Improving the availability of stem cell donors - A letter and email intervention

In a natural field experiment, we randomly vary invitations to an existing activity of a major stem cell donor center, aimed at increasing the willingness of registered donors to be available at the confirmatory typing stage. The activity asks donors to fill out a health questionnaire, and to report any periods longer than three weeks that preclude donation. The experimental interventions comprise of an invitation, in form of a letter and an email, both with consistent framing, as well as the activity. We run two treatments and a baseline group. In the treatments, letters and emails are both framed so donors either belong to (i) a team of “quickly available donors,” or (ii) a team of “highly engaged” donors. The baseline group receives a letter (and email) with neither framing. The activity itself is the same for all invitees, and is voluntary. Not participating has no consequences for the donor. No one can participate in the activity who is not invited.

2022-02-14

2026-02-15

Our primary outcome variable is a potential donor’s availability at the confirmatory typing (CT) stage.

CT availability measures the donor’s availability (i.e., willingness to move forward) at the confirmatory typing stage to follow through with the stem cell donation process, after a matching patient needing a stem cell transplant has been identified, based on genetic characteristics (such as HLA loci, blood group, and immunity to cytomegolavirus). More specifically, CT availability includes going to a clinic for an extra medical screening, and filling out a health questionnaire.

Our secondary outcome variables include the take-up rate of the activities and unavailability reporting behavior.

The take-up rate measures whether an invited registry member has actively agreed to take part in the activity. This measure will be evaluated continually.

Absence reporting behavior measures how often participants in each experimental group reported being absent.

In a natural field experiment, we randomly vary invitations to an existing activity of a major stem cell donor center, aimed at increasing the willingness of registered donors to be available at the confirmatory typing stage.

Not available

For this activity, the donor center determined the frequencies of donor genotypes (for genes that are to date considered relevant for transplantation). For each genotype a certain number of donors are designated with a maximum allowed age. Thereby for more frequent genotypes the maximum age is set lower than for rare genotypes. The number of donors per genotype remains the same, but over the years people drop out of the desired age and so free spots are generated.

On a rolling basis an algorithm selects donors to fill up the free spots in the activity. For the intervention, the generation of a random number was implemented in the algorithm. Based on this number, the donor randomly receives a specific treatment for the intervention.

Individual (potential stem cell donor)

No

n/a

6,000 observations for primary outcome. Sample for analysis: We will send 133,333 invitations each year (comprising all experimental treatments), in 2022, and 133,333 in 2023, and 133,333 in 2024. In addition to our field experiment, the donor center will collect data on a control group of about 133,333 persons who are randomly not invited to the letter intervention, but share same characteristics as experimental subjects. Other data collected by the donor center: Additionally, the donor center will collect data on 133,333 invitations using the letter they used prior to the intervention, only for internal monitoring.

2,000 from each experimental manipulation (plus ca. 2,000 from a control group (without invitation) constructed by the donor center).

For the experimental treatments, we plan to send around 400,000 invitations out over 4 years, which will in expectation lead to 6,000 CT requests in total. Estimated with past donor center data, there is a 1.5% probability of being requested within 2 years of receiving the letter. If we evaluate the results with a two-sided proportion test, and correct for multiple hypothesis testing over the two treatments and a control letter (giving alpha=0.025), we have a power of 0.8, and a minimum detectable effect size of 3.75pp (intention-to-treat), given a baseline availability rate of 80%. We also expect to add 2,000 CT requests to the data from the group that receives no letter in order to test for overall letter effects. The corresponding MDE in this case (comparing treatments to the control group that receives no letter) amounts to 4pp.