Risk Factors for Post-Operative Blood Transfusion Following Total Knee Arthroplasty

Abstract

Background:

As the population ages, rate of total knee arthroplasty increases and thus, it is important to maximize efficiency and minimize risk. Identifying patients who are at higher risk for transfusion can help streamline care provided and minimize superfluous, costly hemoglobin monitoring in low risk patients.

Methods:

Adult patients who underwent total knee arthroplasty (TKA) in 2015 were identified in the National Surgical Quality Improvement Project (NSQIP) database. Patients were divided into two cohorts: those who required transfusion post operatively and those who did not. Patient demographics and comorbidities were compared using univariate analysis; and multivariate analysis was used to determine risk factors for short-term complications.

Results:

Of 48,055 TKA patients, 3.0% required transfusion. The patients who required transfusion were older, had higher BMI, higher rates of comorbidities and were more frequently ASA class 3-4 (p<0.005). Univariate analysis revealed that patients who required transfusion had higher rates of any complication (9.19% v. 4.23%, p<0.001). Multivariate regression analysis identified the following as risk factors for transfusion requirement: Black race (adjusted odds ratio [OR] 1.2, 95% confidence interval [CI] 1.01-1.4), COPD (OR 1.6, 95% CI 1.3-2.0), corticosteroids (OR 1.4, 95% CI 1.1-1.8), bleeding disorder (OR 1.4, CI 1.1-1.9), ASA class 4 (OR 2.3, CI 1.5-4.8), operative time >2 hours (OR 1.3, 95% CI 1.2-1.5) and lack of functional independence (OR 1.6, 95% CI 1.1-2.3).

Conclusions:

In a cohort of patients undergoing primary TKA in 2015, history of COPD, black race, operative time, steroid use, bleeding disorder, lack of functional independence and ASA class 3-4 were independent predictors of need for blood transfusion. Additionally, we found that patients who received transfusion demonstrated a significantly higher rate of the following: any complication, pneumonia, urinary tract infection, septic shock, deep vein thrombosis, renal insufficiency, cardiac arrest, myocardial infarction, unplanned readmission, reoperation and mortality. Presence of these risk factors in TKA patients could represent an indication for hemoglobin monitoring post-operatively.

Level of Evidence: IV

Introduction

As the US population ages, the demand for total knee replacement rises; the rate of total knee arthroplasty increased nearly 3-fold between 1990 and 2002.¹ Surgeons continually seek new methods to streamline perioperative care. This process is also encouraged by the transition from a fee-for-service based design to an episode of care model. An attractive initiative in optimizing safety and efficiency is the potential to offer individualized care to patients based on risk factors. In this regard, an important target focuses on the risk of allogenic blood transfusion.² The rate of red blood cell (RBC) transfusion historically has been reported to be up to 35% following total knee arthroplasty.^3-5 With modern perioperative practices, this rate has been significantly diminished.²

Nevertheless, many surgeons continue to monitor hemoglobin and hematocrit postoperatively as a routine practice. Patient care post-operatively can be made more efficient and less costly by adoption of a restrictive hemoglobin monitoring protocol. In such a protocol, only patients with an elevated risk for transfusion postoperatively would need lab testing. Knowing risk factors for blood transfusion after surgery can help identify patients who could benefit from hemoglobin testing post-operatively. Previous studies have shown that age, ASA class, preoperative hemoglobin, initial postoperative hemoglobin, change in pre to postoperative hemoglobin and adherence to strict transfusion triggers are significant predictors of perioperative blood transfusion.^6-10 Few have evaluated independent risk factors for blood transfusion. The goal of this study was to compare patient characteristics in two cohorts—transfused and not transfused—in order to identify risk factors that may help surgeons identify patients who are at risk of requiring peri-operative blood transfusions.

Methods

Data Collection and Patient Selection

All patients who underwent primary total knee replacement in 2015 were identified using the American College of Surgeons—National Surgical Quality Improvement program (ACS-NSQIP) database. Current procedural terminology codes (CPT) codes for primary total knee arthroplasty (27447) were used. A total of 48,055 patients were identified. Patients were divided into two groups: those who received a blood transfusion during their hospital stay and those who did not. Patients were excluded if the surgery was non-elective, non-primary, or if they discharged on post-operative day zero. Records were available for 46,630 patients who did not receive transfusion and 1,425 patients who did receive transfusion.

