Effect of a near-universal hospitalization-based prophylaxis regimen on annual number of venous thromboembolism events in the US

Venous thromboembolism (VTE), consisting of deep vein thrombosis (DVT) and its complication, pulmonary embolism (PE), is a major health problem.^1-5  The overall age- and sex-adjusted incidence of VTE is 123 (95% confidence interval [CI], 118-127) per 100 000 person-years, and over the 30-year period 1981-2010, the annual incidence of VTE has remained relatively constant.⁶  Despite the importance of this disease, there are few data on the total number of VTE events (incident and recurrent) occurring in the United States per year.^7-9  Moreover, the number of symptomatic VTE events that are potentially preventable and the effects of hospitalization-based prophylaxis on VTE incidence and bleeding are uncertain. In the placebo arms of 2 randomized trials testing anticoagulants as VTE prophylaxis in selected populations of acutely ill medical inpatients, the symptomatic VTE incidence rates were 0.7%, 1.6%, and 2.6% at 14, 15, and 32 days, respectively.^10,11  However, patients at greatest risk for hospitalization-associated VTE (ie, surgery patients and trauma patients) were excluded; asymptomatic DVT detected by mandated imaging was treated, which likely altered the disease natural history; and symptomatic VTE surveillance was truncated at 14 to 32 days while the hospitalization-associated VTE risk period extends to 3 months.²  In the placebo arms of 2 other trials, the 90-day VTE incidence rates were 1.4% and 2.6%, respectively, but again, the study populations were limited to selected acutely ill medical inpatients.^12,13  The 90-day hospitalization-associated VTE rates in the Michigan Hospital Medicine Safety Consortium study ranged from 1.15 to 1.42 per 10 000 patient-days and did not differ by VTE prophylaxis rate, but these VTE rates may be underestimates due to exclusion criteria (ie, any surgery during hospitalization, direct intensive care unit admission).¹⁴  Indirect data also suggest no association between VTE prophylaxis rates and postoperative VTE rates.¹⁵ 

To address these gaps in knowledge, we performed a population-based cohort study in which we identified all incident and recurrent VTE events from a well-defined geographic area over the 6-year period 2005-2010. We estimated age- and sex-specific VTE attack rates, related and unrelated to current or recent hospitalization. Using US census data and the estimated number and duration of hospitalizations per year by age and sex from the National Inpatient Survey (NIS), we estimated the total number of VTE events per year, both overall and related to current or recent hospitalization. Finally, we estimated hospitalization-related VTE attack rates for Mayo Clinic hospitals from 2005 to 2010, where the average VTE prophylaxis rate or indication that prophylaxis was unnecessary increased from 40% to 90% (ie, a near-universal standardized prophylaxis regimen) over the 6-year period.

Using the resources of the Rochester Epidemiology Project (REP) (see supplemental Data, available on the Blood Web site),^16-18  we identified all Olmsted County (2010 census population = 144 248), Minnesota, residents with incident or recurrent DVT PE over the 45-year period 1966-2010, as previously described.^1,2,5,19-21  For this study, we restricted our analyses to residents with incident or recurrent leg DVT alone (including isolated anterior and posterior tibial and peroneal [calf] DVT but excluding isolated gastrocnemius or soleal vein thrombosis) or PE with or without DVT (excluding isolated subsegmental PE) over the 6-year period 2005-2010; 98% of events were objectively diagnosed, and the remainder were clinically diagnosed events treated with anticoagulants. Olmsted County residents with a first VTE while a nonresident and a recurrence while a resident were also included (see supplemental Data). Recurrent VTE was defined as thrombosis of a venous site that was either previously uninvolved or had interval documentation of thrombus resolution.^5,20,21  For deceased patients, all death certificates and autopsy reports were reviewed regardless of the location at death. The study was approved by the Mayo Clinic and Olmsted Medical Center institutional review boards.

