Hospitalizations with blood transfusions and transfusion-related adverse events in US acute care hospitals, 2016–2020

Abstract

Background:

National data on transfusion-related adverse events (TAEs) in the United States are limited. Administrative and payment-related data may augment hemovigilance systems to assess transfusion safety.

Study Design and Methods:

A nationwide administrative database was analyzed to characterize transfusion-related hospitalizations and TAEs by trends, patient/hospital characteristics, and outcomes. Transfusions and TAEs were identified using medical codes and charges. Generalized estimating equations (GEE) modeled transfusion trends, while logistic regression assessed transfusion-associated circulatory overload (TACO) and transfusion-related acute lung injury (TRALI) risk factors.

Results:

During 2016–2020, 8.4% of hospitalizations involved transfusions, with red blood cell (RBC) transfusion being most common (5.2%). In 2020, compared to 2016, hospitalizations with RBC transfusion decreased by 2% (Rate Ratio (RR) 0.98; 95% CI: 0.97–0.99), while plasma transfusion hospitalizations increased by 13% (RR 1.13; 95% CI: 1.08–1.19). TAEs occurred in 0.35% of hospitalizations (3.5/1000 transfusion hospitalizations). Among the TAEs included in the study, TACO, febrile nonhemolytic reactions, and TRALI were most common. In 27% of cases, the specific TAE was unidentified. TAEs were associated with higher inpatient mortality and longer lengths of stay. Variability in TAE rates was observed by patient and hospital characteristics. Risk factors for TACO included age >64, female sex, teaching hospitals, rural location, and Northeast region. TRALI risk was higher in teaching hospitals and those with >200 beds.

Conclusion:

Administrative data provide insights into transfusion practices and associated morbidity and mortality but have limitations. Linking administrative, electronic health record, and blood bank data may enhance TAE identification.

1 |. INTRODUCTION

Blood transfusions are a common procedure in US acute care hospitals.¹ Although widely considered safe, transfusions can cause adverse reactions ranging from mild to life-threatening and, in some cases, fatality. Ensuring the safety of transfusion procedures and adequacy of the blood supply requires monitoring blood utilization and transfusion-related adverse events (TAEs). In the United States, the National Healthcare Safety Network (NHSN) Hemovigilance Module, operated by the Centers for Disease Control and Prevention (CDC), and the National Blood Collection and Utilization Survey (NBCUS), operated by the CDC and the Department of Health and Human Services Office of the Assistant Secretary of Health, collect national data on blood utilization and TAEs.^2,3 Although these systems provide critical insights into blood transfusion safety at a national level, they offer limited information on hospitalizations involving transfusion and TAEs, including patient and hospital characteristics and the impact of adverse events on patient outcomes such as length of stay and all-cause mortality. Additionally, the voluntary nature of the NHSN Hemovigilance system has resulted in low hospital engagement, with only 360 hospitals enrolled, primarily from Massachusetts, where participation is mandatory. This limited participation may result in incomplete representation of TAEs nationwide. The increasing availability of large electronic healthcare databases presents an opportunity to address these limitations and evaluate their potential to complement current hemovigilance efforts. By leveraging these resources, we can better understand the risks associated with blood transfusions, supporting efforts to improve transfusion practices and outcomes.

This study utilized PINC AI^™, a real-world healthcare database representing acute care hospitals nationwide, to identify and characterize hospitalizations involving blood transfusions by blood product type, as well as by patient and hospital characteristics. As a secondary objective, we identified TAEs using clinical diagnostic codes and described the frequency of these events across census divisions and according to patient and hospital characteristics. Lastly, we investigated factors associated with the two leading causes of transfusion-associated fatality: transfusion-associated circulatory overload (TACO) and transfusion-related acute lung injury (TRALI).⁴

2 |. MATERIALS AND METHODS

Hospital discharge and inpatient billing records from over 900 hospitals participating in the PINC AI^™5 healthcare database (Premier hospitals) from 2016 to 2020 were evaluated. The PINC AI^™ database comprises administrative, clinical, and financial data submissions from participating US hospitals representing diverse geographical and practice settings. Hospitalizations with blood transfusion were identified using the International Classification of Diseases, 10th Revision, Procedure Coding System (ICD-10-PCS)⁶ and Current Procedural Terminology (CPT) (Supplementary Tables S1 and S2). To ensure complete identification of blood transfusion procedures, we also reviewed hospital billing records for charge descriptions that clearly identified a transfusion procedure.

For each reporting year, the proportion of hospitalizations with transfusion was calculated by dividing the total number of hospitalizations with transfusion by the total number of hospitalizations among all transfusing hospitals. A transfusing hospital was defined as a hospital with at least one transfusion during the reporting year.

