Visualize Granularities ofVisualize Granularities of Clinical Data …sites.mckesson.com/MCCA/user_forum_downloads/Visualize... · 2017. 6. 16. · Pivotting the data to aggregations

Visualize Granularities ofVisualize Granularities of Clinical Data and Affect ChChange

Thomas Carlough, PharmDPharmacy Services Data Coordinator

Oh the things we’ll talk about

1. Suffer the Past ProcessesWalk through the technology and process before we adopted McKesson Analytics Explorer TM

2. Celebrate the Present SolutionsSee the insights that our team was able to describe with better visualizations

3. Explore the Future PossibilitiesLook at what we are planning for the future of this type of data visualizingg

Atlantic Health SystemNorthern New JerseyNorthern New Jersey- 4 Medical Centers

- Morristown Medical Center- Overlook Medical Center

N t M di l C t- Newton Medical Center- Chilton Medical Center

Should you listen to me?//aboutme//aboutme

Background in Computers• Programming since 1995

Background in Healthcare• Working for hospitals since 2004

Background in Informatics• 4 years Healthcare Informatics experience• Professor of Medical Informatics with Farleigh

Dickenson University in NJ

Other More Tangible Objectives

Understand the value of analytics and data visualization for clinical services

Discover how data visualization can allow you to more quickly identify process issues

Realize the power of combining multiple data sets into a single analysis to achieve greater altitude over a complex situation

So… Where do We Start?

Larisa Koshkina: http://www publicdomainpictures net/view image php?image=57458&picture=roadLarisa Koshkina: http://www.publicdomainpictures.net/view‐image.php?image=57458&picture=road

The Joint Commission (TJC) Center for Transforming Healthcare (CTH):for Transforming Healthcare (CTH):Safe and Effective Insulin Use

Aim to solve critical safety and quality problems in healthcare

Solutions are created using systematic approach to process breakdown (Six Sigma)

9th project by the CTH of which Morristown Medical Center (Atlantic Health System) is 1 of 5 participants

What was the big question this project was trying to answer?

Why, Oh Why?

This is important:• > 1 in 5 patients have diabetes (Common)• Insulin errors can be life threatening (Dangerous)

Why, when surrounded by healthcare professionals, do patients go hyper and hypoglycemic?

Source: Joint Commission Center for Transforming Healthcare CTH Insulin Fact Sheet :: http://www.centerfortransforminghealthcare.org/assets/4/6/CTH_Insulin_Fact_Sheet.pdfRetrieved 6/26/2015

Patients in Defect Status?

It sounds cold; but its just Six Sigma

Any Blood Glucose• > 180• < 70

Paul Brennan : http://www.publicdomainpictures.net/view‐image.php?image=89051&picture=patient‐sign‐at‐hospital

Data We Have, Data We Need

We have:• Laboratory Data: BG Result, Date/Time

We need to identify only:

• Inpatients• Those on Insulin• Adults Only

Data Issues

Lab data and PIS data, but cannot be directly combined

Lab data has BG results

PIS has patient information and medication orders (Inpatient/Adult/Insulin)

No way to use SQL to query both at once since data access was not universal

How can we do a crosswalk?

Excel!

Old School

VLOOKUP

Not a bad option if you are doing it one time

Not sustainable

Petr Kratochvil :: http://www.publicdomainpictures.net/view‐image.php?image=5406&picture=telescope

The Fun Part Take this list of 77 data points and then correlate on defect status

Celebrate the Present

Now you know how it “worked” for version 1

Let’s look at what wecan do with better tools

Starting from the Top

The most important data point was the POC glucose result

Over 16 months: 520,875 results

Now in the improve phase• Were We Improving?• Holding Steady?• Were We Regressing?g g

Defect Rates (Cross Tab)

This doesn’t look like much I know but the raw data lays just below the surfaceraw data lays just below the surface

234,742 blood glucose measurements

Defect Rates (Simple Stratification)

B th By month

Mean POC by STAYMean POC by STAY

It is just as easy as it sounds

Pivotting the data to aggregations (instead of raw data points)Pivotting the data to aggregations (instead of raw data points) More concise population metric for a patient Think of this as their “Final Course Grade”

Mean POC by STAY

Pretty successful over the course of a hospitalizationTime to get down from the 30,000 foot view

Mean POC by DAY My initial thought (remember I was a newbie to this) was “tricky tricky”.

You can pivot with two row IDs Hospital # Lab Date

Mean POC by DAY

Look at the difference

Questions abound, hypotheses?

Patient Trend Mean (Patient Detail)

Patient Joe Smith

Hospital Day Mean

Here’s all of our patients in aggregate.The X Axis is by “Hospital Day”

Sliding Scale Insulin?

