A Cost Analysis of Autologous and Alloge:neic Transhsions in Hip#Replacement Surgery

Neil Blumberg, MD, Scott A. Kirkley, MD, Joanna M. Heal, MRCP, Rochester, New Y&c

PURPOSE: To analyze the cost consequences of autologous versus allogeneic transfusions.

METHODS: Costs were determined when allo- geneic transfusions were given in addition to, or instead of, autologous transfusions. Hospital charges were used to estimate costs for hip-re- placement surgery. The main outcome measure was estimated incremental hospital costs per unit transfused.

RESULTS: Among donors of autologous blood, mean total charges were $7,200 greater for re- cipients of both autologous and allogeneic transfusions than for recipients of autologous transfusion only (P = 0.0001). Each allogeneic transfusion was associated with additional costs of $1,480. In a second cohort of patients receiving identical amounts of either allogeneic or autologous blood (mean = 2.3 units), total hospital charges were a mean of $4,800 greater (P = 0.0001) for allogeneic recipients. The per- unit excess costs associated with each unit of allogeneic blood in this cohort were $1,043.

CONCLUSIONS: Allogeneic transfusions are asso- ciated with incremental hospital costs of about $1,000 to $1,500 per unit transfused when com- pared with costs for similar patients receiving no transfusions or 1 to 5 units of autologous blood. Am J Surg. 1996;171:324-330.

D uring the last decade, the use of predeposit autolo- gous transfusions in elective surgery has grown dra- matically, driven largely by the fear of viral diseases

transmitted via allogeneic blood. Indeed, legislative action in California and New Jersey has mandated that physicians dis- cuss with patients the transfusion option of autologous do- nation prior to surgery. As the use of this modality has in-

From the Transfusion Medicine (NB, SAK), and Hematology (JMH) Units, Departments of Pathology and Laboratory Medicine, and Medicine (JMH), University of Rochester Medical Center and American Red Cross Blood Services (JMH), Rochester Region, Rochester, New York.

Requests for reprints should be addressed to Neil Blumberg, MD, University of Rochester Medical Center, 601 Elmwood Avenue, Box 608, Rochester, New York 14642.

Supported in part by a research grant from Ortho Biotech, Inc., Raritan, New Jersey and the R.W. Johnson Pharmaceutical Research Institute.

Manuscript submitted October 11, 1994 and accepted in revised form March 21, 1995.

creased, so has concern that patients who have little likeli- hood of needing a transfusion are subjected to the unnec- essary inconvenience, small but undetermined risk, and ex- pense of autodonation.‘-’ To assure that the likelihood of requiring allogeneic blood is minimized, patients must donate more units on average than they are likely to require, and thus 50% or more of donated units are discarded.

One theoretic analysis has suggested that au tologous trans- fusion may not be cost effective, given the much greater safety of the blood supply since the advent of additional tests for infectious agents4 However, accumulated data have shown that allogeneic transfusion in the surgical setting is associated with dramatically higher rates of acute morbid- ity, primarily due to postoperative bacterial infection.5-9 These complications have been hypothesized to be due at least in part to transfusion-induced immun~omodulation.‘O This possibility is strengthened by recent data from ran- domized trials demonstrating reductions in postoperative in- fections in patients receiving either autologous blood only or leukocyte-depleted allogeneic transfusiorts.l’~‘Z A theo- retic analysis based upon these new findings suggests that autologous transfusions may be cost effective.13 We con- ducted a study of the costs associated with receipt of autol- ogous transfusion only, allogeneic transfusion only, or both autologous and allogeneic transfusions, in the setting of hip- replacement surgery in our institution to determine whether the additional costs and logistic difficulty of autologous transfusions were offset by the additional mcrbidity and at- tendant costs of allogeneic transfusions.

