American Association for Physician Leadership


Transfusion Utilization and Appropriateness: Thinking Differently at a Tertiary Academic Medical Center

Jennifer Dawson, MBA, MSN, RN, NE-BC

Craig Schwabl, BS, MBA

Peter Pronovost, MD, PhD

Richard Jordan, MD, MS

Keith A. Andrews, DO

Zil Patel, DO

James Hill, Jr., MD, MBA, CPE, FASA, FACHE

July 1, 2022

Volume 9, Issue 4, Pages 18-23


Optimizing the transfusion of blood components reduces risk to patients and costs to organizations. Given the frequency of transfusion and the variation in practice, clinicians desire a way to measure performance in transfusion practices that is relevant for all patient scenarios and accurately attributes to the provider influencing the decision. To identify defects in value, an algorithm was developed that allowed rapid analysis of transfusion events to measure performance against a group of clinical measures focusing on mindful variation. Data from the blood bank inventory and electronic patient record were aggregated into an interactive dashboard and shared transparently with transfusing providers and leaders monthly to enhance awareness. It was hypothesized that an application to raise awareness to unwarranted variation would lead to a decrease in transfusions and an increase in appropriate transfusion.

Blood transfusion is one of the most frequent medical procedures in the United States, with large variability in clinical practices.(1-3) Prudent utilization of blood components enhances patient safety and hospital economics. Transfusions carry risk for direct, immediate consequences, as well as additional consequences that can occur much later.(4,5)

In 2017, there were more than 10 million packed red blood cell (pRBC) transfusions.(2) The cost of pRBC, including storage, labor, and waste, ranges between $522 and $1,183 per unit.(6,7) With the increasing cost of healthcare, it is imperative for organizations to implement strategies that reduce costs, one of which is optimizing the utilization of blood components. While restrictive transfusion guidelines have been proven safe and effective for stable patients, certain comorbidities such as sepsis and clinical criteria such as cardiac history necessitate more liberal transfusion practices.(8-11)

A significant percentage of transfusions are unnecessary or non-beneficial.(9,12) Yet, providers struggle to measure the effectiveness of transfusions due to the variation in the indications for various patient populations and clinical situations. For example, a hemoglobin level might not be measured because the test is not a priority in an emergency or complex situation. Another challenge is attributing responsibility for transfusion decisions when working within a larger multidisciplinary team including residents, and in providing care between multiple locations, including the operating suite.

Given the frequency of transfusion and the variation in practice, clinicians desire a way to measure performance in transfusion practices that is relevant for all patient scenarios and accurately attributes to the provider influencing the transfusion decision.

Considering the key principles for eliminating defects in value, the goal of this initiative was to share data transparently via a dashboard that included an evidence-based transfusion algorithm considering patient-specific criteria.(13) The sophistication of an algorithmic approach allows rapid analysis of transfusion events to measure performance against defined clinical criteria. Including patient-specific variables at the time of each transfusion engages providers to focus on mindful variation, or the variation in practice that is warranted in each clinical scenario.

It was hypothesized that use of such an application to raise awareness would lead to a decrease in total use of pRBC and platelets with an increase in appropriate transfusions.


The campus was large and the resources for the initiative were limited. The 1,032-bed campus included a women’s hospital, pediatric hospital, cancer center, transplant program, and level 1 trauma services for the region. A physician director of patient blood management was appointed as 0.2 full-time equivalent (FTE) to work in collaboration with a registered nurse senior operations engineer committing 0.5 FTE. Additional resource hours were engaged as needed for process evaluation and clinical review.

In addition, the Information Technology Development team completed the initial application build, including both the data collection and visualization components. This was accomplished utilizing 1.0–1.5 FTE resources over the course of one year.

A transfusion appropriateness algorithm (TAA) was developed to analyze blood transfusions for all adult patients at a tertiary academic medical center. A multidisciplinary specialist panel of more than 30 participants was assembled to review current literature and existing practice guidelines to define clinical criteria for appropriate transfusion, with evidence suggesting that such a team approach enhances patient safety.(8)

Specialties involved included anesthesia, cardiac surgery and cardiology, critical care, emergency medicine and trauma, gastroenterology, pathology, hematology, oncology, neurology, interventional radiology, obstetrics, surgery, medicine, orthopedics, and transplant. Each representative completed an evaluation, which included specific patient demographics, transfusion recommendations, and an assessment of the variation in current transfusion practices. The aggregation of data was discussed by the panel to determine patient-specific clinical indications for beneficial transfusions.

