Population Grouping: The Canadian Experience

Similar documents
CIHI s Population Grouping Methodology: Beyond Predicting Costs

CIHI Population Grouping Methodology

Case Mix and Funding. Health Data Users Day May 12, Greg Zinck Manager, Case Mix

TN Bundled Payment Initiative: Overview of Episode Risk Adjustment

Health Quality Ontario

WORKING P A P E R. Comparative Performance of the MS-DRGS and RDRGS in Explaining Variation in Cost for Medicare Hospital Discharges BARBARA O.

TN Bundled Payment Initiative: Overview of Episode Risk Adjustment

HEALTH SYSTEM MATRIX VERSION 8.0 DATA DICTIONARY

Appendix Identification of Study Cohorts

ENROLLMENT : Line of Business Summary

NQF-ENDORSED VOLUNTARY CONSENSUS STANDARD FOR HOSPITAL CARE. Measure Information Form Collected For: CMS Outcome Measures (Claims Based)

TENNCARE Bundled Payment Initiative: Description of Bundle Risk Adjustment for Wave 3 Episodes

MHSPHP Metrics Forum. ACG and Health Services

Supplementary Online Content

EVALUATION AND REFINEMENTS TO THE COMPREHENSIVE AMBULATORY CLASSIFICATION SYSTEM (CACS)

Health Links Target Population Ministry of Health and Long-Term Care

TENNCARE Bundled Payment Initiative: Description of Bundle Risk Adjustment for Wave 2 Episodes

NORTH CAROLINA STATE HEALTH PLAN FOR TEACHERS AND STATE EMPLOYEES

Supporting New Funding Approaches using CIHI s Classification Systems. Health Data Users Day May 27, 2013 Greg Zinck, Manager, Case Mix

Symmetry Episode Treatment Groups

Readmission Analysis Using 3M Methodology

Characteristics and Healthcare Use of COPD Patients Across Multiple Sectors of Care in Alberta CAHSPR Conference 2016

Methodological Issues

This presentation was current at the time it was published or uploaded onto the web. Medicare and commercial payers change their policies frequently.

REIMBURSEMENT AND ICD-10 CODING. December RB Health Partners, Inc.

Appendix 1: Supplementary tables [posted as supplied by author]

The Australian National Subacute and Non acute Patient Classification. AN SNAP V4 User Manual

79 HCCs CMS-HCC Risk Adjustment Model. ICD-10-CM to CMS-HCC Crosswalk. Over 9,500 ICD-10-CM codes map to one or more.

SUMMARY TABLE OF MEASURES, PRODUCT LINES AND CHANGES

ACOFP 55th Annual Convention & Scientific Seminars. How Complicated is Your Panel? Effective Risk Coding in Primary Care. Alison Mancuso, DO, FACOFP

Symmetry Episode Treatment Groups

Agenda. ICD-10 CM ICD-10 PCS Prior Auth Guidelines Claims Processing

TENNCARE Bundled Payment Initiative: Description of Bundle Risk Adjustment for Wave 8 Episodes

Risk of Fractures Following Cataract Surgery in Medicare Beneficiaries

Zhao Y Y et al. Ann Intern Med 2012;156:

OBSERVATIONAL MEDICAL OUTCOMES PARTNERSHIP

Data Fusion: Integrating patientreported survey data and EHR data for health outcomes research

HP Enterprise Services International Classification of Diseases, 10th Edition (ICD-10) Presentation

Needs Assessment and Plan for Integrated Stroke Rehabilitation in the GTA February, 2002

DIAGNOSIS CODING ESSENTIALS FOR LONG-TERM CARE:

Hu J, Gonsahn MD, Nerenz DR. Socioeconomic status and readmissions: evidence from an urban teaching hospital. Health Aff (Millwood). 2014;33(5).

Inpatient Psychiatric Facilities

NCQA Health Insurance Plan Ratings Methodology October 2014

LCD L B-type Natriuretic Peptide (BNP) Assays

Episodes of Care Risk Adjustment

Objectives. Medicare Spending per Beneficiary: Analyzing MSPB Data to Identify Primary Drivers

Efficiency Methodology

TENNCARE Bundled Payment Initiative: Description of Bundle Risk Adjustment for Wave 5 Episodes

Geriatric Emergency Management PLUS Program Costing Analysis at the Ottawa Hospital

Home Health Prospective Payment System. Overview

TENNCARE Bundled Payment Initiative: Description of Bundle Risk Adjustment for Wave 4 Episodes

Leveraging Analytics to Mitigate Financial Risks in ICD-10

Surgical Outcomes: A synopsis & commentary on the Cardiac Care Quality Indicators Report. May 2018

Supplementary appendix

The Risky Business of Claims-Only Risk Adjustments

Registry Highlights. Dale Daniel Symposium Hip Fracture Registry. Overall Volume by Year and Region 3/7/2014

Annual High Claims Survey. Year Ending 31 December 2016

Potential disruption from private exchanges and narrow networks. In 2011, less than 10% of companies used High Performing Networks (narrow networks)

Detecting Anomalous Patterns of Care Using Health Insurance Claims

WHAT ARE THE MOST COMMON CONDITIONS IN PRIMARY CARE?

