Trends in the Incidence of Deep Vein Thrombosis and Pulmonary Embolism:  A 25-Year Population-Based Study FREE

THE REPORTED annual incidence of venous thromboembolism varies widely, ranging from 43.7 to 145.0 per 100000 (published rates were age- and sex-adjusted to the 1980 US white population) for deep vein thrombosis and 20.8 to 65.8 per 100000 for pulmonary embolism.^1- 5 A number of study design factors may have contributed to the wide variation in reported rates. For example, different studies identified cases using data from a variety of sources, including patient questionnaires,¹ the US National Hospital Discharge Survey,² or Medicare claims.⁶ Because of diagnostic uncertainty or misclassification, data from these sources may have underestimated or overestimated the actual incidence rate. Moreover, none of these studies accurately separated initial from recurrent events or included events discovered during autopsy. Studies that identified cases solely by review of inpatient medical records may have underestimated the true incidence since cases not occurring in an acute-care hospital (eg, cases from nursing homes or other long-term care facilities and those in which death occurred suddenly within the community) could be missed.³ Some studies report data from only selected populations, such as patients aged 65 years or older,⁶ patients referred to tertiary care centers for diagnostic evaluation and treatment,⁴ or patients from different geographic districts.⁵ Furthermore, there are no studies of trends in the incidence of venous thromboembolism over time. Due to these limitations in existing data, we performed a study to estimate the incidence of deep vein thrombosis and pulmonary embolism within a well-defined geographic population and describe trends in incidence over time.

Using the data resources of the Rochester Epidemiology Project,⁷ we identified the inception cohort of residents of Olmsted County, Minnesota, with a first-lifetime episode of deep vein thrombosis or pulmonary embolism during the 25-year period from 1966 through 1990. Olmsted County (1990 population, 106470) is located 144 km southeast of Minneapolis, Minn. Population-based epidemiological studies are possible in this setting because medical care is largely self-contained within the community. Most of the care is provided by the Mayo Clinic, Rochester, Minn, with approximately 1000 staff physicians and 2 large affiliated hospitals, and the Olmsted Medical Center, with approximately 75 staff physicians and the affiliated Olmsted Community Hospital, Rochester. The Mayo Clinic and its affiliated hospitals (Saint Marys and Rochester Methodist Hospital, Rochester) have maintained a common medical record system since 1907. The Mayo Clinic unit medical record contains both inpatient and outpatient data that are easily retrievable for review. The medical diagnoses and surgical procedures entered into these medical records have been indexed in an automated form since 1935. The index includes diagnoses made during outpatient office visits and clinic consultations, emergency department care, nursing home care, hospitalizations, autopsy examinations, and death certification.⁸ The medical records of the Olmsted Medical Center and the other medical care providers who serve the residents of Olmsted County are also indexed and retrievable. Thus, details of the medical care provided to the residents of Olmsted County are available for study.

We conducted a retrospective, population-based study to identify all Olmsted County residents with first-lifetime onset of deep vein thrombosis or pulmonary embolism from January 1, 1966, through December 31, 1990. A master list of potential Olmsted County residents with deep vein thrombosis, pulmonary embolism, pulmonary infarction, or similar diagnoses or who had any diagnostic test or procedure used in the diagnosis of deep vein thrombosis or pulmonary embolus, was constructed by searching the computerized indexes of medical diagnoses and surgical procedures, as well as all available databases for diagnostic tests, billing data, death certificates, and autopsy diagnoses for Olmsted County residents. The following data sources were included. First, the Medical Diagnostic Index was searched to identify all patients with diagnoses of (a) deep vein thrombosis, phlebitis, thrombophlebitis, phlegmasia alba dolens, phlegmasia cerulea dolens, and similar terms for venous thrombosis; (b) pulmonary embolism, pulmonary infarction, or similar diagnoses; and (c) varicose veins, varicose ulcer, stasis ulcer, venous insufficiency, postphlebitic syndrome, superficial vein thrombosis, venous stasis, and similar diagnoses; and (d) patients undergoing a radionuclide study for venous thromboembolism, including perfusion or ventilation-perfusion lung scans, radiolabeled fibrinogen leg scanning, or radionuclide venography. Second, the Surgical Procedure Index was searched to identify all patients who underwent a surgical procedure for deep vein thrombosis or pulmonary embolism, including vein ligation, inferior vena cava ligation or interruption, placement of an inferior vena cava filter or similar device, and venous or pulmonary thrombectomy. Third, multiple radiology databases of the Mayo Clinic were searched for patients who underwent venography, pulmonary angiography, duplex ultrasonography, computed tomography, or magnetic resonance imaging to identify any additional patients with diagnoses of deep vein thrombosis, pulmonary embolism, or pulmonary infarction. (During the study period all computed tomography and magnetic resonance imaging studies for Olmsted County residents were performed at the Mayo Clinic and their diagnoses were included in the radiology databases.) Fourth, the echocardiography computer database of the Mayo Clinic was searched to identify patients with diagnoses of pulmonary embolism, pulmonary hypertension, or similar diagnoses. Fifth, data from the Mayo Clinic vascular laboratory logbook were entered into a computer database for this study, and all patients undergoing impedance plethysmography, Doppler ultrasonography, or duplex ultrasonography for venous disease were identified. Finally, computerized billing databases of the Mayo Clinic were searched to identify all patients with charges for noninvasive vascular laboratory tests, including impedance plethysmography and Doppler ultrasonography. The final master list of potential cases contained 9046 individuals whose complete (inpatient and outpatient) medical records from all providers of health care in Olmsted County were reviewed for the study.

