Development and validation of a risk calculator for post-discharge venous thromboembolism following hepatectomy for malignancy

Abstract

Background:

Post-discharge venous thromboembolism (VTE) chemoprophylaxis decreases VTEs following cancer surgery, however identifying high-risk patients remains difficult. Our objectives were to (1) identify factors available at hospital discharge associated with post-discharge VTE following hepatectomy for malignancy and (2) develop and validate a post-discharge VTE risk calculator to evaluate patient-specific risk.

Methods:

Patients who underwent hepatectomy for malignancy from 2014 to 2017 were identified from the ACS NSQIP hepatectomy procedure targeted module. Multivariable logistic regression identified factors associated with post-discharge VTE. A post-discharge VTE risk calculator was constructed, and predicted probabilities of post-discharge VTE were calculated.

Results:

Among 11 172 patients, 95 (0.9%) developed post-discharge VTE. Post-discharge VTE was associated with obese BMI (OR 2.29 vs. normal BMI [95%CI 1.31–3.99]), right hepatectomy/trisegmentectomy (OR 1.63 vs. partial/wedge [95%CI 1.04–2.57]), and several inpatient postoperative complications: renal insufficiency (OR 5.29 [95%CI 1.99–14.07]), transfusion (OR 1.77 [95%CI 1.12–2.80]), non-operative procedural intervention (OR 2.97 [95%CI 1.81–4.86]), and post-hepatectomy liver failure (OR 2.22 [95%CI 1.21–4.08]). Post-discharge VTE risk ranged from 0.3% to 30.2%. Twenty iterations of 10-fold cross validation identified internal validity.

Conclusions:

Risk factors from all phases of care, including inpatient complications, are associated with post-discharge VTE following hepatectomy. Identifying high-risk patients may allow for personalized risk-based post-discharge chemoprophylaxis prescribing.

Introduction

Venous thromboembolism (VTE), including both deep vein thrombosis (DVT) and pulmonary embolism (PE), is a common, but preventable cause of postoperative morbidity and mortality, particularly following major operations.^1-3 Patients with malignancy are at further increased risk.^4,5 Centers for Medicare and Medicaid Services (CMS) has focused efforts on improving inpatient postoperative VTE prophylaxis to decrease this risk.⁶ However, VTE risk extends beyond the index hospitalization, with approximately one third of VTE events occurring after hospital discharge.⁷ Several randomized controlled trials have shown the benefit of post-discharge chemoprophylaxis following abdominal or pelvic resections for malignancy.^8-10 Thus, the American College of Chest Physicians (CHEST) has recommended post-discharge chemoprophylaxis to reach a total of 28 days of postoperative chemoprophylaxis following high-risk surgical resections for malignancy.¹¹

Historically, it was believed that hepatic insufficiency following hepatectomy was protective against VTE, particularly in instances of elevated International Normalized Ratio (INR).¹² However, this theory has since been refuted, with reported VTE event rates ranging from 2.8 to 4.3% following hepatectomy for malignancy.^13-15 In fact, prior studies of VTEs following surgical resections for malignancy revealed that VTE rates following hepatopancreaticobiliary (HPB) resections were second only to esophagogastric resections at 4.2–5.9%.^7,16 Despite the recommendations for post-discharge chemoprophylaxis following high risk operations, and the safety of chemoprophylaxis following hepatectomy, most providers do not prescribe chemoprophylaxis after hospital discharge.^12,17 Overall, 14% of liver surgeons state that they routinely prescribe post-discharge chemoprophylaxis, with prescription rates reaching 39% following major hepatectomy and 26% following minor hepatectomy at HPB fellowship training centers.^18,19

The discrepancy between recommendations for post-discharge chemoprophylaxis and low prescription rates could be related to several factors, including difficulty in predicting patients at elevated risk of post-discharge VTE. While previous studies have identified clinical risk factors for VTE following hepatectomy, no prior study has comprehensively evaluated patient factors that are available at the time of hospital discharge to predict the likelihood of post-discharge VTE development.^15,20,21 The objectives of this study were to (1) identify patient-specific factors available at the time of hospital discharge following hepatectomy that are associated with post-discharge VTE and (2) to develop and validate a post-discharge VTE risk calculator to aid in identifying patients at elevated risk of post-discharge VTE.