METHODS We chose to study hip-replacement surgery because this is

the most common indication for use of autologous blood in our hospital. Also, we had performed a case-control study suggesting that autologous transfusion recipients experience decreased morbidity in this setting compared with allogeneic transfusion recipients receiving identical amounts of blood.’ Our estimate of the incremental cost of collection of each unit of autologous blood ($25 to $50) includes the incre- mental nursing time (0.5 hour at $32/h salary + benefits), clerical time (0.25 hour at $16.00/h), medical coverage (0.05 hour at $100/h), and institutional overhead (0% to 100%) necessary for these donations, as compared with the cost of obtaining a single allogeneic unit from a hea.lthy volunteer donor. This cost estimate does not include increased expense due to increased outdating/wastage, nor the savings (esti- mate: $15 to $20 per unit) realized at some hospitals by not testing for infectious-diseases blood intended only for autol- ogous use. We intentionally have provided particularly gen- erous estimates of the additional resources necessary for au-

124 THE AMERICAN JOURNAL OF SURGERY@ VOLUME 171 MARCH 1996

1 COST ANALYSIS OF TRANSFUSlONlBLUMBERG ET AL

TABLE I Relationship Between Allogeneic Transfusion

and Selected Variables During 1992 in 140 Patients Undergoing Hip-Replacement Surgery

Mean Hospital Mean Length Variable Charges ($) d Stay (d) Allogeneic transfusion 24,798 (n = 19) 13.9

Primary hip replacement No allogeneic transfusion 19,275 (n = 99) 9.4

Primary hip replacement

Allogeneic transfusion 29,411 (n = 11) 17.6 Hip-replacement revision

No allogeneic transfusion 19,475 (n = 11) 9.9 Hip-replacement revision

Allogeneic transfusion 27,132 (n = 12) 15.3 Surgeon number 1

No allogeneic transfusion 19,947 (n = 40) 9.2 Surgeon number 1

Allogeneic transfusion 35,679 (n = 5) 24.8 Surgeon number 2

No allogeneic transfusion 20,414 (n = 16) 8.8 Surgeon number 2

Allogeneic transfusion 22,198 (n = 11) 11.9 Surgeon number 3

No allogeneic transfusion 18,802 (n = 41) 10.2 Surgeon number 3

Allogeneic transfusion 23,272 (n = 2) 9.5 Other surgeons

No allogeneic transfusion 17,473 (n = 13) 8.5 Other surgeons

All allogeneic transfusion recipients also donated and received auto/o- gous blood.

20,000

18,000

16,006

ii z 14,006

% 12,000

10,OW .c s 8,000

$ 4 6,006

4,000

n ~ am

0 Autologous

n Allogeneic

!3..i Total Hospital Total Room Total Ancillary

Charges Charges

Figure 1. The mean (A 95% confidence interval) total hospita charges (P = 0.0001 for the difference between autologous only and allogeneic only), room charges (P = 0.056), and total an- cillary charges (P= 0.0001) for hip-replacement surgical patients receiving of 2 or 3 units of autologous blood only (n = 33), or 2 or 3 units of allogeneic blood only (n = 49).

SW

5 700

= 2

600

z 500

z 400 .c

B $ 200

6 100

0

1 D Autologous m Alloaeneic

hort donating and receiving 2 to 3 units of autologous blood compared with a cohort receiving 2 to 3 units ‘of allogeneic blood. This analysis avoided any difference in number of units used for transfusion as a source of bias. As previously reported,7 the patients in these two cohorts were well matched for clinical and surgical variables such as age, sex, duration of surgery, estimated blood loss during surgery, and preexisting morbidities. We investigated the hospital charges associated with transfusion in these two populations of pa- tients undergoing hip replacement during 1986 to 1988 who received either 2 to 3 units of autologous blood only (n = 33) or 2 to 3 units of allogeneic blood only (n = 49). These two case-control cohorts were restricted to primary hip re- placements in patients without preexisting risk factors for morbidity such as malignancy, autoimmune disease, or chronic infections, and thus represent a relatively homoge- neous group. These 1986 to 1988 cohorts have been reported in detail previously with a finding of reduced lmorbidity as- sociated with autologous transfusions.7 Figures 1 and 2 are based upon data from these two cohorts.