Using explicit criteria, trained and experienced nurse abstractors reviewed all medical records in the community for consenting VTE cases from date first seen by a REP health care provider until incident or recurrent VTE, as previously performed.^1,2,5,19-21  Data were collected on dates and types of incident and recurrent VTE events, baseline characteristics (see supplemental Data), dates of heparin and warfarin initiation and completion and inferior vena cava filter placement, VTE risk factors, and vital status at last clinical contact, as described in detail elsewhere.^5,19,21-23  Using data from the REP Cost Data Warehouse (see supplemental Data), hospital admission and discharge dates were compiled for all Olmsted County residents (2005-2010), including ICD-9-CM discharge diagnosis codes. Individual level in-hospital VTE prophylaxis data were only available for Olmsted County residents admitted to a Mayo Clinic hospital and were captured from their Mayo electronic medical record, including pharmacy lists of medications dispensed in hospital (2005-2007) and hospital order entry data (2008-2010). We also accessed annual 2005-2010 data on number of US hospital discharges and average lengths of stay by age, sex, race, and calendar year from the National (Nationwide) Inpatient Sample (see supplemental Data).²⁴ 

The number of VTE events was summed per year within each Olmsted County resident (a resident could have an incident and recurrent event in a single calendar year), categorized by whether the event was hospitalization related (occurred in-hospital or within 92 days [365 days/4, or ∼3 months] after hospital discharge) or unrelated to recent hospitalization. We calculated total days in hospital for each resident for each calendar year using the REP timeline (see supplemental Data).¹⁷  Total hospital days within Olmsted County were then summed across individuals to obtain total Olmsted County hospital days. To include the full period of VTE risk related to hospitalization, we added 92 days to the end of each hospital stay before summing total days at risk within an individual. If a person had multiple hospital stays within a 92-day window, we summed the total number of days between hospital stays plus the 92 days after the last hospitalization. Using the number of hospital-related incident or recurrent VTE events per year as the numerator and the number of hospital bed–days occupied by Olmsted County residents for that year as the denominator, we estimated annual hospitalization-related VTE attack rates by age and sex expressed as per 10 000 hospital bed–years for the years 2005-2010. VTE attack rates unrelated to hospitalization were obtained by summing VTE events occurring in the community and dividing by total Olmsted County resident time not in hospital (see supplemental Data). To allow for the effect of multiple events per person, standard errors and 95% CIs for estimates of yearly VTE attack rates were determined using the jackknife method.²⁵  The jackknife is nonrandom and asymptotically valid under very general assumptions. Rates for leg DVT alone and for PE with or without DVT were calculated similarly. Applying these rates to the number of US hospital bed–days per year for the years 2005-2010, estimated from NIS, and to US census data, we estimated the total annual number of hospital-related (and in-hospital) and non–hospital-related VTE events among US residents, both overall and separately for whites and blacks. To test for a trend in Olmsted County VTE attack rates over time, we used a generalized linear model of VTE attack rate (counts) with a Poisson error, using a log link and the in-hospital bed–years as an offset. We used total counts per age group (0-39 years, 40-64 years, 65-84 years, and ≥85 years), sex, and calendar years 2005, 2006, 2007, 2008, 2009, and 2010. Bed-years were summed within the above categories. Because of expected clustering within person (multiple VTE events in a given calendar year), we adjusted for overdispersion using the Pearson χ² divided by the degrees of freedom. Modeling was done for Olmsted County residents with a hospitalization-related VTE (in-hospital and in-hospital plus 92 days) for all Olmsted County hospitals, for the subset of residents with a hospitalization-related VTE for Mayo Clinic hospitals only, and for residents with VTE events unrelated to hospitalization.

For VTE prophylaxis analyses, all adult (≥18 years old) Olmsted County residents admitted to a Mayo Clinic hospital (2005-2010) for ≥24 hours were included. Residents hospitalized for chemical dependency or psychiatric evaluation or treatment were excluded. For 2005-2007, any resident inpatient with an electronic dispensed indication for sequential compression devices (SCDs), warfarin, unfractionated heparin (UFH), and/or low-molecular-weight heparin (LMWH) was considered as receiving prophylaxis. For 2008-2010, any resident inpatient with an electronic order stating that VTE prophylaxis was not needed, or who had an electronic order or dispensed medication indication for SCDs, warfarin, UFH, or LMWH, was considered as “receiving indicated VTE prophylaxis.” Annual prophylaxis rates in 2005-2010 by age and sex group were estimated. Duration of prophylaxis was reported in hours; mean and standard deviation (SD) were determined by dividing the duration of prophylaxis by the number of hospitalizations occurring with the given prophylaxis recorded. Prophylaxis failure rates (2008-2010) by prophylaxis type were estimated by dividing the number of VTE events that occurred during or after that prophylaxis type by the total number receiving that prophylaxis type. For patients receiving multiple prophylaxis types, the prophylaxis type assigned for analysis purposes was anticoagulant-based prophylaxis (with or without SCDs), SCDs alone, or no prophylaxis indicated. Prophylaxis failure rates in 2008-2010 by prophylaxis type were calculated with 95% CIs based on an assumed Poisson distribution. Charlson²⁶  and Elixhauser²⁷  comorbidity scores (see supplemental Data) were calculated as respective “number of diseases” counts for each Mayo Clinic hospital stay 2005-2010 for all adult residents; each hospitalization was 1 observation. Models of Mayo Clinic hospital VTE attack rates were then performed using generalized linear regression with a Poisson error and a log link function, adjusting for sex, age-group, and calendar year as well as either the mean unweighted Charlson score or the mean Elixhauser score. SAS version 9.4 was used in all analyses. All tests were 2 sided with an α of 0.05.