Variables

Patient characteristics included the following: demographics, pre-operative health variables and comorbidities, pre-operative laboratory values, and operative variables (Table 1). Demographics included age, sex, race (white, black, other) and smoking status. Pre-operative health variables included the following: body mass index (BMI) and recent weight loss (loss of 10% of total body weight in the previous 6 months). The following comorbidities were also included: diabetes mellitus, chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), hemodialysis use, corticosteroid use, bleeding disorder, pre-operative blood transfusion, and pre-operative sepsis. Pre-operative laboratory values that were analyzed included white blood cell count, hematocrit, platelet count, creatinine, serum albumin and international normalized ratio (INR). Operative variables included American Society of Anesthesiologists (ASA) class, length of operation and length of stay.

Table 1.

Demographic Characteristics, Preoperative Comorbidities, Preoperative Laboratory Values, and Operative Variables

	No Transfusion (%) (N=46,630)	Transfusion (%) (N=1425)	p-value
Demographics
Age (yr)*	66.30(9.54)	68.87(10.25)	<0.0001
Female sex (%)	61.15	69.61	<0.0001
Race (%)			<0.0001
White	87.72	84.61
Black	8.86	12.46
Other	3.43	2.93
Preoperative Comorbidities
BMI (kg/m2)*	26.40(10.13)	25.37(9.87)	0.0063
Recent Weight Loss (%)	0.08	0.21	0.1148
Diabetes Mellitus (%)	18.11	22.18	<0.0001
Smoking (%)	8.76	9.40	0.3949
Chronic Obstructive Pulmonary Disease (%)	3.60	5.40	0.0004
Congestive Heart Failure (CHF) (%)	0.29	1.05	<0.0001
Dialysis (%)	0.15	0.63	<0.0001
Steroids (%)	3.42	5.19	0.0003
Bleeding Disorder (%)	1.98	4.00	<0.0001
Preoperative Blood Transfusion (%)	0.01	0.14	0.0167
Open Wound or Infection (%)	0.16	0.21	0.5092
Pre-op Sepsis (%)	0.21	0.35	0.2391
Preoperative Laboratory Values
White Blood-Cell Count (103 cells/μL)*	7.09(2.25)	6.84(2.10)	<0.0001
Hematocrit (%)*	41.21(3.97)	37.33(4.89)	<0.0001
Platelets (per μL)*	244.5(65.79)	249.8(84.84)	0.0208
Creatinine (mg/dL)*	0.91(0.38)	1.00(0.59)	<0.0001
Serum Albumin (g/dL)*	4.10(0.37)	3.99(0.45)	<0.0001
International Normalized Ratio*	1.02(0.24)	1.04(0.22)	0.0097
Operative Variables
ASA Classification (%)	1.99
1 (no disturbance)	49.07	1.26	<0.0001
2 (mild disturbance)	47.40	36.00
3 (severe disturbance)	1.54	58.95
4 (life-threatening disturbance)		3.79
Length of Operation	2.80(2.42)	108.6(49.55)	<0.0001
Length of Stay		3.97(3.11)	<0.0001

*Presented as mean and associated standard deviation.

Outcomes

The following 30-day complications were analyzed: wound complications (superficial, deep, organ space infection, wound dehiscence), pulmonary complications (pneumonia, unplanned intubation), venous thromboembolism (deep vein thrombosis, pulmonary embolism), cardiac complication (acute myocardial infarction, cardiac arrest requiring resuscitation), renal complications (acute renal failure or insufficiency defined as a rise in Cr >2 mg/dL above baseline), neurologic complication (stroke, coma lasting >24 hours, peripheral nerve injury), urinary tract infection, sepsis, septic shock, unplanned readmission, reoperation, and mortality. ‘Any complication’ was defined as the presence of 1 or more of the above complications (Table 2).

