This study was conducted at The Hospital for Sick Children (SickKids), a large academic pediatric institution in Toronto, Canada. SickKids cares for approximately 350 new patients with cancer and performs 100 hematopoietic stem cell transplants (HSCTs) per year. SickKids has used Epic as its EHR since 2018 [14]. The project consisted of two 1-year periods: November 2023 to October 2024 (preimplementation) and November 2024 to October 2025 (postimplementation) of the transfusion therapy plans. All data were obtained from our institution’s enterprise data warehouse, named SickKids Enterprise-Wide Data in Azure Repository [15].

This project was approved by the Quality Improvement Review Board at SickKids (QIP-2024-01-26T13-52-53). Informed consent was not required, and participants were not compensated for participation. Patient privacy and confidentiality were maintained through anonymization of data.

The workflows for ordering, preparing, and administering blood using both EHR tools (order set and transfusion therapy plan) are detailed in Table S1 in Multimedia Appendix 1. In brief, using the order set for transfusion, orders are entered by a provider at the time a transfusion is needed and cannot be reused, whereas with the therapy plan, the orders can be entered in advance or at the time of transfusion and can be reused. Regardless of how the transfusion orders are entered, a nurse then reviews the orders and “releases” them, rendering them active in the system. One of these orders is a “prepare” order, which specifies the blood product requirements to the blood bank. The blood bank then prepares the requested product and delivers it to the patient’s nurse. The nurse then administers any premedications, uses a “transfuse” order to administer the blood product, monitors for transfusion reactions, and administers the emergency medications if a reaction occurs.

Two transfusion therapy plans were created—one for pRBC orders and one for platelet orders. New sections in our “Treatment” activity, where cancer-specific orders are placed, were created for applying the transfusion therapy plans (Multimedia Appendix 2). Both therapy plans contain the prepare and transfuse orders for their respective blood components (Multimedia Appendices 3 and 4). The blood product orders were configured to maximize adherence to our transfusion volume guidelines. For pRBC orders, if the patient weighs less than 25 kg, providers are presented with order options in mL/kg/dose, and if they weigh 25 kg or more, they are presented with order options in units. For patients less than 25 kg, the maximum mL/kg/dose volume that can be ordered is 10 mL/kg, and for patients more than 25 kg, only 1 unit of pRBC can be ordered at a time. For the platelet orders, if the patient weighs less than 20 kg, providers are presented with order options in mL/kg/dose, and if they weigh 20 kg or more, they are presented with order options in units. For patients less than 20 kg, the maximum dose that can be entered is 10 mL/kg (200 mL), and for patients more than 20 kg, the only dose that can be entered is 1 unit. Hard stops were configured within the orders to prevent ordering transfusion volumes that exceed the recommended maximums. If additional units are required, direct communication with the blood bank is required.

Both plans contain optional premedications to prevent transfusion reactions and emergency medications that can be administered in the case of a transfusion reaction. The pRBC plan also contains a type and screen order. Both plans also include a communication order that indicates in what context the nurse can release the transfusion orders. For the pRBC therapy plan, the default options a provider can select are to transfuse for a hemoglobin level of less than 70, 80, or 90 g/L, and for the platelet therapy plan, the default options a provider can select are to transfuse for a platelet count of 10 or 20 (if febrile), 20, 30, or 50 kcells/µL. However, a provider can manually enter a patient-specific threshold. If a transfusion is needed and the threshold in the communication order is not met, but a transfusion is felt to be clinically indicated by the team (eg, threshold indicates transfuse for hemoglobin of 70 g/dL, and the patient’s hemoglobin is 71 g/dL in a patient with tachycardia and fatigue), an ad hoc one-time nursing communication order can be entered approving the use of the transfusion therapy plan. Dosages of both blood products and supportive care medications automatically update using the patient’s most recent weight. An automated in-basket message to the primary oncology provider is generated every 6 months, indicating that the plan must be reviewed and either renewed or discontinued.

Once the therapy plan is applied and signed, the orders within the therapy plan can be released by a nurse without the need for a provider to enter additional orders. The therapy plan orders can be reused each time a transfusion is required, eliminating the need to reorder blood products each time a transfusion is required.