The Delivery of Radical Prostatectomy to Treat Men With Prostate Cancer

Malignancy ; 191.6; Malignant neoplasm of brain

Quality measures a for measurement year 2016

Transparency The Public Reporting of Physician Performance

TABLE I-1: RESIDENT INFANT DEATHS PER 1,000 LIVE BIRTHS, BY RACE AND ETHNICITY, FLORIDA AND UNITED STATES, CENSUS YEARS AND

Can the categorization of patients with life-limiting conditions help us to

PROGRESS ON HCQI RESEARCH AND DEVELOPMENT WORK

A Pause in the Availability of Risk Adjusted National Benchmarks for AHRQ Indicators and an Alternative Measurement Approach

Changes for Physician Measurement 2018

Australian mortality coding: history, benefits and future directions

University of Bristol - Explore Bristol Research

PFIZER INC. What is the difference in incidence of fracture in women who ever or never used DMPA for contraception?

OP-10: ABDOMEN CT USE OF CONTRAST MATERIAL

Nov FromAtoZCodesMatter

HEDIS/CAHPS 101 August 13, 2012 Minnesota Measurement and Reporting Workgroup

ICD 10 CM Coding and Documentation

Risk Adjustment and Hierarchical Condition Category Coding

Annual Statistics on Organ Replacement in Canada

NQF-ENDORSED VOLUNTARY CONSENSUS STANDARDS FOR HOSPITAL CARE. Measure Information Form Collected For: CMS Outcome Measures (Claims Based)

Yale New Haven Health Services Corporation/Center for Outcomes Research and Evaluation (YNHHSC/CORE)

TECHNICAL NOTES. for Spinal Fusion. June 2016

Efficacy Study of Zoledronic Acid and Teriparatide Combination Therapy in Women With Osteoporosis

Health Quality Ontario

Deconstructing the RADV: The Past, Present, and Future of RADV

NATIONAL QUALITY FORUM

The Johns Hopkins ACG System. Excerpt from Version 11.0 Technical Reference Guide November 2014

Combining Risk Adjustment and HEDIS to Improve Quality of Care. Colleen Gianatasio, CPC, CPC-P, CPMA, CPC-I, CRC

MCO Task Force WELCOME

3/20/2013. "ICD-10 Update Understanding and Analyzing GEMs" March 10, 2013

MEASURING CARE QUALITY

2.1 Numerator: The number of denominator continuous inpatient spells (i.e. spells excluding those with a diagnosis

Arkansas Blue Cross and Blue Shield (ABCBS) Patient Centered Medical Home (PCMH) Specifications Manual

Measure Information Form Collected For: CMS Outcome Measures (Claims Based)

Diagnosis Coding is About to be Much More Important. Matthew Menendez

Collection of Statistics on Causes of Death Azza Badr, PhD, Vital Statistics and Country Support WHO/EMRO

Achieving Quality and Value in Chronic Care Management

A chapter by chapter look at the ICD-10-CM code set Coding Tip Sheet

A Whole Pathway Integrated Approach to Improving Foot Care

Supplementary Online Content

Sheila Rodger, R5 Geriatric Medicine, U of C Supervised by Dr. D. Hogan

Transcription:

Population Grouping: The Canadian Experience PCSI, The Hague October 16, 2015 Douglas Yeo Holly Homan Victoria Zhu Craig Homan 1

Canadian Population Grouping Methodology The Journey so Far Health Condition Categories Cost Weights Mutually Exclusive Classification 2

The Journey So Far Douglas Yeo 3

The Journey So Far What is the POP Grouping Methodology Similar concept to CIHI s other case mix methodologies Some differences from CIHI s other case mix methodologies Does not focus on any one sector or setting Target population includes all persons registered for publicly-funded health care Looks at person over an extensive time period Two major components of the methodology Clinical classification Predictive indicators 4

