Real Numeracy

The National Marrow Donor Program (NMDP) maintains a registry of all the potential bone marrow donors in the U.S. Registering to become a potential bone marrow donor is easy. Just sign up online or at your local blood donation center. They will take a cheek swab to get a sample of your DNA to determine your human leukocyte antigen (HLA) profile and add you to the registry. It’s that easy. Later if you are found to match a patient, you will be asked to undergo a more thorough examination prior to being accepted.

Finding a match

Matching for a bone marrow transplant is more difficult than matching for a kidney transplant. Tissue matching for kidney transplants requires that the patient not have blood type or HLA antibodies for the donor kidney. Most patients have low levels of HLA antibodies, so the chances finding a match are high enough that friends can often donate to them.

Tissue matching for bone marrow is much tougher. It isn’t enough for the patient not carry antibodies for the donor’s bone marrow. The donor and the patient must be exact matches for blood type and HLA profile. The chance of matching with a family member are about 10%. But the chance of matching someone who is not related are less than one-in-100,000, or even lower for minority groups that have less common HLA combinations.

The difficulty of finding a match was not known at the time the registry was created. An excellent story about the discovery of the match difficulty was published in the New York Times Apr 1989.

Encouraging more minorities to participate

Students at many universities now run campaigns to recruit minorities to register to become potential donors. For instance at Dartmouth, Stillman, Univ. Arizona,  Univ. San Francisco, and other campuses nationwide. Vivek Kumar, a software developer in the Bay Area, has led several of these drives and is featured in the video below.

Video from ABC7

Matching software

In a Nov 2010 press release, IBM has announced that NMDP has adopted IBM’s business process management (BPM) software to help automate the matching process.

The NMDP operates a registry of 8 million potential donors. It also cooperates with international registries to access to a total of 14 million potential donors worldwide.

By adding BPM software, the NMDP can take advantage of advanced analytics, social networking and reporting to streamline the record matching process by creating a online dashboard that hospital staff can use to track patients, potential donors, and the results of potential matches by geographic location.

IBM and NMDP hope the new software will lead to faster matches, which hopefully will lead to faster bone marrow transplants and improved medical outcomes for patients.

The U.S. Census has just started to release the 2010 census data at the county level. Until now, only state level data was available. Data for 21 states is currently available with the rest to be released over next two months as it becomes available.

In addition to providing the data in CSV format, the U.S. Census has created several nifty interactive maps that can be viewed at their site or embedded in your site. (Embedding requires support for iframe tags.) The map widget includes three maps, population change, population density, and apportionment for each state for every decade for the past 100 years. It’s pretty nifty and an excellent example of how dense data can be made more approachable.

Image from U.S. Census

[Update: The tool developed by Arbor Research described in this blog post is no longer available.]

If you are a kidney patient considering transplant as a therapy option, you are probably wondering how long you can expect to wait for a deceased donor transplant. There is a free online tool available that you can use to estimate the typical wait time. The tool, developed by Arbor Research, asks four simple questions,

1. Which transplant center are you a patient at
2. What is your blood type
3. Are you a juvenile or an adult
4. If you are an adult, will you (actually your transplant surgeon) accept an expanded criteria donor (ECD)

The tool is fast and easy to use. However, if you have high sensitivity to human leukocyte antigens, then be aware that the estimates will be too low for you, as I explain later.

About the data

The calculator website doesn’t provide any documentation on how it works and what data was used. Luckily, I was able to get detailed information from Arbor Research. Much thanks to Melissa Fava, a coordinator, and Keith McCullough, a data analyst, both at Arbor Research, for taking time to answer all my questions and explaining upcoming updates to the tool.

The typical wait time is the median time to transplant among those who received a transplant in 2008 or 2009. The calculator uses patient data from the Scientific Registry of Transplant Recipients (SRTR) available as of October 2010. Only ‘kidney-alone’ transplant data was used, meaning data for kidney-pancreas, kidney-liver, or other multi-organ transplants were excluded. Transplants using organs from donors who test positive or indeterminate for hepatitis B or hepatitis C were excluded. For all transplant patients, the year they were added to the waiting list was found and the wait time calculated using mid-year convention.

Each transplant was grouped by transplant center, the recipient blood type (O, A, B, or AB), and the donor criteria a candidate is listed for (adult, standard criteria; adult, expanded criteria; or juvenile, standard criteria). The median wait times among recipients within each cell was reported. Cells with fewer than 10 subjects were not reported.

Why do wait times vary among transplant centers?

