Venous thromboembolism mortality and trends in older US adults, 2011–2019
Abstract
Venous thromboembolism (VTE) affects 1.2 million people per year in the United States. With several clinical changes in diagnosis and treatment approaches in the past decade, we evaluated contemporary post-VTE mortality risk profiles and trends. Incident VTE cases were identified from the 2011–2019 Medicare 20% Sample, which is representative of nearly all Americans aged 65 and older. The social deprivation index was linked from public data; race/ethnicity and sex were self-reported. The all-cause mortality risk 30 days and 1 year after incident VTE was calculated in demographic subgroups and by prevalent cancer diagnosis status using model-based standardization. Risks for major cancer types, risk differences by age, sex, race/ethnicity, and socio-economic status (SES), and trends over time are also reported. The all-cause mortality risk among older US adults following incident VTE was 3.1% (95% CI 3.0–3.2) at 30 days and 19.6% (95% CI 19.2–20.1) at 1 year. For cancer-related VTE events, the age-sex-race-standardized risk was 6.0% at 30 days and 34.7% at 1 year. The standardized 30-day and 1-year risks were higher among non-White beneficiaries and among those with low SES. One-year mortality risk decreased 0.28 percentage points per year (95% CI 0.16–0.40) on average across the study period, with no trend observed for 30-day mortality risk. In sum, all-cause mortality risk following incident VTE has decreased slightly in the last decade, but racial and socio-economic disparities persist. Understanding patterns of mortality among demographic subgroups and in cancer-associated events is important for targeting efforts to improve VTE management.
1 INTRODUCTION
Venous thromboembolism (VTE) consists of the interrelated diseases of pulmonary embolism (PE) and deep vein thrombosis (DVT). There are approximately 1.2 million cases of VTE each year in the United States.1 Around half of VTE cases are provoked by events such as immobilization, trauma, surgery, cancer, or hospitalization, whereas the other half of cases are not associated with any triggering events (unprovoked).1 In 2010, the all-cause mortality risk among Medicare beneficiaries following hospitalized DVT was 5.1% at 30 days and 19.6% at 1 year.2 Following hospitalized PE, the all-cause mortality risk was 9.1% at 30 days and 19.6% at 6 months (1-year data were not reported).3 Concomitant cancer is independently associated with increased risk of death in people with VTE.4 As observed in many areas of healthcare, there are disparities in VTE incidence and outcomes by race/ethnicity and possibly by socio-economic status (SES).2, 3, 5, 6 It is important to define the differences in VTE outcomes so that steps can be taken to target efforts to improve VTE management in specific populations.
Though existing published estimates provide valuable insights, changes in VTE diagnosis and treatment over the past decade have impacted VTE management, and, potentially, mortality outcomes. Imaging advances have included the widespread use of spiral computed tomography to diagnose PE and concomitant increased detection of subsegmental PE.7 This more sensitive test, which detects smaller PEs, may result in lower overall mortality rates.7 Conversely, diagnostic tools including age-adjusted D-dimer thresholds and various risk prediction tools have made it possible to rule out low-risk nonspecific scenarios with the goal of minimizing imaging tests that lead to unnecessary anticoagulation therapy,8 however, these tools are variably employed in clinical practice. Perhaps the most impactful change has been the expansion of treatment options for VTE management, with approval of the first of several direct oral anticoagulants (DOACs) in 2012. DOACs have been shown to have lower bleeding risk while being no less effective than established anticoagulant therapies (i.e., warfarin and heparins).8-15 Additionally, outpatient management for acute VTE has gradually increased.16
Because of changes in the diagnosis and management of VTE in the past decade, we sought to estimate the 30-day and 1-year all-cause mortality risk in people newly diagnosed with VTE. Additionally, given known disparities in medical care by race/ethnicity and socio-economic status, we evaluated whether risks differed by race/ethnicity and SES overall, as well as by whether the VTE was cancer–related (overall, and specifically with breast, prostrate, lung or colorectal cancer). We further estimated trends in mortality following VTE for the years 2011–2019.
2 METHODS
2.1 Study cohort
2.1.1 Medicare
We used the Medicare 20% sample for the present analysis. Medicare is a federal health insurance program that covers the vast majority of all US citizens aged 65 and older.17 For context, the 20% sample consisted of 11.4 million individuals in 2020.18 Thirty-six percent of the Medicare-eligible population is enrolled in fee-for-service coverage of Medicare Parts A (hospital insurance), B (medical insurance), and D (prescription pharmacy coverage).17, 18 This 36% of the US population aged 65 and older excludes those enrolled in Medicare Advantage plans (the most common alternative to fee-for-service coverage, also called Medicare Part C) as well as those not enrolled in Part D and therefore is slightly over-representative of people of a lower SES and with a greater burden of comorbidities.18-20 It is also less representative of Black and Hispanic beneficiaries.20 The Medicare 20% sample datasets contain detailed inpatient and outpatient medical claims which are linked to outpatient prescription drug claims and person-level enrollment information including geographic location (ZIP code) and self-reported race/ethnicity. Use of these data was approved by the Institutional Review Board at the University of Minnesota.
