A and B, These rates are performed at the hospital level, and the different sizes of the circles reflect hospital volume. B, These rates reflect postdischarge venous thromboembolism (VTE) rates during the 30 days following discharge after index hospitalization for surgical procedure. The solid lines represent the linear regression fit across all patients. Pearson correlation coefficients and P values are shown.
Holcomb CN, DeRussy A, Richman JS, Hawn MT. Association Between Inpatient Surveillance and Venous Thromboembolism Rates After Hospital Discharge. JAMA Surg. 2015;150(6):520-527. doi:10.1001/jamasurg.2015.35
Venous thromboembolism (VTE) surveillance practices in hospitals, but not adherence to Surgical Care Improvement Program VTE prophylaxis measures, have been reported to explain the variation in VTE rates in hospitals.
To examine the relationship between inpatient surveillance testing for VTE and postdischarge VTE rates at the hospital level to determine whether more frequent inpatient surveillance is associated with reduced occurrence of postdischarge VTEs.
Design, Setting, and Participants
Retrospective study of a US national cohort of Veterans Affairs (VA) patients. National VA Surgical Quality Improvement Program outcome data were linked to VA administrative data on patients undergoing inpatient surgery from 2005 to 2009 and were included in the Surgical Care Improvement Program VTE measurement population.
Main Outcomes and Measures
Surveillance was identified using Current Procedural Terminology codes for diagnostic VTE imaging. Relationships between hospital-level surveillance and VTE rates were assessed with Pearson correlation coefficients, and the postdischarge VTE rate was modeled using linear regression, adjusting for hospital volume, inpatient VTE rate, inpatient surveillance rate, and case mix.
Of 25 975 patients at 79 VA facilities, 296 (1.4%) experienced a VTE during the index hospitalization, and 114 (0.4%) experienced a postdischarge VTE within 30 days after surgery. The median length of stay was 11 days for those with a positive surveillance test result and 9 days for those with a negative test result. There was a positive correlation between inpatient surveillance and inpatient VTE rates (R = 0.33, P = .003) but no significant correlation of inpatient surveillance with either postdischarge surveillance (R = 0.11, P = .29) or postdischarge VTE rates (R = 0.03, P = .76). In an adjusted regression model of the postdischarge VTE rate, only the inpatient VTE rate was significant (β = 0.13, P = .05).
Conclusions and Relevance
Hospitals with higher VTE surveillance rates have higher inpatient VTE rates but not decreased postdischarge VTE rates. However, hospitals with higher inpatient VTE rates have higher postdischarge VTE rates, which suggests that surveillance may be influenced by higher observed rates and not surveillance practices alone.
Venous thromboembolism (VTE) is a recognized occurrence among patients who have undergone surgery and is influenced by both patient risk factors and type of surgical procedure.1- 5 The burden of deep venous thrombosis includes significant long-term morbidity of postthrombotic syndrome and the risk for mortality with associated pulmonary embolism.6,7 In addition, the 1-year financial burden of treating the complications associated with VTE has been estimated at $33 000.8 Studies9,10 support the notion that appropriate prophylaxis can reduce the incidence of postoperative VTE. As such, measures to reduce the rates of VTE have been a target of national efforts to improve the safety and quality of surgical care, including the Surgical Care Improvement Project (SCIP).
The Centers for Medicare and Medicaid Services enacted SCIP in 2006 to reduce the numbers of preventable postoperative complications and included 2 measures for VTE prevention. Adherence to SCIP is accomplished by meeting criteria for both the SCIP VTE-1 process measure (assessing whether appropriate VTE prophylaxis was ordered up to 24 hours after surgery) and the SCIP VTE-2 process measure (measuring whether surgery patients received appropriate VTE prophylaxis within 24 hours prior to and following surgery).11 Despite the use of evidence-based guidelines in the creation of SCIP measures, studies12- 14 have failed to show that adherence to SCIP-VTE measures reduce postoperative VTE rates.
Evidence demonstrates that process measures such as SCIP have not yielded the intended reduction in the numbers of complications, and consequently the Affordable Care Act has shifted the focus toward outcomes as a more relevant measure of surgical quality. Recently, the issue of surveillance bias has emerged as a threat to the validity of the use of outcomes as quality metrics, particularly for VTE. Hospital VTE rates were found to be strongly associated with hospital VTE surveillance practices and not with hospital adherence to SCIP-VTE measures.15,16 The controversy surrounding this concept is whether hospital surveillance practices are due to a higher population risk of VTE or to practice preferences for surveillance to allow for early detection before clinical symptoms. The former practice would suggest increased surveillance as a response to a higher-risk population for VTE, whereas the latter would suggest vigilance with the intent for high-quality care.