Methods

Data source and patient population

Patients who underwent hepatic resection for malignancy between January 1, 2014 and December 31, 2017 were identified from the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) hepatectomy procedure targeted participant use file. ACS NSQIP is a prospectively maintained surgical registry derived from select operative cases at participating hospitals.^22,23 Trained nurse abstractors review patient data for 30 days following surgery, and abstract over 150 variables, including preoperative patient factors such as comorbidities and laboratory values; intraoperative procedure codes; and postoperative complications. Beginning in 2012, ACS NSQIP developed procedure targeted modules that can be merged with the general participant use file to gain additional procedure specific variables including operative approach and specified complications. The hepatectomy procedure targeted module began in 2014, with 120 participating hospitals as of 2017.²⁴ ACS NSQIP data integrity is maintained through regular audits including analysis of inter-rater reliability.²⁵

Patients were considered for analysis if they underwent hepatectomy for any malignancy, either primary malignancy or metastatic disease (n = 11 723). Patients were excluded from post-discharge VTE analysis if they developed an inpatient VTE (n = 263), had a postoperative length of stay longer than 30 days (n = 148), or expired before hospital discharge (n = 140). Thus, the final cohort of 11 172 patients included for analysis had the potential of developing a post-discharge VTE that would be captured by ACS NSQIP.

Primary outcome and covariates

The primary outcome was post-discharge VTE, which was defined as either PE or DVT that was diagnosed after the day of discharge from the index hospitalization, but within the 30-day period through which data is abstracted for ACS NSQIP. PE was defined as a new diagnosis of a blood clot in a pulmonary artery that was confirmed on imaging, while DVT was defined as a new diagnosis of blood clot or thrombus within the venous system which was confirmed on imaging and required (either recommended or actually administered) anticoagulation. Possible predictors of post-discharge VTE were identified a priori based upon a comprehensive review of the literature and clinical face validity.

Preoperative factors evaluated for association with post-discharge VTE included age, sex, race, body mass index (BMI), functional status, American Society of Anesthesia classification (ASA), preoperative albumin, preoperative platelet count, preoperative use of biliary stenting, operative indication (primary malignancy or metastatic disease), and a number of comorbid conditions including bleeding disorder, dyspnea, chronic obstructive pulmonary disease (COPD), hypertension, diabetes, steroid use, and ascites. Intraoperative factors evaluated for association with post-discharge VTE included operative time, extent of operation, concurrent partial resection or ablation, and operative approach. Postoperative factors included a number of postoperative complications, such as surgical site infection (inclusive of superficial, deep, and organ space infections), reintubation, renal insufficiency, transfusion, bile leak, non-operative procedural intervention (inclusive of interventional radiology and gastroenterology procedures), reoperation, and post-hepatectomy liver failure (as defined by the International Study Group of Liver Surgery [ISGLS]).

Only complications that were diagnosed during the index hospitalization were considered for analysis. To do so, the day of complication diagnosis was compared to the day of hospital discharge, and only those complications occurring on or prior to the day of discharge were included. For select complications derived from the hepatectomy procedure targeted participant use file, including bile leak, non-operative procedural intervention, and post-hepatectomy liver failure, no date of diagnosis was provided. In these cases, readmission ICD-10 diagnosis codes were screened for corresponding diagnoses. Any patient readmitted with a diagnosis corresponding to a listed complication was considered to have developed that complication following hospital discharge rather than during the index hospitalization. Therefore, the complication was not considered for analysis.

Statistical analysis

The overall VTE rate was calculated from the original dataset and reported as inpatient and post-discharge VTEs. After applying the exclusion criteria, the frequency of post-discharge VTE based upon preoperative, intraoperative, and postoperative factors was evaluated using separate Chi-squared tests. Predictors with P < 0.05 on bivariate analysis were entered into a multivariable logistic regression model. Patients with missing data were excluded from multivariable logistic regression modelling. Colinear predictors were removed from the model until convergence. Subsequently, the Wald test was used to identify categorical predictors that significantly contributed to the regression model, with categorical variable constructs with P > 0.05 progressively removed from the model until a final model was estimated. The model was evaluated for statistical validity using both the C-statistic, wherein a value closer to 1.0 indicates better discrimination, and the Hosmer and Lemeshow Chi-square, wherein any P < 0.05 indicates poor calibration.^26,27 Internal validation was performed using 10-fold cross validation over 20 iterations, wherein the entire dataset was split into 10 equally sized cohorts. Nine of the cohorts were considered training cohorts, on which the model was estimated, and the model was evaluated on the final testing cohort. This process was repeated with each of the 10 cohorts used as a testing cohort, and the entire sequence was completed 20 times with the resultant C-statistics and Hosmer Lemeshow Chi-squares averaged over each iteration.²⁸

