IST7000 TTU Improving the Patient Experience of Care Analysis The three business goals for the data warehouse in the case study are: Improving the patient

IST7000 TTU Improving the Patient Experience of Care Analysis The three business goals for the data warehouse in the case study are:

Improving the patient experience of care (including quality and satisfaction)
Improving the health of populations
Reducing the per person cost of healthcare

The case study concentrates on the advantages and disadvantages of different architectures for data warehouses. In addition to architecture, the data analyst must also determine what data to capture in the data warehouse to achieve the business goals. For example, one way to improve the patient experience (the first business goal) could be to reduce waits times for appointments. In order to understand what causes wait times, data about the scheduled start time, actual start time, scheduled end time, actual end time, availability of the doctor, and reason for the appointment would have to be analyzed in the data warehouse. Other ways to improve the patient experience could include: getting the diagnosis right the first time, knowing the cost upfront, avoiding second appointments with specialists, providing supplemental information for patients to read at home, etc. Each of those improvements would require different types of data in the data warehouse.

Your assignment is to specify at least one improvement for each of the three business goals above. For the improvement you identify, indicate what specific data would have to be stored in the data warehouse to analyze ways to achieve that improvement. Also specify how that data would be obtained.

Your paper should be 1 – 3 pages in APA format. Include at least one paragraph for each of the three business goals above. White Paper
Comparing the Three Major Approaches to
Healthcare Data Warehousing:
A Deep Dive Review
®
by Steve Barlow
Co-founder &
SVP Client Operations
Health Catalyst®
Health systems are being asked to deliver better care and boost productivity
while simultaneously reducing costs and waste due to the U.S. government’s
Affordable Care Act (ACA). Government mandates, subsidies, and health
insurance marketplaces are all part of the mix of incentives to increase
compliance for the ACA and ultimately improve the delivery of healthcare and
reduce Medicare spending.
The task to improve healthcare presents a significant challenge to providers,
health systems, and payers. But according to the Institute for Healthcare
Improvement (IHI), if health systems focus on achieving the objectives
of a framework called the Triple Aim, they will be able to optimize their
performance and meet the government’s requirements. As stated by the IHI’s
website, the Triple Aim objectives are as follows:
Improve the patient experience of care (including quality and
satisfaction)
Improve the health of populations
Reduce the per capita cost of healthcare
Copyright © 2017 Health Catalyst
Before health systems can focus on achieving the Triple
Aim, however, they need to redesign their system with
the following components listed on the IHI’s website: (1)
focus on individuals and families, (2) redesign primary care
services and structures, (3) focus on population health
management, (4) implement a cost control platform, and
(5) achieve system integration and execution.
A healthcare-specific
enterprise data
warehouse provides
complete and accurate
information from across
an entire organization.
Changing health systems to fit this new model is possible.
But it is critical for health systems to choose the most
appropriate data warehouse for healthcare’s specific
needs as they redesign their systems. This is because
the traditional approach of cobbling together reports that pull in data from
the many various source systems is too costly and time-consuming. A better
approach is for health systems to increase their reliance on complete and
accurate information from across the enterprise-wide data ecosystem of their
organization, which requires a healthcare-specific data warehouse. Once
this data warehouse is implemented, health systems will be able to store and
mine their enormous amounts of data to achieve the Triple Aim.
WHY ACHIEVING THE TRIPLE AIM IS CRITICAL FOR HEALTH SYSTEMS
A 2014 report from the Commonwealth Fund reveals that the U.S. healthcare
system is the most expensive in the world and ranks last for access,
efficiency, and equity in comparison to 10 other nations. The 10 nations from
the report include Australia, Canada, France, Germany, the Netherlands,
New Zealand, Norway, Sweden, Switzerland, and the United Kingdom. In
addition, a 2013 report from the Centers for Disease Control and Prevention
(CDC) states that personal healthcare expenditures in the United States
total $2.3 trillion, with expenditures for hospital care accounting for 31.5
percent and physician and clinical services accounting for 20 percent of all
national healthcare expenditures. Yet despite the high cost of U.S. healthcare,
Americans are not any healthier than citizens of other industrialized nations,
nor do they enjoy greater longevity. Consider these facts about the costs of
healthcare and mortality:
The 2013 report U.S. Health in International Perspective compared
the life expectancy of Americans to the citizens of 17 high-income peer
countries from Western Europe, Australia, Japan, and Canada. The
findings showed that life expectancy for American males ranks last, and
life expectancy for American females ranks next to last.