2.1.2 Study population
The study population for this analysis consisted of individuals enrolled in fee-for-service coverage of Medicare Parts A, B, and D with International Classification of Diseases (ICD) 9 or 10 codes indicating an incident VTE event from 2011 to 2019, confirmed with an anticoagulant prescription. Individuals were required to be enrolled for at least 90 days before their incident VTE to allow for ascertainment of prevalent VTE cases and comorbidities.
The initial sample included 402 852 individuals aged 65–99 with one inpatient or two outpatient ICD codes indicating possible VTE (Figure S1). If individuals had multiple enrollment periods, only the first was considered for inclusion in this study. The first step in defining the analytic sample was to exclude individuals who were never prescribed an anticoagulant between January 1, 2011, and December 31, 2019 (n = 95 288). Next, individuals were excluded if their first anticoagulant prescription was greater than 31 days before or after the first VTE ICD code date (n = 136 417). Then, individuals were excluded if they did not have 90 days of continuous enrollment prior to VTE (n = 42 825). Next, individuals were excluded if they did not live in the 50 US states or Washington DC (n = 207). Lastly, individuals with incident VTE less than 30 days or 365 days before December 31, 2019, were excluded from the respective 30-day and 1-year mortality analyses. Our final analytic sample for 30-day mortality was 127 224, and our final analytic sample for 1-year mortality was 114 284.
2.1.3 Incident VTE
Incident VTE was defined as one inpatient ICD code or two outpatient ICD codes (7–184 days apart) in any position indicating VTE and a new anticoagulant prescription within 31 days of the initial diagnosis. The ICD codes used to define VTE are provided in Table S1. In a validation study using a similar definition, by Sanfilippo et al., this algorithm's specificity was 99%, sensitivity 72% and its positive predictive value was 91%.21 The anticoagulant prescription date was used as the beginning of follow-up time for the identified VTE cases.
2.1.4 Mortality
Death information was available for the years 2011–2019 from the Centers for Medicare and Medicaid Services (CMS) through linkage of Medicare claims to data from the US Social Security Administration, Railroad Retirement Board, and revisions submitted by family members and is 99% complete.22, 23 For the present analysis, all-cause mortality was assessed at 30 days and 1 year. 1-year mortality risk was only assessed for VTE events occurring through 2018 because morality information was not known beyond 2019. For the same reason, individuals with a VTE event after December 1, 2019 were excluded from 30-day mortality risk estimates. These exclusion criteria were applied uniformly for all beneficiaries based on information known when they became eligible for inclusion in this analysis, namely, the date of the VTE event itself and the availability of the data required to observe the presence or absence of the outcome event (i.e., 1-year or 30-day follow-up).
2.1.5 Variables
Information on date of birth, sex, and race/ethnicity was obtained by CMS from Social Security Administration data. A beneficiary's age at the time of incident VTE was calculated from their date of birth. Race/ethnicity was reported within six mutually exclusive categories: White, Black, Asian, North American Native, Hispanic, and Other. Due to small numbers of VTE patients identifying as Asian, North American Native, and Hispanic in our sample, race/ethnicity was operationalized as White, Black, and Other for the present analysis. Race/ethnicity data from the Social Security Administration is known to be accurate for people who identify as White or Black, but frequently misclassifies individuals who identify as Asian or North American Native, or are of Hispanic/Latino origin.24
Beneficiary ZIP code was obtained for the time of incident VTE. Beneficiary ZIP code was linked with the social deprivation index (SDI) from publicly available data25 which combines seven demographic characteristics (percent living in poverty, percent with less than 12 years of education, percent single parent household, percent living in rented housing unit, percent living in overcrowded housing unit, percent of households without a car, and percent non-employed adults under 65 years of age) into a measure of area-level social deprivation.26 The index ranges from 1 to 100, with higher scores indicating greater deprivation. VTE patient SES quartiles were approximated using the area-level SDI, with the first quartile indicating low SES and the fourth quartile indicating high SES.
Prevalent cancer was defined by the presence of any malignant or in situ neoplasm ICD code in any position from inpatient or outpatient visits in the year prior to VTE diagnosis, excluding nonmelanoma skin cancer (Table S1).27 Breast, prostate, lung, and colorectal cancer cases were identified by the presence of any ICD codes in any position from inpatient or outpatient visits that are included in the Chronic Conditions Warehouse (CCW) definitions for these diseases.28 VTE events among beneficiaries with a prevalent cancer diagnosis were considered cancer-related VTE.
