shaun j. grannis, md, ms, facmi, faafp

Data Aggregation, Liquidity, andthe Learning Healthcare System:

Perspectives from the Indiana Experience

Shaun J. Grannis, MD, MS, FACMI, FAAFPBiomedical Informatics Research Scientist,

The Regenstrief InstituteAssociate Professor of Family Medicine,

Indiana University School of Medicine

What We’ll Cover

• The Context• The Problem• Potential Solutions

Regenstrief Institute

• Endowed by Sam Regenstrief - Inventor of the low cost front-loading dishwasher

• Supported the creation of the Institute to apply process improvement to medicine

• A medical informatics “skunkworks”

Regenstrief Informatics - What We Do

• Build medical information systems• Study systems and supporting technologies• Rationalize, organize and standardize health

care data• Pragmatists - needs driven, create solutions to

real-world problems• Describe what works and what doesn’t

High Costs and Inefficiencies• Very large and inefficient information enterprise

that still operates with substantial amounts of paper

• Costs are rising – $2.4 Trillion in 2009, ~$8,000/person – Growth outpacing inflation– Now $2.7 Trillion in annual spending (2012 est.)– May reach $4 Trillion by 2020 (!)

1. RAND Study, Hillestad2. Social Transformation of American Medicine,

Starr

Healthy Life Expectancy versus Expenditure per capita

Total Healthcare Expenditures per Capita $USPPP, 2006 or Latest

Source: OECD Health Database, June 2008 version; WHO World Health Data 2008; EU-15 average is the GDP weighted average

Infant Mortality versus Expenditure per capita

Total Healthcare Expenditures per Capita $USPPP, 2006 or Latest

Source: OECD Health Database, June 2008 version; WHO World Health Data 2008; EU-15 average is the GDP weighted average

Copyright © 2014, The Regenstrief Institute, Inc.

Variation in Medicare Reimbursement Rates

Kocher R, Sahni NR. Rethinking Health Care Labor.N Engl J Med 2011; 365:1370-1372. October 13, 2011

Healthcare Labor Productivity

INPC Data Management and ServicesData Management

Hospital

Data Repository

Health Information

Exchange

Network Applications

Payers

Labs

Outpatient RX

Physician Office

Ambulatory Centers

Public Health

Data Access & Use

Hospitals

Physicians

Labs

PublicHealth

Payer

• Results delivery• Secure document transfer• Shared EMR• Credentialing• Eligibility checking

• Results delivery• Secure document transfer• Shared EMR• CPOE• Credentialing• Eligibility checking

• Results delivery

• Surveillance• Reportable conditions• Results delivery• De-identified, longitudinal

clinical data

• Secure document transfer• Quality Reporting

• De-identified, longitudinalclinical dataResearchers

Clinical Abstract

Overhage JM, Dexter PR, Perkins SM, Cordell WH, McGoff J, McGrath R, McDonald CJ. A randomized, controlled trial of clinical information shared from another institution. Ann Emerg Med. 2002 Jan;39(1):14-23.

Notifiable Condition Detection


System Overview: Notifiable Condition Detector

InboundMessag

e

PotentiallyReportabl

e

Reportable

Condition

ReportableConditionsDatabases

Abnormal flag,Organism name in Dwyer II, Value above threshold

Compare to Dwyer I

Record Countas denominator

E-mailSummar

yRealtime Daily Batch

PrintReport

s

To PublicHealth

To InfectionControl

ELR Completeness

4,785 total reportable casesINPC– 4,625 (97%)Health Dept – 905 (19%)Hospitals – 1,142 (24%)

Timeliness

ELR identified cases 7.9 days earlier than did spontaneous reporting.

Prepopulated Reporting Forms

Clinical Messaging/Public Health Communication

Sample Pre-populated

Reporting Form


Reporting Form

Patient Demographics

Clinical Data

Provider Demographics

Understanding Reporting Workflow

Pre-populated form Information flow

Outcome measures

• Time-to-treatment• Timeliness of reporting to public health• Completeness of reporting data• Level of communication among PH and clinical

providers

Syndromic Surveillance


Over 110 Indiana Emergency Departments

contribute 2.1 million visits to the system each

year.