Patients with active cancer and those undergoing HSCT were included. Patients with active cancer were defined as those seen in an oncology clinic or admitted to an oncology service with an active chemotherapy treatment plan (defined as 1 chemotherapy day within the last 30 days). Patients with HSCT were defined as any patient within 100 days of the infusion of a cellular therapy product.

We evaluated therapy plan usage in 3 ways. Our primary outcome was to evaluate the proportion of eligible blood product orders that came from a transfusion therapy plan on a monthly basis. Secondary outcomes included evaluating the proportion of eligible patients who had a transfusion therapy plan applied and the proportion of eligible patients who received a transfusion during the study period and had a transfusion therapy plan applied. Only transfusions given within an oncology or HSCT department were considered (ie, transfusions in the emergency department, operating room, or intensive care unit were not included). The use of the platelet and pRBC therapy plans was tracked separately.

We evaluated changes in hospital efficiency in 3 ways. First, for patients who received a blood product, we evaluated the time between the complete blood count (CBC) being reported in Epic and the blood being released and administered by the nurse caring for the patient. These metrics were calculated on a per-transfusion basis. If more than one unit of a blood product was included in a single release, these were counted as a single transfusion, and the release of the transfusion was counted as the time for the first unit. Second, we hypothesized that our transfusion therapy plan might decrease the number of times where a CBC and type and screen were drawn in separate blood draws. We evaluated the proportion of transfusions where the CBC and type and screen were drawn within 5 minutes of each other. Since many inpatients might already have one of these premedications ordered for an alternative indication, we only evaluated premedications for outpatient administrations of blood products. Finally, we evaluated the estimated change in overnight transfusion ordering related to eliminating the need for providers to order blood overnight. We calculated the number of blood product orders placed between 7 PM and 7 AM.

To evaluate changes in blood product ordering patterns, we evaluated the mean pretransfusion hemoglobin and platelet levels prior to the transfusion of blood or platelets, respectively. All oncology patients and patients with bone marrow transplants (BMTs) should have their blood ordered as irradiated. We evaluated the proportion of blood products ordered with this precaution during both periods as well. We also evaluated the proportion of patients receiving an appropriate transfusion volume. For blood, the maximum appropriate transfusion volume was defined as 10 mL/kg for patients weighing less than 25 kg and 1 unit for patients weighing more than 25 kg. For platelets, the maximum appropriate transfusion volume was defined as 10 mL/kg up to 200 mL for patients weighing less than 20 kg and 1 unit for patients weighing more than 20 kg. These volumes are in accordance with our internal blood transfusion policy. In the postimplementation period, we also compared the same safety metrics for transfusions that came from a therapy plan compared to those that came from an order set. In addition, we hypothesized that premedications would be less frequently missed with the use of the new therapy plan. We calculated the proportion of patients who had an allergy to blood or platelets listed in Epic and who had a premedication (hydrocortisone, diphenhydramine, cetirizine, or acetaminophen) ordered within 5 minutes of the blood product they were allergic to being ordered.

To assess health care practitioner experiences using the transfusion therapy plan, a survey was administered 6-months post go-live of the transfusion therapy plan to physicians and nurses who care for oncology patients and those with BMT via REDCap (Research Electronic Data Capture; Vanderbilt University). The survey consisted of three parts: (1) demographic information about the respondent, (2) an abbreviated version of the technology acceptance model (TAM) questionnaire, and (3) specific questions related to the impact on transfusion-related practices [16-18]. Demographic information included health care role (oncology attendings, oncology fellows, pediatric residents, clinic nurses, or ward nurses), primary area of practice (leukemia or lymphoma, solid tumor, neuro-oncology, or BMT), and duration of employment at our institution. The TAM is a questionnaire that assesses technology acceptance across two main constructs: (1) perceived usefulness of the technology and (2) perceived ease of use of the technology. The TAM has been shown to be both reliable and valid. To assess the impact of the transfusion therapy plan on specific blood ordering practices, novel questions were developed by a subset of the study team with expertise in questionnaire development, clinical informatics, usability, and transfusion medicine.