Over the 6-year period 2005-2010, 855 residents developed a first lifetime VTE, with a mean (±SD) patient age at VTE onset of 61.7 ± 18.1 years (median, 62.0 years; range, 2-100 years); 449 (52.5%) were female. The mean (±SD) patient age at VTE onset for females and males was 62.7 ± 19.9 (median 64.0 years; range, 2-100 years) and 60.6 ± 15.9 (median, 60.0 years; range, 19-95 years), respectively. The distribution by VTE event type for these 855 residents was 51.4% DVT alone, 28.2% PE, and 20.5% PE and DVT. Over the same time frame, there were 345 recurrent VTE events among 281 residents. The distribution by recurrent VTE event type was 57.4% DVT alone, 26.4% PE, and 16.2% PE and DVT.

The VTE attack rate for current or recently hospitalized residents was 282 per 10 000 person-years (95% CI, 257-308; [in-hospital overall VTE attack rate: 1445 per 10 000 bed-years; 95% CI, 1211-1678]), while the VTE attack rate for community-dwelling residents with no recent hospitalization was 8.1 per 10 000 person-years (95% CI, 7.5-8.7). Attack rates for PE with or without DVT were 133 (117-150) per 10 000 person-years for current or recently hospitalized residents (in-hospital attack rate: 615 [474-757] per 10 000 bed-years), and 3.8 (3.4-4.2) per 10 000 person-years for community-dwelling residents. Attack rates for leg DVT alone were 149 (130-168) per 10 000 person-years for current or recently hospitalized residents (in-hospital attack rate: 829 [644-1015] per 10 000 bed-years), and 4.3 (3.9-4.8) per 10 000 person-years for community-dwelling residents. Annual VTE attack rates, 2005-2010, overall and by VTE event type and by relationship to current or recently hospitalized (in-hospital) vs community-dwelling resident are shown in Table 1. The Olmsted County overall, in-hospital and nonhospitalized VTE attack rates did not change significantly over the calendar years 2005-2010 (P > .25 for all).

Over the 6-year period, an estimated 2.974 million VTE events occurred among US residents (1.547 million hospitalization-related [0.388 million in-hospital]; 1.427 million community-related without recent hospitalization); the average total annual number of VTE events was 495 669 per year. The estimated average total number among whites and African Americans was 391 870 and 67 966 per year, respectively. The estimated total number of VTE events among US residents (2005-2010), overall and by VTE event type, and by relation to current or recently hospitalized [in-hospital] vs community dwelling resident, and calendar year are shown in Table 2. Of the average total annual number of VTE events, 257 783 (52%) were related to current or recent hospitalization (64 747 [25%] of hospitalization-related events were in-hospital), while 237 886 (48%) occurred among residents with no recent hospitalization. Among Olmsted County residents with incident VTE unrelated to current or recent hospitalization, 40.5% were idiopathic, 29.5% were related to recent immobilization, and 18.7% were related to active cancer; among women, 19.6% were related to oral contraceptives (Table 3). Among patients with recurrent VTE unrelated to current of recent hospitalization, 63.5% were idiopathic, 18.2% were related to recent immobilization, and 12.4% were related to active cancer; 9.2% received heparin or warfarin within 3 months prior to the recurrent VTE event (Table 3).