Annual trends in the proportions of hospitalizations with transfusion were examined using three multivariable generalized estimating equation (GEE) models⁷ for each blood product: red blood cells (RBC), plasma, or platelets. The models accounted for autocorrelation in transfusion proportions within each facility and specified a negative binomial distribution of outcome, natural log of monthly discharges as an offset for the proportions, and exchangeable working correlation structure. The models were adjusted for seasonality, categorical annual trend, and facility characteristics (urban vs. rural location, bed size [five categories], teaching status, and census region).

For hospitalizations involving a transfusion, we identified the occurrence of TAEs using International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) codes that specifically indicate an adverse event related to transfusion (Supplementary Table S3).⁸ These ICD-10-CM codes were categorized into 10 groupings: transfusion associated circulatory overload (TACO), febrile nonhemolytic transfusion reaction (FNHTR), transfusion-related acute lung injury (TRALI), hemolytic, transfusion-related anaphylaxis (Anaphylactic), posttransfusion purpura (PTP), transfusion-transmitted infection (TTI), air embolism following transfusion, ABO incompatibility, and adverse event unspecified (Unspecified). As overlap among diagnostic codes existed between acute and delayed hemolytic reactions, both were grouped into a single category. The Unspecified category included ICD-10-CM diagnostic codes that did not specify the type of reaction and were included to fully describe the burden of adverse events related to transfusion.

The proportion of hospitalizations with a TAE was calculated by dividing the total number of hospitalizations with adverse events by the total number of hospitalizations with transfusion. The TAE rates per 1000 hospitalizations with transfusion were also calculated and described by patient and hospital characteristics. The hospitalizations involving transfusion and TAE as well as TAE rates were described by outcomes such as inpatient all-cause mortality and length of stay.

To examine factors associated with the development of TACO and TRALI among all hospitalizations involving transfusion, we fitted two individual-level logistic regression models, one for the outcome of TRALI and one for the outcome of TACO. Each model included all hospitalizations involving transfusion, the outcome of the TRALI or TACO (Yes/No), and the following parameters: year of discharge (categorical, with 2016 as a referent year); patient age (five categories), sex (male, female), and race (White, Black, Asian, Other, Unknown); and facility characteristics (urban vs. rural location, bed size [five categories], teaching status, and census region [four categories]). In addition, we described the transfusion hospitalizations’ primary diagnoses and procedures, using Healthcare Cost and Utilization Project (HCUP) Clinical Classification Software (CCS) categories,^9,10 which were part of the PINC AI^™ database (For results refer to Supplementary Table S4).

This activity was reviewed by CDC and conducted consistent with applicable federal law and CDC policy (See e.g., 45 CFR part 46; 21 CFR part 56; 42 USC §241 (d), 5 USC §552a, 44 USC §3501 et seq.). All data were analyzed using SAS version 9.4 software (SAS Institute, Cary, NC).

3 |. RESULTS

During the study period, blood transfusion procedures occurred in 8.4% of all hospitalizations in the Premier hospitals. Among these, RBC transfusions were the most common (5.2%), followed by plasma, platelet, and whole blood transfusions (0.95%, 0.78%, and 0.02%, respectively) (Table 1). The overall proportion of hospitalizations with a blood transfusion procedure was stable during 2016–2019 (around 8.2%), but then increased to 9.5% in 2020. In the multivariable analysis of temporal trends by specific blood products, we observed a small but statistically significant increase (1%) in the proportion of hospitalizations with RBC transfusions in 2018 and 2019 compared to 2016 (Figure 1). However, in 2020, this proportion decreased by 2% compared to 2016 (rate ratio [RR] 0.98, p < .001). In contrast, the proportion of hospitalizations involving plasma transfusion decreased by 6% in 2019 but subsequently increased by 13% in 2020, relative to 2016. No consistent trend was observed for hospitalizations with platelet transfusion. For full model results see Supplementary Table S5.

TABLE 1.

Hospitalizations with blood transfusion and transfusion-related adverse events occurring among US acute care hospitals participating in the PINC AI^™ Healthcare Database, 2016–2020.