The final regression model was estimated on the logit scale. Predicted probabilities of post-discharge VTE were calculated for each patient and plotted to demonstrate variation. To use this regression model as a post-discharge VTE risk calculator, indicate the beta coefficients applicable to a particular patient and sum those beta coefficients with the model intercept to generate a log probability (LP) of the outcome. The predicted probability of an event can be calculated using the following equation: event probability = exp (LP)/[1 + exp (LP)], in accordance with previously reported methodology.^29-32 Sensitivity and specificity of the risk calculator were calculated for various cut-points and the Youden’s J Index was calculated to determine the cut-point with maximal sensitivity and specificity in predicting post-discharge VTE.³³ Statistical analyses were performed using Stata version 14.2 (Stata Corp, College Station, Texas). This study was determined to be exempt from Institutional Review Board review by Northwestern University.

Results

Cohort characteristics

Among 11 722 patients who underwent hepatectomy for malignancy, 358 (3.1%) developed a VTE. After applying exclusion criteria, 11 172 patients were included in post-discharge VTE analyses, of which 95 (0.9%) were diagnosed with a post-discharge VTE. Overall, 26.5% of VTEs identified in this study were diagnosed following hospital discharge. The mean patient age was 60.8 years, with 56.4% of patients being male and 65.5% white. The majority of patients (62.6%) underwent hepatectomy for metastatic disease, 65.0% underwent partial hepatectomy or wedge resection, and 81.0% of cases were performed via an open surgical approach. Transfusion was the most common complication, which occurred in 17.5% of analyzed patients. Additional demographic information can be found in Table 1.

Table 1.

General cohort characteristics of patients undergoing hepatectomy for malignancy

Patient Characteristic (n = 11 172)	N (%)
Preoperative Factors
Age, years, mean (SD)	60.8 (12.0)
Sex
Male	6300 (56.4)
Female	4872 (43.6)
Race
White	7312 (65.5)
Black	868 (7.8)
Asian	717 (6.4)
Other/Not Reported	2275 (20.4)
Body Mass Index (BMI)
<18.5	195
18.5–24.9	3286 (29.6)
25–29.9	3932 (35.4)
30–34.9	3686 (33.2)
Comorbidities
Bleeding Disorder	419 (3.8)
Dyspnea (moderate exertion/rest)	583 (5.2)
COPD	430 (3.9)
Hypertension	5395 (48.3)
Diabetes	2062 (18.5)
Steroid Use	350 (3.1)
Ascites	66 (0.6)
Functional Status
Independent	11 091 (99.5)
Dependent	61 (0.6)
ASA Class
I/II	2352 (21.1)
III/IV/V	8797 (78.9)
Preoperative Albumin <3 g/dL	806 (16.4)
Preoperative Platelet Count
<150 000	797 (16.5)
150 000–400 000	3819 (79.1)
>400 000	214 (4.4)
Preoperative Biliary Stent	482 (4.4)
Operative Indication
Primary Hepatic Malignancy	4183 (37.4)
Metastatic Disease	6989 (62.6)
Intraoperative Factors
Operative time, min, mean (SD)	247.2 (117.8)
Extent of Operation
Partial/Wedge	7264 (65.0)
Left	1014 (9.1)
Right/Trisegmentectomy	2894 (25.9)
Concurrent Partial Resection or Ablation	6104 (55.5)
Operative Approach
Open	9050 (81.0)
Minimally Invasive	2122 (19.0)
Postoperative Factors
Inpatient Postoperative Complications
SSI	442 (4.0)
Reintubation	97 (0.9)
Renal Insufficiency	69 (0.6)
Transfusion	1950 (17.5)
Bile Leak	783 (7.0)
Non-Operative Procedural Intervention	852 (7.6)
Reoperation	143 (1.3)
Post-Hepatectomy Liver Failure	456 (4.1)
Length of Stay, days, mean, (SD)	6.4 (4.2)
Outcomes
Post-Discharge DVT	66 (0.6)
Post-Discharge PE	33 (0.3)
Any Post-Discharge VTE	95 (0.9)

