- Email: [email protected]
Analysis of Venous Thromboembolism Risk in Patients Undergoing Craniotomy
Original Article
Analysis of Venous Thromboembolism Risk in Patients Undergoing Craniotomy Hanna Algattas, Kristopher T. Kimmell, G. Edward Vates, Babak S. Jahromi
OBJECTIVES: Craniotomy poses a risk for postoperative venous thromboembolism (VTE), but the utility of anticoagulation in this patient population is unclear. We sought to identify risk factors predictive of VTE in patients undergoing craniotomy.
-
METHODS: The American College of Surgeons National Surgical Quality Improvement Project (ACS-NSQIP) database was reviewed for patients undergoing craniotomy. Clinical factors provided by the database were analyzed for association with VTE.
-
CONCLUSIONS: The risk of postoperative VTE after craniotomy can be quantified by a simple risk score, with increasing risk factors conferring increased risk of VTE. On the basis of risk scoring, a subset of patients who would benefit from anticoagulation post craniotomy may be identified.
-
INTRODUCTION
M
-
RESULTS: A total of 10,477 adult patients who underwent craniotomy from 2011e2012 were identified. The rate of VTE was 3.2% (pulmonary embolism [PE] was 1.3%; deep vein thrombosis [DVT] was 2.4%). Several factors were significant in univariate analysis, and a subset persisted after multivariate analysis. Patients were assigned a risk score on the basis of the presence of those variables. Higher risk scores were predictive of VTE risk, as well as increasing time from surgery to discharge and mortality. A receiver operating characteristics curve revealed a significant area under the curve (0.719) for scores being predictive of VTE risk. The model was validated against our similar analysis of 2006e2010 NSQIP data and demonstrated comparable findings.
odern medicine has largely succeeded in treating and extending the course of a number of chronic and degenerative conditions. The current era of medicine has shifted focus more from tertiary prevention to primary prevention and risk reduction. Pursuant with this focus, stakeholders seek to reduce expenses by limiting preventable causes of hospitalization and prolonged admissions. One arena for high-yield prevention and cost reduction is the perioperative period. The Surgical Care Improvement Project (SCIP) was started in 2006 with assistance from the Joint Commission with the goal of diminishing preventable complications before, after, and within the operative window. The instillment of recommendations regarding appropriate antibiotic and medication use, postoperative blood glucose measures, surgical site hair removal, urinary catheterization, temperature control, and prevention of venous
Key words Craniotomy - Venous thromboembolism
PE: Pulmonary embolism PNA: Pneumonia RCT: Randomized controlled trials ROC: Receiver-operating characteristics SCIP: Surgical Care Improvement Project TM&M: Thoracic Morbidity and Mortality System UTI: Urinary tract infection V-Q: Ventilation-perfusion VTE: Venous thromboembolism
-
Abbreviations and Acronyms ACS-NSQIP: American College of Surgeons National Surgical Quality Improvement Project AHRQ: Agency for Healthcare Research and Quality ASA: American Society for Anesthesiologists AUC: Area under the curve CMS: Center for Medicare and Medicaid Services CPR: Cardiopulmonary resuscitation CPT: Current Procedural Terminology CVA: Cerebrovascular accident DVT: Deep vein thrombosis HAC: Hospital-acquired condition ICH: Intracranial hemorrhage No.: Number NQF: National Quality Forum NSQIP: National Surgical Quality Improvement Project OR: Odds ratio
WORLD NEUROSURGERY - [-]: ---, - 2015
Department of Neurosurgery, University of Rochester Medical Center, Rochester, New York, USA To whom correspondence should be addressed: Hanna Algattas, B.A. [E-mail: [email protected]] Citation: World Neurosurg. (2015). http://dx.doi.org/10.1016/j.wneu.2015.06.033 Journal homepage: www.WORLDNEUROSURGERY.org Available online: www.sciencedirect.com 1878-8750/$ - see front matter ª 2015 Elsevier Inc. All rights reserved.
www.WORLDNEUROSURGERY.org
1
ORIGINAL ARTICLE HANNA ALGATTAS ET AL.
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
thromboembolism (VTE) has materialized as a result. The Joint Commission, National Quality Forum (NQF), and Agency for Healthcare Research and Quality (AHRQ) support these quality measures and have created mandates for hospitals to report violations or else suffer financial penalties. In an effort to improve outcomes and avoid fines, hospitals have developed protocols to track these aforementioned measures. Yet there is no standard approach to determining risk of VTE and the benefit of prophylactic anticoagulation in at-risk populations (4). Concerning VTE then, it comes as no surprise that surgeons especially are forced by institutional and financial pressures to tread delicately the line between thrombosis and hemorrhage. This dilemma is most pronounced in neurosurgical procedures. A meta-analysis of 8 randomized controlled trials (RCTs) found that for every 1000 craniotomy patients anticoagulated with heparin, 91 VTEs would be prevented, whereas 7 intracranial hemorrhages (ICH) and 28 minor bleeds would occur (15). Thus the suggested VTE rate is 9.1%, which appears high at face value but is explained in the meta-analysis. Three of 6 studies analyzed were completed before the use of routine mechanical prophylaxis and failed to include such treatment in control groups. Additionally, different methods of detecting VTE were used (including venography and I125 fibrinogen) that have higher sensitivity and likelihood of capturing asymptomatic VTEs. Therefore although the results seem to support anticoagulation, the authors concede that the absolute values may provide false assurance and VTE detection using surrogate markers may detect less worrisome, asymptomatic VTEs. When accounting for these low-risk VTEs the risk reduction of VTE and risk elevation of ICH may be comparable (15). However, those neurosurgical patients with confirmed VTE are more likely to suffer demise from pulmonary embolism (PE) rather than hemorrhage due to anticoagulation, so the decision to anticoagulate in this population is slightly clearer (9). Rates of VTE and the benefit of aggressive prophylaxis protocols have been explored and proven beneficial in numerous subsets of neurosurgery patients, including post spinal surgery and in patients undergoing craniotomy for tumor (8, 9, 12). In the craniotomy for tumor subpopulation, risk factors of VTE have been identified using both national and institutional data (24, 35). Thus identification of risk factors of VTE among all craniotomies would allow anticoagulation to be targeted toward the most at-risk patient populations. The American College of Surgeons National Surgical Quality Improvement Project (NSQIP) (http://site.acsnsqip.org/) database provides a large number of craniotomy cases, which may elucidate populations at risk for VTE (18). Data recorded include comorbid conditions, preoperative, operative, and postoperative complications. We previously analyzed 2006e2010 NSQIP craniotomy data to elucidate factors associated with VTE and identify patients most likely to benefit from prophylactic anticoagulation. In fact, 2011e2012 NSQIP data are the most recent NSQIP data sets currently available and were not available at the time of our previous analysis. By analyzing 2011e2012 NSQIP data, we hope to validate our previous model and refine the risk score but also extend the model with additional analyses (23). A total of 4844 patients were included in NSQIP data from 2006e2010, whereas over 2011e2012 there were 10477 records. Thus identifying factors preserved across multiple years bolsters the analysis and increases the strength of our model.
2
www.SCIENCEDIRECT.com
METHODS Baseline Univariate Analysis A total of 10,477 patients from 2011e2012 underwent craniotomy on the basis of primary Current Procedural Terminology (CPT) procedure code. Ordinal and nominal variables were subjected to univariate two-tailed c2 analysis and Fisher’s exact test when expected frequencies were low to identify clinical factors associated with VTE. All of the factors in the ACS-NSQIP data set were subjected to analysis. For univariate c2 analysis, original a was set at 0.05 and P values were adjusted using Holm-Bonferonni correction. Those factors less than their respective adjusted a were considered statistically significant, and odds ratios (ORs) were calculated when appropriate. Procedures performed for a tumor diagnosis were also identified on the basis of primary CPT code (61510, 61512, 61518, 61519, 61520, 61521, 61526, 61545, 61546, 61575). Many factors were converted to binary variables to assist in calculation of ORs. Preoperative and intraoperative factors with P values <0.05 were subjected to a forward, stepwise, multivariate binary logistical regression analysis (entry level ¼ 0.05, exit ¼ 0.1) for the identification of factors independently correlated with VTE. Covariate interactions were analyzed with chi-square analysis, noted, and included stepwise in iterations of multivariate regression but failed to impact the robustness of the model. Similarly, postoperative factors were subjected to multivariate logistic regression with care taken to exclude patients in which the event of interest occurred after VTE. Again, covariate interactions were analyzed and included in multivariate regression but failed to impact the robustness of the model. Thirteen factors were included in the final multivariate model and were converted to a binary score of 0 or 1 with 1 representing presence of the factor. Goodness of fit of the multivariate logistic regression model was assessed using Hosmer-Lemeshow test where P > 0.05 indicates a model that fits the data well. The VTE risk score was compared with VTE rate, 30day mortality, and days from surgery to discharge using a number of statistical tests including both linear and Cox regression. A receiver operating characteristics (ROC) curve was used to assess predictive strength of VTE risk scores on the basis of reported area under the curve (AUC). AUC values greater than 0.70 represent acceptable discrimination, and values greater than 0.80 indicate excellent discrimination (7, 16, 32). For validation of the present analysis against past data, the cohort of 4844 patients undergoing craniotomy from 2006e2010 NSQIP data was used. The Cox proportional hazards regression model was an additional analysis conducted to examine the association of clinical factors with VTE in regard to time from surgery to discharge. All statistical analyses were performed using SPSS software version 18.0 (IBM, Armonk, New York, USA).
RESULTS The rate of VTE in the cohort of 10,477 patients was 3.2% (PE was 1.3%; DVT was 2.4%). There were 131 patients in the cohort with a PE, of which 77 patients did not have an associated DVT (58.8%). VTE, DVT, and PE were all significantly associated with increased days from operation to discharge (P < 0.001). Additionally, VTE, DVT, and PE were all significantly associated with increased
WORLD NEUROSURGERY, http://dx.doi.org/10.1016/j.wneu.2015.06.033
ORIGINAL ARTICLE HANNA ALGATTAS ET AL.