	2016	2017	2018	2019	2020	Total
No. of facilitiesa	811	850	814	847	789	999
No. of hospitalizationsb	8,353,641	8,753,275	8,365,360	8,387,478	7,118,336	40,978,090
No. (%)c of hospitalizations with blood product transfusion	681,580 (8.16)	718,731 (8.21)	693,724 (8.29)	684,136 (8.16)	672,527 (9.45)	3,450,698 (8.42)
Red Blood Cells	426,793 (5.11)	438,029 (5.00)	425,184 (5.08)	433,585 (5.17)	406,783 (5.71)	2,130,374 (5.20)
Plasma	80,881 (0.97)	79,700 (0.91)	74,477 (0.89)	71,619 (0.85)	84,318 (1.18)	390,955 (0.95)
Platelets	64,386 (0.77)	66,601 (0.76)	64,159 (0.77)	65,183 (0.78)	60,327 (0.85)	320,656 (0.78)
Whole Blood	2132 (0.03)	1825 (0.02)	1708 (0.02)	1653 (0.02)	1545 (0.02)	8863 (0.02)
Blood product unknown	200,783 (2.40)	228,766 (2.61)	220,705 (2.64)	205,351 (2.45)	209,054 (2.94)	1,064,659 (2.60)
No. (%) of transfusion hospitalizations with transfusion-related adverse event	2192 (0.32)	2396 (0.33)	2446 (0.35)	2689 (0.39)	2422 (0.36)	12,145 (0.35)
TACO	694 (0.10)	785 (0.11)	839 (0.12)	920 (0.13)	761 (0.11)	3999 (0.12)
FNHTR	393 (0.06)	457 (0.06)	500 (0.07)	559 (0.08)	526 (0.08)	2435 (0.07)
TRALI	448 (0.07)	472 (0.07)	470 (0.07)	451 (0.07)	396 (0.06)	2237 (0.06)
Hemolytic	66 (0.01)	76 (0.01)	65 (0.01)	80 (0.01)	80 (0.01)	367 (0.01)
Anaphylactic	53 (0.01)	49 (0.01)	62 (0.01)	44 (0.01)	40 (0.01)	248 (0.01)
Posttransfusion Purpura	27 (0.00)	23 (0.00)	26 (0.00)	32 (0.00)	20 (0.00)	128 (0.00)
Transfusion Transmitted Infection	9 (0.00)	10 (0.00)	12 (0.00)	13 (0.00)	11 (0.00)	55 (0.00)
Air embolism	6 (0.00)	4 (0.00)	1 (0.00)	11 (0.00)	2 (0.00)	24 (0.00)
ABO incompatibility	3 (0.00)	4 (0.00)	3 (0.00)	3 (0.00)	2 (0.00)	15 (0.00)
Unspecified	588 (0.09)	623 (0.09)	599 (0.09)	726 (0.11)	739 (0.11)	3275 (0.09)

Abbreviations: TACO, transfusion-associated circulatory overload; FNHTR, febrile nonhemolytic transfusion reaction; TRALI, transfusion-related acute lung injury.

^aFacilities included in the study are those that had at least one blood transfusion during the reported year.

^bAll hospitalizations in the included facilities.

^cPercentage calculated as: number of hospitalizations with specific blood product transfusion/total number of hospitalizations with transfusion * 100.

FIGURE 1.

Adjusted^a annual trends in the proportions of hospitalizations with red blood cell (Model I), plasma (Model II), and platelet (Model III) transfusions in premier hospitals, 2016–2020. CI, confidence interval. ^aEach model adjusted for facility characteristics (number of beds [five categories], teaching status, rural versus urban location, and census region) and autocorrelation of facility-level rates within hospitals through the GEE modeling approach. For the full model results, see Supplement Table S5.

A transfusion-related adverse event was identified in 12,145 (0.35%) of all hospitalizations involving transfusion. Among these events, TACO, FNHTR, and TRALI were the most documented, with rates of 1.16, 0.71, and 0.65 per 1000 transfusion hospitalizations, respectively (Table 2). Posttransfusion purpura and transfusion transmitted infections were rare, occurring at rates of 0.04 and 0.02 per 1000 transfusion hospitalizations, respectively. In 3275 (27%) hospitalizations with a TAE, the specific event could not be identified (Table 1).

TABLE 2.

Characteristics of hospitalizations with blood transfusion and rates of transfusion-related adverse events (per 1000 hospitalizations with transfusion) in US acute care hospitals (N = 999), 2016–2020.

