why we need datamart? - beckershospitalreview.com b/3_tu… · •helps them to "drill...

17
7/8/2015 1 Building Data Warehouse for Research, Reporting and Quality. 6 Years of ICU DataMart Experience Vitaly Herasevich, MD, PhD, MSc Associate Professor of Anesthesiology and Medicine, Department of Anesthesiology, Multidisciplinary Epidemiology and Translational Research in Intensive Care (M.E.T.R.I.C.) [email protected] Jul 2015 Why we need Datamart? EHR Herasevich et al. Medical informatics in ICU., in Principles of Critical Care, 4th, 2015

Upload: phungngoc

Post on 19-Jul-2018

213 views

Category:

Documents


0 download

TRANSCRIPT

7/8/2015

1

Building Data Warehouse for Research, Reporting and Quality. 6 Years of ICU

DataMart Experience

Vitaly Herasevich, MD, PhD, MSc

Associate Professor of Anesthesiology and Medicine,Department of Anesthesiology,

Multidisciplinary Epidemiology and Translational Research in Intensive Care (M.E.T.R.I.C.)

[email protected] 2015

Why we need Datamart?

EHR

Herasevich et al. Medical informatics in ICU., in Principles of Critical Care, 4th, 2015

7/8/2015

2

Data volume before and in ICU

Microbiology, labs, medications, chest X-ray, Nurses flowsheet, Clinical notes (history and impression/plan) – Vitals excluded

Herasevich V, Litell J, Pickering B. Electronic medical records and mHealth anytime, anywhere. Biomed

Instrum Technol. 2012 Fall;Suppl:45-8. PMID: 23039776.

Average data points per day

Per Patient Per 24 bedded ICULabs 60 1440

Drug Orders 10 240Microbiology 2 48

X ray 2 48Vitals 1950 46800

Why we need Datamart?

• 1) So much clinical data

• 2) Physically data stored in different databases

• 3) Make some sense out of data…

©2011 MFMER | slide-5

BIG data, data minng…

7/8/2015

3

©2011 MFMER | slide-7

The market for analytics solutions is not small — more

than 100 vendors currently offer big data tools and products.

©2011 MFMER | slide-8

http://www.tylervigen.com/view_correlation?id=1597

Association is not causation

Big Data = Predictive Analysis = Data Mining

•Data Mining is an analytic process designed to explore data (usually large amounts of data -

typically business or market related - also known as "big data") in search of consistent

patterns and/or systematic relationships

between variables.

•The ultimate goal of data mining is prediction

- and predictive data mining is the most common type of data mining and one that has the most direct business applications.

©2011 MFMER | slide-9

7/8/2015

4

Myths of data mining

• Myth #1: Data mining provides instant crystal ball predictions

• Myth #2: Data mining is not yet viable for medicine

• Myth #3: Data mining requires separate, dedicated

database

• Myth #4: Only PhDs can do data mining

• Myth #5: Data mining is for large companies with lots of customer data

Data mining is not:

• Data mining is a tool, not a magic.

• Data mining will not automatically discover solutions without guidance.

• Data mining will not sit inside of your database and send you an email when some interesting pattern is discovered.

• Data mining may find interesting patterns, but it does not tell you the value of such patterns.

• Data mining does not infer causality.

©2011 MFMER | slide-11

What can data mining do?

• Helps to determine relationships among "internal" factors such as price, product positioning, or staff skills, and "external" factors such as economic indicators, competition, and customer demographics.

• Helps to determine the impact on sales, customer satisfaction, and corporate profits.

• Helps them to "drill down" into summary information

Primarily used today by companies with a strong

consumer focus - retail, financial, communication, and marketing organizations.

©2011 MFMER | slide-12

7/8/2015

5

Common Data Mining Applications

• Market analysis

• Risk analysis and management

• Fraud detection and detection of unusual patterns (outliers)

• Text mining (news group, email, documents) and

Web mining

• Real time data mining

• DNA and bio-data analysis

“Big Data” Applications in Health Science

• Drug discovery and functional genomics

• Analysis of DNA micro-array data

• Gene, disease and drug interaction

• Genomics, proteomics and metabolomics

• Biomedical text mining (finding relations between experimental data and published literature)

• Estimating outcomes of patients

• Epidemiology

• Data mining electronic patient records

ICU Datamart (METRIC Datamart)

7/8/2015

6

Critical Care at Mayo

208 ICU beds

Radiology Reports

RIMS

Nursing Flow Sheet

MICS Lastword

Clinical notes

MCLS Lastword

APACHE

APACHE

Historical

REP

CPOE

Enterprise orders

ICD-9

DSS

Past history

PPI

OR Data mart

HL7

Microbiology Reports

HRBS

Labs

HRBS

Monitored data

Chart+

Transfusion Orders

MYSIS

Drug orders

HRBS

ICU demographics

HRBS

Fluids: in/out

Chart+

Emergency acute area

YES

Surgical schedule

Surgical

ICU Data mart

Key Facts• ~ 15,000 admissions per year• ~ 1,000,000 vital records per week• Data available from 2003• Updated every hour in average (15 min for vitals)

• Near real-time

Herasevich V, et al. ICU data mart: a non-iT

approach. Healthc Inform. 2011;28(11):42,

44–5. PMID: 22121570

METRIC datamart workflow

Li M, Pickering BW, Smith VD, Hadzikadic M, Gajic O, Herasevich V. Medical informatics: an essential tool for health sciences research in acute

care. Bosn J Basic Med Sci. 2009;9 Suppl 1:34–9. PMID: 19912124

7/8/2015

7

Some available data

2001 2002 2003 2004 2005 2006 2011 2015 Monthly average

Drugs orders 90,000

Radiology reports 3,000

Laboratory tests 330,000

Transfusions 6,000

Microbiology tests 3,000

Vital signs (150 var.) 6,000,000

CPOE orders 370,000

Demographics 1,300

Fluids (intake/output) 220,000

Approach

7/8/2015

8

Rule zero

Rule one: lego bricks

7/8/2015

9

Rule two: UNIXsh - no user interface

• No formal web/query Interface

• ODBC connection allows query from any app (JMP, Excel, SAS…)

Approach: technically

• SQL server with institutional support

• Tables divided by years

• In “Current tables” only patients who in currently in ICU

• EAV (entity – attribute – value) structure

• Continuously “Testing – production”

• Test –> production DBs

7/8/2015

10

Real time monitoringStatistical control

Data integrity

• Near 100% accurate

Validation is key.