COPD = chronic obstructive pulmonary disease

ASA = American society of anesthesia

SSI = surgical site infection

DVT = deep vein thrombosis

PE = pulmonary embolism

VTE = venous thromboembolism

Factors associated with post-discharge VTE

Obese BMI was associated with development of post-discharge VTE (1.2% of patients with BMI ≥30.0 vs. 0.6% with BMI 18.5–24.9; OR 2.29 [95% CI 1.31–3.99]). Patient comorbidities, preoperative labs, functional status, and ASA class were not associated with post-discharge VTE (Table 2). Race and preoperative biliary stent placement were associated with post-discharge VTE on bivariate analysis but did not contribute to the logistic regression model.

Table 2.

Characteristics of patients who develop post-discharge VTE following hepatectomy for malignancy

Total Patients n = 11 172	No pdVTE	pdVTE	p-valuea
	N (%)
	11 077 (99.2%)	95 (0.9%)
Preoperative Factors
Age, years
≤50	2137 (99.3%)	15 (0.7%)	0.679
51-60	2921 (99.2%)	23 (0.8%)
61-70	3649 (99.1%)	33 (0.9%)
≥71	2370 (99.0%)	24 (1.0%)
Sex
Male	6243 (99.10%)	57 (0.9%)	0.476
Female	4834 (99.2%)	38 (0.8%)
Race
White	7236 (99.0%)	76 (1.0%)	0.011
Black	861 (99.2%)	7 (0.8%)
Asian	716 (99.9%)	1 (0.1%)
Other/Not Reported	2264 (99.5%)	11 (0.5%)
Body Mass Index (BMI)
<18.5	193 (99.0%)	2 (1.03%)	0.042
18.5–24.9	3268 (99.5%)	18 (0.6%)
25–29.9	3901 (99.2%)	31 (0.8%)
≥30.0	3643 (98.8%)	43 (1.2%)
Comorbidities
Bleeding Disorder	415 (99.1%)	4 (1.0%)	0.813
Dyspnea (moderate exertion/rest)	579 (99.3%)	4 (0.7%)	0.657
COPD	426 (99.1%)	4 (0.9%)	0.854
Hypertension	5346 (99.1%)	49 (0.9%)	0.519
Diabetes	2044 (99.1%)	18 (0.9%)	0.902
Steroid Use	348 (99.4%)	2 (0.6%)	0.564
Ascites	65 (98.5%)	1 (1.5%)	0.555
Functional Status
Independent	10 997 (99.2%)	94 (0.9%)	0.502
Dependent	60 (98.4%)	1 (1.6%)
ASA Class
I/II	2329 (99.0%)	23 (1.0%)	0.455
III/IV/V	8725 (99.2%)	72 (0.8%)
Preoperative Albumin <3 g/dL	801 (99.4%)	5 (0.6%)	0.282
Preoperative Platelet Count
<150 000	790 (99.1%)	7 (0.9%)	0.388
150 000–400 000	3783 (99.1%)	36 (0.9%)
>400 000	210 (98.1%)	4 (1.9%)
Preoperative Biliary Stent	473 (98.1%)	9 (1.9%)	0.014
Operative Indication
Primary Hepatic Malignancy	4142 (99.0%)	41 (1.0%)	0.248
Metastatic Disease	6935 (99.2%)	54 (0.8%)
Intraoperative Factors
Operative time, hours
<2	1124 (99.4%)	7 (0.6%)	<0.001
2-4	5008 (99.4%)	29 (0.6%)
4-6	3253 (99.1%)	31 (0.9%)
>6	1692 (98.4%)	28 (1.6%)
Extent of Operation
Partial/Wedge	7217 (99.4%)	47 (0.7%)	0.001
Left	1006 (99.2%)	8 (0.8%)
Right/Trisegmentectomy	1850 (98.6%)	40 (1.4%)
Concurrent Partial Resection or Ablation	6590 (99.3%)	45 (0.7%)	0.016
Operative Approach
Open	8964 (99.1%)	86(1.0%)	0.018
Minimally Invasive	2113 (99.6%)	9 (0.4%)
Postoperative Factors
Inpatient Postoperative Complications
SSI	439 (99.3%)	3 (0.7%)	0.688
Reintubation	95 (97.9%)	2(2.1%)	0.192
Renal Insufficiency	64 (92.8%)	5 (7.3%)	<0.001
Transfusion	1917 (98.3%)	33(1.7%)	<0.001
Bile Leak	763 (97.5%)	20 (2.6%)	<0.001
Non-Operative Procedural Intervention	829 (97.3%)	23 (2.7%)	<0.001
Reoperation	142 (99.3)	1 (0.7)	0.843
Post-Hepatectomy Liver Failure	440 (96.5%)	16 (3.5%)	<0.001
Length of Stay, days
≤3	1947 (99.5%)	10 (0.5%)	0.061
4-5	3763 (99.2%)	31 (0.8%)
6-8	3469 (99.2%)	29 (0.8%)
9+	1898 (98.7%)	25 (1.3%)