The typical wait time for a deceased donor kidney varies considerably from one transplant center to another. However, it isn’t because some centers are better at finding kidneys than others. Instead, it is based on three factors.

First, transplant centers generally don’t recover their own organs. Instead, they rely on an organ procurement organization (OPO) that recovers organs from a specific geographic region. Most OPOs serve more than one transplant center and have protocols in place to distribute the organs fairly among all the transplant centers within their donation service area (DSA).

Organ recovery rates vary considerable geographically. Recovery rates are lower in large cities. So OPOs that have a large proportion of their population in urban areas have the longest waiting lists. For instance, there are four separate OPOs in California and all contain at least one large city in their DSA. Thus, they all experience low organ donation recovery rates and all transplant centers in California have long wait times.

Second, within a single DSA, children’s hospitals have a shorter wait time than other hospitals. That’s because pediatric patients get priority access to organs.

Finally, some patients are easier to match than others. Patients with AB blood type can accept a donor organ of any other blood type. Patients with O blood type can only accept a donor organ from type O. Some patients have developed antibodies to human leukocyte antigens (HLAs) making it harder to find a matching kidney for them. Some transplant centers have more of these hard-to-match patients than others. So even though the organs are distributed fairly, some hospitals will have patients who end up waiting longer than others. As far as I am aware, transplant centers do not reject patients because they are hard to match or specialize in treating patients who are hard-to-match. The exception is transplant centers that offer intravenous immunoglobulin therapy to remove HLA antibodies prior to transplant.

National, regional, and transplant center median wait times

Let’s take a look at the data. The table below shows the typical wait times for all patients at all transplant centers combined. In general, patients with AB blood type are the quickest to find a match while O and B are the slowest.

The table below shows the typical wait time for adult patients receiving a kidney transplant from a standard criteria donor at three large transplant centers in different parts of the country. Note that although Univ. Wisconsin in Madison is geographically close to Northwestern Memorial Hospital in Chicago, it is in a different  donor service area that excludes the large cities of Chicago and Milwaukee, resulting in significantly shorter wait times.

The table below shows the typical wait times for the five transplant centers within Washington state. All are within the donation service area for LifeCenter Northwest. Excluding Seattle Children’s Hospital, wait times do not vary greatly between hospitals, except Swedish Medical seems to have a longer wait times than the others. (I can’t explain it, nor can I determine if it is statistically significant.)

Typical wait time underestimates average wait time

Calculating a median wait time for patients receiving a transplant during a year is not the same as the median wait time for all patients. That’s because patients who did not receive a transplant during the year are not included in the calculation.

Also note that for open-ended data like wait time, the median will be much lower than the mean. Some hard-to-match patients (specifically, those  with high sensitivity to HLAs) wait much longer than others. They may also be more likely to exit the list prior to transplant, meaning they are not included in the wait time calculation.

The median wait time of patients who receive a transplant understates the expected wait time even more in states like California where the wait time is very long and more than half of patients entering the waiting list never get a transplant. That is, the total patient median wait time is undefined. Patients exit the waiting list for a variety of reasons including death, becoming too ill for transplant surgery, or losing contact with their transplant center.

So why not include all patients in calculating a median wait time? That’s because some patients on the UNOS waiting list are not actually ready to receive a transplant. Some patients are on the so-called inactive list, meaning they are not considered a suitable candidate for a transplant. Including all patients when calculating a median wait time would skew the median wait time too high. But it isn’t desirable to exclude all inactive patients from the calculation either. That’s because patients move back and forth between the active and inactive list. Some patients who receive a transplant are still listed as inactive at the time of surgery. For a detailed discussion of the active and inactive waiting lists, see this Apr 2010 blog post.

A note about cPRA score

The most common measure of the level of sensitivity to HLAs is called the calculated panel reactive antibody (cPRA). It is the percentage of donor organs that would be expected to be incompatible to the patient. The level can vary from 0% to 99%. Most patients have low cPRA level and that is captured in a median wait time. However, high cPRA patients wait much longer. The median wait time for all patients is not a good estimate of their wait times.

One way of improving the accuracy of the wait time calculator would be to include cPRA level as one of the independent variables. However, many patients do not know their cPRA score and Arbor Research did not want to discourage patients from obtaining potentially useful information on wait times.

Also, the median wait times are calculated on individual cells and adding cPRA would make each cell too small to calculate a median. Arbor Research plans to update the wait time calculator soon. It expects to use a random-effects model rather than a simple median for each cell. This should allow estimating a wait time for cells that currently have no estimate and to include additional variables like cPRA so as to produce better wait time estimates for highly sensitized patients.