	All hospitalizations (%) (N = 40,978,090)	Hospitalizations with transfusion (%) (N = 3,450,698)	Hospitalizations with transfusion adverse event (%) (N = 12,145)	Any adverse event (N = 12,145)	TACO (N = 3999)	FNHTR (N = 2435)	TRALI (N = 2237)	Hemolytic (N = 367)	Anaphylactic (N = 248)	PIP (N = 128)	TTI (N = 55)	Unspecified (N = 3275)
Hospitalization outcome
Mean length of stay, days (SD)	4.5 (6.7)	9.6 (13.5)	12.4 (15.5)	12.4 (15.5)	11.7 (13.4)	10.5 (13.4)	16.5 (17.2)	13.3 (17.5)	10.3 (10. 9)	17.8 (40.2)	16.2 (16.9)	11.9 (16.1)
Died during hospitalization or discharged to hospice	3.9%	11.4%	15.5%	15.5%	16.1%	7.9%	32.3%	10.6%	10.5%	12.5%	10.9%	11.7%
Transfusion-related Adverse event rate per 1000 hospitalizations with transfusion
Patient characteristics
Overall				3.52	1.16	0.71	0.65	0.11	0.07	0.04	0.02	0.95
Age group (years)
0–17	13.8%	2.8%	3.0%	3.73	0.45	0.97	0.43	0.18	0.19	0.03	0.01	1.62
18–54	32.4%	23.9%	26.6%	3.93	0.73	1.00	0.78	0.16	0.11	0.05	0.02	1.25
55–64	15.0%	18.3%	17.2%	3.30	0.96	0.65	0.73	0.11	0.06	0.04	0.02	0.92
65–74	17.0%	23.7%	23.6%	3.50	1.22	0.69	0.70	0.10	0.06	0.04	0.01	0.87
75+	22.0%	31.4%	29.7%	3.33	1.61	0.50	0.48	0.06	0.05	0.02	0.02	0.74
Sex
Female	56.8%	53.3%	55.1%	3.64	1.24	0.73	0.64	0.13	0.07	0.04	0.01	0.98
Male	43.2%	46.7%	44.9%	3.38	1.07	0.68	0.65	0.08	0.08	0.04	0.02	0.91
Unknown	0.0%	0.0%	0.0%	0	0	0	0	0	0	0	0	0.00
Race
White	70.8%	69.2%	69.5%	3.54	1.23	0.67	0.66	0.09	0.07	0.03	0.02	0.93
Black	14.4%	17.6%	16.8%	3.36	0.94	0.78	0.57	0.19	0.07	0.04	0.01	0.95
Asian	2.4%	2.7%	2.7%	3.59	1.31	0.70	0.48	0.03	0.09	0.03	0.01	1.06
Other	9.1%	8.1%	8.4%	3.64	1.00	0.82	0.68	0.09	0.09	0.07	0.01	1.06
Unknown	3.3%	2.5%	2.6%	3.75	1.12	0.78	0.86	0.07	0.07	0.07	0.01	0.97
Ethnicity
Hispanic	9.3%	8.8%	9.9%	3.97	0.98	0.89	0.75	0.07	0.09	0.04	0.02	1.38
Non-Hispanic	71.1%	75.9%	74.3%	3.45	1.15	0.68	0.63	0.11	0.07	0.03	0.01	0.92
Unknown	19.6%	15.4%	15.8%	3.63	1.33	0.72	0.67	0.11	0.06	0.06	0.02	0.82
Admission type
Emergency	54.3%	66.0%	69.7%	3.71	1.30	0.76	0.64	0.12	0.06	0.04	0.02	0.97
Elective	19.4%	15.5%	11.1%	2.52	0.61	0.55	0.52	0.05	0.09	0.03	0.01	0.76
Urgent	13.7%	14.9%	16.7%	3.96	1.22	0.74	0.80	0.11	0.10	0.03	0.02	1.11
Trauma Center	0.9%	1.8%	1.6%	3.06	0.77	0.40	0.98	0.18	0.08	0.02	0.00	0.77
Newborn	10.6%	0.9%	0.1%	0.47	0.13	0.00	0.13	0.07	0.00	0.03	0.03	0.10
Unknown	1.1%	1.0%	0.9%	2.99	0.71	0.34	0.68	0.06	0.00	0.03	0.00	1.31
Payor
Medicare	43.4%	60.5%	57.8%	3.36	1.37	0.59	0.58	0.09	0.05	0.03	0.02	0.80
Medicaid	21.8%	14.6%	16.1%	3.89	0.81	1.00	0.75	0.16	0.09	0.06	0.01	1.22
Managed Care	20.7%	14.3%	15.2%	3.74	0.84	0.86	0.75	0.12	0.13	0.05	0.02	1.15
Commercial	7.2%	5.9%	5.4%	3.20	0.88	0.80	0.87	0.16	0.09	0.05	0.01	0.99
Charity or Uninsured	6.9%	4.8%	5.6%	4.10	0.80	0.71	0.66	0.10	0.10	0.03	0.02	1.41
Hospital characteristics
Number of beds
500+	35.1%	39.6%	44.4%	3.95	1.25	0.82	0.77	0.13	0.08	0.04	0.02	1.08
400–499	12.1%	11.8%	11.7%	3.49	1.12	0.65	0.72	0.11	0.07	0.04	0.01	0.92
300–399	17.0%	17.2%	15.7%	3.21	1.12	0.62	0.59	0.08	0.08	0.03	0.01	0.83
200–299	16.7%	15.4%	13.9%	3.18	1.01	0.60	0.60	0.09	0.06	0.03	0.01	0.93
100–199	13.8%	11.6%	10.7%	3.22	1.17	0.63	0.47	0.10	0.07	0.04	0.02	0.86
0–99	5.3%	4.4%	3.7%	2.93	1.08	0.76	0.27	0.10	0.03	0.05	0.01	0.66
Urban/rural local	:ion
Urban	88.1%	88.2%	88.4%	3.53	1.13	0.72	0.66	0.11	0.07	0.04	0.02	0.96
Rural	11.9%	11.8%	11.6%	3.44	1.39	0.58	0.57	0.11	0.07	0.03	0.02	0.84
Teaching status
Yes	49.6%	53.7%	61.4%	4.03	1.31	0.85	0.74	0.13	0.09	0.04	0.02	1.07
No	50.4%	46.3%	38.6%	2.93	0.98	0.54	0.54	0.08	0.05	0.03	0.01	0.81
Census division
South Atlantic (DC, DE, FL, GA, MD, NC, SC, VA, WV)	27.3%	30.5%	29.9%	3.44	1.04	0.62	0.67	0.11	0.07	0.03	0.02	1.07
Middle Atlantic (NJ, NY, PA)	14.7%	14.5%	18.9%	4.59	1.71	1.25	0.59	0.15	0.09	0.06	0.03	0.92
East North Central (IL, IN, MI, OH, WI)	16.1%	14.0%	13.6%	3.43	1.01	0.79	0.65	0.10	0.04	0.03	0.02	0.96
West South Central (AR, LA, OK, TX)	11.4%	13.8%	12.0%	3.07	1.02	0.49	0.67	0.09	0.08	0.04	0.01	0.83
Pacific (AK, CA, HI, OR, WA)	10.0%	8.8%	8.3%	3.33	1.10	0.57	0.72	0.08	0.06	0.03	0.01	0.93
East South Central (AL, KY, MS, TN)	7.3%	8.7%	5.8%	2.38	0.77	0.41	0.46	0.07	0.08	0.04	0.01	0.61
West North Central (IA, KS, MN, MO, ND, NE, SD)	5.6%	4.9%	5.0%	3.63	1.43	0.63	0.63	0.14	0.10	0.03	0.02	0.78
New England (CT, MA, ME, NH, RI, VT)	2.4%	2.4%	3.2%	4.73	2.10	1.02	0.44	0.14	0.10	0.05	0.01	1.01
Mountain (AZ, CO, ID, MT, NM, NV, UT, WY)	5.3%	2.5%	3.2%	4.50	1.04	0.61	1.13	0.06	0.07	0.03	0.00	1.82