Table 2.

Unadjusted and Adjusted 30-day Complications Between TKA with and without Blood Transfusions

	No Transfusion (%) (N=46,630)	Transfusion (%) (N=1425)	p-value
Complications (%)
Any Complication	4.23	9.19	<0.0001
Superficial Wound Infection	0.53	0.49	0.860
Deep Wound Infection	0.14	0.14	1.000
Organ Space Infection	0.15	0.21	0.4891
Wound Dehiscence	0.19	0.28	0.35
Any Wound	0.94	1.05	0.676
Pneumonia	0.32	1.68	<0.0001
Urinary Tract Infection	0.71	1.40	0.002
Sepsis	0.16	0.35	0.083
Septic Shock	0.05	0.56	<0.0001
Deep Venous Thrombosis	0.79	1.54	0.0018
Pulmonary Embolism	0.67	0.84	0.4447
Renal Insufficiency	0.11	0.63	<0.0001
Acute Renal Failure	0.05	0.07	0.484
Stroke	0.08	0.14	0.34
Cardiac Arrest	0.05	0.49	<0.0001
Myocardial Infarction	0.18	0.91	<0.0001
Unplanned Intubation	0.09	0.77	<0.0001
Unplanned Readmission	3.10	5.40	<0.0001
Reoperation	1.16	2.32	<0.0001
Mortality	0.08	0.56	<0.0001
Operative time* (min)	92.67(37.86)	108.6(49.55)	<0.0001
Length of hospital stay* (days)	2.80(2.42)	3.97(3.11)	<0.0001

Statistical Analysis

SAS software, version 9.4 (SAS Institute, Inc. of Cary, North Carolina) was used for data analysis. Two sample independent t-tests were used for between group comparisons of continuous variables. Chi-square test was used to determine difference between categorical variables for the univariate analysis. Next, separate multivariate logistic regression models for transfusion were used to determine the effects of the confounding variables identified from the univariate analysis with a p-value above 0.1. P-values were reported with the level of significance set at p<0.05 in the univariate model. Results from the multivariate logistic regression model are reported as an adjusted odds ratio and its associated 95% confidence interval.

Results

A total of 48,055 total knee arthroplasty patients were identified. Of these, 3% received a transfusion and 97% did not. Patients who received transfusion were older, had higher rates of diabetes mellitus (DM), chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), dialysis, corticosteroid use, bleeding disorders and preoperative blood transfusions. Patients who received transfusion had lower preoperative hematocrit, white blood cell count, and serum albumin and higher creatinine and INR. Those who received a transfusion had longer length of operation and length of stay. Finally, patients who received transfusion were more frequently ASA class 3 or 4 (p<0.001 for all comparisons) (Table 1).

Univariate analysis revealed that patients who received transfusion had a significantly higher rate of the following: any complication (9.19% versus 4.23%, p <0.0001) as well as pneumonia (1.68% versus 0.32%, p<0.0001), urinary tract infection (1.40% versus 0.71%, p= 0.002), septic shock (0.56% versus 0.05%, P<0.001), DVT (1.54% versus 0.79%, p= 0.002), renal insufficiency (0.63% versus 0.11%, <0.0001), cardiac arrest (0.49% versus 0.05%, p<0.0001), myocardial infarction (0.91% versus 0.18%, p <0.0001), unplanned intubation (0.77% versus 0.09%, p<0.0001), unplanned readmission (5.40% versus 3.10%, P <0.0001), reoperation (2.32% versus 1.16%, p <0.0001), mortality (0.56% versus 0.08%, p<0.0001), operative time (108.6 min versus 92.67 minutes, p<0.0001) and length of hospital stay (3.97 days versus 2.80 days, p <0.0001).

Multivariate regression analysis identified the following as risk factors for transfusion requirement: Black race (adjusted odds ratio [OR] 1.2, 95% confidence interval [CI] 1.01-1.4), COPD (OR 1.6, 95% CI 1.3-2.0), corticosteroids (OR 1.4, 95% CI 1.1-1.8), bleeding disorder (OR 1.4, CI 1.1-1.9), ASA class 4 (OR 2.3, CI 1.5-4.8), operative time >2 hours (OR 1.3, 95% CI 1.2-1.5) and lack of functional independence (OR 1.6, 95% CI 1.1-2.3) (Table 3).

