Bending the Cost Curve within a Pioneer ACO: The Role of Care - PowerPoint PPT Presentation

Bending the Cost Curve within a Pioneer ACO: The Role of Care Management John Hsu 28 June 2016 AcademyHealth Annual Research Meeting Boston 1

Study Team and Disclosures John Hsu • Maggie Price • Christine Vogeli • Richard Brand • Michael Chernew • Eric Weil • Sreekanth Chaguturu • Tim Ferris • Funding: NIA P01 AG032952 • Partners Healthcare • Speaker Disclosure: Hsu works at MGH, which is part of the Partners Healthcare System • 2

Background Concerns about medical spending growth • CMS: Alternative Payment Models (APMs) • Movement away from FFS – Changes in the incentive structure – Limited information on provider group strategies • Underlying mixture of underuse and overuse • 3

ACO Summary Several types/levels of ACOs • Initial contracts between 2012-14 • Potential ACO provider group strategies/mechanisms: • Improve matching of service level and need, e.g., clinic vs. ED care – Prevent need for clinically downstream services, e.g., hospitalizations – Reduce unnecessary use – Reduce unit prices – Modest “savings” on average • Comparison group challenges • 4

Study Context Pioneer ACO • Prior experience with Medicare High Risk Care Management Program • Prior program in one large hospital within the system – Revised program for ACO, i.e., the integrated Care Management Program – (iCMP=main ACO intervention) Local market with multiple Pioneer ACOs (5) • State focused on spending growth reduction (Chapter 224) • 5

Research Questions How did alignment to an ACO impact clinical event rates and • spending? Among ACO beneficiaries, how did entry into an integrated Care • Management Program (iCMP) impact clinical event rates and spending? 6

Methods ACO Population Detailed data from one of the largest Pioneer ACOs (>82K beneficiaries • aligned) Study period: 2012-14 • Examined beneficiaries newly aligned in January 2012 or January 2013 – Medicare claims available from 2009-14 – Followed beneficiaries until departures from Traditional Medicare program, • departure from the ACO catchment area, or death iCMP Identification and Entry High risk beneficiaries identified annually based on risk scores • PCP review of annual lists for those with “modifiable” risks/spending – iCMP identification year is the year the beneficiary first appears on this list – Beneficiaries were assessed by a care manager before starting the iCMP • program iCMP analyses focus on those beneficiaries that were on the lists and entered • the iCMP program (i.e. were assessed) in 2012-2014 7

Methods Outcomes • ED Visits: monthly counts – ED Visit Severity – Used the NYU algorithm to classify the severity of ED visits • Algorithm assigns a probability that a diagnosis falls in to each of four categories of • increasing severity; focused on the probabilities of the two lowest severity categories Visit defined as non-emergent of primary care treatable if >50% • Hospitalizations: monthly counts – Medicare Costs – Monthly total costs • Standardized to 2012$ • Models: • Negative binomial models for visit counts with individual-level fixed effects – Linear models for costs with individual-level fixed effects – 8

Sensitivity Analyses “Dose” effect • Historical secular trends • Operational definitions, e.g., non-emergent ED visits • Attrition • Model fit • 9

Baseline Characteristics by ACO Alignment Year ACO Alignment Year 2012 2013 p-value Beneficiaries 42,417 19,649 Mean Age 72.6 71.2 <0.001 Female 61% 60% 0.114 Race: White 89% 89% <0.001 Black 5% 5% Other 6% 7% OREC= Aged 81% 80% 0.029 Mean CMS-HCC Score 1.1 1.2 <0.001 Dual 20% 21% <0.001 10

Baseline Characteristics by iCMP Identification Year iCMP Identification Year 2012 2013 2014 p-value Beneficiaries 2,143 1,917 760 Mean Age 74.3 73.4 73.0 0.0143 Female 59% 62% 61% 0.076 Race: White 89 86% 94% <0.001 Black 6 8% 4% Other 5 6% 3% OREC= Aged 73% 71% 76% 0.027 Mean CMS-HCC Score 2.4 2.5 1.5 <0.001 Dual 24% 31% 25% <0.001 11

Comparable Pre-program Trends ACO Start = 2012 ACO Start = 2013 0.20 0.18 0.16 0.14 0.12 ED Visit Rate 0.10 0.08 0.06 0.04 0.02 0.00 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 2009 2010 2011 iCMP Start: 2014: 1-6 iCMP Start: 2014: 7-12 0.20 0.15 ED Visit Rate 0.10 0.05 0.00 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 2009 2010 2011 12

Modest Changes in Clinical Event Rates with ACO Alignment ACO Alignment (post vs. pre) ED Visits ACO Alignment (vs. pre): 1-6 mos 7-12 mos 13+ mos ACO Alignment (post vs. pre) Non-Emergent ED Visits ACO Alignment (vs. pre): 1-6 mos 7-12 mos 13+ mos Hospitalizaitons ACO Alignment (post vs. pre) ACO Alignment (vs. pre): 1-6 mos 7-12 mos 13+ mos 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 Relative Rate 13 13

Changes in Clinical Event Rates with iCMP Entry iCMP Entry (post vs. pre) ED Visits iCMP Entry (vs. pre): 1-6 mos 7-12 mos 13+ mos iCMP Entry (post vs. pre) Non-Emergent ED Visits iCMP Entry (vs. pre): 1-6 mos 7-12 mos 13+ mos Hospitalizaitons iCMP Entry (post vs. pre) iCMP Entry (vs. pre): 1-6 mos 7-12 mos 13+ mos 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 Relative Rate 14

Changes in Monthly Costs Before and After ACO Alignment and iCMP Entry ACO Alignment (post vs. pre) iCMP Entry (post vs. pre) iCMP Entry (vs. pre): 1-6 mos 7-12 mos 13+ mos -$700 -$600 -$500 -$400 -$300 -$200 -$100 $0 $100 Cost Difference 15 15

Conclusions Modest effects associated with entry into the ACO • Additional, larger effects associated with entry into a care • management program Care management effects initially modest, but larger over time • 16

Limitations Non-random assignment, thus potential selection bias from time-changing • unmeasured covariates Real-world clinical environments, thus while iCMP appears to be the main • program within the ACO, there are other smaller population health programs that could contribute to the observed effects Single health care system, albeit a large system with multiple hospitals and • thousands of physicians, thus unclear generalizability to other settings, e.g., non-ACOs, MSSPs Limited sample sizes • Effect of specific mechanisms with much precision – Potential heterogeneity in effects across system or patient traits – Medicare perspective, with no assessment of total spending including • program costs 17

Implications Promising findings for main “intervention” within the ACO • Evidence consistent with shifts in care delivery/matching need with • site as one potentially effective strategy for reducing spending growth Population stability and time to observe “payoff” • 18

Questions John.Hsu@mgh.harvard.edu 19