Introduction to the Liver Simulated Allocation Model Presenter: - PowerPoint PPT Presentation

Introduction to the Liver Simulated Allocation Model Presenter: John R. Lake, MD 1 Authors: John R. Lake, MD 1 Josh Pyke, PhD 2 David Schladt, MS 2 1. University of Minnesota; SRTR Senior Staff 2. Scientific Registry of Transplant Recipients 1

Topics • Structure of the Liver Simulated Allocation Model (LSAM) • How LSAM is Used in Evaluating Proposed Policies • Strengths and Limitations of LSAM • Share35: LSAM projections and observed results • Current redistricting results: where LSAM is used (and where it is not) • LSAM implications for interpreting the redistricting projections 2

Liver Simulated Allocation Model • We used the Liver Simulated Allocation Model (LSAM) program to simulate the performance of different scenarios. • LSAM uses historical real-world data to estimate the interactions between donors and candidates with new sets of allocation/distribution rules.  LSAM is a discrete-event simulation program  Draws from real donor and candidate data  Models organ offers, organ acceptance, MELD changes over time, waitlist survival, and post-transplant survival  Simulates the uncertainty associated with these events • Used extensively by SRTR/OPTN to predict the impact of many proposed policy changes 3

LSAM Conceptual Flow Chart Rule File Pool of Candidates Pool of Donors Filters and Sorts candidates for Includes candidates Includes actual on the waitlist at start specific donor organs at that time donors with arrival & new candidates varied on each run added throughout period. Offer Acceptance Goes down ranked list until Candidates removed exhausted or accepted. from pool upon Modeled given candidate & removal from waitlist donor characteristics or death Not Transplanted Transplanted Post-Transplant Survival Modeled given candidate & donor characteristics 4

LSAM Strengths and Limitations Limitations Strengths • Predicts direction of change • Draws on real transplant between alternatives, not data necessarily the magnitude of change • Simulates up to 5 years • Cannot account for changes • Multivariable acceptance in listing or acceptance behavior and survival models • Cannot predict outcomes on • Can compare multiple a center-by-center basis allocation and distribution • Most recent input data files systems use data through 2011 5

How LSAM is Used to E valuate Proposed Allocation and Distribution Policies • The committee generates a data request specifying the proposed system and metrics of interest. • LSAM rule files are generated which implement the proposed policy. • LSAM is used to simulate both current and proposed policies.  Each simulation uses multiple iterations to characterize the degree of variability in the results. • The proposed policy results are compared to the current policy results for the committee’s metrics of interest. 6

LSAM Predictions and Share 35

Methods • LSAM was used to simulate 1 year of liver transplants under two sets of allocation rules: pre- and post-Share35. • LSAM simulations were repeated 10 times for each scenario. • Patient inputs were based on candidates and donors from 2010. Both simulations used the same set of donors and candidates in • order to focus on the effects of the allocation rules; this is standard practice with LSAM simulations. • Observed data was extracted from the SRTR SAF for 2 periods, each covering 1 year pre- and post-Share35 implementation on June 18, 2013.

Share35 Comparison: Transplant and Sharing Rates Actual Pre- LSAM Pre-Share35 LSAM Share35 Actual Share35 Share35 Total Livers 6698 6698 6706 7026 6074 (5997-6123) 6102 (6079-6126) Transplanted 6028 (89.9% ) 6361 (90.5% ) 90.7% 91.1% Not Transplanted 624 (575-701) 9.3% 596 (572-619) 8.9% 678 (10.1% ) 665 (9.5% ) (Discarded) Actual Pre- LSAM Pre-Share35 LSAM Share35 Actual Share35 Share35 4079 (4049-4131) 3632 (3585-3660) Local 4505 (74.7% ) 4014 (63.1% ) 67.1% 59.5% 1735 (1687-1783) 2047 (2010-2079) Regional 1229 (20.4% ) 2023 (31.8% ) 28.6% 33.5% National 261 (249-281) 4.3% 423 (408-446) 6.9% 295 (4.9% ) 322 (5.1% ) 9

Share35 Comparison: Waitlist mortality rates LSAM Pre-Share35 LSAM Share35 Actual Pre-Share35 Actual Share35 S tatus 1A 457.7 (359.1-528.7) 502.1 (378.6-653.4) 541.5 395.6 S tatus 1B 47.1 (30.5-65.1) 44.8 (22-61.7) 44.6 43.5 >= MELD/ PELD 35 105.1 (94.3-112.6) 100.1 (88.6-109.5) 156.8 142.8 MELD/ PELD 30-34 28.6 (25.6-34.4) 29.4 (26.6-32.4) 15.5 18.2 MELD/ PELD 25-29 8 (7.2-8.4) 8 (7.5-8.5) 6.4 7.1 MELD/ PELD 15-24 3.4 (3.2-3.5) 3.3 (3.2-3.4) 5.5 6.1 < MELD/ PELD 15 0.5 (0.4-0.5) 0.5 (0.5-0.5) 2.6 2.6 Inactive 6.7 (6.6-6.9) 6.6 (6.4-6.7) 25.4 25.2 10