Two additional cohorts consisted of patients who success- fully donated autologous blood and then either did or did not subsequently receive allogeneic blood transfusions. All patients donating autologous blood for hip-replacement surgery during 1992 were included in this analysis. This sec- ond analysis avoided the potential bias of differences be- tween those who donate autologously and those who do not or cannot. These two cohorts of patients enabled us to ex-

Figure 2. The mean (* 95% confidence interval) pharmacy (P = 0.019 for the difference between autologous only and allo- geneic only), microbiology (P = 0.043), hematology (P= O.OOOl), chemistry (P= 0.062), anatomic pathology (P = OH), blood gas laboratories (P = O.OOOl), blood bank (P = O.OOOl), and respi- ratory therapy (P = 0.98) charges for hip-replacement surgical patients receiving 2 or 3 units of autologous blood only (n = 33) or 2 or 3 units of allogeneic blood only (n = 49).

tologous blood donors and these estimates may substantially overstate the incremental costs.

Given that a randomized trial of autologous versus allo- geneic transfusions has never been performed in the United States due to ethical and medicolegal considerations, we an- alyzed cohort data to estimate the hospital costs associated with autologous versus allogeneic transfusions. To minimize bias and confounding due to possible differences between pa- tients who donate autologous blood and those who do not, and between those donating autologous blood who require additional allogeneic blood and those who do not, we ana- lyzed representative samples of all four populations.

One group consisted of a case-control comparison of a co-

THE AMERICAN JOURNAL OF SURGERY@ VOLUME 171 MARCH 1996 325

:OST ANALYSIS OF TRANSFUSIOtVBLUMBERG ET AL 1

TABLE II Relatlonship Between Hospital Charges and

Length of Stay According to Selected Variables Durlng 1992 in 140 Patients Undergoing Hip-Replacement Surgery

Independent Variables

Dependent Variables Hospital Charges Length of Stay I3 P Value I3 P Value

Revision versus -2032 0.15 -1.99 0.21 primary surgery

Surgeon -1125 0.01 -0.375 0.44

Number of -1210 0.04 -1.068 0.11 autologous units donated

Number of 715 0.11 0.625 0.21

autologous units transfused

Number of 1672 0.0001 1.265 0.0038

allogeneic units transfused

Table values are results of mu/tip/e linear regression. B is the individual vari- able’s coefficient (slope) in the linear regression equation For the continu- ous (transfusion) variables, positive values for I3 mean that as the indepen- dent variable increases, so does the dependent vanable (eg, length of stay}

45,000

40,000

g 35,000 = g 30,000

8 25,000

.f 20,000 I p 15,000

g 10,000

5,000

0

I n None Ml 02

rotal Charges Room Total Ancillary

Figure 3. The mean total hospital charges (* 95% confidence interval), room charges, and total ancillary charges for hip-re- placement surgical patients are shown as a function of number of allogeneic units of blood received. The group receiving zero allogeneic transfusions includes both recipients of autologous blood only (n = 84) and those receiving no transfusions of any sort (n = 26). For total charges (overall P = O.OOOl), the differ- ences of zero versus the other three doses are all P co.05 by post-hoc tests, as is the comparison of 1 unit versus >3 and 2 versus 23. Similar statistical significance (P ~0.0001) was ob- tained for room charges and total ancillary charges.

amine whether increasing amounts of autologous or allo- geneic blood were associated with increased costs. I f a dose- response relationship exists for allogeneic but not autologous transfusions, this would suggest the possibility that some of the increased costs were causally related to allogeneic trans- fusions. Tables I and II and Figures 3 to 6 are based upon these 1992 patients.

Hospital charges were used as a surrogate measure of cost in our initial analysis. Patient names and identifying num-

r 22,500 IT-

20,000

5 17,sOa .

= 2 15,000

; 12,500 v- .E 10,000 .

I P 7,500

6 5,000

2,500

0

I Three or More

large8 Room Total Ancillary

igure 4. The mean total hospital charges (k 95% confidence interval), room charges, and total ancillary charges for hip-re- placement surgical patients are shown as a function of number of autologous units of blood received (n = 26 for 10 units; 34 for 1; 29 for 2; and 21 for 23). The group receiving any allogeneic transfusions (n = 30) is excluded. For total charges (P = 0.48), room charges (P = 0.41), and total ancillary charges (P = 0.79) there was no statistically significant difference between the groups receiving various doses of autologous transfusions.