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
30-day mortality (P < 0.001 for VTE and DVT; P ¼ 0.001 for PE) (Table 1). Many factors in the NSQIP data set were significantly associated with VTE events in univariate analysis and persisted after HolmBonferroni correction (Table 2). Preoperative factors included ventilator dependence (patient required ventilator at any time within 48 hours before surgery), nonelective surgery, estimated probability of mortality or morbidity (NSQIP-derived value based on hierarchical regression analysis of preoperative factors), time from admission to operating room greater than 4 days, emergency case, dependent functional status (based on completion of activities of daily living), age >60 years old, hospital transfer from an acute care facility, hemiplegia, steroid use (regular use within 30 days before surgery), inpatient procedure, body mass index greater than 30, African-American race, and hypertension (requiring medications within 30 days before surgery). Intraoperative factors included ASA scoring of 4 (patient with severe systemic disease that is a constant threat to life) and 5 (moribund patient who is not expected to survive without the operation). Postoperative factors included ventilator dependence greater than 48 hours cumulatively, unplanned reintubation within 30 days, return to operating room within 30 days, infection (included pneumonia, urinary tract infection, sepsis, and septic shock within 30 days), bleeding transfusion (at least unit of red blood cells given from time of surgery start to 72 hours postoperatively), cerebrovascular accident, impaired sensorium (acute confusion or delirium), cardiac arrest requiring CPR, progressive renal insufficiency (rise in creatinine >2 mg/dL from preoperative values), and coma greater than 24 hours. All these factors were subjected to multivariate binary regression analysis. The 13 factors that persisted were then incorporated into a VTE risk score (Table 3). Modifiable preoperative risk factors included steroid use, ventilator use, and time from admission to surgery greater than 4 days. Preoperative factors that are less modifiable to not at all included nonelective cases, BMI >30, age >60, African-American race, inpatient status, and impaired sensorium. Intraoperative and postoperative factors were American Society of Anesthesiologists (ASA) class of 4 or greater, cumulative postoperative ventilation greater than 48 hours, return to the operating room, and infection. The postoperative infection category was a combination of pneumonia, urinary tract infection (UTI), sepsis, and septic shock. The Hosmer-Lemeshow test was used to assess goodness of fit of the multivariate logistic
Table 1. Venous Thromboembolism (VTE) Events and Mortality Rate Group
Total Patients (%)
30-Day Mortality Rate (%)
Overall cohort
10447
4.8
No VTE
10114
4.6
VTE
333 (3.2)
9.6
DVT
246 (2.4)
10.2
PE
131 (1.3)
10.7
DVT, deep vein thrombosis; PE, pulmonary embolism.
WORLD NEUROSURGERY - [-]: ---, - 2015
Table 2. Factors Significant After Univariate Analysis Preoperative
Intraoperative
Postoperative
Ventilator dependence
ASA Class 4e5 Ventilator dependence >48 hours
Nonelective surgery
Unplanned reintubation
Estimated probability of mortality >10%
Infection (PNA, UTI, sepsis, septic shock)
Estimated probability of morbidity >10%
Return to OR
Admission to OR time >4 days
Bleeding transfusion
Emergency case
CVA
Dependent functional status
Impaired sensorium
Age >60
Cardiac arrest requiring CPR
Transfer from acute care facility
Progressive renal insufficiency
Hemiplegia
Coma >24 hours
Steroid use Inpatient BMI >30 African-American race Hypertension ASA, American Society of Anesthesiologists; PNA, pneumonia; UTI, urinary tract infection; OR, operating room; CVA, cerebrovascular accident; CPR, cardiopulmonary resuscitation; BMI, body mass index.
regression model. At each forward step of the regression, the model fit the data (P > 0.05). Each factor was treated as a binary variable, where the presence of the factor was scored as 1 point and the absence as 0 points. Thus with 13 factors included, scores may theoretically range from 0e13. In actuality, scores ranged from 0e11 (median ¼ 2). Increasing VTE score was predictive of VTE incidence, with a rate of 0.0% for a score of 0 and 15.7% for a score 7 (Figure 1; Table 4). The VTE risk score was significantly associated with VTE rate, increasing time from operation to discharge, and mortality (P < 0.001; see Table 4). A receiver operating characteristics (ROC) curve examined the ability for VTE risk score to predict VTE and demonstrated an area under the curve of 0.719 (95% confidence interval [95% CI] 0.691e0.747; P < 0.001). To determine the risk of VTE as it relates to time from operation to discharge, a Cox proportional hazards model was completed. Each variable incorporated into the VTE risk score was analyzed independently, and alphas were adjusted using Holm-Bonferonni correction. Factors displaying significant hazard ratios were steroid use, age >60, elective surgery, and ASA class 4e5 (Table 5). Our previous analysis of NSQIP data from 2006e2010 identified similar factors linked with VTE (23). However, some factors associated with VTE in 2006e2010 data were not associated in the present analysis. Preoperative factors that were consistent among years after univariate and multivariate binary regression
www.WORLDNEUROSURGERY.org
3
ORIGINAL ARTICLE HANNA ALGATTAS ET AL.
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
Table 3. Factors Significant After Multivariate Analysis Factor
Odds Ratio (95% CI)
Ventilator dependence
3.13 (2.23e4.38)
VTE Number of VTE DVT PE Mean Days Mean 30-Day Risk Patients (% Rate Rate Rate from Surgery to Mortality Rate Score Total) (%) (%) (%) Discharge* (n [%])*
Nonelective surgery
2.04 (1.64e2.54)
0
0.4
1 (0.2)
BMI >30
1.51 (1.21e1.88)
1
2010 (20.5) 0.7* 0.4* 0.3*
3.3
5 (1.0)
Age >60
1.61 (1.30e2.01)
2
3117 (29.8) 1.9* 1.1* 1.1
4.1
26 (5.2)
Steroid use
1.76 (1.35e2.31)
3
2419 (23.2) 2.6
1.2
5.3
67 (13.5)
African-American race
1.85 (1.33e2.58)
4
1379 (13.2) 4.4* 3.3* 1.7
7.6
100 (20.1)
Inpatient
9.13 (2.27e36.78)
5
708 (6.8)
6.6* 4.9* 2.5*
11.2
96 (19.3)
Impaired sensorium
2.57 (1.78e3.72)
6
334 (3.2)
9.6* 7.8* 2.1
12.6
82 (16.5)
Admission to OR >4 days
2.00 (1.50e2.66)
7
351 (1.9)
15.7* 13.7* 3.1*
16.9
121 (24.3)
Overall 10447 number
5.8
498
2.37 (1.87e3.02)
Preoperative
Intraoperative ASA 4e5 Postoperative On ventilator >48 hours
5.89 (4.55e7.63)
Infection
5.40 (4.04e7.22)
Return to OR
4.64 (3.53e6.10)
CI, confidence interval; BMI, body mass index; OR, operating room; ASA, American Society of Anesthesiologists.
included age older than 60 and nonelective surgery/emergency case. Postoperative factors preserved across the data sets were infections, ventilator dependence for greater than 48 hours, and return to the operating room. These five factors were incorporated into their own VTE risk score, which was also predictive of VTE incidence, time from surgery to discharge, and mortality (P < 0.001) (Table 6; Figure 2). When applied to combined 2006e2012 data, the risk score of conserved factors was predictive of VTE in an ROC analysis (AUC ¼ 0.693; 95% CI 0.662e0.723, P < 0.001). Hosmer-Lemeshow goodness of fit
Figure 1. Rates of venous thromboembolism, deep vein thrombosis, and pulmonary embolism at each risk score.
4
Table 4. Risk Scores and Associated Venous Thromboembolism Event (VTE) Rates, Days to Discharge, and Mortality
www.SCIENCEDIRECT.com
129 (1.2)
0.0* 0.0
1.9
333* 246
0.0
131
DVT, deep vein thrombosis; PE, pulmonary embolism. *Statistically significant data (P < 0.05).
at the last step of the multivariate regression demonstrated a good fit for both 2006e2010 and 2011e2012 data sets (P ¼ 0.094 and P ¼ 0.272, respectively).
Table 5. Cox Proportional Hazards of Venous Thromboembolism Risk Score Factors Pertaining to Time from Operation to Discharge Factor
P Value
HR (95% CI)
Steroid use
< 0.001 1.89 (1.45e2.46) Yes
Age >60
0.001 1.43 (1.15e1.78) Yes
Nonelective surgery
0.003 1.41 (1.12e1.75) Yes
ASA 4e5
0.003 0.69 (0.54e0.88) Yes
Preoperative Ventilator dependence
0.009 0.64 (0.46e0.90) No
BMI >30
0.019 1.30 (1.04e1.61) No
Postoperative ventilator >48 hours
0.058 0.77 (0.59e1.01) No
Return to OR
0.589 0.93 (0.88e1.06) No
African-American race
0.667 1.08 (0.77e1.49) No
Inpatient
0.683 1.34 (0.33e5.37) No
Impaired sensorium
0.797 0.95 (0.67e1.37) No
Admit to OR >4 days
0.946 0.99 (0.75e1.31) No
Infection
0.983 1.00 (0.75e1.33) No
Significance Postcorrection
HR, hazard ratio; CI, confidence interval; ASA, American Society of Anesthesiologists; OR, operating room.
WORLD NEUROSURGERY, http://dx.doi.org/10.1016/j.wneu.2015.06.033
ORIGINAL ARTICLE
498 240 5.8
6.0 131 (1.3) 69 (1.4) 246 (2.4) 111 (2.3) 331 (3.2) 170 (3.5) 10,447 4844 Totals
WORLD NEUROSURGERY - [-]: ---, - 2015
DVT, deep vein thrombosis; PE, pulmonary embolism. *Statistically significant (P < 0.05). All values in column significant with mark at column head. yMean days from surgery to discharge.
69 (32.7)
y y
97 (24.1) 66 (25.5)
34 (25.2) 17.6
16.3 17.6
24.2 7 (3.3)*
15 (3.7)* 9 (3.5)*
7 (5.2)* 23 (10.9)
52 (12.9) 21 (8.1)*
15 (11.1)* 28 (13.3)
59 (14.6) 28 (10.8)*
20 (14.8)*
259 (5.3)
135 (2.8)
3
403 (3.9)
DISCUSSION
4
211 (2.0)
187 (10.5) 77 (15.1) 8.6 10.4 23 (1.3) 18 (3.5)* 68 (3.8) 18 (3.5)* 83 (4.7) 1781 (17.0) 509 (10.5) 2
37 (7.3)*
28 (0.7) 7 (0.3)
56 (3.1) 5.0
3.8 3.8
5.4 54 (1.3)
32 (0.8)* 15 (0.7)*
20 (1.1) 75 (1.8)
28 (0.7) 19 (0.9)*
38 (2.1) 107 (2.6)
54 (1.4) 31 (1.4)*
54 (3.0) 4155 (39.8)
2158 (44.5)
1783 (36.8)
0
3897 (37.3)
117 (2.8)
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
1
11e12* 06e10* 11e12* 06e10* 11e12 06e10 11e12* 06e10 11e12* 06e10 11e12 06e10 VTE Risk Score
DVT Rate (%) VTE Rate (%) Number of Patients (% Total)
Table 6. Venous Thromboembolism (VTE) Risk Score Conserved Across 2006e2012 Data (23)
PE Rate (%)
Mean Days from Surgery to Discharge
Mean 30-DayMortality Rate (n [%])
HANNA ALGATTAS ET AL.
