Predictors of blood loss during total hip ... - Wiley Online Library

Predictors of Blood Loss During Total Hip Replacement Surgery Jodi M. Grosflam, Elizabeth A. Wright, Paul D. Cleary, and Jeffrey N. Katz Objective. To determine total blood loss and transfusion requirements during primary total hip replacement (THR) surgery; and, to identify predictors of increased blood loss. Methods. This was a prospective cohort study of 295 consecutive patients undergoing primary THR at a 700-bed teaching hospital in Boston, Massachusetts. The main outcome measures were the total blood loss, which was defined (in units) as the initial hematocrit minus final hematocrit divided by 3 plus the number of units transfused, and the number of autologous and homologous units of blood transfused. Results. The mean total blood loss was 3.6 units, and the mean total units transfused was2.1. Univariate predictors of greater total blood loss (at P < 0.05) included being male, receiving general anesthesia, and higher American Society of Anesthesiologists (ASA) class. In multiple linear regression models which controlled for demographic and clinical variables (age, race, marital status, education, smoking history, obesity, underlying disease, payor status, cemented prosthesis, comorbidity, and season of operation), being Jodi M. Grosflam. MD, and Jeffrey N. Katz, MD, MS, Department of Rheumatology/Immunology, and Robert B. Brigham MAMDC. and Elizabeth A. Wright, PhD, Robert B. Brigham MAMDC. Brigham and Women’s Hospital, and Paul D. Cleary, PhD, Departments of Health Care Policy and Social Medicine, Harvard Medical School, Boston, MA. (Dr. Grosflam is now with Internal Medicine Associates, Fort Myers, Florida). Supported in part by NIH grants AR-36308, AR-39921, AR-07530, the Hartford Foundation, and by an Arthritis Foundation Arthritis Investigator Award (to JNK). Address correspondence to Jeffrey N. Katz, MD, MS, Department of Rheumatology/Immunology, Brigham and Women’s Hospital, 75 Francis Street, Boston, MA 02115. Submitted for publication January 27. 1995; accepted in revised form March 28. 1995. 0 1995 by the American College of Rheumatology. 0893-7524/95/$5.00

male, receiving general anesthesia, and having an ASA score of 3 predicted greater total blood loss. Conclusion. Preoperative characteristics can help determine which patients should donate either more or Jess than the customary 2 units of blood. Key words: Joint replacement; Blood transfusion; Arthritis; Blood loss during surgery.

INTRODUCTION Blood loss in total hip replacement (THRJ surgery is substantial, and requires transfusion of homologous or, if available, autologous blood. Patients receiving homologous transfusions remain at risk for infection by both hepatitis B and C viruses, which are characterized by a window period during which serologies and other screening methods may be negative. Homologous transfusion also exposes patients to the risk of cytomegalovirus seroconversion and a small risk of contracting the human immunodeficiency virus (HIV). Homologous transfusion may also cause febrile transfusion reactions. Autologous transfusion has risks (l), but these are fewer than for homologous transfusion. Using autologous, rather than hornohgous, transfusions would therefore diminish the rates of transfusion-related morbidity and mortality. Orthopedic procedures such as primary THR surgery are particularly well suited for autologous blood donation because they are elective and involve generally healthy patients. While much has been written about the development and implementation of autologous donation protocols (2-4), there is little or no information to guide recommendations on the number of units of blood that should be donated prior to surgery (5), nor has a systematic way of predicting transfusion needs been developed. Information on autologous transfusion needs would 167

168 Grosflam et aJ

be particularly valuable to internists, family practitioners, and rheumatologists with patients who are candidates for THR surgery. These practitioners provide an increasing proportion of the preoperative education to patients considering THR, and a discussion of blood loss, transfusion needs, and strategies for decreasing transfusion-associated morbidity are central to the preoperative assessment of the benefits and risks of proceeding with surgery. If possible, guidelines for autologous donation should reflect the amount of blood an individual patient is likely to need. The aim of the present study was to identify preoperative clinical and demographic features that predict the amount of blood loss that occurs during primary total hip replacement surgery. We were especially interested in identifying clinical correlates of total blood loss that could be used to develop a customized autodonation policy.