Abbreviations: ACO, transfusion-associated circulatory overload; FNHTR, febrile nonhemolytic transfusion reaction; PTP, posttransfusion Purpura; TRALI, transfusion-related acute lung injury; TTI, transfusion-transmitted infection.

The transfusion-related adverse event rates varied by patient age, event type, and race/ethnicity (Table 2). The highest TAE rate occurred in patients aged 18–54 years (3.93 per 1000 transfusion hospitalizations), followed closely by pediatric patients (3.73). This elevated rate in pediatric patients was primarily driven by a higher incidence of FNHTR, hemolytic reactions, anaphylactic reactions, and unspecified reactions. Conversely, TACO was more prevalent among older patients, with rates of 1.22 and 1.61 per 1000 transfusion hospitalizations in those aged 65–74 and ≥75 years, respectively. TRALI exhibited a different pattern, being most common among middle-aged adults, with rates of 0.78, 0.73, and 0.70 per 1000 transfusion hospitalizations in the 18–54, 55–64, and 65–74 age groups, respectively. The youngest (0–17 years) and oldest (>75 years) age groups had the lowest TRALI rates. Sex disparities were observed, with higher rates of TACO and FNHTR in females (1.24 and 0.73, respectively) compared to males (1.07 and 0.68). Racial/ethnic differences were evident, with White and Asian patients experiencing the highest TACO rates (1.23 and 1.31, respectively), while Black patients had the highest rates of FNHTR and hemolytic reactions (0.78 and 0.19, respectively). Hispanic patients had the highest rates of FNHTR and TRALI (0.89 and 0.75, respectively).