bers were retrieved from the daily operating room schedule during 1992, as well as from our original database from 1986 to 1988. Data on patient charges and transfusions were re- trieved from hospital and laboratory computer records, re- spectively. For the 1992 cohort, all patients having elective total hip replacements and having autologous blood dona- tions performed were included in the study, regardless of medical history (n = 140). Both first-time an’d repeat joint replacements were included. We also recorded the surgeon’s name; type of operation; the number of autologous units or- dered, donated, and transfused; as well as any allogeneic transfusions administered. Because three studies in our in- stitution,7*9 including two in joint-replacement surgery,7J had demonstrated a relationship between allogeneic blood transfusion and postoperative morbidity that was indepen- dent of clinical variables such as clinical comorbidities, du- ration of surgery, degree of preoperative anemia, surgical blood loss, surgeon, and type of procedure, we did not col- lect detailed clinical information on the causes of increased costs in the 1992 cohort. These data are ptiblished for the 1986-to-1988 cohort7

All dollar amounts represent hospital charges that bear a variable relationship to actual costs. Because there is no pre- cise method for converting charges from 1 year to compara- ble data for another year, we normalized the lcharge data for the 1986-to-1988 patients to 1992 dollars by using the daily room rate for each period to correct for inflation. The data in Figures 1 and 2 are normalized. The data from 1992 pa- tients (Tables I and II and Figures 3 to 6) represent unal- tered hospital charges.

We derived a correction factor for all charges based upon the increase in room rate from the 1986-to-1988 period to 1992. In addition, because the room-rate correction might reflect local changes in charges, but not national changes, we separately calculated a compounded, overall inflation fac-

326 THE AMERICAN JOURNAL OF SURGERY@ VOLUME 171 MARCH 1996

1 COST ANALYSIS OF TRANSFUSlONfBLllMBERG ET AL

1,400 m None

$ 1,200 liil = 02 I I I 0” 1,006

s! I 600

.E 600

0 P 400 m

6 200

0

Figure 5. The mean departmental charges (* 95% confidence interval) for hip-replacement surgical patients are shown as a function of number of allogeneic units of blood received. The group receiving zero allogeneic transfusions includes both re- cipients of autologous blood only (n = 84) and no transfusions of any sort (n = 26). The differences are significant (P ~0.05) for all categories except microbiology, anatomic pathology, and respiratory therapy.

tor based upon data from the Bureau of Labor Statistics, Washington, DC, for the medical-care component of the Consumer Price Index to convert charge data to 1992 dol- lars. The results of the analysis with this second approach were similar to that of the daily room-rate adjustment, and the latter, local figures are presented.

We realized that these approaches could only be an esti- mate of the changes in charge for the thousands of items in- cluded in hospital service and supply bills. We also calcu- lated the dollar value per charge for each category to determine whether the inflation adjustment alone was likely to have led to differences in total charges for individual pa- tients. When calculated on a dollar-per-charge basis, differ- ences between the two groups of patients were neither sta- tistically significant nor substantial, suggesting that, however imprecise the inflation adjustment, it alone could not ac- count for any differences seen.

Finally, we employed hospital accounting information on the relationship between hospital charges and actual costs in 1992 to estimate the actual changes in variable costs for patients receiving allogeneic transfusions in addition to, or instead of, autologous transfusions.

Statistical Analysis All data analysis employed a Macintosh IIcx (Apple

Computer, Cupertino, California) and Statview 4.0 software (Abacus Concepts, Berkeley, California). Differences be- tween means were calculated by two-sided Student’s t-tests (parametric) and Mann-Whitney (nonparametric) methods with similar results; the P values from the former are shown. Both parametric (analysis of variance [ANOVA]) and non- parametric (Kruskal-Wallis) analyses were employed for an- alyzing data with three or more groupings. Two-way ANOVA was used to analyze the role of type of transfusion received

600

5 600

5 0 400 z 2 300 .E 8 200

$ 6 100

:igure 6. The mean departmental charges (* 95% confidence interval) for hip-replacement surgical patients are shown as a function of number of autologous units of blood received (n = 26 for zero units; 34 for 1; 29 for 2 and 21 for Z3). The group receiving allogeneic transfusions (n = 30) is excluded. None of the differences are statistically significant for any categories.

versus other variables, such as year of treatment, transfusion dose, surgeon, and revision versus primary hip replacement. Multiple linear regressions were employed to estimate the sig- nificance of dose of transfusion, type of transfusion, surgeon, and number of units donated in predicting ho:jpital charges and length of stay. For univariate analyses, no correction for multiple comparisons was made, primarily because most com- parisons of interest were significant and Bonferroni-type cor- rections would not add additional information <about the dif- ferences observed. The P values for univarrate analyses, multiply performed, and derived from estimated rather than actual dollar data in some instances, should be viewed as mea- sures of the magnitude of the differences between the means shown rather than traditional probabilistic estimates.