PATIENTS AND METHODS Study sample. Consecutive patients admitted to the Brigham and Women’s Hospital between June 18,1989 and September 30, 1990 for primary elective total hip arthroplasty were studied. Patients were excluded if they were less than 16 years old, had metastatic cancer, were undergoing chemotherapy, had HIV infection, or had received an organ transplant. In addition, patients who had another surgical procedure (including a second, contralateral THR) within 6 weeks of the index primary THR were excluded. Patients were also excluded if a preoperative hematocrit value was unavailable for the 4-week interval preceding the operation. Neither the patients nor the physicians were aware of this analysis of blood loss at the time of the study. Medical record data. A medical record reviewer who was blinded to the hypotheses of this analysis abstracted information about sociodemographic factors, the surgical procedure, and the occurrence of inhospital complications. Questionnaire data. Patients completed questionnaires 6 months postoperatively about their current functional status, height, weight, and education (6). Blood loss data. Initial and final hematocrit and transfusion data were verified by either chart or laboratory computer review. Total blood loss (TBL) was defined as: [(initial hematocrit - final hematocrit)/3] + units of blood transfused. The final hematrocrit was the last recorded hematocrit value that was obtained

Vol. 8, No. 3, September 1995

before discharge from the hospital and at least 3 days after the operation. A change of 3 hematocrit points was considered as equal to 1 unit of blood (7). Predictors of blood loss. The potential predictors of total blood loss considered in this analysis included race, age, sex, marital status, arthritis diagnosis, insurance status, smoking history, comorbid conditions, use of nonsteroidal antiinflammatory drugs (NSAIDs], type of anesthesia, surgeon, American Society of Anesthesiologists (ASA)class (8),use of bone grafting, use of cement on the femoral component, season when operation performed, body mass index, and education level. Comorbid conditions were documented with the instrument described by Greenfield et a1 (9), which records the presence and severity of comorbid conditions in 14 categories. Comorbid conditions (with the exception of hypertension) were infrequent. Therefore, a dichotomous comorbidity variable was created to reflect the presence or absence of any comorbid condition other than hypertension, which was examined separately. Insurance status was categorized as commercial (including Blue Cross/Blue Shield, health maintenance organization], Medicare, or other. Anesthesiologists, blinded to the study hypotheses, assigned an ASA score to each patient. In the analysis, each ASA class was evaluated separately as a predictor, and ASA class I11 was compared to classes I and I1 combined. No patients in the study had an ASA class higher than 111. Use of cement on the femoral component was recorded as present or absent. Use of bone grafting to augment the acetabular region, including either autologous or homologous bone fragments, was recorded as present or absent. Several other variables potentially related to total blood loss, but for which limited information was available, were evaluated during a second chart review. For a randomly generated sample of 30 patients, the highest prothrombin time and lowest hematocrit during hospitalization were obtained. In these same patients, the use of NSAIDs, aspirin, or coumadin either prior to or during the index hospitalization was also noted. If the medication had been used previously, the date it was stopped prior to hospitalization was recorded. Because Brigham and Women’s is a teaching hospital and the operative procedure involved resident physicians who may be less expert early in their residencies, we recorded whether the surgery took place in the first quarter of residency (July through September) or later.

Arthritis Care and Research

NSAID use. To address the issue of NSAID and aspirin use before or during the operative procedure, we sent a questionnaire to the orthopedic surgeons involved in this study. Each surgeon was asked about his (all were male) usual recommendations to patients about NSAID or aspirin usage in the weeks before elective THR. Questions about NSAIDs and aspirin were asked separately. Analyses. Univariate correlates of total blood loss were evaluated with the Student’s t-test for the dichotomous variables, analysis of variance (ANOVA) for categorical predictors, and the Pearson correlation coefficient for continuous variables. Backward and forward stepwise linear regression models were used to identify independent predictors of total blood loss, significant to P 5 0.15. Multiple linear regression was performed with only these predictors to yield a parsimonious model. Two outliers exerting a strong effect on this model were excluded for determination of the final predictive model. Independent variables with P I 0.05 in this final regression model were incorporated into the risk score. The risk score was created by summing, for each patient, the number of these predictors present. The results from this summative risk score, in which each risk factor contributed 1 point, were compared to those using the actual coefficients from the predictive model. The number of patients in each category with at least 4 units of total blood loss was assessed with the Mantel-Haenszel chi-square test. An ANOVA, with the Student-Newman-Keuls test, was performed to test for the differences in mean blood loss between pairs of risk score categories. All analyses were performed with SAS [lo). Linear regressions were performed with the REG and GLM procedures. Exploratory analyses were performed to verify normality of the outcome variable.