Emergency and urgent admissions had the highest rates of TACO (1.30 and 1.22, respectively) and FNHTR (0.76 and 0.74, respectively). Transfusion hospitalizations at trauma centers showed the highest rate of TRALI (0.98). Among insurance groups, Medicare patients had the highest TACO rate (1.37), while Medicaid patients had the highest FNHTR rate (1.00). Patients classified as charity or uninsured had the highest overall rate of TAEs (4.10). This elevated rate was primarily driven by a higher incidence of unspecified TAEs (1.41) in this group.

Adverse event rates also varied by hospital characteristics. The largest hospitals (500+ beds) had the highest overall adverse event rate (3.95 per 1000 transfusion hospitalizations), as well as the highest rates for specific events, including TACO, FNHTR, TRALI, and hemolytic reactions. In contrast, the smallest hospitals (<100 beds) had the lowest overall event rate (2.93); however, their FNHTR rate was the second highest, following the largest hospitals. Rural hospitals had a higher rate of TACO compared to urban hospitals (1.39 vs. 1.13). Conversely, urban hospitals had higher rates of FNHTR and TRALI (0.72 and 0.66, respectively) than rural hospitals (0.58 and 0.57). Teaching hospitals had the highest overall adverse event rate (4.03 vs. 2.93), as well as higher rates across all types of TAEs compared to nonteaching hospitals. Lastly, regional variability was observed, with hospitals in New England recording the highest overall adverse event rate (4.73) and the highest TACO rate (2.10).

Hospitalizations with TAE were associated with longer lengths of stay (12.4 days) compared to all hospitalizations with transfusion (9.6 days) (Table 2). The longest LOS was observed in hospitalizations with PTP, TRALI, and TTI at 17.8, 16.5, and 16.2 days, respectively. Inpatient mortality was also higher in hospitalizations with a TAE (15.5%) compared to the mortality rate (11.4%) in all hospitalizations with transfusion. The highest mortality was observed in hospitalizations with TRALI (32.3%).

In the individual-level multivariable logistic regression analysis of factors associated with TACO (Figure 2), the odds of TACO increased by 16% in 2018 and 30% in 2019 compared to 2016 (odds ratios [ORs]: 1.16, p = .027; and 1.30, p < .001, respectively). A steady increase in the odds of TACO was observed across age groups, with ORs ranging from 0.45 for the youngest group (ages 0–17) to 1.62 in those aged 75 and older, relative to the 55–64 age group. Additionally, Black patients had 14% lower odds of developing TACO compared to White patients (OR 0.86; p < .001). Higher odds of TACO were significantly associated with teaching hospitals, location in the Northeast census region, and rural setting, with ORs of 1.29, 1.38, and 1.35, respectively.

FIGURE 2.

Logistic regression analysis of factors associated with transfusion associated circulatory overload (TACO) among hospitalized patients with blood transfusion, 2016–2020.

In a similar analysis of TRALI (Figure 3), both the youngest (0–17) and the oldest (≥75) age groups had significantly lower odds of developing TRALI compared to those aged 55–64 (ORs: 0.55 and 0.66, respectively). Asian and Black patients also had significantly lower odds of TRALI compared to White patients, with ORs of 0.59 and 0.79, respectively. Teaching hospitals and those with more than 200 beds were associated with higher odds of TRALI. Lastly, hospitals located in the Midwest, Northeast, or South census regions had lower odds of TRALI compared to hospitals in the West.

FIGURE 3.

Logistic regression analysis of factors associated with transfusion-related acute lung injury (TRALI) among hospitalized patients with blood transfusion, 2016–2020.

4 |. DISCUSSION

Blood transfusions are a cornerstone of modern healthcare, with 11% of all hospital stays in the United States in 2010 involving this life-saving procedure.¹¹ Although blood transfusions are generally considered safe, they are not without risk. Continuous national surveillance of transfusion practices and transfusion-related adverse events is essential to improving patient care and identifying trends that may place patient safety at risk. Existing hemovigilance systems in the United States collect valuable information on transfusion practices and adverse events through surveys and both voluntary and mandatory reporting mechanisms. However, these systems currently provide limited granular detail specific to patient and healthcare facility characteristics. To address this gap, our study leveraged a large, nationwide administrative healthcare database to analyze hospitalizations involving blood transfusions. Our primary motivation for this study was to evaluate the feasibility of using administrative data to complement traditional hemovigilance systems to uncover new insights into transfusion safety and utilization.