Preterm-related causes of death accounted for 35 percent of infant
deaths in 2009 as stated on the CDC website.
Copyright © 2017 Health Catalyst
2
At least 44,000 and perhaps as many as 98,000 Americans die in
hospitals each year as a result of medical errors according to the book
To Err Is Human: Building a Safer Health System.
Medicare could save at least $12 billion per year by reducing preventable
readmission cases that are readmitted within 30 days according to the
2007 report to Congress Promoting Greater Efficiency in Medicare.
These facts highlight merely a few of the problems healthcare is facing. To
respond to these issues, there are many data warehouse choices being
developed and marketed to health systems. Knowing which model will be the
most effective and provide the best return on investment (ROI) can be difficult
until the advantages and disadvantages of each option are understood. Then,
with this knowledge, health systems will be able to make a well-informed
decision about their investment.
3 TYPES OF DATE WAREHOUSE MODELS: ENTERPRISE MODEL,
INDEPENDENT DATA MARTS, AND LATE-BINDING™ ARCHITECTURE
Currently there are three main types of data warehouses from which health
systems can choose to store and mine their data. The data warehouse models
are as follows: the enterprise model, the independent data mart model, and
the late-binding™ architecture model. While all three models offer a data
warehouse solution, some have unique attributes that make them ideal for
healthcare.
1. Enterprise Model
Bill Inmon, called the “Father of Data Warehousing” on his website,
developed the enterprise model for data warehouses. This is a complex, topdown design that includes the construction of a big centralized data
warehouse from the outset of the planning stages.
By using the enterprise model approach, it is necessary to
determine in advance all of the data elements anyone would
ever need to use for data analysis, such as safety and
In the healthcare
patient satisfaction data. Analysts are forced to make lasting
analytics environment,
decisions about the data model in the beginning without
the enterprise model
being able to plan for changes in the short- or long-term.
is difficult, expensive,
And then they need to structure the database accordingly,
and time-consuming to
which can take months or even years to complete.
construct.
For certain industries, such as manufacturing, banking,
and retail, or when there is a need to design a new
transaction processing system, this model may be
appropriate. But in the healthcare analytics environment, the enterprise
model is difficult, expensive, and time-consuming to construct because data
Copyright © 2017 Health Catalyst
3
EDW
FINANCIAL SOURCES
(e.g. EPSi, Lawson,
PeopleSoft)
DEPARTMENTAL
SOURCES
(e.g. Apollo)
Patience
Provider
Cost
ADMINISTRATIVE
SOURCES
(e.g. API Time Tracking)
Bad Debt
Provider
Survey
Encounter
Charge
Census
Facility
House
Keeping
Diagnosis
Employee
EMR SOURCES
more transformation
Time
Keeping
less transformation
Procedure
Catha
Lab
PATIENT SATISFACTION
(e.g. NRC Picker)
enforced referential integrity
Figure 1: The traditional enterprise model requires significant transformation of the data model from the source
systems (dark blue ovals) as it is loaded into the enterprise data warehouse (light blue rectangles).
architects must first build a comprehensive blueprint of various data elements,
such as medications, labs, and billing data, for example. To further complicate
the process, the blueprint is not necessarily based on data that has already
been captured — just the forecasted data needs. As a result, the data model
may include superfluous data elements. Additional limitations of the enterprise
model approach are listed below.
Delayed Time to Value
The enterprise model approach creates additional expense and time for
the health system because of the considerable transformation required to
force fit the data into the net new data model. This delayed time to value
is a significant downside of the enterprise model approach. For example,
complex calculations and derivations tend to add increased work and time
to an analyst’s job. In comparison, other data models allow for inputs to a
calculation to be loaded directly into a data warehouse, allowing for greater
flexibility and faster delivery of reports.
Copyright © 2017 Health Catalyst
4
Obscures Data Quality Issues
The enterprise model has rigid acceptance criteria for the data, causing the
need to clean and scrub the data each time it is loaded from the primary
system into the secondary system. This method obscures data quality issues
and delays the improvement of data quality issues at the primary source.
For example, suppose a health system wants to measure gestational age for
mothers because studies have shown that inducing labor before 39 weeks
increases the risk of complications. What the health system will discover as it
goes to pull the data from its EMR is that the data has been captured in many
different ways. Some entries may show “39.2,” “39 weeks and 2 days,” “39W
and 2D,” or “thirty-nine weeks and two days.” The variations go on and on,
making it impossible to easily and accurately use the data without a significant
cleansing effort.