RESULTS Recruitment. Three hundred sixty patients underwent primary total hip replacement during the study period. Nineteen patients were excluded because the procedure was considered nonelective. An additional 19 patients were excluded because administrative demographic data and medical records were not available. Thirty patients were excluded because the first preoperative hematocrit was not available or another surgical procedure was performed within 6 weeks of the operation of interest. After these exclusions, 292 patients remained in this analysis. The age, sex, race, and arthritis diagnosis of these 292 patients were not significantly different from the 68 patients excluded.

Predictors of Blood Loss During THR 169

Table 1. Baseline characteristics of the 292 patients studied Number (%) Female White

184 (63) 282 (97)

Arthritis diagnosis Osteoarthritis Rheumatoid arthritis Other

208 (71) 50 (17) 34 (12)

Age, mean k SD years

64 k 13

Baseline characteristics of the study cohort. The mean age of the population was 64 years (Table 1). Sixty-three percent were female. Ninety-seven percent were white and 71% had osteoarthritis of the hip as the reason for the procedure. Seventeen percent of the population had rheumatoid arthritis. The remaining group had a variety of underlying conditions including juvenile rheumatoid arthritis, spondylarthritis, and avascular necrosis. Characteristics of the operative procedure. During the operative procedure, 118 patients (40%) received either general anesthesia or a combination of general and local anesthesia (Table 2). Bone grafting was utilized in 30% of the surgeries and cementing of the femoral stem was performed in 69% of the procedures. Total blood loss. Only 35 patients (12%) received no transfusion, while the remaining 257 patients received from 1 to 15 units. In total, 623 units were transfused, of which 269 units (43%)were homologous. One hundred and thirteen (44%) of the transfused patients received at least 1 homologous unit. The mean change in hematocrit from the preoperative measurement to the final measurement was 4.4% (SD 4.4) [Table 3). The average number of units of blood transfused was 2.1 units (SD 1.81. This included units transfused intraoperatively and postoperatively, Table 2. Characteristics of the operative procedure in 292 patients who underwent total hip replacement Number (TO) Anesthesia General Local and general Local

48 (16) 70 (24) 174 (60)

Bone graft Cemented prosthesis

87 (30) 200 (69)

170 Grosflam et a1


Table 5. Number of patients in each risk score category and mean units of total blood loss

Table 3. Elements of total blood loss Mean

-

SD

4.4 & 4.4 2.1 k 1.8 3.6 2 2.2

Hematocrit change (70) Total units transfused Total blood loss (units)* * Calculated as: (initial hematocrit

&

final hematocrit)/3

+ no. of units trans-

fused.

Risk score

NO.[TO)of patients

0 1

71 (24) 157 (54)