pdVTE = post-discharge venous thromboembolism

COPD = chronic obstructive pulmonary disease

ASA = American society of anesthesiologists

SSI = surgical site infection

^aChi-squared test

With regard to intraoperative factors, post-discharge VTE occurred more frequently as operative time increased (1.6% of patients with operating time >6 h vs. 0.6% if < 2 h; p < 0.001), and operative extent increased (1.4% for right hepatectomy/trisegmentectomy vs. 0.7% for partial/wedge; p = 0.001). These two predictors were colinear, with only extent of resection remaining in the final model, wherein right hepatectomy/trisegmentectomy had an increased odds of post-discharge VTE compared with partial/wedge resection (OR 1.63 [95% CI 1.04–2.57]; Table 3). Patients who underwent concurrent partial resection or ablation less frequently developed post-discharge VTE on bivariate analysis (0.7% vs. 1.0%; p = 0.016), but did not remain in the final adjusted model. Similarly, open operative approach (1.0% vs. 0.4% of minimally invasive; p = 0.018) led to more frequent post-discharge VTEs on bivariate analysis, but did not significantly contribute to the final model.

Table 3.

Association between patient characteristics and post-discharge VTE following hepatectomy for malignancy

	Adjusted Odds Ratio (95% Confidence Interval)	p-value
Body Mass Index (BMI)
<18.5	1.80 (0.41–7.88)	0.438
18.5–24.9	1.00	REF
25.0–29.9	1.47 (0.81–2.65)	0.203
≥30	2.29 (1.31–3.99)	0.004
Extent of Operation
Partial/Wedge	1.00	REF
Left	1.01 (0.45–2.25)	0.987
Right/Trisegmentectomy	1.63 (1.04–2.57)	0.034
Renal Insufficiency	5.29 (1.99–14.07)	0.001
Transfusion	1.77 (1.12–2.80)	0.014
Non-Operative Procedural Intervention	2.97 (1.81–4.86)	<0.001
Post-Hepatectomy Liver Failure	2.22 (1.21–4.08)	0.010

11,036 patients included in multivariable logistic regression analysis. N = 136 patients excluded based on missing data including 73 based on missing BMI data and 63 based on missing non-operative procedural intervention data.

A number of inpatient complications were associated with post-discharge VTE events, including renal insufficiency (7.3%, OR 5.29 [95% CI 1.99–14.07]), transfusion (1.7%, OR 1.77 [95% CI 1.12–2.80]), non-operative procedural intervention (2.7%, OR 2.97 [95% CI 1.81–4.86]), and post-hepatectomy liver failure (3.5%, OR 2.22 [95% CI 1.21–4.08]). Although patients with bile leak more frequently developed post-discharge VTE on bivariate analysis (2.6% vs. 0.7%, p < 0.001), this construct fell out of the final multivariable logistic regression model. There was no difference in post-discharge VTE events based upon post-operative surgical site infection (0.7% vs. 0.9% p = 0.688) or reoperation (0.7% vs. 0.9%, p = 0.843) in the inpatient setting. There was no statistically significant difference in post-discharge VTE rates based on inpatient length of stay (Table 2).