Data Is Bound Early
Top-down data warehouse models require early binding of the data. (Data
binding is a technique in which raw data elements are mapped to conceptual
definitions.) Early binding means the data is mapped into a predefined
data model as it is brought into the warehouse, which limits the ability to
make changes to the data in the future. For situations where data rules are
relatively static, nonvolatile, and do not frequently change, early binding
may be appropriate. Industries that employ early binding models include
manufacturing, communications, retail, and financial services.
When health systems try to bind every data element to business rules early,
however, they face a time-consuming and expensive approach to data
warehousing. It is also difficult
to make changes to the data.
This is because business rules
and vocabulary standards
in healthcare are among the
most complex in any industry,
and they undergo almost
constant change, resulting in
high volatility.
In fact, there are only a limited
number of core data elements
that should be bound early
because they are fundamental
to almost all analytic use
cases. Because they are
fundamental and not volatile,
it is appropriate to bind those
Copyright © 2017 Health Catalyst
Figure 2: This illustration details the six points in a data warehouse
where data can be bound to rules and vocabularies. Rules and vocabularies with low volatility can be bound at points 1 and 2 (early binding);
points 4 and 5, are appropriate for those with high volatility.
5
data elements early. All other data elements should be bound late (i.e.,
when clinicians are trying to solve a problem) because of their volatility.
For example, length of stay (LOS) in a hospital may sound straightforward
on paper, but surgeons might define LOS as point of incision to discharge
from the post-anesthesia care unit (PACU), and cardiologists might define
it as emergency department (ED) arrival to discharge. Because the LOS
definition will change for different use cases, the objective is to bind it later.
The following examples show which types of data are volatile and which
types are not volatile.
Volatile data that should be bound late:
Calculating length of stay (LOS)
Attributing a primary care provider to a particular patient with a chronic
disease
Calculating revenue (or expense) allocation and projections to a
department or physician
Data definitions of general disease states for patient registries
Defining patient exclusion criteria for disease and/or population
management
Defining patient admission, discharge, and transfer rules
Nonvolatile data that may be bound early:
Facility identifier
Provider identifier
When there is boundless
scope and complexity
with a data warehouse
project, the final product
may miss the mark
or completely fail to
deliver the functionality
leadership expects.
Patient identifier
Gender
Date
Time of arrival
Boundless Scope
Healthcare business processes can be complex, and
teams can spend anywhere from six months to multiple
years mapping their organization’s information systems
to a single enterprise-wide data model. To account for this intricacy, the data
model can become enormous in scope and complexity and may miss the
mark in terms of having the functionality leadership expects. In fact, Gartner
estimated in 2005 that as many as 50 percent of data warehouse projects
Copyright © 2017 Health Catalyst
6
would have only limited acceptance or fail entirely through 2007. When a
project fails, a lot of money may be spent on something that is never used or
never even gets launched.
2. Independent Data Marts
The independent data mart approach to data warehouse design is a bottomsup approach to data modeling.
With this data model approach, the organization starts small, building
individual data marts as places to store specific information for each hospital
department. Then the independent data mart draws further departmentspecific data from the various primary source systems to provide the data that
each department needs.
FINANCIAL SOURCES
(e.g. EPSi, Lawson,
PeopleSoft)
DEPARTMENTAL
SOURCES
(e.g. Apollo)
Regulatory
Labor
Productivity
ADMINISTRATIVE
SOURCES
(e.g. API Time Tracking)
Revenue Cycle
Pregnancy
Oncology
Heart
Failure
Asthma
Redundant
data
extracts
Diabetes
Census
EMR SOURCE
more transformation
PATIENT SATISFACTION
SOURCES
(e.g. NRC Picker)
less transformation
Figure 3: With the independent data mart model, an organization builds an analytic data mart for
a particular department — such as heart failure — gathers the data it needs directly from
the source systems, and maps it to different areas.
Using the independent data mart model approach, analysts can create and
build analyses and reports, which are then combined to form a large data
warehouse. For example, if there is a need to analyze revenue cycle or
oncology, then a separate data mart is built for that specific department.
Copyright © 2017 Health Catalyst
7
Data is then loaded into this new independent data mart from the primary
source systems to support this new subject area.