LSAM and Share35 summary • LSAM correctly predicted the direction of change for Share35 results in most categories of interest. • The projected magnitude of change was smaller than observed in some cases—LSAM projections were conservative. • LSAM overestimated the proportion of regional transplants pre-Share35 but still predicted an increase in regional sharing as seen in Share35. • LSAM underestimated transplant rates and overestimated death rates for candidates in the MELD 25-34 range. 11

Where is LSAM used? LSAM used? District map design No Supply/demand analysis No Proximity circles analysis Yes (disparity and summative metrics) Cost analysis Yes (simulated transplants and outcomes) 12

LSAM Strengths and Limitations Limitations Strengths • Predicts direction of change • Draws on real transplant between alternatives, not data necessarily the magnitude of change • Simulates up to 5 years • Cannot account for changes • Multivariable acceptance in listing or acceptance behavior and survival models • Cannot predict outcomes on • Can compare multiple a center-by-center basis allocation and distribution • Most recent input data files systems use data through 2011 13

LSAM Offer Acceptance Model • Higher travel time is correlated with lower acceptance for all patients • Nonlocal transplant is correlated with lower acceptance for all but Status 1A patients • This results from current organ allocation policy: most organs traveling beyond the local DSA have been turned down by all candidates in the DSA • All of the modeled scenarios use full regional or district-wide sharing as the first level of allocation, so this effect is not likely to persist 14

Transplant counts and organ sharing Overall Transplant Counts Local Transplant Percentage Transplant counts correlate highly with local transplant percentage in the LSAM projections 15

Implications for redistricting • The simulated scenarios share high-quality organs more broadly than current policy because the first level of distribution is regional sharing. • The LSAM acceptance model assumes that organ acceptance probability declines with distance, but this effect is likely to decrease as acceptance behavior responds to wider sharing. • LSAM projects a slight decrease in transplants due to acceptance model effects. • If transplant rates remain at historical levels, slightly more transplants will be performed than LSAM predicts, and some outcome metrics will improve. 16

Overall Summary • LSAM simulates the liver allocation process. Uncertainty in offer acceptance and patient outcomes is modeled using historical data. • LSAM correctly predicted the direction of change in most categories of interest for Share35. • The offer acceptance model is limited by its reliance on historical behavior, especially the relationship between organ quality and transplant distance. This may cause LSAM to underestimate the number of transplants under redistricting and so to underestimate the benefits. • Despite its limitations LSAM has been used extensively to evaluate proposed liver allocation policies. 17

Supplemental Slides

Share35 Comparison: Recipient Age Groups LSAM Pre-Share35 LSAM Share35 Actual Pre-Share35 Actual Share35 <1 105 (99-110) 1.7% 101 (94-113) 1.7% 112 (1.9%) 128 (2%) 1-5 Years 151 (139-163) 2.5% 157 (145-168) 2.6% 199 (3.3%) 179 (2.8%) 6-11 Years 70 (65-75) 1.2% 72 (67-74) 1.2% 71 (1.2%) 86 (1.4%) 12-17 Years 91 (83-98) 1.5% 92 (82-98) 1.5% 90 (1.5%) 88 (1.4%) 18-34 Years 361 (336-384) 5.9% 369 (351-381) 6% 295 (4.9%) 316 (5%) 35-49 Years 1010 (982-1021) 1030 (1002-1061) 910 (15.1%) 926 (14.6%) 16.6% 16.9% 50-64 Years 3603 (3547-3652) 3600 (3567-3645) 3514 (58.3%) 3624 (57%) 59.3% 59% 65+ Years 684 (657-703) 11.3% 681 (670-692) 11.2% 838 (13.9%) 1012 (15.9%)

Share35 Comparison: Recipient Race/ E thnicity Actual Pre- LSAM Pre-Share35 LSAM Share35 Share35 Actual Share35 White 4183 (4141-4220) 4182 (4134-4205) 4204 (69.7%) 4347 (68.4%) 68.9% 68.5% African American 723 (698-741) 733 (718-761) 12% 634 (10.5%) 689 (10.8%) 11.9% Hispanic/Latino 816 (788-838) 832 (818-852) 848 (14.1%) 939 (14.8%) 13.4% 13.6% Asian 277 (263-295) 279 (263-288) 262 (4.3%) 301 (4.7%) 4.6% 4.6% Other 76 (70-82) 1.2% 76 (72-80) 1.2% 81 (1.3%) 83 (1.3%)