We chose to quantify VTE risk on the basis of a number of patient factors provided by the NSQIP database to identify a subset of patients at highest risk for VTE and in whom the benefit of anticoagulation exceeds the risk of hemorrhage. Our data reveal a set of factors highly predictive of VTE, which may aid in distinguishing those patients at increased risk for VTE and for whom the risk-benefit ratio favors the use of pharmacologic prophylaxis. Increasing VTE risk score was significantly associated with VTE incidence, time to discharge, and mortality. The AUC utilizing factors isolated from multivariate analysis of 2011e2012 data demonstrated a predictive model showing acceptable discrimination (16). When validating our 2006e2010 analysis with the present findings, approximately half of the associated factors persisted but still effectively discriminated cases of VTE from the present sample. Factors associated with VTE from the previous analysis of 2006e2010 NSQIP craniotomy cases were transfer from an acute care hospital, preoperative sepsis, emergency cases, dependent functional status, age older than 60, tumor involving central nervous system, craniotomy for tumor, operation time greater than 4 hours, pneumonia, UTI, return to operating room, and postoperative ventilator dependence for >48 hours (23). When comparing the present results with our previous analysis of 2006e2010 data, about half of the factors persisted. Notably, all of the postoperative factors were represented in both 2006e2010 and 2011e2012 data sets. In addition, postoperative factors demonstrated the highest ORs compared with preoperative and intraoperative factors. A VTE risk score with only these factors performed similarly in ROC analysis and suggests that postoperative, relative to preoperative and intraoperative, factors are most predictive of VTE. However, postoperative factors are less modifiable and more difficult to identify early with the aim of preventing VTE. To note, although 2011e2012 data were not available during completion of our 2006e2010 analysis, the present study remains novel. Novel factors are introduced as associated with VTE, despite similar rates of VTE in both cohorts. The present analysis was corrected for multiple statistical comparisons using the HolmBonferonni correction method, increasing the reliability of the current data, and AUC values were examined using HosmerLemeshow goodness of fit. Introduction of both statistical methods increases the reliability of the current data. Also novel, a Cox proportional hazards regression model was used to determine if the association of risk factors with VTE changed with regard to time from surgery to discharge. Steroid use, age >60, nonelective surgery, and ASA class 4e5 all provided significant hazard ratios. The results demonstrate these factors increase risk for VTE for each unit increase in time after surgery. Care must be taken in interpreting such results because some may have effects on total LOS alone. The introduction of time into the risk model could be interesting because it appears some factors lend to increasing risk with longer time spent in the hospital. NSQIP has been used across a broad array of surgical specialties (13, 21, 28, 30). For DVT to be counted in NSQIP, the clot requires either anticoagulation or vena cava filter placement and also diagnosis by either duplex, venogram, or computed tomography scan; thrombosis must occur along the vein and not only the
www.WORLDNEUROSURGERY.org
5
ORIGINAL ARTICLE HANNA ALGATTAS ET AL.
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
Figure 2. Venous thromboembolism (VTE) rates across VTE risk score between 2006 and 2010 and 2011 and 2012 data (23).
catheter (2). PE has similar requirements per NSQIP and must be demonstrated with either a highly probable V-Q scan or definitively with another imaging modality (2). Similar approaches to data collection, risk stratification, and statistical analysis using NSQIP have been taken elsewhere (17). Pharmacologic VTE prophylaxis in certain specialties has more clear clinical utility and has been welcomed, but, in the setting of neurosurgery, a complex balancing act between complications on opposing sides of the spectrum remains. We identified the overall VTE incidence as 3.2% (DVT ¼ 2.4%, PE ¼ 1.3%); importantly, NSQIP only reports symptomatic VTEs, which may underestimate total VTE rates. Rates of VTE among neurosurgical cohorts vary among studies. One study identified DVT risk post neurosurgery to range from 9%e16% and PE as 4%; however, that study contained a small cohort of patients deemed “high risk” for DVT depending on degree of mechanical or pharmacologic prophylaxis (22). Our previous study using 2006e2010 NSQIP craniotomy data identified the DVT and PE rates as 2.6% and 1.4%, respectively (23). Of note, rate of PE was reduced at VTE risk scores of six (2.1%) compared with scores of 5 (2.5%) (see Table 4). Rationale behind this trend is the low amount of PEs reported, which only occurred in 131 cases (1.3%). At risk scores 7, however, rates of PE increase to 3.1%, a statistically significant finding. Separately, the 30-day mortality rate for patients with VTE post craniotomy was 9.6%, which compares favorably with existing literature regarding mortality rate of patients with VTE after general and vascular surgery (11.19%) (33). Anticoagulation is proven to reduce VTE events in a number of trials; Hamilton and colleagues (15) analyzed 6 RCTs and demonstrated a 42% reduction in symptomatic and asymptomatic VTE with heparin. Pooled results for hemorrhage were statistically insignificant, except for more minor bleeds, but the authors concede hemorrhage is still “clinically significant.” Both symptomatic and asymptomatic VTEs were documented in those studies, which begs the question if asymptomatic VTE are comparable with risk of hemorrhage; comparing symptomatic VTE to hemorrhage events was roughly similar in their study (15). Rates of hemorrhage in neurosurgery
6
www.SCIENCEDIRECT.com
patients have been cited as 2.1% in those receiving lowmolecular-weight heparin and 1.1% in those without anticoagulation (19). In a study using D-dimer screening post craniotomy, expectedly, the group screened regularly was diagnosed with VTE more often than a nonscreened group (7.0% vs. 3.7%, respectively). However, the screened group suffered more hemorrhagic complications compared with the nonscreened group (2.2% vs. 0.5%, respectively). According to the authors, the increased VTE events among the screened group likely were asymptomatic, nonelife-threatening VTEs arising distally and the risk of these VTEs must be carefully weighed with the hemorrhagic risk (1). Thus overtreating asymptomatic VTE becomes a growing concern (31). Large and more robust analyses are needed to identify true risk for secondary hemorrhage. The identification of risk factors associated with reoperation for hemorrhage may present an opportunity to balance the risk of thrombosis and hemorrhage in perioperative craniotomy patients. However, increased screening is not necessarily the answer to reducing VTE risk. In fact, hospitals with high-quality ratings often have increased risk-adjusted VTE rates even in the setting of increased prophylaxis rates. This paradox is associated with increased VTE imaging rates at said hospitals, revealing a significant surveillance bias caused by increased VTE screening (3). To improve the use of VTE as a quality measure, imaging rates and the capture of symptomatic versus asymptomatic VTE should be considered. Many of the identified factors were consistent with our original hypotheses. Ventilator dependence, postoperative infections, and return to the operating room are all factors that presumably lengthen hospital stay and reduce mobility, increasing the likelihood of VTE. Additionally, inflammation has been linked to hypercoagulable states and may be related to cytokine and interleukin-mediated changes (26, 27). Along with those chemical imbalances, central hypovolemia has been associated with increased coagulation (27), which is interesting in respect to perioperative care, especially those cases where significant blood loss becomes an issue. Impaired sensorium was strongly associated with VTE relative to other preoperative factors. Gajdos and colleagues (11) identified impaired sensorium to be associated with increased risk for 22 out of 23 postoperative complications recorded in NSQIP, including pneumonia, ventilator dependence, and DVT. Patients with impaired sensorium are less likely to appropriately follow an aggressive pulmonary toilet regimen or ambulate and are more likely to require urinary catheterization, making them susceptible to pneumonia, ventilator dependence, VTE, and UTI among other complications (11). Many of the factors associated with VTE postcraniotomy in the present analysis have been identified elsewhere and among different and similar types of procedures, including craniotomy for tumor but also in surgery for inflammatory bowel disease (1, 22, 24, 34-36). In the current study, covariate interactions failed to significantly impact the multivariate model; however, association between factors should be taken into consideration when interpreting the findings. Postoperative infections including pneumonia were associated with VTE. Several preoperative factors are predictive of pneumonia including functional status, sepsis, lung disease, smoking, and ASA classification (14). Notably, functional
WORLD NEUROSURGERY, http://dx.doi.org/10.1016/j.wneu.2015.06.033
ORIGINAL ARTICLE HANNA ALGATTAS ET AL.
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
status and ASA class were associated with VTE after univariate and multivariate analysis, respectively. There is also concern regarding duplicity of factors isolated after multivariate analysis; inpatient status and nonelective cases are one such link. Inherently, inpatients are more likely to have nonelective (i.e., emergency) cases. However, the association between VTE and inpatients was the strongest on the basis of OR (OR ¼ 9.13) and remains a warranted measure. Stemming from this idea, the possibility of weighting the VTE risk score based on OR was thoroughly explored because different factors, especially postoperative events, varied in strength of association to VTE. Multiple attempts at weighting the risk score were examined, but they only marginally improved ROC discrimination and thus were not presented. The notion of a VTE risk score is not novel, being introduced by pioneers such as Dr. Joseph Caprini, and has been used to varying extents across surgical specialties (5). Cassidy et al. created a mandatory electronic VTE risk assessment, which needed to be completed by surgeons before placing an order. Institutional NSQIP data were analyzed before and after the program was put in place, and the group realized significant reductions in DVT (1.9%e0.3%) and PE (1.1%e 0.5%). However, there was no mention of bleeding episodes possibly introduced by more rigorous VTE prophylaxis (6). Factors included in our VTE score were compared with those found in previous analysis of 2006e2010 craniotomy cases to determine coinciding factors with stronger association to VTE (23). The presented risk model is imperfect on the basis of ROC analysis and also due to the overall rarity of VTE, especially PE events; however, stratification of patients based on score does represent a clinically relevant and significant distinction. The present analysis and existing NSQIP-focused research consistently identify preoperative factors that offer a large contribution to the incidence of hospital-acquired conditions (HAC). Because many preoperative factors are nonmodifiable, Molena and colleagues (29) suggest nonpayment policies implemented by the Centers for Medicare and Medicaid Services (CMS) may be excessively harsh. This argument has been echoed in surgical oncology and bariatric surgery (25, 29). The present findings agree with these statements. Interestingly, factors with the strongest association to VTE based on OR were the postoperative factors (infection, return to OR, ventilation >48 hours). However,
REFERENCES 1. Aishima K, Yoshimoto Y: Screening strategy using sequential serum D-dimer assay for the detection and prevention of venous thromboembolism after elective brain tumor surgery. Br J Neurosurg 27: 348-354, 2013. 2. American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP): User guide for the 2012 ACS NSQIP Participant Use Data File. Available at: www.acsnsqip.org 3. Bilimoria KY, Chung J, Ju MH, Haut ER, Bentrem DJ, Ko CY, Baker DW: Evaluation of surveillance bias and the validity of the venous thromboembolism quality measure. JAMA 310: 1482-1489, 2013.
across studies, these findings need to be carefully reviewed to ensure postoperative events occurring after HACs are eliminated. Our analysis is not without limitation, including its retrospective design and generalizability to participating hospitals, which are often large academic centers. Additionally, although anticoagulation or vena cava filter placement is required for DVT or PE events per NSQIP, NSQIP does not record specific prophylaxis or ensuing treatments, which may skew data. Although NSQIP represents a growing and robust surgical database, one issue is the possibility of random sampling error. When compared with the Thoracic Morbidity and Mortality System (TM&M) surgical database system, NSQIP captured 21.1% of patients included in TM&M (20). With the capture of fewer patients came significant differences in the rates of some procedures (20). In a study examining pancreatectomy NSQIP data, discordance between NSQIP-reported data and reviewed cases was 27.3%. Certain events, such as postoperative bleeding requiring transfusion, represent common and often misclassified outcomes (10). However, comparing the more robust results of the 2011e2012 analysis with previously acquired 2006e2010 data significantly strengthens the model and identifies factors consistently associated with VTE. Using those factors for VTE risk modeling in additional non-NSQIP data sets would be useful in bringing a clinical score to practice. Thus future studies should be aimed at validating the risk score in a separate patient population. In fact, with a more refined VTE risk model, additional studies that balance risk of VTE with risk of hemorrhage can be combined to determine the risk of two separate complications on opposing sides of the coagulation spectrum.