There is a benefit to the independent data model approach: it takes less time
for the organization to build the data warehouse, and analysts can start to
analyze data quickly — a big difference from the two- to five-year lifecycle of
the enterprise model. However, it grows very quickly, as do the data streams,
until several redundant streams exist. This creates a challenge for those
trying to maintain the model. If one underlying source system changes, they
have to change each extraction routine that uses that particular source.
The many isolated data marts also means there is not an atomic-level data
warehouse from which to build additional data marts in the future. There also
several other major drawbacks to this model, which are explained below.
Lacks Granular Data
Often, independent data marts do not contain data at the lowest level of
granularity or patient-level detail. (In data warehousing, granularity refers to the
level of detail stored in a database and how that level relates to other data. For
example, one database list might store patient names; another list might store
individual patient encounters, a finer level of detail.) Instead, data transformed
in a data mart is usually summarized up one or more levels from the lowest
level of granularity. This means the data mart may present information about a
certain metric falling below the benchmark, but it does not contain the granular
data that enables the analyst to dig down and determine why that metric is low.
Without that more detailed information, it is difficult to make the data actionable
and to determine how to bring that metric up to the benchmark.
Lacks Efficiency
The independent data model causes source systems to be repetitiously and
unnecessarily bombarded by data extracts, which slows down the system.
Redundant feeds from each source system need to be built to feed each
independent data mart. This creates a challenge for those trying to maintain
the model. If one underlying source system changes, then each extraction
routine that uses that particular source needs to be changed. It also results
in a significant drain on the system. Imagine building a new feed from the
electronic health record (EHR) into every data mart built: heart failure,
pregnancy, asthma, diabetes, oncology … and the list could go on and on.
The single-purpose “point” solution may have addressed the immediate need,
but as the demand for analytics grows, the mass of redundant feeds from the
same primary systems creates a solution that is difficult to maintain.
Requires Early Binding of Data
Like the enterprise model, the independent data source model requires early
binding of the data. As data is brought into each independent data mart, the
Copyright © 2017 Health Catalyst
8
data is mapped into the predefined data model, a process called conformance
and normalization. The terms imply exactly what is required: data that was
modeled and captured in disparate source transaction systems must conform
to a new data model in the data warehouse.
While at first this approach might appear reasonable, in
practice, it leads to major problems when applied to the
healthcare industry. This is because the data environment
is much more complicated than a sales receipt in the
retail industry, and the analytic use cases are constantly
changing. For example, the process of mapping and
conforming data to these early binding models in a
healthcare delivery data warehouse typically takes 18 to
24 months or longer. When new data sources are added to
the data warehouse — as occurs in mergers, acquisitions,
and ACO partnerships — this lengthy time- to-value is
repeated again and again.
The healthcare data
environment is much
more complicated than a
sales receipt in the retail
industry, and the analytic
use cases are constantly
changing.
Likewise, as the complexity of analytic use cases inevitably matures in an
organization, the early binding data model must be modified and the source
system data must be conformed and mapped again. These early binding data
models cannot keep pace with the changes in the analytic environment and
the data warehouse subsequently fails to deliver.
3. Late-Binding™ Architecture
A new approach to data modeling to address healthcare’s
unique data needs is Health Catalyst’s Late-Binding™
architecture (Figure 4). Binding data later means delaying
The Late-Binding™
the application of business rules, such as data cleansing,
architecture is ideal for
normalization, and aggregation for as long as possible.
what-if scenario analysis
This provides health systems with time to review and
and is best suited for everrevise data, form hypotheses, and determine optimal
changing healthcare data
analytic uses. In addition, there is no longer a need to
and evolving use cases.
make lasting decisions about the data model upfront,
which is useful, since it is difficult to see what new
information will come down the road in two, three, or five
years. By binding late, analysts only need to bind the
data when there is an actual clinical or business problem to solve. The LateBinding™ architecture is especially ideal for what-if scenario analysis and is
best suited for ever-changing healthcare data and evolving use cases. Similar
to just-in-time binding, Late-Binding™ works like this:
1
First, data in its atomic form is brought from the source systems to
the source marts of the data warehouse. There is very little data
Copyright © 2017 Health Catalyst
9
transformation that occurs at this point, but some transformation, such
as the variation in naming standards by the source systems, will be
resolved to a single standard. For example, the term “patient” from
one source system may be titled “PT_NAME” and from anothe…
Purchase answer to see full
attachment