Most studies examining surveillance have been limited to the index hospitalization, and data are lacking on how SCIP adherence and inpatient VTE surveillance affect postdischarge VTE rates. In the present study, we sought to determine if inpatient surveillance practices affected outpatient VTE outcomes at the hospital level. We hypothesized that hospitals with higher levels of inpatient surveillance will have higher VTE rates due to surveillance bias but will have lower rates of postdischarge VTE because preclinical VTE would be detected prior to discharge.
This is a retrospective study of a national cohort of Veterans Affairs (VA) patients with data on SCIP-VTE measures matched to outcomes from the VA Surgical Quality Improvement Program (VASQIP) database from 2006 to 2009. Patient-level SCIP data were merged with VASQIP data, which included information on patient comorbidities and demographics; surgical characteristics such as emergency vs elective case status, work relative value unit, and surgical subspecialty; and postoperative outcomes including VTE. The study protocol was reviewed and approved by the Birmingham VA Medical Research and Development Committee and the institutional review board, as well as by the VA Surgical Quality Data Use Group and the VA Office of Information and Analytics in the VA central office, Washington, DC. Requirement for informed consent was waived by the institutional review board in the setting of a retrospective study using previously collected data.
The SCIP population for the SCIP-VTE measures includes patients who underwent major surgical procedures, including orthopedic, colorectal, gynecologic, neurosurgical, and urologic surgical procedures. The SCIP-VTE module excludes vascular procedures, and our study excluded cardiac cases because the VASQIP collects different variables for cardiac procedures vs noncardiac procedures. The SCIP patient population used for this analysis has been previously described.13 Hospitals with no reported inpatient VTEs and patients with missing descriptions of surgical data and outcomes were excluded from our study.
The Veterans Health Administration Office of Information and Analytics External Peer Review Program contracts with the West Virginia Medical Institute to collect VA hospital SCIP measures. Beginning in 2006, these data have been collected by the VA according to guidelines established by the Joint Commission and the Centers for Medicare and Medicaid Services. Abstractor reliability is assessed regularly.
The VASQIP uses trained clinical nurse reviewers to collect demographic data, preoperative laboratory data, and data on surgical characteristics and 30-day postoperative outcomes for a sample of VA patients undergoing major surgery. The quality of the data at a sample of VA medical centers has previously been shown to be reliable.3
Administrative data from both inpatient and outpatient databases of the VA Corporate Data Warehouse were used to identify inpatient and postdischarge surveillance tests.
The independent variables of interest were facility rates of index hospitalization surveillance and adherence to the SCIP VTE-2 process measure. Inpatient surveillance was defined using Current Procedural Terminology codes (71260, 71250, 71270, 71275, 71550, 71551, 71552, 71555, 75741, 75743, 75746, 75820, 75822, 78445, 78456, 78457, 78458, 78585, 78584, 78580, 78599, 78598, 78597, 78582, 93970, and 93971) and International Classification of Diseases, Ninth Revision, Clinical Modification codes (8877, 8843, 8866, and 9215) to identify venous duplex ultrasonographic images of the extremities, ventilation-perfusion scans, and/or computed tomographic angiography scans of the chest that occurred before discharge and at least 1 day prior to any diagnosed VTE. Facility rates were calculated with the number of patients who had a VTE surveillance test as the numerator and the total number of SCIP patients as the denominator. Adherence to the SCIP VTE-2 process measure was reported by the SCIP.
The dependent variable of interest was the occurrence of deep venous thrombosis or pulmonary embolism after hospital discharge but within 30 days of surgery, as reported by the VASQIP. These complications were combined to create a composite VTE outcome variable with only the first event included in our study for those patients who may have incurred both outcomes. Facility rates were calculated with the number of VTE-diagnosed patients as the numerator and the total number of SCIP patients as the denominator. Venous thromboembolisms that were identified prior to or on the day of hospital discharge were considered “index” events, and VTEs that occurred after discharge from the index hospitalization were considered “postdischarge” events, even if it resulted in a readmission to the hospital.
Additional patient-level covariates were included, such as demographics, functional status, lifestyle variables (eg, tobacco use), and comorbidities collected in the VASQIP. Facility-level analyses used the facility-specific mean values or rates for these covariates to account for different patient populations at different hospitals.
The unadjusted relationships among patient-level variables were assessed using the χ2 test and the Wilcoxon rank sum test. The unadjusted relationships of hospital-level rates were assessed using simple linear regression. Linear regression analysis was used to examine the adjusted relationship between facility rates of index VTE diagnosis, SCIP VTE-2 adherence, and postdischarge VTE rates.