2.2 Statistical analysis
All-cause mortality risks and confidence intervals were calculated using logistic regression models with 30-day or 1-year mortality as the outcome and with state-level cluster robust standard errors, to account for correlation of individuals within states.29 Crude estimates of mortality risk across the entire study period were calculated using an intercept-only model. Standardized estimates of mortality risk across the study period for beneficiaries aged 65–74, beneficiaries aged 75+, men, women, Black beneficiaries, White beneficiaries, and beneficiaries of some other race/ethnicity were estimated from logistic models adjusted for age, sex, and race/ethnicity, as appropriate. Standardized estimates for each SES quartile risk were adjusted for age, sex, and race/ethnicity, as were standardized estimates for cancer-related and non-cancer-related VTE events. Standardized estimates were obtained by fitting a logistic regression model with the main predictor of interest and standardization variables. Stata's margins command was used to post-process the regression estimates and calculate standardized risks. Similarly, separate models for each subtype of cancer (breast, prostate, lung, colon, and other) were fit to obtain standardized all-cause mortality risk estimates for each cancer type.
Risk differences (RD) between demographic subgroups were estimated from the same models described above.29 We also assessed effect modification on an additive scale between cancer-relatedness and demographic variables using the models described above with an added interaction term for the association of interest. In order to evaluate 30-day and 1-year all-cause mortality risk time trends from 2011 to 2019, we assessed year of VTE (continuous) as a predictor of mortality in linear regression models with state-level cluster robust standard errors. Year of VTE was the sole predictor in the model to calculate the trend in the overall population. Trends within demographic subgroups and by cancer-relatedness were estimated using separate models for each respective subset of the sample, each adjusted for age, sex, and race/ethnicity, as appropriate. Given that a binary outcome was used in the linear regression models, we verified that the fitted models did not predict risks less than 0 or greater than 1 among the analytic sample. Analyses were conducted in SAS version 9.4 and STATA version 16.1.
3 RESULTS
3.1 Study population
The analytic sample included 127 224 people with VTE aged 65 and older. The mean age was 76.9 years (SD = 8.0), and 60.6% were female. Eighty-four percent of the VTE beneficiaries identified as White, 10.4% as Black, and 5.2% were grouped into the “other” race/ethnicity category. Cancer-related thrombosis accounted for 30.6% of VTE events (Table 1). Cancer-related VTE events were more prevalent at younger ages and in men, whereas no distinct pattern of prevalence for cancer-related VTE events was observed across racial/ethnic groups or SES quartiles.
| Cancer relatedness | Overall | Cancer-related | Non-cancer-related | ||||
|---|---|---|---|---|---|---|---|
| Non-cancer-related | N | % | N | % | N | % | |
| 88 243 | 69.4% | ||||||
| Cancer-related | 38 981 | 30.6% | |||||
| Age | 65–74 | 59 517 | 46.8% | 19 808 | 50.8% | 39 709 | 45.0% |
| 75+ | 67 707 | 53.2% | 19 173 | 49.2% | 48 534 | 55.0% | |
| Continuous | Mean = 76.9 | SD = 8.0 | Mean = 75.9 | SD = 7.3 | Mean = 77.4 | SD = 8.3 | |
| Sex | Men | 50 120 | 39.4% | 17 772 | 45.6% | 32 348 | 36.7% |
| Women | 77 104 | 60.6% | 21 209 | 54.4% | 55 895 | 63.4% | |
| Race/ethnicity | White | 107 368 | 84.4% | 33 145 | 85.0% | 74 223 | 84.1% |
| Black | 13 268 | 10.4% | 3959 | 10.2% | 9309 | 10.6% | |
| Otherb | 6588 | 5.2% | 1877 | 4.8% | 4711 | 5.3% | |
| Socio-economic Status | First quartile (SDI 69–100) | 32 980 | 25.9% | 9624 | 24.7% | 23 356 | 26.5% |
| Second quartile (SDI 45–68) | 31 606 | 24.8% | 9258 | 23.8% | 22 348 | 25.3% | |
| Third Quartile (SDI 23–44) | 30 942 | 24.3% | 9683 | 24.8% | 21 259 | 24.1% | |
| Fourth Quartile (SDI 0–22) | 31 696 | 24.9% | 10 416 | 26.7% | 21 280 | 24.1% | |
- a Based on 30-day mortality cohort.
- b Includes Asian, North American Native, Hispanic, and Other.