Hospital

Interface

Engine(Routing)

Information Flow: Clinical

Network ConnectionHL7 ADT

message

Hospital ED

Registration

Hospital Firewall(Encryption)

Firewall(Decryption)

Message Listene

rMessage Processor

Imported into

Clinical Reposito

ryClinical Repositor

y

Hospital

Interface

Engine(Routing)

Information Flow: PH Surveillance

Network ConnectionHL7 ADT

message

Hospital ED

Registration

Hospital Firewall(Encryption)

Firewall(Decryption)

Message Listene

rMessage Processor

Batched, delivered to ISDH

every 3 hours

Public Health

Neuro Event

GI Event

Natural Disaster

H1N1 Surveillance

Flu CC

Pneumonia ICD9

Pneumonia CC

ILI ICD9

ILI CC

All Flu Tests

Positive Flu Tests

Positive Rate

Flu ICD9

H1N

1, A

pril

2009

H1N

1,O

ct

2009

All Health Care is Not Local: An Evaluation of the Distribution of

Emergency Department Care Delivered in Indiana

All Health Care is Not Local

• Over 3 years, 2.8 million patients totaled 7.4 million visits for an average of 2.6 visits per patient.

• More than 40% of ED visits during the study period were for patients having data at multiple institutions.

• This population analysis suggests a pull model is necessary, and helps inform the ongoing dialog regarding the merits of peer-to-peer (push) and federated aggregate HIE (pull) NwHIN architectures.

Leveraging Analytics to Enable Accountable Care

A network diagram illustrating the connectedness among Indiana EDs that participate in PHESS. Circular nodes represent EDs; node size indicates the visit volume; node color indicates the centrality of the ED. The gray edges connecting nodes indicate where patient crossover occurs. EDs that share proportionally larger number of patients are clustered together. While general clusters of “medical trading areas” emerge, the myriad gray edges clearly illustrate how interconnected all EDs are to one another.

• Patients receive healthcare from multiple providers and across organizations

• More than 40% of ED visits are for patients having data at multiple institutions

Distribution of patients stratified by the total number of ED visits. Note that six patients visited the ED more than 300 times and a single patient accumulated 385 visits for the 3-year study period.


• Shifting nonurgent visits from ED to primary care: During the 6-month trial with 9 Central Indiana hospitals, the 320,000-member managed health plan reduced nonurgent ED visits among members served by these hospitals by 53 percent, while simultaneously increasing primary care office visits by 68 percent.

• Cost savings: The shift from ED to primary care visits that occurred during the pilot test saved the health plan an estimated $2 to $4 million over the 6-month period.

http://www.innovations.ahrq.gov/content.aspx?id=3988


84% PPV for predicting which patients who will use ED > 16 times in two years.

Supporting ACO Services

• Care management support– HIE Information matched to CMS defined ACO

population– Patient care summaries extracted– Delivered to ACO care management systems via

CDA documents• Readmission risk stratification (LACE model)

– Adaptation underway

Patient Address Change

ADT Processor

Update person_address table with new address information

person_address table

Address Update Detector

In real-time, Address Update Detector detects and writes address changes to the post_processing table

post_processing table

Geo-Coding Application

Geo-Coding app reads the post_processing table

5

Call Polis Center web service which returns geo-coded addresses

Polis Web Service

3

4

6

1

2

Integrating socio-behavioral determinants of health using geospatial information

Comer KF, Grannis S, Dixon BE, Bodenhamer DJ, Wiehe SE. Incorporating geospatial capacity within clinical data systems to address social determinants of health. Public Health Rep. 2011 Sep-Oct;126 Suppl 3:54-61.

Improving Efficiency of Data Integration

A Cautionary Note: the Era of “Big Medical Data”, Analytics, and Data

Quality

Aggregate Data Example:Diabetes and Obesity Cohort“Coders should pay attention to the BMI because

it makes a difference in terms of reimbursement […]. A BMI of 40 or higher - diagnosis code V85.4 - is considered a complicating condition, meaning higher reimbursement when reporting this code along with the appropriate principal diagnosis.”

(http://medicalcodingpro.wordpress.com/page/2/)

Data often reflect financial incentives, not the true population distribution.

“Using Information Entropy to Monitor Chief Complaint Characteristics and Quality”

Data Quality• Supplemental data is often necessary to enhance

practice based population health processes

Missing data rate for a sample of clinical transactions received by the INPC in 2008.

Conclusion

shaun j. grannis, md, ms, facmi, faafp

Documents

health care labor

health care wages

health aff

health care datapragmatists

world health data

oecd health database

health spending projections

annual expenditures