Restrictive transfusion indications were set as a hemoglobin (Hgb) of < 7g/dL for pRBC, and a platelet count of < 10K/µL for platelets.(8,12,14) In addition, less restrictive threshold parameters and alternate clinical measures were chosen to address common clinical situations that necessitate transfusion without meeting the restrictive indications; for example, measuring cardiac patients to a Hgb of 8, measuring lactate levels to identify inadequate tissue perfusion, or considering transfusion appropriate for bleeding patients.(8,10,11,14) Standard and alternate metrics for each blood component were aggregated into the TAA, choosing to assign the more liberal measurements in multidimensional recommendations.

Transfusion data were pulled from the blood bank inventory software, WellSky Transfusion, and the electronic medical record (EMR). Given the distributed nature of the blood bank and EMR systems, the blood bank inventory software was selected as the source of truth for blood used. Data were initially collected from the blood bank system and then enriched with clinical data collected from the EMR.

Based on the selected blood component, the algorithm evaluated specific lab values and ranges, clinical conditions, procedures, activated protocols, physician attributes, and temporal boundaries to determine transfusion appropriateness. Various combinations of these criteria were invoked until a specific threshold was exceeded or all relevant combinations were exhausted. In the event of multiple transfusion events, the algorithm was re-evaluated with the most up-to-date relevant data. Results were stored and included in enterprise metrics and key performance indicators measuring the overall score of the assessment.

To allow provider-specific feedback and attribute responsibility, the blood order was enhanced to include an auto-populating attending provider field. The attending was changed when care was coordinated between multiple specialties to accurately reflect the source of the transfusion request. Further attribution was enhanced by aggregating orders entered by residents to the attending provider. Data were attributed to the provider chosen in the attending provider field. This eliminated the focus on the discharging provider and moved the focus to the provider making each transfusion decision throughout each patient’s encounter.

All collected data were then aggregated into a dashboard presenting the laboratory thresholds for each blood component alongside the appropriateness score derived from the TAA as seen in Figure 1. The application housed multiple reports and allowed for reviewing the provider or patient lists for comparison or further study. A consolidated report for each patient encounter showed individual transfusion data across time, including each provider who prescribed the treatments.

Figure 1. Dashboard Sample. Note. Dashboard copied from the analytic software cropping additional wrap and report links from the image. The live screen includes direct links to focused reports, and each screen offers interactive drill through capabilities to access data behind each metric.

Data were loaded starting in August 2017 to apply the algorithm and sort data for analysis. Mined data included the blood component given, specialty service assigned to the patient, associated procedures, active medications, pertinent lab values, blood loss, vital signs, other blood products released during a particular patient encounter, the responsible provider, and that person’s credentialed department.

In August 2018, the dashboards were shared monthly with the transfusing providers with the goal of enhancing awareness of their transfusion practices. Transfusions with clinical indications falling inside the TAA parameters were reported as “appropriate” on the dashboard, along with standard data of lab values at the time of transfusion.

After signing into the application, providers could select a patient to review and see the clinical data points that were collected across the entire encounter. To improve awareness, communications were multifaceted and focused on reaching the largest number of departments between August 2018 and March 2019. Email introductions and reminders were sent to all providers by senior leadership at the start of the implementation period and when changes were made to the EMR blood order.

Presentations were made by project leaders and department leaders to introduce the measurement system and to discuss its implications. Internal and medical community publications showcased progress mid-implementation to encourage participation. Leaders used features of the dashboard as the basis for focused discussions on practice variation and to monitor ongoing transfusion activity.

An interrupted time-series study design was used to compare two 12-month periods, including the time prior to implementation, from August 2017 through July 2018, and the time following implementation, from April 2019 through March 2020. To minimize the influence of discussions versus the availability of data impacting transfusion decisions, the period of implementation was not included. Due to the large volume, the greatest opportunity and highest focus of the initiative were pRBC, followed by platelets, due to their high cost in comparison to the other components.