3.2 30-day mortality risk
From 2011 to 2019, we observed 3930 deaths within 30 days of a VTE event, which is equivalent to a 3.1% (95% CI 3.0%–3.2%) 30-day mortality risk. Standardized 30-day mortality risk estimates for demographic subgroups, overall and stratified by association with a prevalent cancer diagnosis, are included in Table 2. The 30-day age-sex-race-standardized mortality risk was 6.0% (95% CI 5.7%–6.2%) in cancer-related VTEs and 1.9% (95% CI 1.8%–2.0%) in non-cancer-related VTEs. The standardized 30-day all-cause mortality risk following VTE was 11.4% (95% CI 10.7%–12.2%) among beneficiaries with lung cancer, 4.9% (95% CI 4.2%–5.7%) among colon cancer patients, 3.5% (95% CI 3.2%–3.8%) among breast cancer patients, and 3.3% (95% CI 2.9%–3.7%) among prostate cancer patients (Table 3).
| 30-day mortality | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Overall | Cancer-related | Non-cancer-related | ||||||||
| N | n deaths | Risk (95% CI) | N | n deaths | Risk (95% CI) | N | n deaths | Risk (95% CI) | ||
| Crude overall | 127 224 | 3930 | 3.1% (3.0%-3.2%) | |||||||
| Cancer relatednessa,b,c | Non-cancer-related | 88 243 | 1658 | 1.9% (1.8%-2.0%) | ||||||
| Cancer-related | 3 981 | 2272 | 6.0% (5.7%-6.2%) | |||||||
| Ageb,c | 65–74 | 59 517 | 1501 | 2.5% (2.4%–2.6%) | 19 808 | 1090 | 4.7% (4.4%–5.0%) | 39 709 | 411 | 1.5% (1.4%–1.5%) |
| 75+ | 67 707 | 2429 | 3.6% (3.4%-3.7%) | 19 173 | 1182 | 7.1% (6.7%-7.4%) | 48 534 | 1247 | 2.2% (2.1%–2.3%) | |
| Sexa,c | Men | 50 120 | 1540 | 3.2% (3.0%–3.3%) | 17 772 | 987 | 5.7% (5.4%–6.0%) | 32 348 | 553 | 1.8% (1.6%–1.9%) |
| Women | 77 104 | 2390 | 3.0% (2.9%–3.2%) | 21 209 | 1285 | 6.1% (5.8%–6.5%) | 55 895 | 1105 | 1.9% (1.8%–2.0%) | |
| Race/ethnicitya,b | White | 107 368 | 3236 | 3.0% (2.9%–3.1%) | 33 145 | 1890 | 5.8% (5.5%–6.1%) | 74 223 | 1346 | 1.8% (1.7%–1.9%) |
| Black | 13 268 | 475 | 3.6% (3.3%–3.9%) | 3959 | 261 | 7.0% (6.4%–7.5%) | 9309 | 214 | 2.2% (2.0%–2.4%) | |
| Otherd | 6588 | 219 | 3.4% (3.1%–3.6%) | 1877 | 121 | 6.7% (6.1%-–7.3%) | 4711 | 98 | 2.1% (1.9%–2.3%) | |
| Socio-economic statusa,b,c | First Quartile | 32 980 | 1185 | 3.5% (3.3%–3.8%) | 9624 | 663 | 6.9% (6.5%–7.4%) | 23 356 | 522 | 2.2% (2.0%–2.3%) |
| second Quartile | 31 606 | 1001 | 3.2% (3.0%–3.4%) | 9258 | 546 | 6.3% (5.9%–6.6%) | 22 348 | 455 | 1.9% (1.8%–2.1%) | |
| Third Quartile | 30,942 | 928 | 3.0% (2.8%-3.3%) | 9683 | 558 | 5.8% (5.3%-6.3%) | 21 259 | 370 | 1.8% (1.7%–1.9%) | |
| Fourth Quartile | 31 696 | 816 | 2.6% (2.4%–2.8%) | 10416 | 505 | 4.9% (4.5%–5.3%) | 21 280 | 311 | 1.5% (1.4%–1.6%) | |
| 1-year mortality | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Overall | Cancer-related | Non-cancer-related | ||||||||
| N | n deaths | Risk (95% CI) | N | n deaths | Risk (95% CI) | N | n deaths | Risk (95% CI) | ||
| Crude overall | 114 284 | 22 447 | 19.6% (19.2%–20.1%) | |||||||
| Cancer relatednessa,b,c | Non-cancer-related | 79 521 | 10 611 | 13.2% (12.8%–13.6%) | ||||||
| Cancer-related | 34 763 | 11 836 | 34.7% (34.0%–35.3%) | |||||||
| Ageb,c | 65–74 | 52 938 | 8584 | 16.2% (15.8%–16.6%) | 17 552 | 5885 | 28.6% (28.0%–29.2%) | 35 386 | 2699 | 10.2% (9.8%–10.5%) |
| 75+ | 61 346 | 13 863 | 22.6% (22.2%–23.1%) | 17 211 | 5951 | 39.9% (39.1%–40.7%) | 44 135 | 7912 | 15.8% (15.4%–16.2%) | |
| Sexa,c | Men | 44 704 | 8665 | 19.9% (19.4%–20.4%) | 15 765 | 5161 | 33.4% (32.6%–34.3%) | 28 939 | 3504 | 12.6% (12.2%–13.0%) |
| Women | 69 580 | 13 782 | 19.5% (19.1%–19.9%) | 18 998 | 6675 | 35.5% (34.8%–36.1%) | 50 582 | 7107 | 13.6% (13.2%–14.0%) | |