Post-discharge VTE risk calculator

Beta coefficients for use as a post-discharge VTE risk calculator are shown in Table 4. The predicted probabilities of post-discharge VTE ranged from 0.003 to 0.302 in the analyzed sample (Fig. 1). The Youden’s J Index was identified at a risk cut-point of 0.75%; 27.2% of patients had a predicted probability of post-discharge VTE greater than 0.7%, 21.2% had a predicted probability of post-discharge VTE greater than 1.0% and 3.4% a predicted probability greater than 3.0% (Table 5). The calculator’s C-statistic was 0.713 indicating good model discrimination and the Hosmer and Lemeshow Chi-square p = 0.369 indicating good model calibration. Twenty iterations of 10-fold cross validation yielded a mean C-statistic of 0.700 and Hosmer and Lemeshow Chi-square p = 0.236.

Table 4.

Beta coefficients for post-discharge VTE risk calculator following hepatectomy for malignancy

	Beta Coefficient	95% CI
Body Mass Index (BMI)
<18.5	0.59	−0.89–2.06
18.5–24.9	0.00	REF
25.0–29.9	0.38	−0.21–0.97
≥30	0.83	0.27–1.38
Extent of Operation
Partial/Wedge	0.00	REF
Left	0.01	−0.80–0.81
Right/Trisegmentectomy	0.49	0.04–0.94
Renal Insufficiency	1.67	0.69–2.64
Transfusion	0.57	0.11–1.03
Non-Operative Procedural Intervention	1.09	0.59–1.58
Post-Hepatectomy Liver Failure	0.80	0.19–1.41
Constant	−5.84	−6.37–−5.30

Summed beta coefficients and Constant = LP.

Predicted Probability = exp(LP)/[1 + exp(LP)].

Figure 1.

Predicted probability of post-discharge VTE following hepatectomy for malignancy

Table 5.

Sensitivity and specificity of post-discharge VTE risk calculator at various risk thresholds

Cut-Point	Patients Included	Sensitivity	Specificity	Youden’s Ja
>0.25%	100.0%	100.0%	0.0%	0.0
>0.50%	62.0%	86.0%	38.2%	24.2
>0.75%	27.2%	60.2%	73.1%	33.3
>1.00%	21.2%	53.8%	79.1%	32.8
>1.25%	11.7%	39.8%	88.6%	28.4
>1.50%	9.7%	33.3%	90.5%	23.9
>1.75%	8.5%	33.3%	91.7%	25.0
>2.0%	5.6%	25.8%	94.6%	20.4
>3.0%	3.4%	19.4%	96.7%	16.1

^aYouden’s J Index is a composite measure of the accuracy of a test which is used to identify the best diagnostic cut-point. It is calculated as sensitivity + specificity-1.

Discussion

Approximately one third of VTEs occur after hospital discharge in patients undergoing hepatectomy for malignancy, which in absolute terms equates to less than 1% of at-risk patients. Despite randomized controlled trials demonstrating the effectiveness of post-discharge chemoprophylaxis in preventing VTE in this population, most providers are not routinely, or are only selectively prescribing post-discharge chemoprophylaxis. Using only information available at the time of hospital discharge, we identified patient factors associated with the development of post-discharge VTE, including BMI, extended hepatectomy, and a number of inpatient postoperative complications inclusive of renal insufficiency, transfusion, non-operative procedural interventions, and post-hepatectomy liver failure. A post-discharge VTE risk calculator was constructed to identify patients at elevated risk of post-discharge VTE. With a continuing emphasis on personalized medicine, this post-discharge calculator may be beneficial in identifying patients with an increased probability of developing post-discharge VTE. Future efforts should prospectively evaluate whether risk-based tailored prescribing can achieve the same VTE prevention efficacy as is currently achieved by universal post-discharge chemoprophylaxis recommendations that have had selective uptake.