All pRBC and platelets released from the blood bank to patients older than 18 were included in analysis. Utilization included all pRBC and platelet components transfused during the measured time periods. The appropriateness score was determined by the results of the measures captured by the TAA for each transfusion. Parametric inferential statistics using the two-sample t-test were performed for the means and differences between the time periods. Statistical significance was set as a p value < 0.05.

The mean and standard deviations were found for each variable by time period. The mean difference with confidence intervals and p values were analyzed to identify change between the time periods. Transfusions were analyzed by total transfused units, and appropriateness was analyzed as the percent of appropriate units of the total transfused units.

Additional descriptive data were collected, including the quantity and types of hospital encounters; the case mix index; and patient outcomes of geometric mean length of stay (GMLOS), mortality, and readmissions. A two-sample t-test was performed for each variable to identify differences across the time periods. The median and interquartile range (IQR) were found to identify variability over time and in totality.

Transfusion costs were calculated by multiplying the number of transfused units in each time period by 2.5 times the average acquisition cost. Calculating the cost of transfusion was complex and required accounting for acquisition costs and the additional costs of storage, administration, and system and societal expenditures related to complications.

The consensus on cost estimates from the Cost of Blood Consensus Conference applied activity-based costing methods as high as seven times the acquisition cost of blood components.(15) The calculation factor used for this evaluation was the result of internal evaluation culminating in the costs of the items and services that would not be used if a unit was not transfused, realizing that many additional indirect costs are positively impacted by reduced utilization and increased appropriateness.


Over the two compared 12-month periods, 63,287 patient discharges were included: 31,390 in the pre-implementation group and 31,897 in the post-implementation group (Table 1). There were 46,835 pRBC transfusions included, decreasing from 24,607 to 22,228 between groups, and 15,132 platelet transfusions included, decreasing from 7,989 to 7,143 between groups (Table 2).

Transfusion of pRBC decreased from a mean 2,051 (±109) to a mean 1,852 (±89) units per month (p = <.001) with a mean difference of 198 [CI = 114, 283]. Platelet transfusions decreased from a mean 665 (±105) to a mean 595 (±47) units per month (p = .025) with a mean difference of 71 [CI = 13, 140]. The measure of appropriateness increased for both pRBC and platelets. The appropriate percent of pRBC increased from a mean .67 (± .03) to a mean .81 (± .02) per month (p = < .001) with a mean difference of − .14 [CI = − .16, − .12]. The appropriate percent of platelets increased from a mean .69 (± .04) to a mean .73 (± .03) per month (p = < .022) with a mean difference of − .04 [CI = − .07, − .01] (Table 2).

The types of hospital patient encounters remained stable, averaging 41% surgical, 26% medical, 18% oncology, and 15% obstetric (Figure 2). The case mix index increased slightly (p = <.001) from a median 1.95 with an interquartile range (IQR) of 0.083 to a median 2.04 with an IQR of 0.084 (Table 1). Patient outcomes remained stable. The GMLOS was measured in days with a median 4.16 and IQR of 0.039. Readmissions had a median 14.6% with an IQR of 0.009, and mortality had a median 5% with an IQR of 0.002 (Table 1). Patient outcomes remained the same while transfusions decreased. The 12-month direct cost of pRBC and platelet transfusions decreased by $2,532,594.

Figure 2. Hospital Patient Encounter Types by Month


Implementing the dashboard with the TAA and accurate transfusion attribution was associated with fewer transfused pRBC and platelet units and increased appropriate utilization. Analyzing the status of each patient at the time of transfusion assessed the opportunity for improvement differently from measuring one lab value alone. Relevant data shared transparently influenced self-regulated change in transfusion practice by focusing providers on mindful variation.

Strategies to reduce transfusion have resulted in decreased healthcare costs and lower associated patient risks.(5,8,12) Our findings demonstrate a significant reduction in transfusions and cost savings following implementation of the dashboards without negatively impacting patient outcomes. The change represents an additional 10% reduction in pRBC and platelet transfusions during these compared 12-month time periods, and this reduction is incremental to previous work focused on restrictive standards.