Table 3.

Multivariable Regression Analysis for Blood Transfusion

	Adjusted Odds Ratio	95% Confidence Interval
Age (years)
<50	Ref	Ref
≥80	1.319	0.992-1.753
Female Sex	1.000	0.903-1.107
Race
White	Ref	Ref
Black	1.194	1.0137-1.402
Other	0.576	0.406-0.819
Diabetes	1.079	0.951-1.224
COPD	1.608	1.311-1.972
Congestive Heart Failure	1.050	0.525-2.101
Corticosteroids	1.434	1.149-1.789
Bleeding Disorder	1.415	1.081-1.854
ASA Classification
1	Ref	Ref
2	1.440	0.857-2.419
3	1.920	1.142-3.227
4+	2.626	1.447-4.765
Operative Time
<2 hours	Ref	Ref
>2 hours	1.341	1.188-1.513
Functional Status
Independent	Ref	Ref
Not Independent	1.594	1.116-2.275

Discussion

While modern TKA produces excellent clinical results, surgeons continually seek adaptations that will enhance their efficiency while maintaining patient safety and reducing cost of care. Given the cost associated with reflex lab testing in post-operative patients, we sought to identify independent risk factors that place specific patients at increased risk for transfusion in the ACS-NSQIP database. Previous studies have shown that age, ASA class, preoperative hemoglobin, initial postoperative hemoglobin, change in pre to postoperative hemoglobin and adherence to transfusion triggers are significant predictors of perioperative blood transfusion.^6-10 We found that while controlling for confounding demographics and comorbidities, Black race, history of COPD, corticosteroid use, and bleeding disorder are independent risk factors for transfusion requirement. Additionally, we found that patients who received transfusion demonstrated a significantly higher rate of the following: any complication, pneumonia, urinary tract infection, septic shock, deep vein thrombosis, renal insufficiency, cardiac arrest, myocardial infarction, unplanned infarction, unplanned intubation, unplanned readmission, reoperation and mortality. We are unable to comment on whether this is a cause and effect relationship.

Risk stratification has repeatedly been cited as an important tool for differentiating the quality of outcomes following frequently performed procedures, specifically, costly procedures such as TKA. Individual risk assessment models have been created for DVT prophylaxis, allotting relative weights to risk factors (hypercoagulability, metastatic cancer, stroke, sepsis, COPD) and assigning a relative risk to individual patients.¹¹ Similarly, a simple risk assessment model can be generated to identify patients who need post-operative hemoglobin monitoring. Our findings, including these independent risk factors, can be used to generate such a model. Our results indicate that the following are independent risk factors for postoperative transfusion requirement: Black race, history of COPD, corticosteroid use, history of bleeding disorder, operative time >2 hours, lack of functional independence and ASA class 3 and 4.

Several of the limitations of our study are common to large database research: the data were collected retrospectively; the data are also dependent on accuracy of coding. An additional limitation is the inclusion of only 1 year of NSQIP data; while this decreased variability in transfusion protocols, it also limits the volume of data. There are likely factors that place a patient at higher risk for transfusion that we did not consider or that are not included in the NSQIP database. Finally, our cohort does not include patients who were readmitted for acute blood loss anemia or those who had asymptomatic acute blood loss anemia and received transfusion outside of the immediate peri-operative period. Further research is needed to develop and test the proposed risk assessment model for post-operative hemoglobin monitoring restricted to high risk patients.

Conclusions

In a cohort of patients undergoing primary TKA in 2015, history of comorbidities including COPD, corticosteroid use, bleeding disorder, black race, lack of functional independence, operative time >2 hours and ASA class 3-4 were independent predictors of need for blood transfusion. The results of this study indicate that risk factors exist which can help guide surgeons in selective hemoglobin monitoring in the post-operative period. Judicious use of laboratory testing can be an adjunct practice for cost-containment after total joint arthroplasty.