2 3

59 (20) 5 (21

-~

Mean blood loss (units)*

70of patients with 2 4 units transfusedt

2.8 3.5 4.5 5.9

18 27

41 60

~~~~

but did not include any blood collected by a cell saver during the operation or recycled from the patient during the operative period. Autologous units were collected prior to patient admission to the hospital for the elective procedure. The mean total blood loss, calculated with the formula shown in Table 3, was 3.6 units (SD 2.2). Univariate predictors of total blood loss. Greater blood loss was associated with being male (P = 0.0041, use of general or combination anesthesia (P = 0.02), ASA class I11 versus class I and I1 ( P = 0.031, surgeon (P = 0.05], and presence of a comorbidity aside from hypertension (P = 0.07). Total blood loss was not associated with age, race, martial status, insurance status, education level, body mass index, arthritis diagnosis, hypertension, smoking history, bone grafting, epidural anesthesia, cemented femur, or season of operation (all P L 0.19). Multiple linear regression models of total blood loss. The final regression model contained 4 predictors of total blood loss: being male, receiving general or combined anesthesia, ASA class 111, and surgeon. All predictors were significant at P I0.05. The model explained 13% of the total variance in total blood loss. For construction of the risk score, we utilized only generalizable and measurable predictors, namely sex, anesthesia type, and ASA class. The removal of the surgeon variable from the model (which had P = 0.04) changed the coefficients of the remaining predictors very slightly and decreased the explained variation to 8%. This model is presented in Table 4. To test the Table 4. Independent predictors of total blood loss in multiple linear regression P Predictor

coefficient

SE

P

Sex Anesthesia type ASA class*

0.89 0.59 0.69

0.22 0.22 0.26

0.0001

* ASA

=

American Society of Anesthesiologists.

0.0071

0.0083

* P = 0.0001 for difference between groups, by ANOVA. t P = 0.001 for linear trend, by Mantel-Haenszel chi-square.

robustness of the model without surgeon, this model was run on 8 subsets of the data, each one excluding the patients of 1surgeon. All predictors remained significant, and in each case, the model predicted between 7% and 10% of the variation in total blood loss.

Risk score and total blood loss. The risk score was constructed by assigning 1 point for the presence of each risk factor shown in Table 4, and summing the number of points for each patient. As the risk score increased from 0 to 3, the mean total blood loss increased from 2.8 to 5.9 units (Table 5). Mean total blood loss was significantly greater in higher-risk categories (P = O.OOOl), and the means of the 2 highest risk groups were significantly different from the mean of the lowest risk score category. A higher risk score also was associated with a greater likelihood of receiving 4 or more units of blood (P = 0.001, Mantel-Haenszel chisquare]. An alternative risk score was calculated using the predicted blood loss calculated from the regression equation coefficients. Predicted blood loss was converted to an integer value by rounding up for any decimal value, since this represented a need for an additional unit of blood. The Spearman correlation between the risk score and the total blood loss predicted by the regression coefficients was 0.92, indicating that the approximations inherent in calculation of the risk score did not compromise accuracy of prediction. Table 6 illustrates that the average difference between the amount of blood autodonated and the amount transfused increased from 0.2 units at a risk score of 0, to 4.4 units at a risk score of 3. The percentage of patients in each risk group who donated less than the mean number of units transfused for their group increased with the risk score, from 30% to 100%. Patients in the 2 lower risk categories were significantly more likely to autodonate than were those in the higher categories (P = 0.007).



Table 6. Excess number of units transfused, by risk score 70of risk

No. of Risk score patients 0 1 2 3

71 157 59

5

Units transfused [mean 4 SD)

*

1.9 1.0 2.0 k 1.3 2.6 k 2.9 5.2 4.1

*

group Units donating less than transfused minus units the mean donated transfused (mean _t SD) to group 0.2 k 1.1 0.5 t 1.3 1.4 t 3.1 4.4 i 4.9

30 36 90 100

Surgeon effects, NSAID use, and other miscellaneous predictors of total blood loss. Seven of the 8 orthopedic surgeons who performed the procedures in this study completed a questionnaire on routine instructions regarding aspirin and NSAID use. Each surgeon had a single recommendation that he provided for all patients undergoing THR surgery. Five of 7 surgeons responded that they generally advised their patients to stop NSAID usage 1week prior to surgery, whereas 2 surgeons advised discontinuing the day before. A review of thirty randomly selected charts showed no consistent relationship between the use of NSAIDs and total blood loss. The highest prothrombin time did not correlate with greater total blood loss.