The adjusted relationship between facility rates of index surveillance, SCIP VTE-2 adherence, and postdischarge VTE was assessed with ordinary least squares regression models. Factors identified to have significant bivariate associations with postdischarge VTE rates were included as covariates. Backward stepwise selection was used to identify a parsimonious model. Analyses were completed using SAS version 9.2 (SAS Institute Inc). The significance level was set at P < .05.
The study cohort included 25 975 patients meeting the criteria for the SCIP-VTE measures at 79 VA facilities. The characteristics of the cohort were stratified by surveillance and VTE occurrence and are shown in Table 1. The overall VTE surveillance rate of the cohort was 18.8%, with a 1.6% VTE rate. Being older than 60 years of age and the comorbidities of insulin-dependent diabetes mellitus, peripheral vascular disease, a history of myocardial infarction, acute renal failure, or a malignant neoplasm were associated with higher surveillance and VTE rates. Surgical characteristics associated with increased surveillance and VTE rates included orthopedic and neurosurgical procedures, as well as the presence of other postoperative complications.
Factors associated with positive and negative surveillance test results are shown in Table 2. Thoracic procedures had the highest rates of negative surveillance test results, while gynecologic procedures had a higher likelihood of positive surveillance compared with other types of procedures. The percentage of VTE surveillance practices performed varied widely at the hospital level, ranging from 5.5% to 50.4% with a median surveillance rate of 17.6% (Table 3). The hospital-level VTE rate ranged from 0.3% to 4.6%.
Correlations between surveillance practices, hospital characteristics, and VTE occurrence are displayed in Table 4. The proportion of patients meeting the SCIP-VTE measures and the surgical volume of the hospital were not significantly associated with surveillance or VTE occurrence. The postoperative hospital length of stay was positively correlated with both surveillance (R = 0.45, P < .05) and index VTE rate (R = 0.25, P < .05) but negatively correlated with postdischarge VTE rate (R = −0.11, P = .31), in part because of the duration of exposure time in the hospital. Inpatient surveillance was positively correlated with inpatient VTE rates (R = 0.33, P = .003) (Figure, A) but not with either postdischarge surveillance or VTE occurrence (R = 0.03, P = .76) (Figure, B). A multiple linear regression model assessing predictors of hospital postdischarge VTE rates adjusted for patient risk factors and predischarge complications found that the index hospitalization VTE rate was the strongest predictor of postdischarge VTE occurrence (β = 0.13, P = .05). Neither predischarge nor postdischarge surveillance rates were associated with postdischarge VTE rates.
Using national VA data from the SCIP-VTE measure population, we found that hospitals with higher inpatient surveillance rates had higher inpatient VTE rates, but we found no correlation between inpatient surveillance and postdischarge surveillance or VTE rates. However, higher inpatient VTE rate was a significant predictor of outpatient VTE rate, which suggests that surveillance rates may be more influenced by higher observed rates and less by facility-level surveillance practice preferences. One of the strengths of our study is that both surveillance and VTE outcomes were assessed for the same population and that information on postdischarge VTE outcomes was available. In addition, we confirmed that adherence to the SCIP VTE-2 process measure was not associated with VTE outcome, even after adjusting for hospital VTE surveillance rates.
Our study is congruent with a previous report15 of the strong association between inpatient surveillance rates and the rate of inpatient VTEs diagnosed. However, the prior report15 was limited to only inpatient VTE outcomes that were determined from administrative data. A follow-up study16 using Medicare data to assess hospital surveillance patterns and National Surgical Quality Improvement Program data to determine VTE rates confirmed that surveillance was associated with VTE occurrence. However, an important limitation of this follow-up study16 is that hospital surveillance was measured for a different surgical population than that for VTE outcome. If increased surveillance fully explains increased inpatient VTE rates, then we would expect that postdischarge VTE rates would decrease as more VTEs would be diagnosed prior to hospital discharge. In our study, we found a positive association between inpatient VTE and postdischarge VTE rates. This suggests that the majority of inpatient surveillance may be ordered appropriately and that hospitals with higher rates of at-risk patients have higher inpatient and postdischarge VTE rates. Interestingly, inpatient surveillance was not correlated with postdischarge surveillance, which suggests that surveillance practices are limited to the inpatient setting.
Currently, the Center for Medicaid and Medicare Services uses administrative claims data to compose the Patient Safety Indicators used in assessing quality measures. In 2015, the Patient Safety Indicators, including VTE, will be included in the Value-Based Purchasing program. Although administrative data are convenient and accessible, the reliability for capture of actual events is questionable. One analysis17 found that studies using administrative data overestimated VTE rates by 40% compared with those using clinical data. One of the strengths of our study is the use of VASQIP data, which include both inpatient and postdischarge events occurring within 30 days of surgery accessed from medical records. Both the VASQIP and the National Surgical Quality Improvement Program (which is modeled after the VASQIP) have been shown to be superior in validity and reliability to Patient Safety Indicators for assessing VTE outcomes.17,18 Our study, which assessed surveillance and outcomes in the same surgical population for each hospital, did find a positive association between the use of surveillance tests and VTE outcomes using VASQIP-assessed events.