| Race/ethnicitya,b | White | 96 403 | 18 434 | 19.1% (18.7%–19.5%) | 29 528 | 9884 | 33.8% (33.2%–34.5%) | 66 875 | 8550 | 12.7% (12.4%–13.1%) |
| Black | 12 057 | 2844 | 23.7% (22.9%–24.4%) | 3575 | 1,367 | 40.8% (39.5%–42.1%) | 8482 | 1477 | 16.5% (15.7%–17.2%) | |
| Otherd | 5824 | 1169 | 20.3% (19.4%–21.2%) | 1660 | 585 | 36.4% (34.9%–37.9%) | 4164 | 584 | 14.1% (13.4%–14.8%) | |
| Socio-economic statusa,b,c | First quartile | 29 918 | 6519 | 21.2% (20.6%–21.7%) | 8656 | 3,221 | 37.5% (36.6%–38.3%) | 21 262 | 3298 | 14.6% (14.1%–15.0%) |
| Second quartile | 28 404 | 5689 | 20.1% (19.6%–20.6%) | 8285 | 2,837 | 35.9% (34.9%–36.9%) | 20 119 | 2852 | 13.8% (13.3%–14.2%) | |
| Third quartile | 27 799 | 5210 | 19.0% (18.5%–19.5%) | 8627 | 2,852 | 33.6% (32.8%–34.4%) | 19 172 | 2358 | 12.5% (12.2%–12.9%) | |
| Fourth quartile | 28 163 | 5029 | 18.2% (17.6%–18.7%) | 9195 | 2,926 | 31.8% (31.0%–32.7%) | 18 968 | 2103 | 11.7% (11.3%–12.1%) | |
- Note: Differences in the association between cancer-related VTE and 30-day and 1-year mortality risks were tested by age (30-day mortality p < .001; 1-year mortality p < .001), sex (p = .30; p = .05), race/ethnicity (p = .44; p = .60), and SES (p = .001; p = .44).
- a Adjusted for age.
- b Adjusted for sex.
- c Adjusted for race/ethnicity.
- d Includes Asian, North American Native, Hispanic, and other.
| 30-day mortality | 1-year mortality | ||||||
|---|---|---|---|---|---|---|---|
| N | n deaths | Risk (95% CI) | N | n deaths | Risk (95% CI) | ||
| Cancer type | Lunga,b,c | 6842 | 747 | 11.4% (10.7%–12.2%) | 6113 | 3507 | 58.9% (57.8%–59.9%) |
| Colona,b,c | 4571 | 224 | 4.9% (4.2%–5.7%) | 4133 | 1397 | 34.0% (32.6%–35.5%) | |
| Breasta,c | 7061 | 242 | 3.5% (3.2%–3.8%) | 6318 | 1614 | 26.2% (25.1%–27.3%) | |
| Prostatea,c | 6667 | 230 | 3.3% (2.9%-3.7) | 5970 | 1383 | 22.3% (21.3%–23.3%) | |
| Othera,b,c | 15 911 | 987 | 6.3% (5.9%–6.7%) | 14 098 | 4865 | 35.1% (34.3%–35.9%) | |
- a Adjusted for age.
- b Adjusted for sex.
- c Adjusted for race/ethnicity.
RDs by demographic subgroups are listed in Table S2. The standardized 30-day risk was similar for men and women, higher for non-White than for White beneficiaries, higher with increasing age, and higher with lower SES. The pattern of greatly increased risk of death in cancer-related VTEs compared to non-cancer-related VTEs persisted among subgroups. Of note, the difference in 30-day mortality risk between the lowest SES quartile and the highest SES quartile was larger among individuals with cancer-related VTE (RD 1.9, 95% CI 1.3–2.6) as compared to those with non-cancer related VTE (RD 0.7, 95% CI 0.4–0.9) (Figure 36, p-interaction = .001).
3.3 1-year mortality risk
From 2011 to 2019, we observed 22 447 deaths within 1 year of an incident VTE (1-year mortality risk 19.6%; 95% CI 19.2%–20.1%). Standardized 1-year mortality risk estimates for demographic subgroups, overall and stratified by association with a prevalent cancer diagnosis, are included in Table 2. The 1-year age-sex-race-standardized mortality risk was 34.7% (95% CI 34.0%–35.3%) in cancer-related VTEs and 13.2% (95% CI 12.8%–13.6%) in non-cancer-related VTEs. The standardized 1-year all-cause mortality risk following VTE was 58.9% (95% CI 57.8%–59.9%) among beneficiaries with lung cancer, 34.0% (95% CI 32.6%–35.5%) among colon cancer patients, 26.2% (95% CI 25.1%–27.3%) among breast cancer patients, and 22.3% (95% CI 21.3%–23.3%) among prostate cancer patients (Table 3).