Among Olmsted County residents hospitalized at a Mayo Clinic hospital (2005-2010), the in-hospital rate of either receiving VTE prophylaxis or having an indication that prophylaxis was unnecessary increased from ∼15% to 40% in 2005 to ∼90% (ie, a near universally applied standardized VTE prophylaxis regimen) by 2010 (Figure 1); most of the rate increase was due to VTE prophylaxis. Over this same time frame, the annual current or recent hospitalization-related VTE attack rate (which ranged from 251 to 306 per 10 000 person-years) did not change significantly (P = .57; Table 4) after adjusting for age and sex. Similarly, the annual in-patient VTE attack rate, which ranged from 1155 to 1751 per 10 000 bed-years, did not change significantly (P = .37) after adjusting for age and sex. Both the Charlson²⁶  and Elixauser²⁷  comorbidity indices, which indicate severity of illness, increased significantly over this timeframe (P < .0001), and VTE was associated with each index (Charlson P value = .08; Elixhauser P value <.0001). However, in models adjusting for age, sex, and either the Charlson or Elixhauser comorbidity index, annual Mayo Clinic hospitalization–related VTE attack rates did not change significantly (Charlson: nonsignificant 2.1% rate increase [95% CI, −3.1% to 7.4%; P = .42] per year; Elixhauser: nonsignificant 1.8% rate increase [95% CI, −3.5% to 7.0%; P = .51] per year, respectively).

Detailed electronic VTE prophylaxis data were available for the 3-year period 2008-2010, during which 15 533 unique Olmsted County residents ≥18 years of age were hospitalized at a Mayo Clinic hospital for a total of 25 617 hospitalizations; 66% and 21% received pharmacologic (with or without SCD) prophylaxis and SCD prophylaxis alone, respectively, while for the remainder prophylaxis was judged to be not indicated. The median (interquartile range [IQR]) durations of in-hospital anticoagulant-based (ie, UFH, LMWH, or warfarin) and/or SCD prophylaxis, and of only anticoagulant-based prophylaxis were 70 (40, 122) and 63 (40, 112) hours, respectively. The median (IQR) duration of hospitalization was 3 (2, 5) days. Of those hospitalizations resulting in a live patient discharge, 7.4%, 0.9%, and 2.3% received a discharge prescription for warfarin, UFH, and LMWH, respectively. Over this 3-year period, 267 hospitalizations were complicated by an incident (n = 171) or recurrent (n = 96) VTE event, either in-hospital (n = 68 [25%]) or within 92 days after hospital discharge (n = 199 [75%]). The median (IQR) time from hospital discharge to VTE event was 19.5 (7, 46) days. The prophylaxis failure rates (95% CI) for warfarin, UFH, LMWH, and SCDs were 1.11% (0.77% to 1.55%), 1.17% (0.95% to 1.43%), 1.35% (0.97% to 1.83%), and 0.72% (0.48% to 1.03%), respectively. Among patients where prophylaxis was judged to be not indicated, 0.06% (0.02% to 0.15%) developed VTE. Additionally, over this 6-year period, 108 unique patients received an inferior vena cava filter as VTE “prophylaxis”; 10 (9.3%) patients developed VTE (8 leg DVT and 2 PE) after filter placement. These 108 patients were not included in the overall number of patients categorized as having received prophylaxis unless they also received pharmacologic and/or SCD prophylaxis. In a subgroup analysis of Olmsted County residents ≥18 years of age who were hospitalized at a Mayo Clinic hospital (2005-2010) and for whom sufficient data were available, there was no apparent change in bleeding due to changes in VTE prophylaxis rates, although this should be interpreted with caution (see supplemental Data).

Data on the total number of VTE events occurring in the United States annually are sparse and conflicting.^7-9  Using the longitudinal resources of the REP, we identified all symptomatic incident and recurrent DVT and PE events (including events discovered at autopsy) among residents from a well-defined geographic area (Olmsted County, Minnesota) over the 6-year period 2005-2010, and we estimated annual VTE attack (incident and recurrent VTE) rates related and unrelated to current or recent (within 92 days of discharge) hospitalization. Applying these age- and sex-adjusted rates to NIS and US census data, we estimated an average total number of VTE events among US residents of almost 500 000 per year; the estimated average total number among whites and African Americans was 391 870 and 67 966 per year, respectively (total number of US residents in 2010 = 308 745 538 [241 937 061 whites; 40 250 635 African Americans]).