What are the Timelines for the Population Grouping Methodology Project? September 2014 Interim release Mapping tables of ICD codes to health conditions April 2015 Alpha version Clinical classification: health condition flags Predictive indicators for health resources Grouping software October 2015 Beta version Functional status classification, for continuing care sub-population Socioeconomic effects incorporated into indicators March 2016 version 1.0 Mutually exclusive classification Alpha and beta for expert group; 1.0 for public release 5

Some Population Grouping Methodology Applications Profiling and planning Coordinated care and targeted care management High users Risk adjustment Health system indicators Funding Physician Regions 6

Concurrent and Prospective Periods Clinical classification Concurrent period clinical data used to build clinical profiles 2012-13 2013-14 2014-15 Predictive indicators (e.g. cost, mortality) Predicted cost during concurrent period Predicted cost during prospective period 7

Health Condition Categories Holly Homan 8

Evaluation Criteria Criteria Types of Questions Clinical Relevancy Are the disease categories useful to clinicians? Is it useful for managers and administrators in describing populations? Explanation of Variation How well do the categories differentiate costs? How well do the categories differentiate utilization needs? What is the R 2? Logical Hierarchy / Transparency Manageable Number of Groups Does the methodology apply hierarchies where appropriate? Is there a large number of terminal categories? Do certain categories not contribute to identifying chronic population? Sensitivity to Data Inputs Does the methodology address variations that might exist in data inputs? Reliability of Predictive Indicators Does the methodology yield similar results to population estimates? What is the stability of indicators over time? 9

Health Condition Categories Started with medical Case Mix Groups (CMG+) Vetted through a clinical panel Modified to fit a population methodology Input from expert advisory group and physician panel Malignancy teased out Physician claims data Diagnostic, supportive, ongoing 10

Assigning Diagnoses to a Person CIHI source data Hospital inpatient Day surgery Physician claims CCRS assessments Diagnosis codes ICD-10-CA ICD-9 G81.99 Hemiplegia unspecified S32.090 Fx lumbar vertebra J20.9 Acute bronchitis, unspecified U99.043 Snowboarding Z13.5 Special screening eye & ear Martin 11

Assigning Health Condition Categories Diagnosis codes Apply to persons Look at all encounters over the concurrent period Certain POP health (e.g. 2 years) conditions tagged to a Map each diagnosis to person one of the 214 health conditions A07 Paralytic Syndrome / Spinal Cord Injury H42 Fracture / Dislocation Vertebrae, Pelvis D44 Acute & Other Respiratory Diseases Martin 12

Health Condition: Tagging Rules Tagging Rules > 1 time rule for hospital stays >Physician claims data 1 time and 2 times rule A07 Paralytic Syndrome / Spinal Cord Injury H42 Fracture / Dislocation Vertebrae, Pelvis D44 Acute & Other Respiratory Diseases Martin 13

Health Condition: Clinical Overrides Tagging Rules > 1 time rule for hospital stays >Physician claims data 1 time and 2 times rule Clinical Overrides Health Condition overrides address redundancies that may exist on a profile after the assignment of health conditions is completed A07 Paralytic Syndrome/Spinal Cord Injury H42 Fracture/Dislocation Vertebrae, Pelvis D44 Acute & Other Respiratory Disease Martin 14

0-1 2-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-89 90+ Number of Conditions Present During the Concurrent Period, by Age Group, Alberta 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 5+ Conditions 4 Conditions 3 Conditions 2 Conditions 1 Condition 0 Conditions Non-Users Age Group (in Years) 15

Cost Weights Victoria Zhu 16

Assigning Cost Weights to Persons (Illustration) Q04 Depression Q07 - Drug/Alcohol Abuse/Dependence Cost Weights (hospital + physician costs) Cost Indicator Effect Concurrent Prospective Q04 0.26 0.42 Q07 0.98 0.96 Total 1.24 1.38 Joe Joe s concurrent cost weight is 1.24 Joe s prospective cost weight is 1.38 17

Modeled Cost for a Person (Illustration) Joe s concurrent cost weight is 1.24 Joe s prospective cost weight is 1.38 Q04 Depression Q07 - Drug/Alcohol Abuse/Dependence Population average concurrent cost is $3,258 Population average prospective cost is $1,752 Joe Joe s expected concurrent cost: $4,040 = $3,258 x 1.24 Joe s expected prospective cost: $2,418= $1,752 x 1.38 Mock data 18

Population distribution Populations Non Users Users W/O HCs Users With HCs Total Population Volume 444,312 212,894 3,264,548 3,921,754 * 2010-2011 and 2011-2012 Alberta data used in methodology development 19