DISCUSSION In this study, total blood loss during primary, elective total hip replacement surgery varied considerably and could be predicted with 3 easily determined variables: sex, ASA class, and anesthesia type. These results are consistent with findings of other investigators. In a study of 212 THRs, blood loss was higher for males and was positively associated with weight and height (11).In another report involving 27 patients who received different types of anesthesia, including epidural and general anesthesia, the group receiving epidural anesthesia had the least blood loss (12). Toy et a1 studied 324 unilateral elective primary or revision THR surgeries (13)and documented total blood loss of 3.2 units in primary hip replacements, similar to our mean of 3.6 units. In another study (111,mean total blood loss was 1,090 ml (about 2 units], significantly less than we found, However, that was a retrospective study and blood loss was calculated from drainage bags and sponges. Our findings may also be different because we studied all THR patients, including those who did not donate autologous blood

(who may be sicker] and because of our more objective method of estimating blood loss. In our study, patients received a mean of 2.1 units of blood during their hospitalization. This is consistent with other reports, in which patients received a mean of 2 units for primary and 2.9 units for revision procedures (13). In an additional study of autologous donation in THR surgery, 50 patients donated an average of 2.5 units, and the mean transfusion need was 4 units (14).

Fifty-six percent of the patients we studied received autologous blood only. In one retrospective analysis of total joint replacement surgery, 74% of the patients enrolled in the autologous transfusion program received only autologous blood (4). In other studies, 4979% of the patients enrolling in an autologous donation program received only autologous blood (2,15,16]. In one study of primary THR, patients predeposited an average of 2.6 units of blood, and 92% of the surgeries in which autologous blood had been predonated were performed without the transfusion of homologous blood. Homologous blood was needed for 15 of 21 cases in which no blood was deposited, but in only 1 of 87 cases in which 3 or more units were predeposited (171. In some of these studies, particularly the latter, less homologous blood was used because patients were able to donate a greater number of units of blood. It could be that patients who need to donate the most blood are the least able to do so. This may explain why patients in lower-risk groups were significantly more likely to autodonate in our studies. However, at all risk scores, many patients donated less blood than they received by transfusion. The risks of excess autodonation are small. They include physiologic consequences and inconvenience for the patient and financial costs to the health care system for processing this blood. Only 46 (16%) of the patients in this study donated excess autologous blood. In our study, patients who donated blood received slightly more transfused blood than did those who did not donate (2.2 versus 1.9 units], but this difference was not statistically significant. Nevertheless, we cannot rule out a bias toward transfusion if blood has been autodonated. Our results are consistent with findings by other investigators that patients do not donate enough autologous blood. Several studies have shown that the reasons for underdonation are unrelated to patients’ overall ability to donate blood. Thomson et a1 reported that the most common reason for not enrolling in the autologous donation program was that the patient lived outside the area (18). Hull et a1 found that surgeons did not request adequate amounts of autologous blood,

172


Grosflam et a1

failed to schedule elective orthopedic procedures far enough in advance to allow patients to give the requested amount of blood, and overtransfused some patients with the autologous blood (19). Studies aIso show that anemia is not a major barrier to autologous donation. Another cause of underdonation of autologous blood is purported to be a lack of awareness on the part of the physician of the patients’ need to autodonate blood (20). Programs to improve physician awareness of an autologous program have been shown to increase enrollment (3,s). Varied ad hoc recommendations have been made regarding the amount of autologous blood which should be donated and ways to increase the autologous donations (16). In one report it was recommended that patients donate 3 units of blood after it was found that a 3-unit donation was associated with no adverse events and that 74% of those who did so received autologous blood only (4). Similarly, in a study of autologous blood donation in 44 patients undergoing orthopedic procedures, it was argued that 10 of 17 patients could have avoided receiving homologous blood if each had stored 4 units (21). Goodnough suggested that orthopedic patients are particularly well suited for autologous donation and recommended that 1 unit of autologous blood can be obtained from donors every 72 hours up to 72 hours before surgery as long as the hematocrit is 233% (22). Limitations to autodonating blood preoperatively relate to limitations in how long the blood can be stored and the erythropoietic response to phlebotomy. We believe that the number of units of blood which should be donated should be individualized and can be estimated from the risk score we have created. We found a borderline effect (P = 0.04) of individual surgeon upon total blood loss, but we omitted the surgeon effect from our risk score because it is not generalized to other settings. Nevertheless, this finding suggests that surgeon-specific aspects of the process of care may influence blood loss. This phenomenon should be investigated in other institutional settings and prompt discussion among surgeons about technical and other factors that might explain these intersurgeon variations. Our study has several potential limitations. First, there is no optimal way to measure blood loss precisely. We assume that blood loss is a linear function of the change in hematocrit and amount transfused (7). This may be imprecise, but probably is more accurate than surgeons’ estimates based upon drainage bags and sponges. We acknowledge that hemoglobin may have been a more accurate measure of blood loss. Another limitation to our study is that we used hematocrit change instead of initial or final hematocrit