Hasler, et al.,(1) experienced a 10% savings of $330,000 in comparative direct cost savings after implementing restrictive transfusion standards for medicine patients with the intervention of grand rounds and email reminders of the guidelines. An initiative by Frank, et al.,(16) experienced a reduction of $2.5M comparative direct cost savings across five hospitals by implementing a multimodal approach to patient blood management. It is unclear which tactics were most influential in these changes. There were no comparative all-inclusive, patient-focused initiatives. Improvement efforts differed in the described interventions or by exclusions. Utilizing data in an innovative management system, organized around each patient at the time of transfusion, may hardwire mindful variation in transfusion decisions.

The dashboard application was successful for several reasons, including involving specialists from the institution to better define which patient variables are considered when treating with transfusion, incorporating evidence-based knowledge into the TAA, and attributing responsibility thoughtfully. Including specialties with high use and complex patients, such as oncology, obstetrics, and cardiac surgery, was especially important in developing the most relevant analysis to then apply to practice.

By utilizing patient variables to identify clinical indications and assess transfusion appropriateness, the focus of administering a transfusion shifts from treating a lab value (i.e., hemoglobin) to treating a patient, thus providing better and more holistic, patient-centered care.

We recognize this study has limitations. This was an observational study design and thus cannot make a causal inference between our intervention and improved transfusion practices. Yet there were no other interventions during the post-intervention period to account for the improvement. We implemented the software at one large tertiary academic medical center, and the results may not be generalizable to other hospitals.

What began as an internal solution transformed into an innovative method to address transfusions on a larger scale. Future development of the TAA will include both measurement enhancements and reporting enhancements with standard application programming interface queries for data using standard data labels for compatibility with any application. Access may be opened to all system users with live links to the software and dashboard communications. Access to the application may be expanded to other facilities within the larger healthcare system and to hospitals outside the system.

Future studies may enhance the TAA for various population demographics and in the management of additional medical conditions. The appropriateness metric may be shared with the patient to enhance the transparency of quality metrics. Another option may be to incorporate the TAA into the EMR software systems to provide relevant, evidence-based information to providers at the time of order entry paired with the automated data reporting and analysis currently provided.


While standardized restrictive transfusion criteria have been well-established, the decision to treat with transfusion is far more complex. In light of data demonstrating that nearly half of all transfusions in this country may result in more harm than good, posing significant risk and costs to both patients and hospital systems, this effort was put forth to provide relevant data in the pursuit of highly reliable transfusion practice.

The sophistication of the algorithmic approach, using patient-specific variables to analyze transfusion events against evidence based clinical criteria, provided value by raising awareness to unwarranted variation. We were able to demonstrate a reduction in pRBC and platelet transfusions and an increase in appropriate utilization despite the blood management efforts undertaken prior to this initiative. Sharing relevant data transparently helped providers optimize transfusions to reduce patient risk and cost.


  1. Hasler S, Kleeman A, Abrams R, et al. Patient Safety Intervention To Reduce Unnecessary Red Blood Cell Utilization. Am J Manag Care. 2016;22(4):295–300.

  2. Jones JM, Sapiano MRP, Savinkina AA, et al. Slowing Decline in Blood Collection and Transfusion in the United States — 2017. Transfusion. 2020;60(Suppl 2):S1–S9.

  3. Ramsey G, Wagar EA, Grimm EE, et al. Red Blood Cell Transfusion Practices: A College of American Pathologists Q-Probes Study of Compliance with Audit Criteria in 128 Hospitals. Arch Pathol Lab Med. 2015;139(3):351–355.

  4. Aubron C, Aries P, Le Niger C, et al. How Clinicians Can Minimize Transfusion-related Adverse Events. Transfus Clin Biol. 2018;25(4):257–261.

  5. Redding N, Piews D, Dodds A. Risks of Perioperative Blood Transfusions. Anaesth Intens Care. 2019;20(3):74–78.

  6. McQuilten ZK, Higgins AM, Burns K, et al. The Cost of Blood: A Study of the Total Cost of Red Blood Cell Transfusion in Patients with β-thalassemia Using Time-driven Activity-based Costing. Transfusion. 2019;59(11):3386–3395.

  7. Shander A, Hofmann A, Ozawa S, et al. Activity-based Costs of Blood Transfusions in Surgical Patients at Four Hospitals. Transfusion. 2010;50(4):753–765.

  8. Carson JL, Guyatt G, Heddle NM, et al. Clinical Practice Guidelines from the AABB: Red Blood Cell Transfusion Thresholds and Storage. JAMA. 2016;316(19):2025–2035.

