended changes. This work would likely
be done by a Nurse Informaticist, if the
role and
function existed in one’s organization. If that
role does not exist, get ready….this is your
ball to
carry.
What is the best that can happen? Not
only is more beneficial care provided, the
organization
can understand potential workload more precisely
to ensure adequate numbers of competent
staff are available to perform this work and
other work. Simplifications occurs as individual
tracking of large variables sets is dramatically
reduced and tracking is done on an exception
basis. With or without a Nurse informaticist
this an analytics task of the BI/Analytics
practitioner.
15. Business Functional Areas of Focus
These are color coded as to whether they
are typically possessed currently, are being
developed, or recommended to be developed.
The lists are presented to encourage
consideration of the myriad of possibilities.
Finance
● Financial statements analysis- ratios, as simple
as cost per visit
● Financial ratios
● Statistics and Ration Medians.
● Labor compensation and supply cost ratios
● Key Performance Indicators
● Benchmarks-external/internal
Operations
● Operation management(core business processes):
○ Resource utilization design capacity versus
effective capacity
○ On time first case in surgery-actual incision
time to scheduled start
○ Benchmarks-external/internal
○ Processing flow/throughput/queues length or time
○ Capacity
○ Staff scheduling efficiency/effectiveness
○ Visit, procedure, etc. scheduling
efficiency/effectiveness
○ Unit specific key performance indicators
Clinical
16. ● Healthcare user satisfaction-HCAHPS
● Evidence based medicine pathways
utilization/effectiveness
● 30-day readmission
● DRG/ICD/demographic groupings
● Risk management for HAIs, falls, decubitus
ulcers, etc.
● Clinical specific key performance indicators
● benchmarks-external/internal