CONCLUSIONS A number of factors have been associated with VTE events in postcraniotomy patients, and a subset of these have been validated across our previous data set, mostly postoperative factors. The development of a VTE risk score successfully predicted VTE incidence, increased time to discharge, and mortality. The results suggest a subset of postcraniotomy patients who are more likely to benefit from prophylactic anticoagulation postoperatively may be identified.
4. Camden R, Ludwig S: Prophylaxis against venous thromboembolism in hospitalized medically ill patients: update and practical approach. Am J Health Syst Pharm 71:909-917, 2014. 5. Caprini JA, Arcelus JI, Hasty JH, Tamhane AC, Fabrega F: Clinical assessment of venous thromboembolic risk in surgical patients. Semin Thromb Hemost 17 (Suppl 3):304-312, 1991. 6. Cassidy MR, Rosenkranz P, McAneny D: Reducing postoperative venous thromboembolism complications with a standardized riskstratified prophylaxis protocol and mobilization program. J Am Coll Surg 218:1095-1104, 2014. 7. Cohen ME, Dimick JB, Bilimoria KY, Ko CY, Richards K, Hall BL: Risk adjustment in the American College of Surgeons National Surgical Quality Improvement Program: a comparison of
WORLD NEUROSURGERY - [-]: ---, - 2015
logistic versus hierarchical modeling. J Am Coll Surg 209:687-693, 2009. 8. Collen JF, Jackson JL, Shorr AF, Moores LK: Prevention of venous thromboembolism in neurosurgery: a metaanalysis. Chest 134:237-249, 2008. 9. Cote LP, Greenberg S, Caprini JA, Stone J, Arcelus JI, Lopez-Jimenez L, Rosa V, Schellong S, Monreal M: Outcomes in neurosurgical patients who develop venous thromboembolism: a review of the RIETE registry. Clin Appl Thromb Hemost 20:772-778, 2014. 10. Epelboym I, Gawlas I, Lee JA, Schrope B, Chabot JA, Allendorf JD: Limitations of ACSNSQIP in reporting complications for patients undergoing pancreatectomy: underscoring the need for a pancreas-specific module. World J Surg 38:1461-1467, 2014.
www.WORLDNEUROSURGERY.org
7
ORIGINAL ARTICLE HANNA ALGATTAS ET AL.
11. Gajdos C, Kile D, Hawn MT, Finlayson E, Henderson WG, Robinson TN: The significance of preoperative impaired sensorium on surgical outcomes in nonemergent general surgical operations. JAMA Surg 150:30-36, 2015.
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
21. Kester BS, Merkow RP, Ju MH, Peabody TD, Bentrem DJ, Ko CY, Bilimoria KY: Effect of postdischarge venous thromboembolism on hospital quality comparisons following hip and knee arthroplasty. J Bone Joint Surg Am 96:1476-1484, 2014.
12. Goldhaber SZ, Dunn K, Gerhard-Herman M, Park JK, Black PM: Low rate of venous thromboembolism after craniotomy for brain tumor using multimodality prophylaxis. Chest 122:1933-1937, 2002.
22. Khaldi A, Helo N, Schneck MJ, Origitano TC: Venous thromboembolism: deep venous thrombosis and pulmonary embolism in a neurosurgical population. J Neurosurg 114:40-46, 2011.
13. Grant DW, Mlodinow A, Ver Halen JP, Kim JY: Catastrophic outcomes in free tissue transfer: a six-year review of the NSQIP database. Plast Surg Int 2014:704206, 2014.
23. Kimmell KT, Jahromi BS: Clinical factors associated with venous thromboembolism risk in patients undergoing craniotomy. J Neurosurg 122: 1004-1011, 2015.
14. Gupta H, Gupta PK, Schuller D, Fang X, Miller WJ, Modrykamien A, Wichman TO, Morrow LE: Development and validation of a risk calculator for predicting postoperative pneumonia. Mayo Clin Proc 88:1241-1249, 2013.
24. Kimmell KT, Walter KA: Risk factors for venous thromboembolism in patients undergoing craniotomy for neoplastic disease. J Neurooncol 120: 567-573, 2014.
15. Hamilton MG, Yee WH, Hull RD, Ghali WA: Venous thromboembolism prophylaxis in patients undergoing cranial neurosurgery: a systematic review and meta-analysis. Neurosurgery 68:571-581, 2011. 16. Hosmer DW, Lemeshow S: Applied Logistic Regression. 2nd ed. New York: John Wiley & Sons; 2000. 17. Hyder JA, Wakeam E, Habermann EB, Hess EP, Cima RR, Nguyen LL: Derivation and validation of a simple calculator to predict home discharge after surgery. J Am Coll Surg 218:226-236, 2014. 18. Ingraham AM, Richards KE, Hall BL, Ko CY: Quality improvement in surgery: the American College of Surgeons National Surgical Quality Improvement Program approach. Adv Surg 44: 251-267, 2010. 19. Iorio A, Agnelli G: Low-molecular-weight and unfractionated heparin for prevention of venous thromboembolism in neurosurgery: a meta-analysis. Arch Intern Med 160:2327-2332, 2000. 20. Ivanovic J, Seely AJ, Anstee C, Villeneuve PJ, Gilbert S, Maziak DE, Shamji FM, Forster AJ, Sundaresan RS: Measuring surgical quality: comparison of postoperative adverse events with the American College of Surgeons NSQIP and the Thoracic Morbidity and Mortality classification system. J Am Coll Surg 218:1024-1031, 2014.
8
www.SCIENCEDIRECT.com
25. Lidor AO, Moran-Atkin E, Stem Magnuson TH, Steele KE, Feinberg Schweitzer MA: Hospital-acquired conditions ter bariatric surgery: we can predict, but can prevent? Surg Endosc 28:3285-3292, 2014.
M, R, afwe
26. Lipinski S, Bremer L, Lammers T, Thieme F, Schreiber S, Rosenstiel P: Coagulation and inflammation. Molecular insights and diagnostic implications. Hamostaseologie 31:94-102; 104, 2011. 27. Meyer MA, Ostrowski SR, Overgaard A, Ganio MS, Secher NH, Crandall CG, Johansson PI: Hypercoagulability in response to elevated body temperature and central hypovolemia. J Surg Res 185:e93-e100, 2013. 28. Moghadamyeghaneh Z, Hanna MH, Carmichael JC, Nguyen NT, Stamos MJ: A nationwide analysis of postoperative deep vein thrombosis and pulmonary embolism in colon and rectal surgery. J Gastrointest Surg 18: 2169-2177, 2014. 29. Molena D, Mungo B, Stem M, Feinberg RL, Lidor AO: Prevalence, impact, and risk factors for hospital-acquired conditions after major surgical resection for cancer: a NSQIP analysis. J Gastrointest Surg 19:142-151; discussion 151, 2015. 30. Orr NT, Davenport DL, Roth JS: Outcomes of simultaneous laparoscopic cholecystectomy and ventral hernia repair compared to that of
laparoscopic cholecystectomy alone. Surg Endosc 27:67-73, 2013. 31. Prell J, Rachinger J, Smaczny R, Taute BM, Rampp S, Illert J, Koman G, Marquart C, Rachinger A, Simmermacher S, Alfieri A, Scheller C, Strauss C: D-dimer plasma level: a reliable marker for venous thromboembolism after elective craniotomy. J Neurosurg 119:1340-1346, 2013. 32. Ramanan B, Gupta PK, Gupta H, Fang X, Forse RA: Development and validation of a bariatric surgery mortality risk calculator. J Am Coll Surg 214:892-900, 2012. 33. Rogers SO Jr, Kilaru RK, Hosokawa P, Henderson WG, Zinner MJ, Khuri SF: Multivariable predictors of postoperative venous thromboembolic events after general and vascular surgery: results from the patient safety in surgery study. J Am Coll Surg 204:1211-1221, 2007. 34. Rolston JD, Han SJ, Bloch O, Parsa AT: What clinical factors predict the incidence of deep venous thrombosis and pulmonary embolism in neurosurgical patients? J Neurosurg 121:908-918, 2014. 35. Smith TR, Nanney AD 3rd, Lall RR, Graham RB, McClendon J Jr, Lall RR, Adel JG, Zakarija A, Cote DJ, Chandler JP: Development of venous thromboembolism (VTE) in patients undergoing surgery for brain tumors: results from a single center over a 10-year period. J Clin Neurosci 22: 519-525, 2015. 36. Wallaert JB, De Martino RR, Marsicovetere PS, Goodney PP, Finlayson SR, Murray JJ, Holubar SD: Venous thromboembolism after surgery for inflammatory bowel disease: are there modifiable risk factors? Data from ACS NSQIP. Dis Colon Rectum 55:1138-1144, 2012. Conflict of interest statement: The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper. Received 1 June 2015; accepted 16 June 2015 Citation: World Neurosurg. (2015). http://dx.doi.org/10.1016/j.wneu.2015.06.033 Journal homepage: www.WORLDNEUROSURGERY.org Available online: www.sciencedirect.com 1878-8750/$ - see front matter ª 2015 Elsevier Inc. All rights reserved.