RDs by demographic subgroups are listed in Table S2. The standardized 1-year risk was also similar for men and women, higher for non-White versus White beneficiaries, higher with increasing age, higher with lower SES, and the pattern of greatly increased risk of death in cancer-related VTEs compared to non-cancer-related VTEs persisted among subgroups. The difference in 1-year mortality risk between the lowest and highest SES quartiles was not different among individuals with cancer-related VTE (RD 4.1, 95% CI 2.6–5.6) as compared to those with non-cancer related VTE (RD 3.5, 95% CI 3.0–4.1) (Figure S2b, p-interaction = .44).
3.4 Trends in mortality post-VTE from 2011 to 2019
There was no evidence of increasing or decreasing trends in 30-day mortality risk overall during the study period (Table 4; Figure S3a). A slight increase in 30-day mortality of 0.11 percentage points per year (95% CI 0.01–0.22), on average, was identified among Black beneficiaries. One-year mortality showed a decreasing risk over time overall, as well as among some subgroups, however, the magnitudes of change were modest (Table 4; Figure S3b). We observed differences in 1-year mortality trends for beneficiaries aged 65–74 compared to those aged 75 and older (p = .01) and for men compared to women (p < .001). After adjusting for age and race/ethnicity, the 1-year mortality risk for men decreased by 0.39 points per year (95% CI 0.21–0.56), on average, whereas for women there was no meaningful change.
| 30-day mortality | 1-year mortality | ||||||
|---|---|---|---|---|---|---|---|
| Yearly percentage-point change | 95% CI | p-value | Yearly percentage-point change | 95% CI | p-value | ||
| Crude overall | 0.01 | (−0.03, 0.05) | .60 | −0.28 | (−0.40, −0.16) | <.001 | |
| Cancer relatednessa,b,c | Non-cancer-related | −0.02 | (−0.05, 0.02) | .34 | −0.24 | (−0.34, −0.14) | <.001 |
| Cancer-related | 0.06 | (−0.05, 0.15) | .18 | −0.37 | (−0.65, −0.09) | .01 | |
| Ageb,c | 65–74 | 0.00 | (−0.04, 0.04) | .96 | −0.31 | (−0.47, −0.15) | <.001 |
| 75+ | 0.05 | (−0.01, 0.10) | .11 | −0.07 | (−0.20, 0.07) | .35 | |
| Sexa,c | Men | 0.01 | (−0.05, 0.07) | .76 | −0.39 | (−0.56, −0.21) | <.001 |
| Women | 0.04 | (−0.01, 0.09) | .16 | −0.05 | (−0.18, 0.08) | .49 | |
| Race/ethnicitya,b | White | 0.01 | (−0.03, 0.06) | .63 | −0.21 | (−0.34, −0.08) | .001 |
| Black | 0.11 | (0.01, 0.22) | .03 | 0.02 | (−0.26, 0.31) | .88 | |
| Otherd | 0.09 | (−0.08, 0.25) | .30 | −0.06 | (−0.58, 0.45) | .81 | |
| Socio-economic Statusa,b,c | First quartile | 0.07 | (−0.03, 0.16) | .17 | −0.05 | (−0.25, 0.16) | .66 |
| Second quartile | −0.02 | (−0.11, 0.06) | .59 | −0.15 | (−0.37, 0.07) | .18 | |
| Third quartile | 0.01 | (−0.05, 0.08) | .73 | −0.17 | (−0.38, 0.04) | .10 | |
| Fourth quartile | 0.07 | (−0.00, 0.13) | .06 | −0.28 | (−0.48, −0.09) | .004 | |
- Note: Differences in 30-day and 1-year mortality time trends were tested by cancer relatedness (30-day mortality p = .07; 1-year mortality p = .74), age (p = .16; p = .01), sex (p = .51; <.001), race/ethnicity (p = .19; p = .34), and SES (p = .99; p = .06).
- a Adjusted for age.
- b Adjusted for sex.
- c Adjusted for race/ethnicity.
- d Includes Asian, North American Native, Hispanic, and Other.
4 DISCUSSION
Using data from the Medicare 20% Sample (2011–2019), the all-cause mortality risk among adults 65 years and older following incident VTE was 3.1% at 30 days and 19.6% at 1 year. Mortality risk post-VTE was similar for men and women but was elevated among individuals ≥75 versus 65–74, among those with lower versus higher SES, and among non-White versus White people. Mortality risk was substantially higher among VTE beneficiaries with cancer; individuals with incident cancer-related VTE had mortality risks of 6.0% at 30 days and 34.7% at 1 year. There were also differences in mortality post-VTE according to associated cancer type. Nonetheless, mortality was not uncommon among beneficiaries with incident VTE without concomitant cancer. The risk of death was 1.9% at 30 days and 13.2% at 1 year in this population. Based on yearly trends from 2011 to 2019, there was slight evidence of declining 1-year all-cause mortality post-VTE, with no evidence of improvement in 30-day all-cause mortality. Importantly, any observations regarding all-cause mortality risk might very well be due to reasons other than changes in the VTE-related care, such as changes in the burden of underlying risk factors. Altogether, these results show there has been minimal improvement in mortality following incident VTE events in the last decade.
