Using GIS for Analyzing Optimal Organ Allocation for Liver Transplantation

Size: px

Start display at page:

Download "Using GIS for Analyzing Optimal Organ Allocation for Liver Transplantation"

Isaac Casey
5 years ago
Views:

1 Using GIS for Analyzing Optimal Organ Allocation for Liver Transplantation September 7, Presenters: Naoru Koizumi 1, Amit Patel 1,Yang Xu 1, Nigel Waters 1, Monica Gentili 2, Ammar Malik 1, Dennis Nicholas 1 1. George Mason University 2. University of Salerno

2 Organ Allocation Problems Disparities in the US Organ Transplant System Due to Race & Ethnicity: African Americans tend to have low access to liver transplants despite their higher risk of infection from the Hepatitis C Virus (HCV). Lack of insurance and delayed referral are potential causes. Due to Socio-economic Status Individuals living in higher income ZIP codes are more likely to receive transplantation. Individuals with a higher socio-economic status are more likely to be registered at multiple Organ Procurement Organizations (OPO). 2 References: Ong J.P., Collantes R., Pitts A., Martin L., Sheridan M., Younoussi Z. (2005), High Prevalence of Uninsured Among HCV-Positive Individuals, Journal of Clinical Gastroenterology, 39(9): Tuttle-Newhall, J.E., et al. (1997), A Statewide, Population-based, Time Series Analysis of Access to Liver Transplantation, Transplantation, 63: Merion, R.M., et al. (2004), Prevalence and Outcomes of Multiple-listing for Cadaveric Kidney and Liver, Transplantation, American Journal of Transplantation, 4(1):

3 Organ Allocation Problems Spatial Disparities in Organ Transplant (our focus) Due to Geography where candidates live matters A candidate whose residence is close to a transplantation center tends to have a higher access. The access to transplant differs significantly depending on where you live. Mortality rates tend to be lower for those who live in areas with a higher number of transplants and a larger population size. 3 References: Ellison, M.D., et al. (2003) Geographic differences in access to transplantation in the United States, Transplantation, 76(9): Klassen, A.C., et al. (1998), Factors influencing waiting time and successful receipt of cadaveric liver transplant in the United States: 1990 to 1992, Medical Care, 36(3): Brown, K.A. and Moonka, D. (2004), Liver Transplantation, Current Opinion in Gastroenterology, 20(3): Brown, R.S. and Lake, J.R. (2005), The Survival Impact of Liver Transplantation in the MELD Era and the Future of Organ Allocation and Distribution, American Journal of Transplantation, 5(2):

4 Liver (as opposed to Kidney) transplants The number of liver transplants is on the rise in recent years (13% in 1988 to 23% in 2008). Serious cases require urgent action (e.g. potential death within 7 days puts candidates into a Status 1 category those who receive a priority in the waiting line); Also there are no medical alternatives to a liver transplant. 4 Reference: Institute of Medicine Committee on Organ Procurement and Transplantation Policy, (1999), Organ Procurement and Transplantation: Assessing Current Policies and the Potential Impact of the DHHS Final Rule, Washington, DC, National Academic Press.

5 Main Questions Main questions: Is there a nationwide disparity problem in access to liver transplant? If so, where? Data: SRTR database from United Network of Organ Sharing (UNOS) for the period between Approach: Spatial analysis using ArcGIS

6 Analytical Approach 6 Step 1: Geocode transplant recipients (T) and wait listed candidates (W) at zip code centroids Step 2: Create a regularly spaced grid of points on the US map Step 3: Calculate Standardized Transplant Ratio (STR) at each grid point with specific search radius Step 4: Interpolate a continuous surface based on the values of the STR grid points

7 Step 1: Geocoding Recipients and Waitlisted Candidates Map of the patients in the US(2008) 7

8 8 Step 2: Creating regularly spaced grid points

9 Step 3: Calculating the Standardized Transplant Ratio (STR) STR = (t i /n i )/(T/N) The radius of the circle is the kernel size STR is defined as a function of the risk of receiving a liver transplant incurred for a randomly selected patient residing at a given grid point relative to the average risk in the whole of the US Where: t i : # of transplant cases in a circle with a specific radius r around grid point i 9 t 1 =5 n 1 =12 Reference: Samuel Soret, Karl j.mccleary etal n i : # of waiting and recipient patients in a circle with a specific radius r around grid point i T: total # of transplant cases in the US N: total # of waiting and recipient patients in the US