Several studies19- 21 have been published on individual hospital VTE prevention strategies with subsequent improvement in VTE outcomes, whereas increased SCIP-VTE adherence has not been shown to be associated with decreased VTE rates.13,15 The lack of robustness for the SCIP-VTE measures likely explains the lack of improved VTE outcomes despite high adherence rates. For example, a patient who underwent major abdominal surgery and received chemical prophylaxis within 24 hours after surgery without receiving any other prophylactic medication throughout the duration of the hospitalization would meet the criteria for SCIP-VTE compliance. Successful reductions in hospital VTE rates have been achieved by focusing on risk stratification and an appropriate duration of VTE prophylactic measures that are not limited to prophylaxis alone.19- 21 This scenario suggests that the definition for SCIP-VTE adherence should be expanded beyond the initial 24 hours after surgery if it is to be a useful quality measure. Perhaps, a threshold proportion of hospital days covered by VTE prophylaxis (eg, ≥80%) would be a more informative measure of hospital quality regarding VTE prevention.
Where do we go from here with respect to VTE as a metric for surgical quality of care? Given that adherence rates to current SCIP measures are more than 95%, the opportunity to achieve further benefit must be questioned. The limited focus of the SCIP-VTE metric to the first hospital day may have the unintended consequence of decreasing emphasis on the need for prevention strategies throughout the hospitalization. Hospitals seeking to further reduce VTE rates should implement individual evidence-based VTE prophylaxis guidelines above those required by SCIP compliance. Furthermore, if surveillance is associated with an increased number of VTE diagnoses during the index hospitalization without reduction in the number of postdischarge diagnoses, the role of VTE outcomes as a valid measure of hospital quality must be reconsidered. It does appear that some hospitals have higher-risk surgical populations that cannot be completely accounted for by adjusting for patient and procedure factors and cannot be mitigated by current SCIP measures.
Our study has several limitations, including a study population composed of mostly older men, which limits the generalizability of our results to other populations. Furthermore, we only assessed VTE prevention as measured by SCIP, and other hospital VTE prevention practices could be associated with hospital surveillance practices, thus confounding our observation. We did not pursue access to data after 2009 and cannot account for any changes occurring in the past 5 years regarding VTE surveillance. However, SCIP adherence was approaching 100% by the end of the study period, and it is unlikely that prevention measures or VTE outcomes changed significantly in the past few years. Finally, only variables assessed by VASQIP medical record review were available for our analyses, leaving the possibility of bias or confounding due to unobserved factors.
Postdischarge VTE rates in the 30 days after surgery are not decreased by higher inpatient surveillance rates but are associated with higher inpatient VTE rates. Thus, patient risk factors and case mix likely contribute to hospital VTE rates, and surveillance bias may reflect the underlying at-risk population. Using its current definition, we find that SCIP-VTE adherence is an inadequate assessment of hospital quality because it is not associated with VTE outcomes. Further research is needed to determine whether refined VTE prophylaxis measures or outcome assessment can be reliable measures of surgical quality.
Accepted for Publication: October 24, 2014.
Corresponding Author: Mary T. Hawn, MD, MPH, Section of Gastrointestinal Surgery, Department of Surgery, University of Alabama at Birmingham, 1922 7th Ave S, KB 428, Birmingham, AL 35294-0016 (firstname.lastname@example.org).
Published Online: April 1, 2015. doi:10.1001/jamasurg.2015.35.
Author Contributions: Dr Hawn had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Holcomb, DeRussy, Hawn.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Holcomb, DeRussy.
Critical revision of the manuscript for important intellectual content: Richman, Hawn.
Statistical analysis: DeRussy, Richman.
Obtained funding: Hawn.
Administrative, technical, or material support: DeRussy, Hawn.
Study supervision: Hawn.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study is supported by a VA Health Services Research and Development Grant. Dr Holcomb is supported by grant T32 HS013852-11 from the Agency for Healthcare Research and Quality, Rockville, Maryland. Dr Richman is supported by a VA Career Development Award.
Role of the Funder/Sponsor: The funders/sponsors had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The opinions expressed are those of the authors and not necessarily those of the Department of Veterans Affairs or the US government.
Previous Presentation: This paper was presented at the 10th Annual Academic Surgical Congress; February 5, 2015; Las Vegas, Nevada.