Regression Models Additive models with linear regression Cost as the response variable Separate models for concurrent and prospective Separate models applied to each sub-population Non-users: OLS, Age/Sex Users without any health conditions: OLS, Age/Sex Users with at least one health conditions: WLS, 214 Health Conditions 20

Goodness of Fit (Concurrent) Concurrent Models Goodness of Fit Model Partition Vol Avg Cost ($) Avg Predicted Cost ($) Bias ($) Prediction Error R 2 (%) Non Users Users with No Health Conditions Users with Health Conditions Overall Est 310,674 4 4 0 8 0.5 Val 133,638 4 4 0 8 0.5 Est 149,106 263 263 0 163 42.1 Val 63,788 265 265 0 164 43.9 Est 2,284,872 3,897 3,897 0 2,607 37.2 Val 979,676 3,896 3,904-8 2,599 41.2 Est 2,744,652 3,259 3,259 0 2,180 37.7 Val 1,177,102 3,257 3,264-7 2,173 41.7 * 2010-2011 and 2011-2012 Alberta data used in methodology development 21

Goodness of Fit (Prospective) Prospective Models Goodness of Fit Model Partition Vol Avg Cost ($) Avg Predicted Cost ($) Bias ($) Prediction Error R 2 (%) Non Users Users with No Health Conditions Users with Health Conditions Overall Est 310,621 299 299 0 491 0.4 Val 133,418 295 299-4 489 0.3 Est 149,101 544 544 0 732 0.6 Val 63,682 523 543-20 712 0.8 Est 2,272,722 2,029 2,029 0 2,401 7.6 Val 972,655 2,032 2,028 4 2,403 7.2 Est 2,732,444 1,751 1,751 0 2,092 7.7 Val 1,169,755 1,751 1,750 2 2,092 7.3 * 2010-2011 and 2011-2012 Alberta data used in methodology development 22

Profiling of Population (Concurrent) Decile sss Volume Average Actual Cost Average Modeled Cost Proportion of Total Costs Average Number of Health Conditions Average Age 1 392,175 4 4 0.01% 0.0 33 2 392,175 203 171 0.6% 0.5 26 3 392,176 363 360 1.1% 1.0 33 4 392,175 529 587 1.6% 1.7 31 5 392,176 782 932 2.4% 2.4 32 6 392,175 1,101 1,387 3.4% 3.0 34 7 392,175 1,623 2,096 5% 3.8 38 8 392,176 2,573 3,380 8% 4.5 43 9 392,175 4,453 5,720 14% 5.1 44 10 392,176 20,952 17,966 64% 8.2 53 Overall 3,921,754 3,258 3,260 100% 3.0 37 * 2010-2011 and 2011-2012 Alberta data used in methodology development 23

Profiling of Population (Prospective) Decile Volume Average Actual Cost Average Modeled Cost Proportion of Total Costs Average Number of Health Conditions Average Age 1 390,219 342 164 1.95% 0.6 22.1 2 390,220 394 261 2.25% 0.6 27.5 3 390,220 539 378 3.08% 1.1 30.8 4 390,220 618 521 3.53% 1.4 32.2 5 390,220 756 704 4.32% 2.1 33.8 6 390,220 969 974 5.53% 2.8 34.1 7 390,220 1,290 1,372 7.37% 3.4 37.2 8 390,220 1,754 1,966 10.02% 4.2 41.7 9 390,220 2,647 3,045 15.11% 5.3 48.3 10 390,220 8,204 8,123 46.85% 8.2 59.7 Overall 3,902,199 1,751 1,751 100.00% 3.0 36.7 * 2010-2011 and 2011-2012 Alberta data used in methodology development 24

What s New in Beta Health Conditions 225 health conditions instead of 214 Clinical overrides applied Any changes in classification are applied Health Condition Interactions OLS used for users with health conditions November 2015 release 25

Plan for Version 1.0 & Future Release Reflects Mutually Exclusive Rule More Predictors Functional Status SES etc. More Predictive Indicators Version 1.0 release: March 2016 26

Questions? Comments? Thank you! 27

Mutually Exclusive Classification Craig Homan 28

Identification of Conditions Additive methodology originally contained 214 conditions Maintain same tagging rules as additive Every evolution (SES, over-ride rules, functional status) in additive model ripples through this model Mutually Exclusive methodology Rolled 214 conditions up to 111 conditions Similar cost and clinical characteristics e.g. Merged CAD/Arrhythmia/Other Heart Disease Clinically validated Create room for splits while maintaining clinical meaningfulness Age/gender tested and not implemented 29