value to determine blood loss. Identical changes in hematocrit may or may not be deemed indicators for transfusion, depending on the physician and on the initial hematocrit. If the initial hematocrit is low, it is more likely that a transfusion will be given. In 97% of a sample of 211 of our patients, the initial hematocrit was greater than 3070, indicating that the initial hematocrit is seldom low in this patient population. Also, we were unable to study explicitly the use of NSAIDs (including aspirin) during and before the orthopedic procedure. Aspirin and NSAIDs prolong bleeding time and could be important predictors of blood loss. However, future studies could better address this issue. The sample consisted primarily of Caucasian patients at a tertiary hospital. Our findings were derived in a single patient population and should be validated in other settings (23). Finally, the models explained just 10% of the total variance in total blood loss. Nevertheless, as shown in Table 5, our approach successfully stratifies patients according to blood loss and transfusion need. We conclude that blood loss in elective total hip replacements is variable and often exceeds 3 units. A subset of patients at risk for increased transfusion requirements can be identified with our risk score and counseled to autodonate more blood. Future studies should validate the risk score we have proposed and investigate methods to increase autologous donation. We thank Drs. Matthew Liang, Elizabeth Karlson, and David Gutstein for reviewing a draft of this manuscript, and Dr. Barbara McNeil for her role in the design and execution of this study and review of a draft of this manuscript.

REFERENCES 1. Simpson MB, Georgopoulos G, Orsini E, Eilert RE: Au-

2.

3.

4.

5.

6.

tologous transfusions for orthopedic procedures at a children’s hospital. J Bone Joint Surg 74A:652-658, 1992 Axelrod FB, Pepkowitz SH, Goldfinger D: Establishment of a schedule of optimal preoperative collection of autologous blood. Transfusion 29:677-680, 1989 Rebulla P, Giovanetti AM, Petrini G, Mezzetti M, Sirchia G: Autologous blood predeposit for elective surgery: a program for better use and conservation of blood. Surgery 463-466,1985 MacFarlane BJ, Marx L, Anquist K, Pineo G, Chenger J, Cassol E: Analysis of a protocol for an autologous blood transfusion program for total joint replacement surgery. Can J Surg 31:126-129, 1988 Moore SB, Swenke PK, Foss ML, Rand JA, Cabanela ME, Kavanagh B, Taswell H F Simplified enrollment for autologous transfusion: automatic referral of presurgical patients for assessment for autologous blood collections. Mayo Clin Proc 67:323-327, 1992 Cleary PD, Reilly DT, Greenfield S, et al: Using patient


reports to assess health related quality of life after total hip replacement. Qua1 Life Res 23-11, 1993 7. Menitove JE: Blood Transfusion. In, Cecil Textbook of Medicine. Nineteenth edition. Edited by JB Wyngaarden, LH Smith Jr, JC Bennett. Philadelphia, WB Saunders, 1992 8. Tinker JH, Roberts SL: Anesthesia risk. In, Anesthesia. Second edition. Edited by RD Miller. New York, Churchill Livingstone, 1986 9. Greenfield S, Apolone G, McNeil BJ, Cleary PD: The importance of co-existent disease in the occurrence of postoperative complications and one-year recovery in patients undergoing total hip replacement: comorbidity and outcomes after hip replacement. Med Care 31:141154, 1993 10. SAS User’s Guide: Basics, Version 5. Cary, NC, SAS Institute, 1985 11. Flordal PA, Neander G: Blood loss in total hip replacement. Arch Orthop Trauma Surg 111:34-38, 1991 12. Keith I: Anaesthesia and blood loss in total hip replacement. Anaesthesia 32:444-450, 1977 13. Toy PTCY, Kaplan EB, McVay PA, Lee SJ, Strauss RG. Stehling LD: Blood loss and replacenent in total hip arthroplasty: a multicenter study. Transfusion 32:63-67, 1992 14. Mallory TH, Kennedy M: The use of banked autologous blood in total hip replacement surgery. Clin Orthop Relat Res 117254-257, 1976


15. Woolson ST, Marsh JS, Tanner JB: Transfusion of pre-

viously deposited autologous blood for patients undergoing hip replacement surgery. J Bone Joint Surg 69:325328,1987 16. Eckardt JJ, Gossett TC, Amstutz HC: Autologous transfusion and total hip arthroplasty. Clin Orthop Relat Res 132:39-45,197a 17. Woolson ST, Watt JM: Use of autologous blood in total hip replacement. J Bone Joint Surg 73:76-80, 1991 18. Thomson JD, Callaghan JJ, Savory CG, Stanton RP, Pierce RN: Prior deposition of autologous blood in elective orthopaedic surgery. J Bone Joint Surg 69:320-324, 1987 19. Hull AL, Wasman J, Goodnough LT: Effects of a CME program on physicians’ transfusion practices. Academic Med 54:681-685,1989 20. Goodnough LT, Wasman J, Corlucci K, Chernosky A: Limitations to donating adequate autologous blood prior to elective orthopedic surgery. Arch Surg 124:494-496, 1989 21. Wasman J, Goodnough L T Autologous blood donation for elective surgery. JAMA 258:3135-3137.1987 22. Goodnough LT, Shuck JM: Risks, options, and informed consent for blood transfusion in elective surgery. Am J Surg 159:602-609, 1990 23. Wasson JH, Sox HC, Neff RK, Goldman J: Clinical prediction rules: applications and methodological standards. N Engl J Med 313:793-799, 1985