Overall, our findings support the feasibility of using administrative data to evaluate blood transfusion practices and safety in the United States, though certain limitations remain. By analyzing medical procedure codes and hospital charges, we identified 3.5 million hospitalizations involving transfusion out of a total of 40.9 million hospitalizations across 999 transfusing facilities. This extensive dataset enabled us to uncover critical trends and identify rare TAEs. We found that red blood cell transfusion, the most common type, was involved in 5.2% of hospitalizations during 2016–2020, aligning with the 4% reported for 2015–2018 in the National Inpatient Sample, the largest all-payer inpatient database.¹² Additionally, when compared with the published estimate of hospitalizations with transfusion in 2010 (11%), our estimate from 2016 to 2020 (8.4%) reflects a 24% decline. This reduction aligns with the widespread adoption of patient blood management programs, advancements in surgical techniques, and improved management of clinical conditions that previously required transfusion.^13,14

Our characterization of hospitalizations involving transfusions revealed that transfusions were more frequent among older adults, males, Black, and non-Hispanic patients. They were also more common in emergent, urgent, or trauma center admissions, patients with Medicare coverage, and in larger (≥500 beds) or teaching hospitals. Furthermore, we observed a 16% increase in the overall proportion of hospitalizations involving transfusion in 2020 compared to 2019, driven largely by a 13% rise in plasma transfusions. This increase aligns with the heightened use of COVID-19 convalescent plasma during the early pandemic period.^15,16 Interestingly, when we excluded patients with a COVID-19 diagnosis, plasma transfusion trends showed a decline, emphasizing the unique impact of the pandemic on transfusion practices (Supplementary Figures S1 and S2). Despite the surge in plasma transfusions, our individual-level risk factor analysis for TACO and TRALI showed no elevated risks for these severe adverse events in 2020 compared to 2016 (Figures 2 and 3). These findings suggest that while transfusion practices and overall healthcare utilization shifted during the pandemic, the underlying risk factors for severe adverse events remained stable.

Our analysis of TAE incidence relied on ICD-10 diagnostic codes and showed that TAEs remained relatively rare, being observed in 0.35% of all hospitalizations involving transfusions. Previous studies using alternative methods, such as clinical reports and symptom-based assessments, have reported TAE incidence rates ranging from 0.2% to 3%.^17–20 These variations reflect differences in detection methods, populations studied, and the types of adverse events included. For instance, a prospective six-hospital study, which relied on clinical reports and other data, estimated the incidence of transfusion reactions at 0.67%.²⁰ This aligns with our finding of 0.35%, especially since our analysis excluded mild allergic reactions, which represent the majority of TAEs.^20,21 When considering all TAEs, pediatric patients in our study demonstrated a higher overall rate (3.73 per 1000 hospitalizations with transfusion), compared to older adults (3.5 and 3.3 per 1000 in 65–74 and >75 age groups, respectively). Pediatric patients also had the highest rate of hemolytic and anaphylactic reactions and a high rate of FNHTR, consistent with previously reported analyses.^20,22 These findings highlight the necessity of vigilant monitoring and targeted preventive strategies to mitigate the risks of these reactions in the pediatric population.

Our evaluation of the incidence of TRALI showed that it was coded in 0.07% of all hospitalizations involving transfusion, similar to previous studies.^23,24 However, the incidence of TACO was 0.12%, which is significantly lower than the 1%–10% reported previously.^23,25–27 Data from the NHSN HM and other international systems also demonstrate that TACO is more common than TRALI.^28–30 The likely under-recognition of TACO in ICD-10 codes can be attributed to several factors, one of which is the overlap of TACO symptoms (e.g., shortness of breath, cough, pulmonary edema) with those of other medical conditions, such as heart failure, acute respiratory distress syndrome, or fluid overload from non-transfusion causes. TACO may be underrecognized as a distinct transfusion-related complication, leading to underdocumentation or miscoding. Given TACO’s role as a leading cause of transfusion-associated morbidity and mortality, accurate identification and reporting are essential for patient safety and national policy guidance.

Several other important findings emerged regarding hospital and patient characteristics and outcomes. Specifically, we observed higher rates of TAEs in larger hospitals and those with academic affiliations compared to smaller and nonacademic facilities. Several factors may explain this variability: (1) greater awareness of transfusion reactions and the presence of dedicated transfusion medicine specialists in teaching hospitals, (2) targeted training on diagnosing, managing, and reporting adverse reactions in facilities with formal hemovigilance programs, and (3) the more complex patient populations in academic centers, who may be at higher risk for TAEs. Our finding of the highest overall TAE rate among uninsured/charity patients, primarily driven by a high rate of unspecified events, warrants further investigation. This elevated rate may suggest deficiencies in documentation and coding quality in facilities serving underserved populations, or greater disease complexity and severity among uninsured patients.