4.1 Overall mortality among people with VTE
When compared to prior decades, our results show that all-cause mortality risk following incident VTE may be gradually improving. A prior study based on the Medicare population in 2010 reported that all-cause mortality following incident DVT was 5.1% at 30 days and 19.6% at 1 year after adjusting for age, sex, race/ethnicity, and comorbidities.2 In the same population, all-cause mortality risk following incident PE was 9.1% at 30 days and 19.6% at 6 months.3 However, caution is needed when comparing our findings to those from 2010 due to the different methodologies employed. Specifically, the 2010 study included only hospitalized VTEs whereas we also included outpatient events. Events managed in an outpatient setting would likely be less severe, and thus associated with lower mortality risk. Additionally, we calculated 1-year mortality for both DVT and PE, whereas the prior study calculated 6-month mortality following PE. It is also possible that any observed changes are due to reasons other than VTE-related developments given that these analyses looked at all-cause mortality.
Wide variation in estimates of 30-day mortality after VTE has been reported in prior publications. In data from ~100 000 patients from 26 countries who were included in the RIETE registry, 30-day mortality following VTE was 2.6% for distal DVT, 3.3% for proximal DVT, and 5.2% for PE.30 In the Norwegian-based Tromsø study, 30-day mortality after VTE was only 1.7%.31 However, this paper was based on 710 VTE events and 12 deaths, the cohort had a wide age range at enrollment (25–97 years), and incident VTE was ascertained from the mid-1990s to early 2010s.31 The Tromsø study also evaluated 1-year mortality post-VTE and estimated it to be 7.3%.31 In the LITE study, 30-day mortality following VTE was 11%.4 However this study consisted of participants at least 45 years of age at enrollment, 265 VTE events and 29 deaths contributed to the analysis, and incident VTE was ascertained from the late 1980s to the late 1990s.4 Overall, compared to the earlier data, our results indicate that mortality risks following VTE have improved during the same time period in which earlier detection of emboli, diagnosis based on better assessment of risk, reduced risk of major bleeding with the use of DOACs, and other improvements in VTE-related medical care also arose.8
4.2 Mortality according to cancer relatedness
Mortality risk differs considerably by whether cancer was prevalent at the time of VTE. After cancer itself, thrombosis is the next leading cause of death in people with cancer and complicates their care.32 In the present analysis, the 30-day mortality risk was 1.9% in non-cancer-related VTEs and 6.0% in cancer-related VTEs. The 1-year mortality risk was 13.2% in non-cancer-related VTEs and 34.7% in cancer-related VTEs. While the historical risks of death are greater than what we see in this analysis of the Medicare 20% sample, the pattern for cancer-relatedness risk is similar.4, 31, 33 Our study additionally reported risk of death was greatest among VTE patients with underlying lung cancer, with colorectal, breast and prostate cancer having lower 30-day and 1-year mortality risks, respectively. Findings from our trend analysis do suggest that cancer-related VTE 1-year mortality risk decreased from 2011 to 2019. Importantly, we assessed all-cause mortality, so reduced mortality risks among VTE patients with cancer may be a function of improvements in cancer treatment, VTE management, or both. Preventing VTE among cancer patients and optimizing outcomes among VTE patients who develop cancer is a major clinical and research priority.34, 35
4.3 Mortality according to beneficiary demographic characteristics
Our study is among the first to examine mortality risk following VTE by SES. We found that mortality was higher among individuals with lower SES, as defined by a high area SDI. For 30-day risk, this trend was stronger for individuals with cancer-related VTEs than for individuals with non-cancer-related VTEs. This finding is not unexpected since cancer survival in the United States in general is higher for those of higher SES in most broad racial/ethnic groups.36 Reasons for the cancer mortality disparities are complex and include factors such as poorer access to timely screening and treatment among individuals of lower SES and lifestyle differences which can lead to poorer cancer outcomes.36 Differences in the pace of adopting new guidelines in VTE management and DOAC prescriptions also vary by demographic characteristics and by geographic location and may contribute to this finding.16, 37 However, as this analysis considers the general all-cause mortality risk, differences may very well be due to reasons other than VTE-specific mortality.
We also observed differences by race/ethnicity. In our study sample, the 30-day mortality risk was 3.6% for Black beneficiaries and 3.0% for White beneficiaries. The 1-year mortality risk was 23.7% and 19.1%, respectively. Beneficiaries of other racial/ethnic groups also had a pattern of elevated risk compared to White beneficiaries. Prior research has shown that reasons for racial disparities in VTE mortality are nuanced and may be related to structural racism leading to differences in treatment approach, comorbidities, insurance status, and severity of clinically-recognized VTE events.5, 38, 39
In our analysis, men showed slightly greater age-race-standardized mortality risk than women at both 30 days and 1 year. The LITE study did not find male sex to be an independent predictor of 28-day mortality risk, although power to detect an effect was low.4 Minges et al. examined risks by sex in DVT and PE separately and found conflicting patterns, with women showing a greater risk of death following incident DVT and men showing a greater risk of death following incident PE.2, 3 Overall, mortality 30 days and 1 year following VTE do not appear to vary by sex.