The data are reported as means rather than medians, as these are more representative of the actual total costs of car- ing for the patients reported, given that outliers may repre- sent the primary comparisons of interest in the context of costs. In addition, by displaying the mean, the overall total difference in charges can be readily obtained by multiplying the number of patients by the average charges. Data, where outliers are prominent, are readily seen in the graphs, as the 95% confidence limits shown are relatively wide. Because the variability of patients and treatments may have led to undetected differences in intensity, success, or length of treatment, we also calculated all costs and re:source utiliza- tion on a per-day basis, as well as per-hospital admission. Similar results were found as those when total hospital ad- mission charges were analyzed.

RESULTS Costs During 1992 in Autologous Donors A.lso Receiving Small Numbers of Allogeneic Transfusions

Among patients donating autologous blood prior to surgery, patients who received both autologous and allo- geneic transfusions (n = 30) spent an average of 15.2 days

THE AMERICAN JOURNAL OF SURGERY@ VOLUME 171 MARCH 1996 327

in the hospital as compared with 9.4 days for those who re- ceived only autologous transfusions or no transfusions (n = 110) (P = 0.0001). The mean stay increased from 13.2 (n = 9) to 14.3 (n = 13) to 23.2 (n = 4) days for patients re- ceiving 1, 2, or 3 allogeneic transfusions (P = O.OOOl), re- spectively, in addition to autologous blood. The corre- sponding length of stay for those receiving only 1, 2, or 3 autologous transfusions was 9.2 (n = 34), 8.9 (n = 29), and 10.8 (n = 15) days (P = OSO), respectively. There were no significant differences between recipients of autologous blood or no transfusions in length of stay or hospital charges. While even 1 or 2 units of additional allogeneic blood were associated in a dose-dependent manner with longer hospi- tal stays, increasing doses of autologous blood were not as- sociated with longer stays up to a dose of 5 units. When the analysis was restricted to a total transfusion dose of 3 to 5 units, the length of stay in recipients of any allogeneic blood was 14.7 days versus 10.5 days in recipients of only autolo- gous blood (P = 0.12).

Total mean hospital charges were $26,490 for recipients of both autologous and allogeneic transfusions versus $19,295 for recipients of autologous only or no transfusions (P = 0.0001). The total hospital charge, room charge, and total ancillary charge data on the 30 patients who received au- tologous and allogeneic transfusions (Figure 3) and the 84 who received only autologous transfusions (Figure 4) are shown along with data on the 26 patients receiving no trans- fusions. The data are presented according to the number of units transfused of that type of blood. There was a statisti- cally significant dose response for allogeneic, but not for au- tologous, transfusions. Similar data from individual hospital charge categories are shown in Figures 5 and 6.

Table I presents hospital charges and length of stay ac- cording to receipt of allogeneic blood, type of surgery, and surgeon. Two-way ANOVA demonstrated that the effect of allogeneic transfusions was seen independent of surgeon or type of surgery. For length of stay, both allogeneic transfu- sion (P = 0.0001) and surgeon (P = 0.005 1) were significant predictors, but type of hip replacement was not (,P = 0.159). For hospital charges, both surgeon (P = 0.0001) and allo- geneic transfusion (P = 0.0001) were significant predictors, but type of hip replacement was not (I’ = 0.07). The inter- action terms for allogeneic transfusion/surgeon were signifi- cant for both total charges (P = 0.003) and length of stay (P = 0.0008). Neither interaction term was significant for allop geneic transfusion/type of hip replacement (P = 0.10; P = 0.30, respectively).