In terms of patient outcomes, we found that patients with TAEs had a longer average LOS (12.4 days) compared to those without (9.6 days), consistent with previous research.^31,32 Prolonged stays increase resource utilization and the risk of healthcare-associated complications, such as infections.³³ Additionally, inpatient mortality was significantly higher among patients with TAE (15.5%) than those without (11.4%), with the highest mortality observed in TRALI cases (32.3%). To fully understand the impact of TAEs on patient outcomes, it is essential to include data on the timing of transfusion procedures and TAEs. While large administrative data offer significant potential for analyzing the impact of adverse events across facilities and regions, their utility will be limited without detailed information on timing.

Another limitation of administrative data is its reliance on ICD-10 codes in capturing TAEs. Previous research highlighted the misalignment between adverse events recorded in electronic health records and those reported and confirmed by blood bank investigation.^34,35 These inconsistencies may be due to hospitals’ variations in adverse event documentation practices. Inpatient coders typically rely on clinical provider notes from admission, daily progress, and discharge summaries when assigning ICD-10 codes. However, provider notes may not consistently document transfusion reaction diagnoses in the medical record even if these reactions are reported to and investigated by transfusion services. Generally, mild adverse reactions, such as simple allergic or febrile nonhemolytic events, which resolve quickly and often appear clinically insignificant, may be under-recorded in provider diagnoses. Even when documented and reported to hemovigilance systems, these events may not be routinely reviewed during the hospitalization coding process. Consequently, healthcare databases may be skewed toward documenting more severe reactions. To address this, efforts should focus on improving clinician awareness, standardizing documentation practices, integrating transfusion service data, and leveraging electronic health record technology to flag potential TAEs based on clinical parameters, laboratory findings, or transfusion-related entries.

In our study, 27% of hospitalizations with TAEs were coded as “Unspecified transfusion reaction,” potentially contributing to both underestimation and overestimation of specific reactions. Additionally, the structure of administrative data poses limitations. Unlike hemovigilance systems that express reaction rates relative to total units transfused, our dataset lacked information on number of products transfused per hospitalization, preventing trend analysis by transfusion volume. Furthermore, the lack of chronological data linking adverse events to transfusion procedures limited our ability to attribute these events to specific types of transfused products. Lastly, our analysis of inpatient service charges enabled identification of an additional 31% of hospitalizations with transfusion that were not captured by ICD-10 and CPT procedure codes alone (Supplementary Table S6). We validated our method of identifying transfusion procedures using charges and billing records by comparing TAE rates between hospitalizations identified through codes and those identified through charge review, finding similar rates across both methods. However, due to the lack of detailed blood product information in charge descriptions, our estimates of hospitalizations by blood product type are likely lower than the actual numbers.

In conclusion, this study highlights the value of utilizing large administrative and billing records to supplement hemovigilance systems, thereby enhancing our understanding of transfusion safety in the United States. These data provide detailed patient- and facility-level information, making them particularly useful for identifying and characterizing transfusion practices and their impact on patient outcomes. However, reliance on ICD-10 codes limits the ability to reliably and accurately identify transfusion-related adverse events. Improving data completeness and accuracy will require linking administrative datasets with blood bank records and clinical documentation from electronic health records. Lastly, the observed variation in rate estimates for reactions such as TACO underscores the ongoing need for hospitals to engage in formal hemovigilance to accurately capture the risks of transfusion and better inform patients and providers. CDC continues to collaborate with partners, including the Association for the Advancement of Blood and Biotherapies, to streamline and improve hemovigilance reporting in the United States.

Supplementary Material

Additional supporting information can be found online in the Supporting Information section at the end of this article.

Abbreviations:

CCS

clinical classification software

CDC

Centers for Disease Control and Prevention

COVID-19

Coronavirus Disease 2019

CPT

Current Procedural Terminology

FNHTR

febrile non-hemolytic transfusion reaction

GEE

generalized estimating equation

HCUP

healthcare cost and utilization project

ICD-10-CM

International Classification of Diseases, 10th Revision, Clinical Modification

ICD-10-PCS

International Classification of Diseases, 10th Revision, Procedure Coding System

LOS

length of stay

NBCUS

National Blood Collection and Utilization Survey

NHSN

National Healthcare Safety Network

PTP

post-transfusion purpura

RBC

red blood cells

TACO

transfusion-associated circulatory overload

TAE

transfusion-related adverse event

TRALI

transfusion-related acute lung injury

TTI

transfusion-transmitted infection