Physiological parameters

Herasevich V, Pickering BW, Dong Y, et al. Informatics

infrastructure for syndrome surveillance, decision support,

reporting, and modeling of critical illness. Mayo Clin Proc

2010;85(3):247-254. (PMID: 20194152)

Herasevich V, Kor D, Li M, et al. ICU Data Mart: A Non-IT

Approach. Healthcare Informatics 2011;28(11):42-45. (PMID:

22121570)

7/8/2015

11

Areas of implementation

APACHE replacement project

APACHE replacement1994 - 2009

7/8/2015

12

Free text search for medical admission diagnoses

Chandra S, et al. Mapping physicians’ admission diagnoses to structured concepts towards fully automatic calculation of acute physiology

and chronic health evaluation score. BMJ Open. 2011;1(2):e000216. PMID: 22102639

Clinical reports

Effective management

Joint Commission on Healthcare Organizations (JCAHO) measurement

of ICU performance.

• Mortality report

• Length of Stay Review

• ICU Death Review

• ICU admission Low Risk Monitor Review

• ICU Readmission Review

7/8/2015

13

METRIC Reports

• Monthly reports• Ad-hock reports

• Customized reports

1. Hospital Length of Stay for ICU Graduates –Unadjusted

2. ICU Length of Stay – Unadjusted

3. ICU Length of Stay – Adjusted

4. ICU Readmission Rate

5. ICU Admissions

6. ICU Admission Source and Service

7. Duration of Mechanical Ventilation

8. ICU Mortality Rate – Unadjusted

9. Hospital Mortality Rate – Adjusted

10. ICU Admissions for Low-Risk Monitoring

11. ICU Census - Hourly Utilization

AWARE real time administrative

dashboard

© 2014 Mayo Foundation for Medical Education and Research

Sniffers

7/8/2015

14

Sniffers – rule based DSS

Herasevich V, Pickering BW, Dong Y, Peters SG, Gajic O. Informatics infrastructure for syndrome surveillance, decision support, reporting, and modeling of critical illness. Mayo Clin Proc. 2010;85(3):247–54. PMID: 20194152

SniffersOR Datamart

1. No Temp: If the temperature recordings were not started after 45 minutes into OR location.

2. Hypothermia: If there are three consecutive low temps < 35 after 60 minutes into OR location.

3. VILI – High Tidal volume: If there are three consecutive high tidal volumes after 60 minutes into OR.

4. VILI – Peak airway pressure: if there are three consecutive peak airway pressures > 35 cm H20 after 60 minutes into OR.

5. Hypoglycemia: If the patient glucose level dropped < 70

6. Glucose Check alert: If the patient had insulin administered and had no glucose test done within 2 hours from the last glucose test.

7. Pressor alert: to alert the covering anesthesiologist when thresholds are exceeded for intravenous pressor support.

8. IV Administration of greater than 10 ml pressor (100 mcg/ml phenylephrine and/or 50 mg/ml ephedrine) in a 30 minutes period.

ICU datamart

1. Hip Arthroplasty Study Alert : patients list who scheduled the hip arthroplasty surgery. Knee Arthroplasty Study Alert: patients list who scheduled the Knee arthroplasty surgery.

2. Surgery Glucose Study Alert: Patient list who scheduled Thoracic surgery with glucose problems.

3. Pepsin Study Alert: Patients list who scheduled Thoracic and other surgeries. For Blood Draw study.

4. Septic-sniffer alert: the patient list who are the ICU patients and suspicious developed septic.

5. Neuromyopathy-alert : Basically the septic patients for the pediatric ICU(s) patients.

6. Pectus excavatum repair alert: Patients list who scheduled pectus excavatum repair surgery.

7. QTC sniffer: new born lists for the QTC >475 ms. PI:

Notable sniffers

Herasevich V, Pieper MS,

Pulido J, et al. Enrollment into a time sensitive clinical study in the critical care

setting: results from computerized septic shock sniffer implementation. J Am

Med Inform Assoc 2011. (PMID: 21508415)

Herasevich V, Tsapenko M,

Kojicic M, et al. Limiting ventilator-induced lung injury through individual electronic

medical record surveillance. Crit Care Med 2011;39(1):34-39. (PMID: 20959788)

Herasevich V, Yilmaz M,

Khan H, et al. Validation of an electronic surveillance system for acute lung injury. Intensive

Care Med 2009;35(6):1018-1023. (PMID: 19280175)

ALI VILI Septic Shock

7/8/2015

15

Data retrieval for research

Olmsted county – unique for

population based research

Olmsted county

admission

METRIC datamart

Clinical studies

• Enrollment to time sensitive trials

• Retrospective studies for Quality Improvement an research

7/8/2015

16

Data mining

Visual mining

In conclusion

7/8/2015

17

Datamart usage

1. Administrative reporting

2. Clinical research, including population based

3. Quality improvement: Point of care novel user interfaces, alerts and decision supports tools

4. Predictive analytics

Chest. 2014;145(6):1190. doi:10.1378/chest.145.6.1190

[email protected]