4.4 Trends in VTE mortality from 2011 to 2019
There was no trend in overall 30-day mortality risk and a slight decrease in overall 1-year mortality risk from 2011 to 2019. For cancer-related VTE events, there was a modest decrease in 1-year mortality and no change in 30-day mortality. This change in 1-year mortality could be due to improved tertiary VTE prevention or improved cancer treatment in general. Several demographic subgroups showed a decrease in 1-year mortality, including SES quartile 4, people aged 65–74, White beneficiaries, and men. Furthermore, men had a more rapid decrease in 1-year mortality risk than women, although men also had a slightly higher average risk across the study period. Statistical significance is largely influenced by sample size, and actual rates of change in this analysis of ~120 000 VTE patients were small. Observed trends also presented some potential concerns of widening racial disparities, however, the magnitudes were small and therefore vulnerable to residual confounding that may alter conclusions. Additionally, random chance is a potential concern given the number of subgroup comparisons made. Prior analyses of the Medicare 20% sample data for earlier years (1999–2010) found similar overall trends among hospitalized DVT patients,2 while among hospitalized PE patients they found modestly decreasing mortality overall and across all demographic subgroups both at 30 days and 6 months.3 An analysis of the US Nationwide Inpatient Sample, where the unit of study is hospitalization discharges (not unique individuals), reported that the PE mortality rate increased slightly from 2008 to 2018.40 This finding suggests that increased sensitivity of detecting PEs is not the sole driver of observed all-cause mortality risk trends in our analysis.
4.5 Strengths and limitations
There are some limitations to this study that should be considered. The Medicare fee-for-service 20% sample does not include people who are enrolled in Medicare Advantage plans, and we excluded those not enrolled in Medicare Part D. This leads to people with lower SES and greater comorbidity burden being slightly over-represented in the study, and a slight underrepresentation of Black and Hispanic beneficiaries.20 The sample also does not include anyone younger than 65, which limits the generalizability of our findings. Categorizing race/ethnicity into only three groups hides the reality that not all minority groups in the United States have the same experiences with VTE and healthcare. Misclassification is also an important issue when using administrative data. We attempted to mitigate any major errors by using validated and previously published algorithms when available.21, 27 Although there are currently no validation studies for cancer-type definitions, we used definitions from CMS and the CCW.28 We also did not attempt to differentiate PE and DVT events, due to concerns about misclassification. Specific cause of death is unable to be ascertained in robust ways from administrative data which limits our ability to determine VTE-specific mortality and mortality trends. Over the time frame of this study, there was adoption of more sensitive methods to detect DVTs and PEs which likely would not have been diagnosed in the past. Whether these events have an impact on VTE mortality over time is unclear. Any secular trends in mortality could represent differences in the VTE events being diagnosed or treatment approaches for VTE or comorbid conditions.7 Regardless, estimates of clinically recognized VTE, and their association with mortality, are valuable from a resource planning perspective. Lastly, it is important to keep in mind that several statistical tests were conducted without correction for multiple comparisons, thereby inflating the Type I error rate.
Despite these limitations, the study has notable strengths. First, it has a large sample size, which allowed us to examine trends and observe differences between groups. The sample also covers the entire United States population. Second, approximately 98% of beneficiaries' death information was linked from the Social Security Administration which is a nearly perfect source for mortality data, and overall, 99% of the deaths were validated.22, 23 Third, we linked a measure of area-level SES with each beneficiary based on ZIP code. Risk in SES subgroups has not been considered in prior studies of mortality following incident VTE.
5 CONCLUSION
Among US adults aged 65 and older, the all-cause mortality risk following incident VTE was 3.1% at 30 days and 19.6% at 1 year, and there was evidence of slight improvements in 1-year mortality over the last decade. Risk varies by demographic characteristics, being higher among patients who are older, identify as a racial/ethnic minority, or have low SES. Concomitant cancer was a potent mortality risk factor across all VTE patient groups considered. Improving health equity and optimizing management of VTE, particularly cancer-associated VTE, should continue to be areas of focus.
FUNDING INFORMATION
This work was supported by National Institutes of Health, National Heart, Lung and Blood Institute grants NIH R01 HL131579, NIH K24 HL159246 (PL), and T32 HL007779-29 (WW).
CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest.
PATIENT CONSENT STATEMENT
Administrative data were retrospectively used, per policies of the Centers for Medicare & Medicaid Services.
Open Research
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from Centers for Medicare & Medicaid Services. Restrictions apply to the availability of these data, which were used under license for this study.