Indeed, patients undergoing hip appliance revisions with only autologous transfusions had shorter lengths of stay and lesser hospital charges than those undergoing primary hip replacement with 1 to 2 units of allogeneic blood (data not shown). Multiple linear regressions were performed using length of stay, total hospital charges, and total ancillary charges, room charges, and departmental charges as depen- dent outcome variables; and surgeon, type of surgery (revi- sion versus primary), number of autologous units donated preoperatively, number of autologous units transfused, and number of allogeneic units transfused as predictor-indepen, dent variables. Both overall regressions were significant: P = 0.0007 (r = .381) for length of stay and P = 0.0001

COST ANALYSIS OF TRANSFUSION/BLUMBERG ET/

(r = 519) for hospital charges. When only the transfusion variables were included in the regression, the sole signifi- cant predictor was allogeneic transfusion dose (P = 0.0007 for length of stay and P = 0.0001 for total hospital charges). That is, autologous units donated or transfused did not achieve significance at P ~0.05 as predictors of either length of stay or hospital charges when allogeneic transfusion dose was part of the regression. The number of units of allogeneic blood transfused was both the most consistently statistically significant predictor of length of stay and hospital charges, and quantitatively the most important. Typical multiple lin- ear regression data are shown in Table IL

At our hospital, actual variable costs average about 50% of hospital charges. Using this 50% figure, we estimate that an additional $107,925 in actual hospital costs were associ- ated with the 7.3 allogeneic units transfused to 30 patients, as opposed to the $9,650 to $19,300 incremental expense ($25 to $50 per unit) for the 386 autologous units donated by all 140 patients in the study. Thus, on the mean, each allogeneic transfusion was associated with adclitional actual costs of about $1,480, as compared with a maximum of per- haps $50 per autologous unit collected.

Allogeneic Versus Autologous Transfusion in Historical Cohorts

A potential source of bias in the previous analysis is that all recipients of allogeneic blood also received autologous blood, and thus may represent patients with particularly dif- ficult operative courses, and thus higher transfusion require- ments. To deal with this issue, it would be desl.rable to om- pare recipients of autologous blood with patients receiving identical doses of allogeneic blood only. Such a comparison is not currently feasible due to the almost universal use of au- tologous blood for orthopedic surgery in our institution. However, a previously reported case-control cohort of pa- tients undergoing primary hip replacement in 1986 to 1988 was available for analysis.7 These patients were selected to be free of chronic infection, autoimmune or malignant diseases, and had only osteoarthritis as the cause of their joint disease.

The hospital charges for 33 patients receiving 2 or 3 units of autologous blood only and 49 patients reoeiving 2 or 3 units of allogeneic blood only during 1986 tcl 1988 in our hospital are shown in Figures 1 and 2. As previcsusly reported, the rate of infection or suspected infection was 32% in those receiving allogeneic blood versus 3% in those receiving au- tologous blood (P = 0.0029). Wound infections accounted for 6 of the 17 morbid events, which also included 4 (non- wound) cutaneous infections, 4 urinary tract infections, and 3 unexplained persistent fevers treated with extended courses of antibiotics. These two groups of patients were well matched in all clinical variables. There were no significant differences between the autologous and allogeneic blood recipients in variables such as mean age (58 versus 59 years), sex (62% versus 60% male), units transfused (2.3 versus 2.3), duration of surgery (148 versus 143 minutes), blood loss (890 versus 1,000 mL), or days of wound drainage (2.9 versus 3.0). The autologous recipients were significantly more anemic before surgery (mean hematocrit 38%) than those not donating pre- operatively (41%) (P = 0.0008), but mean discharge hema- tocrits were similar (autologous 32%; allogeneic 33%).

328 THE AMERICAN JOURNAL OF SURGERYe VOLUME 171 MARCH 1996

1 COST ANALYSIS OF TRANSFUSION/BLU-1

The recipients of autologous blood had reduced hospital charges compared with recipients of allogeneic blood (a mean difference of $4,800; P = O.OOOl), primarily accounted for by reduction in ancillary service charges (a mean reduction of $3,250; P = 0.0001). Employing an estimate that hospital variable costs average about 50% of charges, the estimated excess cost of each allogeneic transfusion was $4,800, multi- plied by 0.5, divided by 2.3 (the mean number of units trans- fused) to equal $1,043. This is smaller, but comparable to the excess costs associated with each allogeneic unit given to pa- tients who also had received autologous blood ($1,480). Mean length of stay in these two cohorts of patients, who d’ff d 1 h h h h d donated and been trans- 1 ere on y m w et er t ey a fused with autologous blood, was 12.1 days for autologous ver- sus 13.5 days for allogeneic blood recipients (P = 0.043). The allogeneic recipients received a mean of 1.3 more days of an- tibiotic therapy than the autologous recipients (P = 0.06).