Overall VTE rates

In this retrospective cohort study, 3.1% of patients developed a VTE within 30 days following hepatectomy. Of these VTEs, 26.5% were diagnosed after discharge from the index hospitalization. This overall VTE rate is consistent with multiple prior observational studies that have identified elevated VTE rates following hepatectomy for malignancy, with approximately 30% of events occurring after hospital discharge.^7,14,20 The American College of Chest Physicians CHEST guidelines categorizes this level of risk as “moderate” (VTE rate of ~3%) whereas high risk is ~6%.¹¹

Factors associated with post-discharge VTE

Patient factors from all phases of care were associated with post-discharge VTE following hepatectomy for malignancy. Obese BMI was associated with a greater than two-fold increased odds of post-discharge VTE after adjusting for other covariates. This is consistent with other studies that have demonstrated an association between BMI and VTE in post-operative patients, those undergoing abdominal or pelvic resections for malignancy, and even specifically in the post-discharge setting following hepatectomy.^14,15,33-35

Although post-discharge VTEs occurred more frequently based upon several intraoperative factors on bivariate analysis, including operative time, extent of operation, concurrent partial resections or ablations, and operative approach, these factors were largely collinear. Ultimately only extent of operation remained in the final model, with right hepatectomies and trisegmentectomies associated with increased odds of post-discharge VTE. A prior study from MD Anderson that identified risk factors for both inpatient and post-discharge VTEs demonstrated an association between prolonged operative time and post-discharge VTE, but extended and right hepatectomies were associated with VTEs overall.²⁰ Clearly operative extent and operative time are related, and although in isolation both are likely important considerations when discerning VTE risk, varying model compositions may ultimately lead to preferential inclusion of one or the other of these factors.

Several prior studies have evaluated factors associated with post-discharge VTE following hepatectomy, and have identified both surgical site infection and bile leak as complications associated with post-discharge VTE.^20,21 Interestingly, when considering only complications that were diagnosed during the index hospitalization, and therefore providing information available for consideration at the time of hospital discharge, neither surgical site infections nor bile leak were associated with post-discharge VTE. However, patients who underwent inpatient non-operative procedural interventions had a nearly 3-fold increased odds of developing a post-discharge VTE. These findings further refine the previously understood risk factors, indicating that infectious complications and bile leak pose a spectrum of risk, with only those complications serious enough to require procedural intervention being associated with post-discharge VTE.

Inpatient reoperation was not associated with post-discharge VTE in this study, which is contrary to prior studies that have identified this association following hepatectomies for malignancy.^14,20,36 However, hemorrhage is the most common indication for reoperation following hepatectomy, and transfusion is also known to be associated with VTE development.^14,36,37 Transfusion was associated with post-discharge VTE in this study. In addition, renal insufficiency and post-hepatectomy liver failure were both associated with an increased risk of post-discharge VTE. Renal insufficiency has been identified as a risk factor for VTE in postoperative patients in general, as well as those following hepatectomy.^38-40 Postoperative renal insufficiency is common following hepatectomy, with an incidence as high as 15%.^41,42 This increased incidence may allow for the identification of this statistically significant association between postoperative renal insufficiency and VTE, which has not previously been identified. Additionally, no prior study to our knowledge has identified an association between post-hepatectomy liver failure and VTE. Considering the long-held belief that post-hepatectomy liver dysfunction is protective against VTE, these patients may not be receiving appropriate VTE chemoprophylaxis (either inpatient or post-discharge), despite research demonstrating the existence of a procoagulant state and safety of VTE chemoprophylaxis in these patients.^12,17,21,43