WORLD NEUROSURGERY, http://dx.doi.org/10.1016/j.wneu.2015.06.033
Analysis of Venous Thromboembolism Risk in Patients Undergoing Craniotomy Hanna Algattas, Kristopher T. Kimmell, G. Edward Vates, Babak S. Jahromi
OBJECTIVES: Craniotomy poses a risk for postoperative venous thromboembolism (VTE), but the utility of anticoagulation in this patient population is unclear. We sought to identify risk factors predictive of VTE in patients undergoing craniotomy.
-
METHODS: The American College of Surgeons National Surgical Quality Improvement Project (ACS-NSQIP) database was reviewed for patients undergoing craniotomy. Clinical factors provided by the database were analyzed for association with VTE.
-
CONCLUSIONS: The risk of postoperative VTE after craniotomy can be quantified by a simple risk score, with increasing risk factors conferring increased risk of VTE. On the basis of risk scoring, a subset of patients who would benefit from anticoagulation post craniotomy may be identified.
-
INTRODUCTION
M
-
RESULTS: A total of 10,477 adult patients who underwent craniotomy from 2011e2012 were identified. The rate of VTE was 3.2% (pulmonary embolism [PE] was 1.3%; deep vein thrombosis [DVT] was 2.4%). Several factors were significant in univariate analysis, and a subset persisted after multivariate analysis. Patients were assigned a risk score on the basis of the presence of those variables. Higher risk scores were predictive of VTE risk, as well as increasing time from surgery to discharge and mortality. A receiver operating characteristics curve revealed a significant area under the curve (0.719) for scores being predictive of VTE risk. The model was validated against our similar analysis of 2006e2010 NSQIP data and demonstrated comparable findings.
odern medicine has largely succeeded in treating and extending the course of a number of chronic and degenerative conditions. The current era of medicine has shifted focus more from tertiary prevention to primary prevention and risk reduction. Pursuant with this focus, stakeholders seek to reduce expenses by limiting preventable causes of hospitalization and prolonged admissions. One arena for high-yield prevention and cost reduction is the perioperative period. The Surgical Care Improvement Project (SCIP) was started in 2006 with assistance from the Joint Commission with the goal of diminishing preventable complications before, after, and within the operative window. The instillment of recommendations regarding appropriate antibiotic and medication use, postoperative blood glucose measures, surgical site hair removal, urinary catheterization, temperature control, and prevention of venous
Key words Craniotomy - Venous thromboembolism
PE: Pulmonary embolism PNA: Pneumonia RCT: Randomized controlled trials ROC: Receiver-operating characteristics SCIP: Surgical Care Improvement Project TM&M: Thoracic Morbidity and Mortality System UTI: Urinary tract infection V-Q: Ventilation-perfusion VTE: Venous thromboembolism
-
Abbreviations and Acronyms ACS-NSQIP: American College of Surgeons National Surgical Quality Improvement Project AHRQ: Agency for Healthcare Research and Quality ASA: American Society for Anesthesiologists AUC: Area under the curve CMS: Center for Medicare and Medicaid Services CPR: Cardiopulmonary resuscitation CPT: Current Procedural Terminology CVA: Cerebrovascular accident DVT: Deep vein thrombosis HAC: Hospital-acquired condition ICH: Intracranial hemorrhage No.: Number NQF: National Quality Forum NSQIP: National Surgical Quality Improvement Project OR: Odds ratio
WORLD NEUROSURGERY - [-]: ---, - 2015
Department of Neurosurgery, University of Rochester Medical Center, Rochester, New York, USA To whom correspondence should be addressed: Hanna Algattas, B.A. [E-mail: [email protected]] Citation: World Neurosurg. (2015). http://dx.doi.org/10.1016/j.wneu.2015.06.033 Journal homepage: www.WORLDNEUROSURGERY.org Available online: www.sciencedirect.com 1878-8750/$ - see front matter ª 2015 Elsevier Inc. All rights reserved.
www.WORLDNEUROSURGERY.org
1
ORIGINAL ARTICLE HANNA ALGATTAS ET AL.
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
thromboembolism (VTE) has materialized as a result. The Joint Commission, National Quality Forum (NQF), and Agency for Healthcare Research and Quality (AHRQ) support these quality measures and have created mandates for hospitals to report violations or else suffer financial penalties. In an effort to improve outcomes and avoid fines, hospitals have developed protocols to track these aforementioned measures. Yet there is no standard approach to determining risk of VTE and the benefit of prophylactic anticoagulation in at-risk populations (4). Concerning VTE then, it comes as no surprise that surgeons especially are forced by institutional and financial pressures to tread delicately the line between thrombosis and hemorrhage. This dilemma is most pronounced in neurosurgical procedures. A meta-analysis of 8 randomized controlled trials (RCTs) found that for every 1000 craniotomy patients anticoagulated with heparin, 91 VTEs would be prevented, whereas 7 intracranial hemorrhages (ICH) and 28 minor bleeds would occur (15). Thus the suggested VTE rate is 9.1%, which appears high at face value but is explained in the meta-analysis. Three of 6 studies analyzed were completed before the use of routine mechanical prophylaxis and failed to include such treatment in control groups. Additionally, different methods of detecting VTE were used (including venography and I125 fibrinogen) that have higher sensitivity and likelihood of capturing asymptomatic VTEs. Therefore although the results seem to support anticoagulation, the authors concede that the absolute values may provide false assurance and VTE detection using surrogate markers may detect less worrisome, asymptomatic VTEs. When accounting for these low-risk VTEs the risk reduction of VTE and risk elevation of ICH may be comparable (15). However, those neurosurgical patients with confirmed VTE are more likely to suffer demise from pulmonary embolism (PE) rather than hemorrhage due to anticoagulation, so the decision to anticoagulate in this population is slightly clearer (9). Rates of VTE and the benefit of aggressive prophylaxis protocols have been explored and proven beneficial in numerous subsets of neurosurgery patients, including post spinal surgery and in patients undergoing craniotomy for tumor (8, 9, 12). In the craniotomy for tumor subpopulation, risk factors of VTE have been identified using both national and institutional data (24, 35). Thus identification of risk factors of VTE among all craniotomies would allow anticoagulation to be targeted toward the most at-risk patient populations. The American College of Surgeons National Surgical Quality Improvement Project (NSQIP) (http://site.acsnsqip.org/) database provides a large number of craniotomy cases, which may elucidate populations at risk for VTE (18). Data recorded include comorbid conditions, preoperative, operative, and postoperative complications. We previously analyzed 2006e2010 NSQIP craniotomy data to elucidate factors associated with VTE and identify patients most likely to benefit from prophylactic anticoagulation. In fact, 2011e2012 NSQIP data are the most recent NSQIP data sets currently available and were not available at the time of our previous analysis. By analyzing 2011e2012 NSQIP data, we hope to validate our previous model and refine the risk score but also extend the model with additional analyses (23). A total of 4844 patients were included in NSQIP data from 2006e2010, whereas over 2011e2012 there were 10477 records. Thus identifying factors preserved across multiple years bolsters the analysis and increases the strength of our model.
2
www.SCIENCEDIRECT.com
METHODS Baseline Univariate Analysis A total of 10,477 patients from 2011e2012 underwent craniotomy on the basis of primary Current Procedural Terminology (CPT) procedure code. Ordinal and nominal variables were subjected to univariate two-tailed c2 analysis and Fisher’s exact test when expected frequencies were low to identify clinical factors associated with VTE. All of the factors in the ACS-NSQIP data set were subjected to analysis. For univariate c2 analysis, original a was set at 0.05 and P values were adjusted using Holm-Bonferonni correction. Those factors less than their respective adjusted a were considered statistically significant, and odds ratios (ORs) were calculated when appropriate. Procedures performed for a tumor diagnosis were also identified on the basis of primary CPT code (61510, 61512, 61518, 61519, 61520, 61521, 61526, 61545, 61546, 61575). Many factors were converted to binary variables to assist in calculation of ORs. Preoperative and intraoperative factors with P values <0.05 were subjected to a forward, stepwise, multivariate binary logistical regression analysis (entry level ¼ 0.05, exit ¼ 0.1) for the identification of factors independently correlated with VTE. Covariate interactions were analyzed with chi-square analysis, noted, and included stepwise in iterations of multivariate regression but failed to impact the robustness of the model. Similarly, postoperative factors were subjected to multivariate logistic regression with care taken to exclude patients in which the event of interest occurred after VTE. Again, covariate interactions were analyzed and included in multivariate regression but failed to impact the robustness of the model. Thirteen factors were included in the final multivariate model and were converted to a binary score of 0 or 1 with 1 representing presence of the factor. Goodness of fit of the multivariate logistic regression model was assessed using Hosmer-Lemeshow test where P > 0.05 indicates a model that fits the data well. The VTE risk score was compared with VTE rate, 30day mortality, and days from surgery to discharge using a number of statistical tests including both linear and Cox regression. A receiver operating characteristics (ROC) curve was used to assess predictive strength of VTE risk scores on the basis of reported area under the curve (AUC). AUC values greater than 0.70 represent acceptable discrimination, and values greater than 0.80 indicate excellent discrimination (7, 16, 32). For validation of the present analysis against past data, the cohort of 4844 patients undergoing craniotomy from 2006e2010 NSQIP data was used. The Cox proportional hazards regression model was an additional analysis conducted to examine the association of clinical factors with VTE in regard to time from surgery to discharge. All statistical analyses were performed using SPSS software version 18.0 (IBM, Armonk, New York, USA).
RESULTS The rate of VTE in the cohort of 10,477 patients was 3.2% (PE was 1.3%; DVT was 2.4%). There were 131 patients in the cohort with a PE, of which 77 patients did not have an associated DVT (58.8%). VTE, DVT, and PE were all significantly associated with increased days from operation to discharge (P < 0.001). Additionally, VTE, DVT, and PE were all significantly associated with increased
WORLD NEUROSURGERY, http://dx.doi.org/10.1016/j.wneu.2015.06.033
ORIGINAL ARTICLE HANNA ALGATTAS ET AL.