COMMENTS Autologous transfusion, insofar as it reduces the use of al-

logeneic blood, reduces the risks of short-term immunologic and long-term infectious complications of transfusion, such as alloimmunization, white cell or plasma-related fever, ur- ticaria, and viral transmission. However, while preoperative donation is apparently quite safe, it is not risk free, and ex- tensive logistic support is necessary for collection of autolo- gous blood. In over a decade of experience with autologous donations, primarily for orthopedic surgery, we have not had any complications that required hospitalization. However, the theoretic and practical questions of added risk and com- plexity have led to concerns about the cost effectiveness of autodonation, and whether the benefits (avoidance of allo- geneic blood) are worth the uncertain risks of preoperative anemia and transient hypovolemia.‘.3 Because the short- term morbidity of postoperative infection associated with al- logeneic transfusions may be more common than the more severe but much rarer complications of human immunode- ficiency virus-l or hepatitis B infections, theoretic analysis of autologous transfusion that only addresses these latter risks may underestimate the cost effectiveness of this procedure. Autologous transfusions have been shown in most,7~8~12,14-16 but not all, studiesI to reduce the short-term morbidity as- sociated with allogeneic transfusions.

Our cost analysis, based upon hospital charges, demon- strates the reduction in resource consumption one would ex- pect from such a reduction in morbidity, including that pre- viously demonstrated’ in one of the two patient cohorts reported here. When allogeneic transfusions are given in- stead of autologous blood, or in addition to it, sibmificant in- creases in hospital charges occur. Inference regarding a causal relationship between allogeneic transfusions and increased costs would be inappropriate if the data reported here were considered in isolation. It is possible that subtle differences in patients who chose to donate autologously, or who re- quire a small number of allogeneic units in addition to au- tologous blood, account for some of the differences in re- source consumption we observed. It is very unlikely that they account for all the difference in our setting.

Given the increase in postoperative infections after allo- geneic blood transfusions reported in most retrospective

studies 5-‘o~34,‘6~1s,‘9 two randomized clinical studies,“,‘* and in an&al models,20*21 some of the difference we report is likely explained by reduced morbidity when allogeneic blood transfusions are avoided. It is clear from these previous stud- ies that allogeneic transfusion dose is not merely acting as a surrogate marker for other variables such as age, duration of surgery, anemia, blood loss, or comorbid illness. Whether our results can be generalized will need to be determined by ad- ditional studies in other settings. Some studies have failed to detect differences in postoperative infection rates between recipients of autologous and allogeneic transfusions,” or have found more modest differences.15

In hip-replacement surgery in our institution, the addi- tional logistic difficulties and expense of each unit of autol- ogous blood (perhaps $50 maximum) appear to be offset by hospital costs that increase approximately $l,C43 to $1,480 per unit of allogeneic blood transfused. The additional costs associated with transfusion of 1 unit of allogeneic blood are more than an order of magnitude greater than the additional costs associated with collecting 1 unit of auto‘logous blood. Thus, unless our estimates are inaccurate by approximately 20- to 30-fold, or autologous blood discard rates approach or exceed 95%, use of autologous transfusion for hip-replace- ment surgery appears to reduce, rather than increase, the cost of patient care in our center.

Caveats include the fact that this is the first comprehen- sive study of short-term allogeneic transfusion costs utiliz- ing hospital data and conceivably may reflect some unique aspect of our setting. These data do not necessarily imply that collection of autologous units for patients who have little likelihood of transfusion would be cost effective. Furthermore, additional risks of autologous donation may exist that have not yet been identified, especia‘ily in patients with high-risk cardiovascular or pulmonary disease. Further research into the clinical efficacy, risks, and cost effective- ness of autologous transfusion is warranted by the findings in this study.

This timely article addresses the cost of different types of trcms- fusions and the reader should see the already published supple- ment to this Journal entitled “Blood management practice guide- lines” fur a broader examination of contemporary review of transfusion practices.