  9. Goodnough LT, Panigrahi AK. Blood Transfusion Therapy. Med Clin North Am. 2017;101(2):431–447.

  10. Minasyan, H. Sepsis and Septic Shock: Pathogenesis and Treatment Perspectives. Journal of Critical Care. 2017;40:229–242.

  11. Sadana D, Pratzer A, Scher LJ, et al. Promoting High-Value Practice by Reducing Unnecessary Transfusions with a Patient Blood Management Program. JAMA Internal Medicine. 2018;178(1):115–122.

  12. Burns CD, Brown JP, Corwin HL, et al. Special Report from the Society for the Advancement of Blood Management: The Choosing Wisely Campaign. Anesth Analg. 2019; 129(5):1381-1386.

  13. Pronovost PJ, Urwin JW, Beck E, et al. Making a Dent in the Trillion-dollar Problem: Toward Zero Defects. NEJM Catalyst Innovations in Care Delivery. 2021;2(1):1–23.

  14. Kaufman RM, Djulbegovic B, Gernsheimer T, et al. Platelet Transfusion: A Clinical Practice Guideline from the AABB. Ann Intern Med. 2015;162(3):205–213.

  15. Abraham I, Sun D. The Cost of Blood Transfusion in Western Europe as Estimated from Six Studies. Transfusion. 2012;52(9):1983–1988.

  16. Frank SM, Thakkar RN, Podlasek SJ, et al. Implementing a Health System-wide Patient Blood Management Program with a Clinical Community Approach. Anesthesiology. 2017;127(5):754–764.

Jennifer Dawson, MBA, MSN, RN, NE-BC

Jennifer Dawson, MBA, MSN, RN, NE-BC, is an operations engineer and high-reliability medicine strategist at University Hospitals Health System in Cleveland, Ohio.

Craig Schwabl, BS, MBA

Craig Schwabl, BS, MBA, is the vice president of digital solutions and enterprise analytics at University Hospitals Health System in Cleveland, Ohio.

Peter Pronovost, MD, PhD

Peter Pronovost, MD, PhD, is the chief clinical transformation officer at University Hospitals Health System and professor for the schools of medicine, nursing, and management at Case Western Reserve University in Cleveland, Ohio. He co-chairs the Healthcare Quality Summit with the Deputy Secretary of the Department of Health and Human Services.

Richard Jordan, MD, MS

Richard Jordan, MD, MS, is an anesthesiology resident physician (CA-1/PGY-2) at Case Western Reserve University–University Hospitals in Cleveland, Ohio.

Keith A. Andrews, DO

Keith A. Andrews, DO, is an anesthesiology resident physician (CA-2/PGY-3) at Case Western Reserve University–University Hospitals in Cleveland, Ohio.

Zil Patel, DO

Zil Patel, DO, is an anesthesiology resident physician (CA-1/PGY-2) at Case Western Reserve University–University Hospitals in Cleveland, Ohio.

James Hill, Jr., MD, MBA, CPE, FASA, FACHE

James Hill, Jr., MD, MBA, CPE, FASA, FACHE, is the chief operating officer and critical care anesthesiologist at University Hospitals Parma Medical Center and an assistant professor for the school of medicine at Case Western Reserve. He previously was the chief medical officer of University Hospitals Parma Medical Center and the system medical director of transfusion services and blood management and division chief of trauma anesthesiology at University Hospitals Cleveland Medical Center in Cleveland, Ohio.

Interested in sharing leadership insights? Contribute

This article is available to AAPL Members.

Log in to view.

For over 45 years.

The American Association for Physician Leadership has helped physicians develop their leadership skills through education, career development, thought leadership and community building.

The American Association for Physician Leadership (AAPL) changed its name from the American College of Physician Executives (ACPE) in 2014. We may have changed our name, but we are the same organization that has been serving physician leaders since 1975.


Mail Processing Address
PO Box 96503 I BMB 97493
Washington, DC 20090-6503

Payment Remittance Address
PO Box 745725
Atlanta, GA 30374-5725
(800) 562-8088
(813) 287-8993 Fax



AAPL providers leadership development programs designed to retain valuable team members and improve patient outcomes.

American Association for Physician Leadership®

formerly known as the American College of Physician Executives (ACPE)