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
30-day mortality (P < 0.001 for VTE and DVT; P ¼ 0.001 for PE) (Table 1). Many factors in the NSQIP data set were significantly associated with VTE events in univariate analysis and persisted after HolmBonferroni correction (Table 2). Preoperative factors included ventilator dependence (patient required ventilator at any time within 48 hours before surgery), nonelective surgery, estimated probability of mortality or morbidity (NSQIP-derived value based on hierarchical regression analysis of preoperative factors), time from admission to operating room greater than 4 days, emergency case, dependent functional status (based on completion of activities of daily living), age >60 years old, hospital transfer from an acute care facility, hemiplegia, steroid use (regular use within 30 days before surgery), inpatient procedure, body mass index greater than 30, African-American race, and hypertension (requiring medications within 30 days before surgery). Intraoperative factors included ASA scoring of 4 (patient with severe systemic disease that is a constant threat to life) and 5 (moribund patient who is not expected to survive without the operation). Postoperative factors included ventilator dependence greater than 48 hours cumulatively, unplanned reintubation within 30 days, return to operating room within 30 days, infection (included pneumonia, urinary tract infection, sepsis, and septic shock within 30 days), bleeding transfusion (at least unit of red blood cells given from time of surgery start to 72 hours postoperatively), cerebrovascular accident, impaired sensorium (acute confusion or delirium), cardiac arrest requiring CPR, progressive renal insufficiency (rise in creatinine >2 mg/dL from preoperative values), and coma greater than 24 hours. All these factors were subjected to multivariate binary regression analysis. The 13 factors that persisted were then incorporated into a VTE risk score (Table 3). Modifiable preoperative risk factors included steroid use, ventilator use, and time from admission to surgery greater than 4 days. Preoperative factors that are less modifiable to not at all included nonelective cases, BMI >30, age >60, African-American race, inpatient status, and impaired sensorium. Intraoperative and postoperative factors were American Society of Anesthesiologists (ASA) class of 4 or greater, cumulative postoperative ventilation greater than 48 hours, return to the operating room, and infection. The postoperative infection category was a combination of pneumonia, urinary tract infection (UTI), sepsis, and septic shock. The Hosmer-Lemeshow test was used to assess goodness of fit of the multivariate logistic
Table 1. Venous Thromboembolism (VTE) Events and Mortality Rate Group
Total Patients (%)
30-Day Mortality Rate (%)
Overall cohort
10447
4.8
No VTE
10114
4.6
VTE
333 (3.2)
9.6
DVT
246 (2.4)
10.2
PE
131 (1.3)
10.7
DVT, deep vein thrombosis; PE, pulmonary embolism.
WORLD NEUROSURGERY - [-]: ---, - 2015
Table 2. Factors Significant After Univariate Analysis Preoperative
Intraoperative
Postoperative
Ventilator dependence
ASA Class 4e5 Ventilator dependence >48 hours
Nonelective surgery
Unplanned reintubation
Estimated probability of mortality >10%
Infection (PNA, UTI, sepsis, septic shock)
Estimated probability of morbidity >10%
Return to OR
Admission to OR time >4 days
Bleeding transfusion
Emergency case
CVA
Dependent functional status
Impaired sensorium
Age >60
Cardiac arrest requiring CPR
Transfer from acute care facility
Progressive renal insufficiency
Hemiplegia
Coma >24 hours
Steroid use Inpatient BMI >30 African-American race Hypertension ASA, American Society of Anesthesiologists; PNA, pneumonia; UTI, urinary tract infection; OR, operating room; CVA, cerebrovascular accident; CPR, cardiopulmonary resuscitation; BMI, body mass index.
regression model. At each forward step of the regression, the model fit the data (P > 0.05). Each factor was treated as a binary variable, where the presence of the factor was scored as 1 point and the absence as 0 points. Thus with 13 factors included, scores may theoretically range from 0e13. In actuality, scores ranged from 0e11 (median ¼ 2). Increasing VTE score was predictive of VTE incidence, with a rate of 0.0% for a score of 0 and 15.7% for a score 7 (Figure 1; Table 4). The VTE risk score was significantly associated with VTE rate, increasing time from operation to discharge, and mortality (P < 0.001; see Table 4). A receiver operating characteristics (ROC) curve examined the ability for VTE risk score to predict VTE and demonstrated an area under the curve of 0.719 (95% confidence interval [95% CI] 0.691e0.747; P < 0.001). To determine the risk of VTE as it relates to time from operation to discharge, a Cox proportional hazards model was completed. Each variable incorporated into the VTE risk score was analyzed independently, and alphas were adjusted using Holm-Bonferonni correction. Factors displaying significant hazard ratios were steroid use, age >60, elective surgery, and ASA class 4e5 (Table 5). Our previous analysis of NSQIP data from 2006e2010 identified similar factors linked with VTE (23). However, some factors associated with VTE in 2006e2010 data were not associated in the present analysis. Preoperative factors that were consistent among years after univariate and multivariate binary regression
www.WORLDNEUROSURGERY.org
3
ORIGINAL ARTICLE HANNA ALGATTAS ET AL.
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
Table 3. Factors Significant After Multivariate Analysis Factor
Odds Ratio (95% CI)
Ventilator dependence
3.13 (2.23e4.38)
VTE Number of VTE DVT PE Mean Days Mean 30-Day Risk Patients (% Rate Rate Rate from Surgery to Mortality Rate Score Total) (%) (%) (%) Discharge* (n [%])*
Nonelective surgery
2.04 (1.64e2.54)
0
0.4
1 (0.2)
BMI >30
1.51 (1.21e1.88)
1
2010 (20.5) 0.7* 0.4* 0.3*
3.3
5 (1.0)
Age >60
1.61 (1.30e2.01)
2
3117 (29.8) 1.9* 1.1* 1.1
4.1
26 (5.2)
Steroid use
1.76 (1.35e2.31)
3
2419 (23.2) 2.6
1.2
5.3
67 (13.5)
African-American race
1.85 (1.33e2.58)
4
1379 (13.2) 4.4* 3.3* 1.7
7.6
100 (20.1)
Inpatient
9.13 (2.27e36.78)
5
708 (6.8)
6.6* 4.9* 2.5*
11.2
96 (19.3)
Impaired sensorium
2.57 (1.78e3.72)
6
334 (3.2)
9.6* 7.8* 2.1
12.6
82 (16.5)
Admission to OR >4 days
2.00 (1.50e2.66)
7
351 (1.9)
15.7* 13.7* 3.1*
16.9
121 (24.3)
Overall 10447 number
5.8
498
2.37 (1.87e3.02)
Preoperative
Intraoperative ASA 4e5 Postoperative On ventilator >48 hours
5.89 (4.55e7.63)
Infection
5.40 (4.04e7.22)
Return to OR
4.64 (3.53e6.10)
CI, confidence interval; BMI, body mass index; OR, operating room; ASA, American Society of Anesthesiologists.
included age older than 60 and nonelective surgery/emergency case. Postoperative factors preserved across the data sets were infections, ventilator dependence for greater than 48 hours, and return to the operating room. These five factors were incorporated into their own VTE risk score, which was also predictive of VTE incidence, time from surgery to discharge, and mortality (P < 0.001) (Table 6; Figure 2). When applied to combined 2006e2012 data, the risk score of conserved factors was predictive of VTE in an ROC analysis (AUC ¼ 0.693; 95% CI 0.662e0.723, P < 0.001). Hosmer-Lemeshow goodness of fit
Figure 1. Rates of venous thromboembolism, deep vein thrombosis, and pulmonary embolism at each risk score.
4
Table 4. Risk Scores and Associated Venous Thromboembolism Event (VTE) Rates, Days to Discharge, and Mortality
www.SCIENCEDIRECT.com
129 (1.2)
0.0* 0.0
1.9
333* 246
0.0
131
DVT, deep vein thrombosis; PE, pulmonary embolism. *Statistically significant data (P < 0.05).
at the last step of the multivariate regression demonstrated a good fit for both 2006e2010 and 2011e2012 data sets (P ¼ 0.094 and P ¼ 0.272, respectively).
Table 5. Cox Proportional Hazards of Venous Thromboembolism Risk Score Factors Pertaining to Time from Operation to Discharge Factor
P Value
HR (95% CI)
Steroid use
< 0.001 1.89 (1.45e2.46) Yes
Age >60
0.001 1.43 (1.15e1.78) Yes
Nonelective surgery
0.003 1.41 (1.12e1.75) Yes
ASA 4e5
0.003 0.69 (0.54e0.88) Yes
Preoperative Ventilator dependence
0.009 0.64 (0.46e0.90) No
BMI >30
0.019 1.30 (1.04e1.61) No
Postoperative ventilator >48 hours
0.058 0.77 (0.59e1.01) No
Return to OR
0.589 0.93 (0.88e1.06) No
African-American race
0.667 1.08 (0.77e1.49) No
Inpatient
0.683 1.34 (0.33e5.37) No
Impaired sensorium
0.797 0.95 (0.67e1.37) No
Admit to OR >4 days
0.946 0.99 (0.75e1.31) No
Infection
0.983 1.00 (0.75e1.33) No
Significance Postcorrection
HR, hazard ratio; CI, confidence interval; ASA, American Society of Anesthesiologists; OR, operating room.
WORLD NEUROSURGERY, http://dx.doi.org/10.1016/j.wneu.2015.06.033
ORIGINAL ARTICLE
498 240 5.8
6.0 131 (1.3) 69 (1.4) 246 (2.4) 111 (2.3) 331 (3.2) 170 (3.5) 10,447 4844 Totals
WORLD NEUROSURGERY - [-]: ---, - 2015
DVT, deep vein thrombosis; PE, pulmonary embolism. *Statistically significant (P < 0.05). All values in column significant with mark at column head. yMean days from surgery to discharge.
69 (32.7)
y y
97 (24.1) 66 (25.5)
34 (25.2) 17.6
16.3 17.6
24.2 7 (3.3)*
15 (3.7)* 9 (3.5)*
7 (5.2)* 23 (10.9)
52 (12.9) 21 (8.1)*
15 (11.1)* 28 (13.3)
59 (14.6) 28 (10.8)*
20 (14.8)*
259 (5.3)
135 (2.8)
3
403 (3.9)
DISCUSSION
4
211 (2.0)
187 (10.5) 77 (15.1) 8.6 10.4 23 (1.3) 18 (3.5)* 68 (3.8) 18 (3.5)* 83 (4.7) 1781 (17.0) 509 (10.5) 2
37 (7.3)*
28 (0.7) 7 (0.3)
56 (3.1) 5.0
3.8 3.8
5.4 54 (1.3)
32 (0.8)* 15 (0.7)*
20 (1.1) 75 (1.8)
28 (0.7) 19 (0.9)*
38 (2.1) 107 (2.6)
54 (1.4) 31 (1.4)*
54 (3.0) 4155 (39.8)
2158 (44.5)
1783 (36.8)
0
3897 (37.3)
117 (2.8)
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
1
11e12* 06e10* 11e12* 06e10* 11e12 06e10 11e12* 06e10 11e12* 06e10 11e12 06e10 VTE Risk Score
DVT Rate (%) VTE Rate (%) Number of Patients (% Total)
Table 6. Venous Thromboembolism (VTE) Risk Score Conserved Across 2006e2012 Data (23)
PE Rate (%)
Mean Days from Surgery to Discharge
Mean 30-DayMortality Rate (n [%])
HANNA ALGATTAS ET AL.