Post-discharge VTE risk calculator

Although several randomized controlled trials have demonstrated the benefits of post-discharge VTE chemoprophylaxis in patients undergoing abdominal/pelvic resections for malignancy, these studies have limitations.^8,9,44 Most importantly, all of the major studies screened asymptomatic patients for VTEs, which calls into question whether diagnosed VTEs were clinically significant. Therefore, some providers believe that these studies over-estimated the clinical benefit of post-discharge chemoprophylaxis, which may contribute to low prescribing rates.⁴⁵ The CHEST guidelines recommend post-discharge VTE chemoprophylaxis until postoperative day 28 for all patients at high risk for VTE’ while the American Society of Clinical Oncology (ASCO) guidelines recommend chemoprophylaxis for up to 28 days for patients who have high risk features, such as restricted mobility, obesity, history of VTE, or other risk factors.^11,46 The Americas Hepato-Pancreato-Biliary Association (AHPBA) practice guidelines on VTE Prophylaxis in Liver Surgery could not definitively recommend post-discharge chemoprophylaxis following hepatectomy noting that high volume institutions have conflicting practice patterns for prescribing post-discharge chemoprophylaxis, with low VTE rates regardless of prescribing practices.¹² This lack of institutional consensus on the appropriateness of post-discharge VTE chemoprophylaxis has also been reflected in provider surveys, which have indicated that the majority of providers, even in HPB training programs, do not prescribe post-discharge chemoprophylaxis.^18,19

Given the lack of consistent prescribing, combined with the recommendations that high-risk patients should be prescribed chemoprophylaxis, we developed a post-discharge VTE risk calculator to aide in the identification of patients at elevated risk for post-discharge VTE. This technique is similar to that used by prior studies that have developed patient specific risk calculators derived from ACS NSQIP data.^31,35,47 Using this calculator, an individual patient’s risk of post-discharge VTE can be calculated by summing the applicable beta coefficients found in Table 4 corresponding with the patient’s risk profile. Overall, examined patients had a wide range of risk, with the majority of patients having <1% risk of post-discharge VTE. Calculating an individual patient’s predicted probability of post-discharge VTE can be considered in light of the CHEST recommendations, as well as economic analyses that have indicated that post-discharge chemoprophylaxis is only cost effective when the post-discharge VTE risk exceeds 2.4%.⁴⁸ Although these data do not indicate a strict risk cut-off below which it is safe to proceed without post-discharge VTE chemoprophylaxis, there is potential for tailored risk-based prescribing dependent upon provider and patient risk aversion.

Limitations

This study has limitations. First, this study utilized observational data, and therefore causal inference cannot be derived from these data. Second, patients were identified from the ACS NSQIP procedure targeted program for hepatectomy, which requires additional abstraction resources above those of the general participant use file. Thus, there may be a selection bias toward high-resource hospitals that participant in this program, which may limit the generalizability of these findings. Third, medication prescribing and adherence data is not available in ACS NSQIP, meaning that we could not adjust for preoperative anticoagulant usage nor receipt of inpatient or post-discharge chemoprophylaxis in this analysis. Thus, this study likely underestimates the post-discharge VTE rate that would occur in the absence of any chemoprophylaxis. However, prior studies have demonstrated the relatively low utilization of post-discharge chemoprophylaxis, thus the majority of analyzed patients likely did not receive post-discharge chemoprophylaxis and the associations between patient risk factors and post-discharge VTE are unlikely to change based upon prescribing patterns. Fourth, ACS NSQIP lacks data related to prior VTE diagnoses or pro-thrombotic conditions, which are associated with postoperative VTE and post-discharge VTE development. Although we were unable to evaluate those specific risk factors in this study, prior research has demonstrated the importance of post-discharge prophylaxis in these patients following major abdominal operations for malignancy.^49,50 Fifth, VTE risk extends beyond 30-days, but such events could not be identified in this study. To that end, patients with longer hospitalizations had a shorter at-risk interval in this study, as post-discharge VTEs were only recorded until postoperative day 30. These are limitations of the ACS NSQIP data, and again, likely underestimate the true post-discharge VTE rate.

Conclusions

The overall VTE incidence is 3.1% following hepatectomy for malignancy, with 0.9% of patients developing a post-discharge VTE within 30 days postoperatively. Preoperative, intraoperative, and postoperative factors were associated with post-discharge VTE, including BMI, extent of operation, and select inpatient complications of renal insufficiency, transfusion, non-operative procedural intervention, and post-hepatectomy liver failure. Based on a post-discharge VTE risk calculator, there was a wide range of predicted post-discharge VTE risk, ranging from 0.3% to 30.2%. This calculator may be helpful in identifying high-risk patients who should be prescribed post-discharge chemoprophylaxis, and future research will focus on prospectively evaluating the efficacy of risk-based post-discharge VTE prophylaxis prescribing compared with that which is currently achieved by universal prescribing guidelines that are selectively followed.