Compared with community attack rates, hospitalization-related attack rates were >35-fold higher, and in-hospital rates were over 180-fold higher. However, because the duration of time that residents lived in the community with no recent hospitalization was much higher than the time residents occupied a hospital bed or were recently discharged from hospital, nearly 50% of the total estimated annual number of VTE events were unrelated to current or recent hospitalization. This observation has important implications for attempts to reduce the occurrence of VTE. If in-hospital VTE prophylaxis was universally provided and effective, only half of all annual VTE events in the United States would be prevented. In order to further reduce the occurrence of VTE, the individual patient at high risk for VTE unrelated to hospitalization must be identified. While high VTE-risk populations (ie, active cancer, trauma/fracture, leg paresis, and nursing home residents) can be identified, these characteristics have low predictive value for the individual; better risk-prediction tools for the individual are needed.

We also found that current or recent hospitalization-related (and in-hospital) VTE attack rates (adjusted for age, sex, and comorbidity) among Olmsted County residents admitted to a Mayo Clinic hospital (2005-2010) did not change significantly despite an increase in the rate of VTE prophylaxis (or an indication that prophylaxis was unnecessary) to ∼90% near the midpoint of the study period. These data support prior findings that The Joint Commission and the Centers for Medicare and Medicaid Services (US) Clinical Quality (performance) measures (available 1 May 2009) regarding in-hospital VTE prophylaxis have failed.¹⁴  Our data also suggest that this failure may have been due to an inadequate duration of prophylaxis; the overall median duration of in-hospital prophylaxis was only ∼3 days, and virtually no patients were discharged with a prophylaxis prescription. Most (75%) VTE events occurred after discharge, with a 19.5-day median time to VTE. In general, crude prophylaxis failure rates were low. Any apparent difference in crude prophylaxis failure rates by prophylaxis type should be interpreted with caution since failure rates were not adjusted for indication or comorbidity. Of note, VTE attack rates were very low (0.06%) for patients judged not to require prophylaxis. While extending the duration of prophylaxis to all patients judged to be at risk for VTE could reduce hospitalization-related VTE events,^13,28,29  the risk for anticoagulation-related major bleeding would correspondingly increase.^13,28,30  We suggest a better strategy would be to identify the individual high VTE-risk patient and target longer-duration prophylaxis to that individual.

Our study has several important strengths. Due to the unique features of the REP, our study avoids referral bias and other potential distortions from including a too-healthy population. We believe that this cohort better reflects “real world” clinical practice in contrast to a highly selected population participating in clinical trials. We included the entire spectrum of VTE disease occurring in the community, including persons with rapidly fatal and chronic care facility (eg, nursing home) VTE events who did not reach the hospital. Our study also has important limitations. Prophylaxis regimens may have changed after the end of our study period (2010), and VTE attack rates may have changed accordingly. We have no data on the number of acutely ill medical patients who received SCD alone prophylaxis or whether SCDs were used as prescribed. The age, sex, and racial distribution of Olmsted County (83% white in 2010) is similar to that for Minnesota, the upper Midwest, and the US white population; however, residents of Olmsted County exhibit higher median income and education level compared with these geographic regions.¹⁸  While no single geographic area is representative of all others, the underrepresentation of minorities may compromise the generalizability of our findings to different racial and ethnic groups.

In conclusion, almost 500 000 incident or recurrent VTE events occur in the United States annually, and almost 50% of these events are unrelated to current or recent hospitalization. Despite the achievement of near universal in-hospital VTE prophylaxis among those in whom prophylaxis was indicated, hospitalization-related VTE attack rates did not change significantly, suggesting that such prophylaxis measures have failed. Additional efforts are needed to identify the individual inpatient and outpatient at high risk for incident and recurrent VTE and target (longer duration) primary and secondary VTE prophylaxis to that high-risk individual who would benefit most.

Research reported in this publication was supported in part by the National Institutes of Health, National Heart Lung and Blood Institute (grant R01HL66216) (J.A.H.) and was made possible by the Rochester Epidemiology Project (National Institutes of Health, National Institute on Aging [grant R01AG034676]). Research support also was provided by Mayo Foundation.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. All authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analyses.

Contribution: J.A.H. conceived of the study, collected the data, interpreted the analyses, and wrote the manuscript; D.J.C. and T.M.P. participated in data collection, performed the statistical analyses, and participated in manuscript preparation; A.A.A. participated in data collection, analyses interpretation, and manuscript preparation; and K.R.B. participated in data collection, directed the statistical analyses, and participated in manuscript preparation.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: John A. Heit, Stabile 6 Hematology Research, Mayo Clinic, 200 First St SW, Rochester, MN 55905; e-mail: heit.john@mayo.edu.