We chose to quantify VTE risk on the basis of a number of patient factors provided by the NSQIP database to identify a subset of patients at highest risk for VTE and in whom the benefit of anticoagulation exceeds the risk of hemorrhage. Our data reveal a set of factors highly predictive of VTE, which may aid in distinguishing those patients at increased risk for VTE and for whom the risk-benefit ratio favors the use of pharmacologic prophylaxis. Increasing VTE risk score was significantly associated with VTE incidence, time to discharge, and mortality. The AUC utilizing factors isolated from multivariate analysis of 2011e2012 data demonstrated a predictive model showing acceptable discrimination (16). When validating our 2006e2010 analysis with the present findings, approximately half of the associated factors persisted but still effectively discriminated cases of VTE from the present sample. Factors associated with VTE from the previous analysis of 2006e2010 NSQIP craniotomy cases were transfer from an acute care hospital, preoperative sepsis, emergency cases, dependent functional status, age older than 60, tumor involving central nervous system, craniotomy for tumor, operation time greater than 4 hours, pneumonia, UTI, return to operating room, and postoperative ventilator dependence for >48 hours (23). When comparing the present results with our previous analysis of 2006e2010 data, about half of the factors persisted. Notably, all of the postoperative factors were represented in both 2006e2010 and 2011e2012 data sets. In addition, postoperative factors demonstrated the highest ORs compared with preoperative and intraoperative factors. A VTE risk score with only these factors performed similarly in ROC analysis and suggests that postoperative, relative to preoperative and intraoperative, factors are most predictive of VTE. However, postoperative factors are less modifiable and more difficult to identify early with the aim of preventing VTE. To note, although 2011e2012 data were not available during completion of our 2006e2010 analysis, the present study remains novel. Novel factors are introduced as associated with VTE, despite similar rates of VTE in both cohorts. The present analysis was corrected for multiple statistical comparisons using the HolmBonferonni correction method, increasing the reliability of the current data, and AUC values were examined using HosmerLemeshow goodness of fit. Introduction of both statistical methods increases the reliability of the current data. Also novel, a Cox proportional hazards regression model was used to determine if the association of risk factors with VTE changed with regard to time from surgery to discharge. Steroid use, age >60, nonelective surgery, and ASA class 4e5 all provided significant hazard ratios. The results demonstrate these factors increase risk for VTE for each unit increase in time after surgery. Care must be taken in interpreting such results because some may have effects on total LOS alone. The introduction of time into the risk model could be interesting because it appears some factors lend to increasing risk with longer time spent in the hospital. NSQIP has been used across a broad array of surgical specialties (13, 21, 28, 30). For DVT to be counted in NSQIP, the clot requires either anticoagulation or vena cava filter placement and also diagnosis by either duplex, venogram, or computed tomography scan; thrombosis must occur along the vein and not only the
www.WORLDNEUROSURGERY.org
5
ORIGINAL ARTICLE HANNA ALGATTAS ET AL.
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
Figure 2. Venous thromboembolism (VTE) rates across VTE risk score between 2006 and 2010 and 2011 and 2012 data (23).
catheter (2). PE has similar requirements per NSQIP and must be demonstrated with either a highly probable V-Q scan or definitively with another imaging modality (2). Similar approaches to data collection, risk stratification, and statistical analysis using NSQIP have been taken elsewhere (17). Pharmacologic VTE prophylaxis in certain specialties has more clear clinical utility and has been welcomed, but, in the setting of neurosurgery, a complex balancing act between complications on opposing sides of the spectrum remains. We identified the overall VTE incidence as 3.2% (DVT ¼ 2.4%, PE ¼ 1.3%); importantly, NSQIP only reports symptomatic VTEs, which may underestimate total VTE rates. Rates of VTE among neurosurgical cohorts vary among studies. One study identified DVT risk post neurosurgery to range from 9%e16% and PE as 4%; however, that study contained a small cohort of patients deemed “high risk” for DVT depending on degree of mechanical or pharmacologic prophylaxis (22). Our previous study using 2006e2010 NSQIP craniotomy data identified the DVT and PE rates as 2.6% and 1.4%, respectively (23). Of note, rate of PE was reduced at VTE risk scores of six (2.1%) compared with scores of 5 (2.5%) (see Table 4). Rationale behind this trend is the low amount of PEs reported, which only occurred in 131 cases (1.3%). At risk scores 7, however, rates of PE increase to 3.1%, a statistically significant finding. Separately, the 30-day mortality rate for patients with VTE post craniotomy was 9.6%, which compares favorably with existing literature regarding mortality rate of patients with VTE after general and vascular surgery (11.19%) (33). Anticoagulation is proven to reduce VTE events in a number of trials; Hamilton and colleagues (15) analyzed 6 RCTs and demonstrated a 42% reduction in symptomatic and asymptomatic VTE with heparin. Pooled results for hemorrhage were statistically insignificant, except for more minor bleeds, but the authors concede hemorrhage is still “clinically significant.” Both symptomatic and asymptomatic VTEs were documented in those studies, which begs the question if asymptomatic VTE are comparable with risk of hemorrhage; comparing symptomatic VTE to hemorrhage events was roughly similar in their study (15). Rates of hemorrhage in neurosurgery
6
www.SCIENCEDIRECT.com
patients have been cited as 2.1% in those receiving lowmolecular-weight heparin and 1.1% in those without anticoagulation (19). In a study using D-dimer screening post craniotomy, expectedly, the group screened regularly was diagnosed with VTE more often than a nonscreened group (7.0% vs. 3.7%, respectively). However, the screened group suffered more hemorrhagic complications compared with the nonscreened group (2.2% vs. 0.5%, respectively). According to the authors, the increased VTE events among the screened group likely were asymptomatic, nonelife-threatening VTEs arising distally and the risk of these VTEs must be carefully weighed with the hemorrhagic risk (1). Thus overtreating asymptomatic VTE becomes a growing concern (31). Large and more robust analyses are needed to identify true risk for secondary hemorrhage. The identification of risk factors associated with reoperation for hemorrhage may present an opportunity to balance the risk of thrombosis and hemorrhage in perioperative craniotomy patients. However, increased screening is not necessarily the answer to reducing VTE risk. In fact, hospitals with high-quality ratings often have increased risk-adjusted VTE rates even in the setting of increased prophylaxis rates. This paradox is associated with increased VTE imaging rates at said hospitals, revealing a significant surveillance bias caused by increased VTE screening (3). To improve the use of VTE as a quality measure, imaging rates and the capture of symptomatic versus asymptomatic VTE should be considered. Many of the identified factors were consistent with our original hypotheses. Ventilator dependence, postoperative infections, and return to the operating room are all factors that presumably lengthen hospital stay and reduce mobility, increasing the likelihood of VTE. Additionally, inflammation has been linked to hypercoagulable states and may be related to cytokine and interleukin-mediated changes (26, 27). Along with those chemical imbalances, central hypovolemia has been associated with increased coagulation (27), which is interesting in respect to perioperative care, especially those cases where significant blood loss becomes an issue. Impaired sensorium was strongly associated with VTE relative to other preoperative factors. Gajdos and colleagues (11) identified impaired sensorium to be associated with increased risk for 22 out of 23 postoperative complications recorded in NSQIP, including pneumonia, ventilator dependence, and DVT. Patients with impaired sensorium are less likely to appropriately follow an aggressive pulmonary toilet regimen or ambulate and are more likely to require urinary catheterization, making them susceptible to pneumonia, ventilator dependence, VTE, and UTI among other complications (11). Many of the factors associated with VTE postcraniotomy in the present analysis have been identified elsewhere and among different and similar types of procedures, including craniotomy for tumor but also in surgery for inflammatory bowel disease (1, 22, 24, 34-36). In the current study, covariate interactions failed to significantly impact the multivariate model; however, association between factors should be taken into consideration when interpreting the findings. Postoperative infections including pneumonia were associated with VTE. Several preoperative factors are predictive of pneumonia including functional status, sepsis, lung disease, smoking, and ASA classification (14). Notably, functional
WORLD NEUROSURGERY, http://dx.doi.org/10.1016/j.wneu.2015.06.033
ORIGINAL ARTICLE HANNA ALGATTAS ET AL.
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
status and ASA class were associated with VTE after univariate and multivariate analysis, respectively. There is also concern regarding duplicity of factors isolated after multivariate analysis; inpatient status and nonelective cases are one such link. Inherently, inpatients are more likely to have nonelective (i.e., emergency) cases. However, the association between VTE and inpatients was the strongest on the basis of OR (OR ¼ 9.13) and remains a warranted measure. Stemming from this idea, the possibility of weighting the VTE risk score based on OR was thoroughly explored because different factors, especially postoperative events, varied in strength of association to VTE. Multiple attempts at weighting the risk score were examined, but they only marginally improved ROC discrimination and thus were not presented. The notion of a VTE risk score is not novel, being introduced by pioneers such as Dr. Joseph Caprini, and has been used to varying extents across surgical specialties (5). Cassidy et al. created a mandatory electronic VTE risk assessment, which needed to be completed by surgeons before placing an order. Institutional NSQIP data were analyzed before and after the program was put in place, and the group realized significant reductions in DVT (1.9%e0.3%) and PE (1.1%e 0.5%). However, there was no mention of bleeding episodes possibly introduced by more rigorous VTE prophylaxis (6). Factors included in our VTE score were compared with those found in previous analysis of 2006e2010 craniotomy cases to determine coinciding factors with stronger association to VTE (23). The presented risk model is imperfect on the basis of ROC analysis and also due to the overall rarity of VTE, especially PE events; however, stratification of patients based on score does represent a clinically relevant and significant distinction. The present analysis and existing NSQIP-focused research consistently identify preoperative factors that offer a large contribution to the incidence of hospital-acquired conditions (HAC). Because many preoperative factors are nonmodifiable, Molena and colleagues (29) suggest nonpayment policies implemented by the Centers for Medicare and Medicaid Services (CMS) may be excessively harsh. This argument has been echoed in surgical oncology and bariatric surgery (25, 29). The present findings agree with these statements. Interestingly, factors with the strongest association to VTE based on OR were the postoperative factors (infection, return to OR, ventilation >48 hours). However,
REFERENCES 1. Aishima K, Yoshimoto Y: Screening strategy using sequential serum D-dimer assay for the detection and prevention of venous thromboembolism after elective brain tumor surgery. Br J Neurosurg 27: 348-354, 2013. 2. American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP): User guide for the 2012 ACS NSQIP Participant Use Data File. Available at: www.acsnsqip.org 3. Bilimoria KY, Chung J, Ju MH, Haut ER, Bentrem DJ, Ko CY, Baker DW: Evaluation of surveillance bias and the validity of the venous thromboembolism quality measure. JAMA 310: 1482-1489, 2013.
across studies, these findings need to be carefully reviewed to ensure postoperative events occurring after HACs are eliminated. Our analysis is not without limitation, including its retrospective design and generalizability to participating hospitals, which are often large academic centers. Additionally, although anticoagulation or vena cava filter placement is required for DVT or PE events per NSQIP, NSQIP does not record specific prophylaxis or ensuing treatments, which may skew data. Although NSQIP represents a growing and robust surgical database, one issue is the possibility of random sampling error. When compared with the Thoracic Morbidity and Mortality System (TM&M) surgical database system, NSQIP captured 21.1% of patients included in TM&M (20). With the capture of fewer patients came significant differences in the rates of some procedures (20). In a study examining pancreatectomy NSQIP data, discordance between NSQIP-reported data and reviewed cases was 27.3%. Certain events, such as postoperative bleeding requiring transfusion, represent common and often misclassified outcomes (10). However, comparing the more robust results of the 2011e2012 analysis with previously acquired 2006e2010 data significantly strengthens the model and identifies factors consistently associated with VTE. Using those factors for VTE risk modeling in additional non-NSQIP data sets would be useful in bringing a clinical score to practice. Thus future studies should be aimed at validating the risk score in a separate patient population. In fact, with a more refined VTE risk model, additional studies that balance risk of VTE with risk of hemorrhage can be combined to determine the risk of two separate complications on opposing sides of the coagulation spectrum.