Categories Each condition linked to a Category Major Moderate Minor Other Chronic Cancer Newborn Acute Mental Health Chronic Acute Chronic Acute Cancer Mental Health Obstetrics 30

Hierarchy in Mutually Exclusive Methodology Hierarchy identifies single most significant condition for each person Categorized to the highest condition of the 111 on the list Cost, clinical considerations Concurrent and prospective costs Cell # Description Hierarchy Category S01 Palliative State (Acute) 1 Major Acute S41 Transplant Complication 2 Major Acute S43 Ostomy Complication 3 Major Acute N41 Extremely Low Birth Weight or Immaturity 4 Major Newborn D41 Respiratory Failure 5 Major Acute D04 Pulmonary Hypertension 6 Major Chronic A07 Paralytic Syndrome/Spinal Cord Injury 7 Major Chronic JT1A Diabetes/Hypoglycemia with Chronic Kidney Disease/Failure 8 Major Chronic J01 Cystic Fibrosis 9 Major Chronic I05 Skin Ulcer (incl. Decubitus) 10 Major Chronic 31

Effect of Comorbidities Identified significant comorbidities Major/moderate Categories Split where sufficient volume and cost distinctions seen No cost difference where inside/outside same body system 32

177 Cell Methodology Significant comorbidity splits Category Rank 111 Cond. 177 cells Description 177 Cell Logic Major Chronic 6 D04 D04 Pulmonary Hypertension Major Chronic 6 D04 D04_mm Pulmonary Hypertension, with major/moderate comorbidity Major Chronic 7 A07 A07 Paralytic Syndrome/Spinal Cord Injury Major Chronic 7 A07 A07_mm Paralytic Syndrome/Spinal Cord Injury, with major/moderate comorbidity Major Chronic 9 J01 J01 Cystic Fibrosis Major Chronic 9 J01 J01_mm Cystic Fibrosis, with major/moderate comorbidity Major Chronic 15 E01 E01 Heart Failure Major Chronic 15 E01 E01_mm Heart Failure, with major/moderate comorbidity 33

443 Cell Methodology Built on 177 cells Added total condition count splits where appropriate Volume, cost distinction Category Rank 111 Cond. 177 cells Description 177 Cell Logic Total Condition Count Splits 443 cells Major Chronic 6 D04 D04 Pulmonary Hypertension no split Major Chronic 6 D04 D04_mm Pulmonary Hypertension, with major/moderate comorbidity 2-4,5-10,11-14,15+ Major Chronic 9 J01 J01 Cystic Fibrosis no split Major Chronic 9 J01 J01_mm Cystic Fibrosis, with major/moderate comorbidity no split Major Chronic 15 E01 E01 Heart Failure 1-2,3-4,5+ Major Chronic 15 E01 E01_mm Heart Failure, with major/moderate comorbidity 2-3,4-6,7-9,10-14,15+ 34

689 Cell Methodology Rolled back to 111 Conditions Added comorbidity splits where appropriate Major/moderate/minor comorbidities, Total Condition Count A09 Moderate Chronic Epilepsy no moderate, no minor comorbidities A09a Moderate Chronic Epilepsy no moderate, with <3 Minor Comorbidities and Total Condition Count =2 A09b Moderate Chronic Epilepsy no moderate, with <3 Minor Comorbidities and Total Condition Count 3+ A09c Moderate Chronic Epilepsy no moderate, with 3+ Minor Comorbidities A09d Moderate Chronic Epilepsy with Moderate Comorbidity and < 2 Minor Comorbidities A09d Moderate Chronic Epilepsy with Moderate Comorbidity and 2+ Minor Comorbidities 35

Overview of Model Evolution and R 2 Concurrent Data Prospective Data Development 214 Cond. 111 Cond. 177 Cells 443 Cells 689 Cells Estimation (70%) 37.2 26.3 29.7 41.1 45.3 Validation (30%) 41.2 28.9 32.6 45.8 45.0 Estimation (70%) 7.6 5.3 6.0 7.3 8.1 Validation (30%) 7.2 5.0 5.6 6.7 8.5 Additive model Increased number of cells for greater clinical distinction Added Major/ Moderate splits where applicable Added cellspecific count splits where possible Comorbidity and condition count splits 36

Next steps Chronic vs acute Cost not different Clinical perspective Socioeconomic status (SES) (R 2 +4%) Functional Status Compare goodness of fit statistics for all models Additive and mutually exclusive quite similar Provide recommendations for further development 37

38