REFERENCES 1. Bove IR. Too much of a good thine? Transfwion. 1993:33:542.

2. Goldfinger D, Haimowitz-M. Is autilogous l&d transfusion worth the cost? Pro. Trunsfusion. 1994;34:75-78.

3. AuBuchon JP, Birkmeyer JD. Is autologous blood transfusion worth the cost? Con. Transfusion. 1994;34:79-83. 4. Birkmeyer JD, Goodnough LT, AuBuchon JP, et al. The cost-ef- fectiveness of preoperative autologous blood donation for total hip and knee replacement. Transfusion. 1993;33:544-551.

5. Tartter PI. Blood transfusion and postoperative infections. Transfusion. 1989;29:456-458. 6. Tartter PI, Driefuss RM, Malon AM, et al. Relationship of postop-

erative septic complications and blood transfusions in patients with Crohn’s disease. Am J Surg. 1988;155:43-47. 7. Murphy P, Heal JM, Blumberg N. Infection or suspected infection after hip replacement surgery with autologous or homologous transfu- sions. Tran.s&.sion. 1991;31:212-217.

8. Triulzi DJ, Vanek K, Ryan DH, Blumberg N. A clinical and im-

THE AMERICAN JOURNAL OF SURGERYm VOLUME 171 MARCH 1996 329

munologic study of blood transfusion and postoperative bacterial in- fection in spinal surgery. Trunsfu.sion. 1992;32:517-524. 9. Murphy PJ, Cannery C, Hicks CL Jr., Blumberg N. Homologous

blood transfusion as a risk factor for postoperative infection after coro- nary artery bypass graft operations. J Thorac CuroYiovosc Surg. 1992; 104:1092-1099. 10. Blumberg N, Heal JM. Effects of transfusion on immune func-

tion-cancer recurrence and infection. Arch Padlo1 Lab Med. 1994; 1 l&371-379. 11. Jensen LS, Andersen AJ, Christiansen PM, et al. Postoperative

infection and natural killer cell function following blood transfusion in patients undergoing elective colorectal surgery. Br j Surg. 1992; 79:513-516.

12. Heiss MM, Mempel W, Jauch K-W, et al. Beneficial effect of au- tologous blood transfusion on infectious complications after colorec- tal cancer surgery. Lancet. 1993;342:1328-1333.

13. Healy JC, Frankforter SA, Graves BK, et al. Pre-operative autol- ogous blood donation in total hip arthroplasry: a cost-effectiveness analysis. Arch Pathol Lab Med. 19941 l&465-470. 14. Mezrow CK, Bergstein 1, Tamer PI. Postoperative infections fol-

lowing autologous and homologous blood transfusions. Trnsfusion. 1992;32:27-30.

15. Femandez MC, Gottlieb M, Menitove JE. Blood transfusion and postoperative infection in orthopedic patients. TrurLsfwion. 1992;32:

318-322. 16. Howard HL, Rushambuza FG, Martlew VJ, et al. Clinical bene- fits of autologous blood transfusion: an objective assessment. Clin Lab Haem&. 1993;15:165-171.

17. Busch ORC, Hop WCJ, Hoynck van Papendrecht MAW, et al. Blood transfusions and prognosis in colorectal cancer. NEJM. 1993;328:1372-1376. 18. Triulzi DJ, Blumbqg N, Heal JM. Association of transfusion with

postoperative bacteria1 infection. Crit Reu Clin Lab Sci. 1990;28:95-107. 19. Blumberg N, Heal JM. Transfusion-associated immunomodula-

tion. In: Anderson KC, Ness PM, eds. Scientific Baris of Transftuion Medicine. Philadelphia: W.B. Saunders; 1994:58&598. 20. Waymack JP, Gallon L, Barcelli U, Alexander JW’. Effect of blood transfusions on macrophage function in a burned animal model. Curr

Surg. 1986;43:305-307. 21. Gianotti L, Pyles T, Alexander JW, et al. Identification of the blood component responsible for increased susceptibility to gut-de- rived infection. Transfusion. 1993;33:45&465.

COST ANALYSIS OF TRANSFUSlON/BLUMBERG ET AL 1

330 THE AMERICAN JOURNAL OF SURGER’P VOLUME 171 MARCH 1996