The goal of this study is to test whether emails that inform patients they have a blood type in need are more effective at encouraging patients to schedule and attend blood donation appointments, compared to email messages that do not mention the patient has a blood type in need.

There has been a years-long national shortage of several blood types in the U.S., including in the Geisinger community. Previously, the study team collaborated with Miller Keystone, where Geisinger refers patients who wish to donate blood and from whom Geisinger receives blood for clinical purposes, on an outreach study to encourage blood donation in patients with needed blood types. The study demonstrated that, compared to a no-message control group, patient portal messages sent to patients with needed blood types increase patients' likelihood of attending donation appointments. However, results were ambiguous with respect to which of two message versions was most effective. One version, which stated that the patient had a needed blood type (blood-type message), caused a numerically, but not significantly, higher number of patients who attended appointments compared to the other version, which did not state that the patient had a blood type in need but rather informed the recipient of a general blood shortage (no-blood-type message).

Because Miller Keystone particularly values reaching new donors, the team ran a preregistered exploratory analysis to test whether the messages were differently effective for new donors compared to those who had previously donated at a Miller Keystone site. There was a significant interaction between previous donor status (previous donor, previous non-donor) and message type, such that previous non-donors were relatively more responsive to the blood-type message, while previous donors were more responsive to the no-blood-type message. However, the groups were uneven with respect to the number of patients that had previously donated. Moreover, when the analysis was limited to patients who opened their messages, this interaction effect disappeared: blood-type messages were still most effective in previous non-donors, and there was no difference in message effectiveness among previous donors. These follow-up analyses, and the unevenness of previous donors across groups, call into question the robustness of the interaction effect.

The present study will again test whether the blood-type message is more effective than the no-blood-type message for patients with needed blood types overall, and separately for previous donors and non-donors. Messages will be sent via email rather than via patient portal in the present study. Randomization will occur at the email-address level to one of the two message types to ensure everyone using the same email address receives the same message (although each patient will be sent an individualized message with their name; there will be no no-contact control condition this time). Email addresses will be excluded if they are shared by patients with different blood types. Randomization will be stratified by whether all patients using the same email address are previous donors or not (email addresses shared by previous donors and non-donors will be excluded).

Inclusion Criteria:

• Documented blood type in short supply
• Age 18+
• Can be contacted via email

Exclusion Criteria:

• Hemoglobin test result < 12.5 within the 3 months prior to list creation
• Shares an email address with a patient of a different needed blood type
• Email address associated with at least one previous donor and one non-previous donor

Data with no personally identifiable information will be made available to other researchers on the Open Science Framework for transparency.

The data will become available after publication of study results in a scientific journal and will be available as long as the Open Science Framework hosts the data.

The data on the Open Science Framework will be open to anyone requesting that information.

The patients in the study will not know that other messages are being sent to other patients, although they will see the text of their own message. Providers will be blind to patient conditions.