10 Step 4: Spatial Interpolation (Kriging) Parameters used in kriging: -Grid Point Distance: 10 Miles -Radius (kernel size): 7 miles 10

11 Smoothed map of Liver Transplantation Standardized Transplant Ratio (2006) 11 Fixed Kernel size:7 miles

12 Smoothed map of Liver Transplantation Standardized Transplant Ratio (2007) 12 Fixed Kernel size:7 miles

13 Smoothed map of Liver Transplantation Standardized Transplant Ratio (2008) 13 Fixed Kernel size:7 miles

14 14 Map of the locations of the liver transplant centers in the US

15 Spatial Interpolation again using kriging but with a larger kernel size Parameters used in kriging: -Grid Point Distance: 10 miles -Radius (kernel size): 303 miles, 289 miles and 287 miles for 2008, 2007 and 2006, respectively We fixed the kernel size for each year to reach the minimum sample size (n) requirement for each grid point based on the normal approximation rule -Binomial distribution -np and n(1-p) are both greater than 10 (where p=t/n) -n=37 for 2008, n=37 for 2007 and n=36 for

16 Smoothed map of Liver Transplantation Standardized Transplant Ratio (2006) 16 Fixed Kernel size: 303 miles

17 Smoothed map of Liver Transplantation Standardized Transplant Ratio (2007) 17 Fixed Kernel size: 289 miles

18 Smoothed map of Liver Transplantation Standardized Transplant Ratio (2008) 18 Fixed Kernel size: 287 miles

19 19 Map of the locations of the liver transplant centers in the US

20 Spatial Interpolation with variable kernel size -This is an alternative to the previous approach where we fixed the radius to get the minimum sample at each grid point. -In this analysis we varied the size of the circle. The larger circles were drawn where patients were sparsely populated and smaller circles were drawn where patients were clustered. -In this approach, we fixed the sample size but we did not fix the radius. The radius was determined individually at each grid point. 20

21 21 Smoothed map of Liver Transplantation Standardized Transplant Ratio (2006)

22 22 Smoothed map of Liver Transplantation Standardized Transplant Ratio (2007)

23 23 Smoothed map of Liver Transplantation Standardized Transplant Ratio (2008)

24 24 Map of the locations of the liver transplant centers in the US

25 Variable vs. Fixed Kernel Sizes Fixed kernel size : -Radius = 7miles -Shows localized measure -but lacks sufficient sample size -Larger radius -Shows the regional measure -But loses local variability Variable kernel size: 25 -Avoid drawing large circles in the areas where we had sufficient samples within smaller radius. -capture the local variation in dense areas and at the same time ensured sufficient sample size in sparse areas. -But if the disk is larger, it is not reflective of local situation and hence might be misleading especially in the areas where patients are sparsely populated.

26 Main Findings There is a distinctive high access location in the Florida area (more organs are available). STR tends to be higher around transplant centers (patients are located closer to transplant centers) 26

27 Other on going analysis Location Allocation Analysis Good follow-up care leads to a successful outcome of a transplant. Good geographic access to follow-up care (at transplant center) is an important issue for transplant recipients. Where are the good locations to add more transplant centers that can maximize the access to follow-up care among transplant recipients? 27

28 28 Map of Location-Allocation Analysis(2008)

29 29 Map of Location-Allocation Analysis (2008)

30 30 Map of Location-Allocation Analyst(2008)

31 31 Map of Location-Allocation Analyst(2008)

32 32 Map of Location-Allocation Analyst(2008)

33 Future work 33 We will create similar maps to show disparities in terms of waiting time, number of deaths while waiting, number of post-graft failures, among many other possibilities. Spatial and Geographically Weighted Regressions County level regression Dependent variable: STR, Waiting time, plus others. Independent variables: county race composition, median income, average age (FL), and number of transplant programs (MN)

34 Thank you. Any Questions and Comments?

FAIRNESS/EQUITY UTILITY/EFFICACY EFFICIENCY. The new kidney allocation system (KAS) what has it done? 9/26/2018. Disclosures

FAIRNESS/EQUITY UTILITY/EFFICACY EFFICIENCY. The new kidney allocation system (KAS) what has it done? 9/26/2018. Disclosures The new kidney allocation system (KAS) what has it done? Disclosures No financial disclosure Ryutaro Hirose, MD Professor in Clinical Surgery University of California San Francisco Objectives Describe