CONCLUSIONS A number of factors have been associated with VTE events in postcraniotomy patients, and a subset of these have been validated across our previous data set, mostly postoperative factors. The development of a VTE risk score successfully predicted VTE incidence, increased time to discharge, and mortality. The results suggest a subset of postcraniotomy patients who are more likely to benefit from prophylactic anticoagulation postoperatively may be identified.
4. Camden R, Ludwig S: Prophylaxis against venous thromboembolism in hospitalized medically ill patients: update and practical approach. Am J Health Syst Pharm 71:909-917, 2014. 5. Caprini JA, Arcelus JI, Hasty JH, Tamhane AC, Fabrega F: Clinical assessment of venous thromboembolic risk in surgical patients. Semin Thromb Hemost 17 (Suppl 3):304-312, 1991. 6. Cassidy MR, Rosenkranz P, McAneny D: Reducing postoperative venous thromboembolism complications with a standardized riskstratified prophylaxis protocol and mobilization program. J Am Coll Surg 218:1095-1104, 2014. 7. Cohen ME, Dimick JB, Bilimoria KY, Ko CY, Richards K, Hall BL: Risk adjustment in the American College of Surgeons National Surgical Quality Improvement Program: a comparison of
WORLD NEUROSURGERY - [-]: ---, - 2015
logistic versus hierarchical modeling. J Am Coll Surg 209:687-693, 2009. 8. Collen JF, Jackson JL, Shorr AF, Moores LK: Prevention of venous thromboembolism in neurosurgery: a metaanalysis. Chest 134:237-249, 2008. 9. Cote LP, Greenberg S, Caprini JA, Stone J, Arcelus JI, Lopez-Jimenez L, Rosa V, Schellong S, Monreal M: Outcomes in neurosurgical patients who develop venous thromboembolism: a review of the RIETE registry. Clin Appl Thromb Hemost 20:772-778, 2014. 10. Epelboym I, Gawlas I, Lee JA, Schrope B, Chabot JA, Allendorf JD: Limitations of ACSNSQIP in reporting complications for patients undergoing pancreatectomy: underscoring the need for a pancreas-specific module. World J Surg 38:1461-1467, 2014.
www.WORLDNEUROSURGERY.org
7
ORIGINAL ARTICLE HANNA ALGATTAS ET AL.
11. Gajdos C, Kile D, Hawn MT, Finlayson E, Henderson WG, Robinson TN: The significance of preoperative impaired sensorium on surgical outcomes in nonemergent general surgical operations. JAMA Surg 150:30-36, 2015.
ANALYSIS OF VTE RISK IN CRANIOTOMY PATIENTS
21. Kester BS, Merkow RP, Ju MH, Peabody TD, Bentrem DJ, Ko CY, Bilimoria KY: Effect of postdischarge venous thromboembolism on hospital quality comparisons following hip and knee arthroplasty. J Bone Joint Surg Am 96:1476-1484, 2014.
12. Goldhaber SZ, Dunn K, Gerhard-Herman M, Park JK, Black PM: Low rate of venous thromboembolism after craniotomy for brain tumor using multimodality prophylaxis. Chest 122:1933-1937, 2002.
22. Khaldi A, Helo N, Schneck MJ, Origitano TC: Venous thromboembolism: deep venous thrombosis and pulmonary embolism in a neurosurgical population. J Neurosurg 114:40-46, 2011.
13. Grant DW, Mlodinow A, Ver Halen JP, Kim JY: Catastrophic outcomes in free tissue transfer: a six-year review of the NSQIP database. Plast Surg Int 2014:704206, 2014.
23. Kimmell KT, Jahromi BS: Clinical factors associated with venous thromboembolism risk in patients undergoing craniotomy. J Neurosurg 122: 1004-1011, 2015.
14. Gupta H, Gupta PK, Schuller D, Fang X, Miller WJ, Modrykamien A, Wichman TO, Morrow LE: Development and validation of a risk calculator for predicting postoperative pneumonia. Mayo Clin Proc 88:1241-1249, 2013.
24. Kimmell KT, Walter KA: Risk factors for venous thromboembolism in patients undergoing craniotomy for neoplastic disease. J Neurooncol 120: 567-573, 2014.
15. Hamilton MG, Yee WH, Hull RD, Ghali WA: Venous thromboembolism prophylaxis in patients undergoing cranial neurosurgery: a systematic review and meta-analysis. Neurosurgery 68:571-581, 2011. 16. Hosmer DW, Lemeshow S: Applied Logistic Regression. 2nd ed. New York: John Wiley & Sons; 2000. 17. Hyder JA, Wakeam E, Habermann EB, Hess EP, Cima RR, Nguyen LL: Derivation and validation of a simple calculator to predict home discharge after surgery. J Am Coll Surg 218:226-236, 2014. 18. Ingraham AM, Richards KE, Hall BL, Ko CY: Quality improvement in surgery: the American College of Surgeons National Surgical Quality Improvement Program approach. Adv Surg 44: 251-267, 2010. 19. Iorio A, Agnelli G: Low-molecular-weight and unfractionated heparin for prevention of venous thromboembolism in neurosurgery: a meta-analysis. Arch Intern Med 160:2327-2332, 2000. 20. Ivanovic J, Seely AJ, Anstee C, Villeneuve PJ, Gilbert S, Maziak DE, Shamji FM, Forster AJ, Sundaresan RS: Measuring surgical quality: comparison of postoperative adverse events with the American College of Surgeons NSQIP and the Thoracic Morbidity and Mortality classification system. J Am Coll Surg 218:1024-1031, 2014.
8
www.SCIENCEDIRECT.com
25. Lidor AO, Moran-Atkin E, Stem Magnuson TH, Steele KE, Feinberg Schweitzer MA: Hospital-acquired conditions ter bariatric surgery: we can predict, but can prevent? Surg Endosc 28:3285-3292, 2014.
M, R, afwe
26. Lipinski S, Bremer L, Lammers T, Thieme F, Schreiber S, Rosenstiel P: Coagulation and inflammation. Molecular insights and diagnostic implications. Hamostaseologie 31:94-102; 104, 2011. 27. Meyer MA, Ostrowski SR, Overgaard A, Ganio MS, Secher NH, Crandall CG, Johansson PI: Hypercoagulability in response to elevated body temperature and central hypovolemia. J Surg Res 185:e93-e100, 2013. 28. Moghadamyeghaneh Z, Hanna MH, Carmichael JC, Nguyen NT, Stamos MJ: A nationwide analysis of postoperative deep vein thrombosis and pulmonary embolism in colon and rectal surgery. J Gastrointest Surg 18: 2169-2177, 2014. 29. Molena D, Mungo B, Stem M, Feinberg RL, Lidor AO: Prevalence, impact, and risk factors for hospital-acquired conditions after major surgical resection for cancer: a NSQIP analysis. J Gastrointest Surg 19:142-151; discussion 151, 2015. 30. Orr NT, Davenport DL, Roth JS: Outcomes of simultaneous laparoscopic cholecystectomy and ventral hernia repair compared to that of
laparoscopic cholecystectomy alone. Surg Endosc 27:67-73, 2013. 31. Prell J, Rachinger J, Smaczny R, Taute BM, Rampp S, Illert J, Koman G, Marquart C, Rachinger A, Simmermacher S, Alfieri A, Scheller C, Strauss C: D-dimer plasma level: a reliable marker for venous thromboembolism after elective craniotomy. J Neurosurg 119:1340-1346, 2013. 32. Ramanan B, Gupta PK, Gupta H, Fang X, Forse RA: Development and validation of a bariatric surgery mortality risk calculator. J Am Coll Surg 214:892-900, 2012. 33. Rogers SO Jr, Kilaru RK, Hosokawa P, Henderson WG, Zinner MJ, Khuri SF: Multivariable predictors of postoperative venous thromboembolic events after general and vascular surgery: results from the patient safety in surgery study. J Am Coll Surg 204:1211-1221, 2007. 34. Rolston JD, Han SJ, Bloch O, Parsa AT: What clinical factors predict the incidence of deep venous thrombosis and pulmonary embolism in neurosurgical patients? J Neurosurg 121:908-918, 2014. 35. Smith TR, Nanney AD 3rd, Lall RR, Graham RB, McClendon J Jr, Lall RR, Adel JG, Zakarija A, Cote DJ, Chandler JP: Development of venous thromboembolism (VTE) in patients undergoing surgery for brain tumors: results from a single center over a 10-year period. J Clin Neurosci 22: 519-525, 2015. 36. Wallaert JB, De Martino RR, Marsicovetere PS, Goodney PP, Finlayson SR, Murray JJ, Holubar SD: Venous thromboembolism after surgery for inflammatory bowel disease: are there modifiable risk factors? Data from ACS NSQIP. Dis Colon Rectum 55:1138-1144, 2012. Conflict of interest statement: The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper. Received 1 June 2015; accepted 16 June 2015 Citation: World Neurosurg. (2015). http://dx.doi.org/10.1016/j.wneu.2015.06.033 Journal homepage: www.WORLDNEUROSURGERY.org Available online: www.sciencedirect.com 1878-8750/$ - see front matter ª 2015 Elsevier Inc. All rights reserved.
WORLD NEUROSURGERY, http://dx.doi.org/10.1016/j.wneu.2015.06.033