Descriptive statistics were used to summarize demographic data, transfusion history, and therapy plan usage. To determine factors associated with the use of a therapy plan and differences in outcomes between order set and therapy plan–derived transfusions, the chi-square test and Fisher exact test were used for categorical variables, and the Wilcoxon rank-sum test was used for continuous, nonnormally distributed variables (RStudio version 1.4.1717; Posit PBC).

The demographics of the preimplementation and postimplementation cohorts are shown in Table 1. The preimplementation cohort included 558 patients and the postimplementation cohort included 521 patients. No statistical differences in the baseline characteristics of the 2 cohorts were identified.

During the 1-year postimplementation period, 26% (n=138) and 24% (n=125) of eligible patients had a pRBC or platelet therapy plan applied, respectively. Of the 238 patients with either transfusion therapy plan applied, 59% (n=141) had both plans applied. The proportion of eligible patients who had either transfusion therapy plan applied ranged from 32% (97/307) for patients with leukemia or lymphoma to 100% for patients with neuro-oncology (75/75) and BMT (95/95; Table S2 in Multimedia Appendix 1). Of the 220 patients who received pRBCs, 58% (n=127) had a pRBC transfusion plan, and of the 158 patients who received platelets, 67% (n=106) had a platelet transfusion plan. Overall, during the postimplementation period, 59% (n=1634) of transfusion orders originated from a transfusion therapy plan. The proportion of eligible patients with either therapy plan, the proportion of patients who received a transfusion who had a therapy plan, and the proportion of transfusion orders coming from a therapy plan, stratified by month, are shown in Figures 1-3. In the last month of our evaluation period (October 2025), 39% (60/156) of eligible patients had at least 1 therapy plan, 60% (35/58) of patients who were transfused had the corresponding transfusion therapy plan, and 72% (88/122) of all transfusion orders came from a therapy plan.

Changes in efficiency metrics preimplementation and postimplementation of the transfusion therapy plans are shown in Table 2. There was a significant difference in the timing between the transfusion and premedication orders (median 14, IQR 0-159 vs 10, IQR 0-145 min; P=.04). In the postimplementation period, there was a significant difference in the median time between the CBC being reported and the transfusion being released (178, IQR 35-394 vs 129, IQR 15-315 min; P<.001) and between the CBC being reported and transfusion administration (227, IQR 92-443 vs 181, IQR 76-367 min; P<.001) when using the therapy plan compared to the order set (Table 3). In the postintervention period, 1453 transfusions from therapy plans were released between 7 PM and 7 AM, suggesting an estimated avoidance of 4 overnight orders per night by the overnight resident or fellow.

The impact of the transfusion therapy plans on safety metrics is shown in Table 4. There was a significant difference in the proportion of patients with allergies who were ordered premedications (273/724, 38% vs 292/651, 45%; P=.008). No significant differences were identified in pretransfusion thresholds between the preimplementation and postimplementation cohorts. There was a significant increase in the proportion of patients receiving guideline-concordant transfusion volumes in the postimplementation period compared to the preimplementation period (58% vs 85%; P<.001). In the postimplementation period, there was a significant difference in the mean hemoglobin value prior to pRBC transfusion (73, SD 11 g/dL vs 70, SD 5 g/dL; P<.001), the mean platelet value prior to platelet transfusion (24, SD 17 kcells/µL vs 19, SD 13 kcells/µL; P<.001), the proportion of blood products where the volume ordered was guideline-concordant (87/1143, 76% vs 1484/1634, 91%; P<.001) and the proportion of blood products that were ordered as irradiated (814/1143, 71% vs 1595/1634, 98%; P<.001; Table 3).

The health care practitioner experience survey was completed by 57 individuals (27 nurses and 30 providers). This represents a response rate of approximately 15% of our oncology workforce. Of the 27 nurses, 19 worked primarily in the inpatient setting, and 8 worked primarily in the outpatient setting. Of the 30 providers, 4 were attending physicians, 8 were nurse practitioners or physician assistants, 9 were hematology or oncology fellows, and 9 were general pediatrics residents. Results of the abbreviated TAM questionnaire are presented in Figure 4 and Table S3 in Multimedia Appendix 1. Self-reported preferences related to ordering tools for blood products are shown in Table 5. Overall, 95% (n=54) of respondents felt we should continue to use the transfusion therapy plans.