Glycemic Events

Hyperglycemic Events >>>(frequency)>>>

Hypoglycemic Events

Extreme Hyperglycemic Events• >300 BG – Potential clinical sequelae

Easily Maintained within the Data Set

The use of the “Calculated Column” feature was utilized to produce four columns• Hyperglycemic Events: If(Lab.Result>180,1,0)• Hypoglycemic Events: If(Lab.Result<70,1,0)• Extreme Hyperglycemic Events: If(Lab.Result>300,1,0)• Severe Hypoglycemic Events: If(Lab.Result<50,1,0)

Easily Maintained within the Data Set

The result was an integer between 0 and 1, 1 if the event existed and 0 if it did not

Using this method allowed us to add a summation to our pivot for our populations

Time to Expand Our Reach

Encounter Data• Includes bed management / registration• Includes diagnoses• Includes prescriber detail

Clinical Data• Includes home med list• Includes MAR• Includes electronic form usage statistics• Meal Data

Encounter Data – Patient Location

Encounter Data – Service

Encounter Data – Diabetic

Diabetic Patients Non‐Diabetic PatientsDiabetic Patients Non Diabetic Patients

Encounter Data – Prescriber

Physician Group

Deployment of a Dashboard toDeployment of a Dashboard to Nursing Management

Service Line Performance

How Do You Ask Why?How Do You Ask Why?Measuring Process

Understanding outcomes and affecting change requires measuring the process

Investigations• Prandial Insulin use• Basal Insulin use• Basal Insulin dose changes• Overuse of sliding scale

Finding the WhyFinding the WhyDigging into the Detail

Importing raw MAR and patient medication profile data turns this analytics software into an investigation tool

Matching account numbers on pivoted aggregation data allows diving in

George Hodan :: http://www.publicdomainpictures.net/view‐image.php?image=36643&picture=games‐on‐the‐beachGeo ge oda ttp // pub cdo a p ctu es et/ e age p p age 366 3&p ctu e ga es o t e beac

Investigation Panel for Clinicians

Future Process Measures

Electronic form Usage (data already available and integrated)

Appropriate Basal dose changes after correctional insulin use• Challenge: Capture correctional dose for day N and

prepare boolean (T/F) if basal dose is the same Timely administration of insulin

A i i i li ti l ?• Are we giving our insulin timely? Timely capture of BG measurement

• Are our BG measurements being performed too close to l ?meals?

Are we correcting insulin with Sliding Scale too often?• Percent administrations vs. POC checks

The Road Ahead

Data Governance• Finalizing definitions and populations across the health

system• Organizing data sources so we are all using the same

information Analytics Infrastructure

• New hires• Reorganization• Culture of analyticsy• Very exciting!

Success Measures

Initiative Measure Outcome

Cost Cost of Adverse Drug Events Not yet Cost Cost of Adverse Drug Events quantified

Quality Defect Rate Reduced!!y

Utilization Excess Days associated with poor glycemic control

Not yet quantified

Efficiency Data at everyone's fingertips when they want it Priceless!

Q estions?Questions?

ugh ©

t :: M

LinkedIn: www.linkedin.com/in/thomascarlough

LinkedIn: www.linkedin.com/in/thomascarloughTwitter: @EtherAnalyticsEmail: Thomas.Carlough@AtlanticHealth.org

Images in this Presentation

All images in this presentation are within the public domain, many are sourced through http://www.publicdomainpictures.net

• I have no affiliation with this website but appreciate public domain and creative commons artists.

All graphs and visuals otherwise included were created by McKesson Performance Analytics ExplorerTMy p

Thank YoThank You

LinkedIn: www.linkedin.com/in/thomascarloughTwitter: @EtherAnalyticsEmail: Thomas.Carlough@AtlanticHealth.org

Visualize Granularities ofVisualize Granularities of Clinical Data …sites.mckesson.com/MCCA/user_forum_downloads/Visualize... · 2017. 6. 16. · Pivotting the data to aggregations

Documents

How to Visualize

Reading Strategy: Visualize

Visualize your success

Intelligy...SOLIDWORKS Visualize Top 5 Professional-Only...

pdf (visualize/download)

LucidShape Visualize Module datasheet - Synopsys€¦ ·...

CIS630 1 Penn Different Sense Granularities Martha Palmer,.....

Visualize This!

Supporting Imprecision in Multidimensional Databases Using....

SOLIDWORKS 2016 Visualize

How to Visualize Your GPU-Accelerated Simulation...

SOLIDWORKS SOLIDWORKS VISUALIZE

Integrating Physician Pti DtitPractice Data into...

SOLIDWORKS VISUALIZE

Why Visualize?

PA Tues 1050 Model Builder Custom Spotlight Brian Belkin...