Each episode of deep vein thrombosis or pulmonary embolism was categorized into the highest of 3 levels of diagnostic certainty (definite, probable, or possible) based on the following criteria. A deep vein thrombosis was categorized as follows: definite when confirmed by venography, computed tomography, magnetic resonance imaging, or pathologic examination of thrombus removed during surgery or autopsy; probable if testing for the definite level of diagnostic certainty was either not performed or results were indeterminate and the results of at least 1 of the following noninvasive tests were positive: impedance plethysmography, continuous wave Doppler ultrasonographic examination performed in the Mayo Clinic vascular laboratory, compression duplex ultrasonography, radionuclide venography, or radiolabeled fibrinogen leg scanning; and possible if confirmatory tests were not done or results were indeterminate and (a) the medical record indicated that a physician made a diagnosis of deep vein thrombosis (or possible deep vein thrombosis), (b) signs and symptoms consistent with deep vein thrombosis were present, and (c) the patient underwent therapy with anticoagulants (heparin sodium, warfarin sodium, or a similar agent) or a surgical procedure for deep vein thrombosis. A pulmonary embolism was categorized as follows: definite, when confirmed by pulmonary angiography, computed tomography, magnetic resonance imaging, or pathologic examination of thrombus removed during surgery or autopsy; probable if testing for the definite level of diagnostic certainty was either not performed or results were indeterminate and a perfusion or ventilation-perfusion lung scan was interpreted as high probability for pulmonary embolism; and possible if confirmatory tests were either not done or results were indeterminate and (a) the medical record indicated that a physician made a diagnosis of pulmonary embolism, (b) signs and symptoms consistent with pulmonary embolism were present, and (c) the patient underwent therapy with anticoagulants (heparin, warfarin, or a similar agent) or a surgical procedure for pulmonary embolism such as placement of an inferior vena cava filter. A short period of anticoagulation therapy while awaiting completion of diagnostic evaluation for either suspected deep vein thrombosis or pulmonary embolism was insufficient grounds for inclusion as possible venous thromboembolism. An episode of venous thromboembolism consisting of both deep vein thrombosis and pulmonary embolism was categorized into the highest level of diagnostic certainty present for either manifestation. Because pulmonary embolism is a complication of deep vein thrombosis, the results are presented as deep vein thrombosis alone or pulmonary embolism with or without deep vein thrombosis.

Venous thromboembolism events that were first identified during the patient's lifetime and met our criteria, and events confirmed during autopsy that had been objectively verified by an invasive or noninvasive test on the date of death were classified as events discovered before death. Events confirmed during autopsy meeting only our diagnostic criteria for a possible deep vein thrombosis or pulmonary embolism on the date of death, or events discovered during autopsy for which medical record review could not establish the date of onset, were classified as events discovered after death.

Mayo Clinic pathologists performed all autopsy examinations and completed the respective death certificates of persons dying within Olmsted County during the study period. For purposes of analysis, pulmonary embolism events discovered after death were classified as a "cause of death" if the pathologist labeled it as such in the autopsy report or the death certificate listed pulmonary embolism as an immediate or underlying cause of death or included pulmonary embolism in part 1 of the death certificate. Pulmonary embolism events discovered after death were classified as a contributory cause of death if the pathologist completing the death certificate included pulmonary embolism as a contributing cause or other significant condition on part 2 of the death certificate. Pulmonary embolism events first identified on postmortem examination but not specifically labeled as a cause of death in the autopsy report or listed on the death certificate were categorized as not a cause of death.

Eligibility for this analysis required that patients have a first-time episode of deep vein thrombosis or pulmonary embolism while residents of Olmsted County during the study period from January 1, 1966, through December 31, 1990. Residency in Olmsted County at the time of first diagnosis of venous thromboembolism was confirmed by review of each patient's medical record to verify the patient's address on the date of diagnosis. In making these determinations, we had access to a mean (±SD) of 30.5 ± 19.2 years (median, 30 years) of documented medical history prior to the first diagnosis of venous thromboembolism in these patients.

Four experienced nurse abstractors reviewed the complete inpatient and outpatient medical records from all providers of care in the community for each potentially eligible patient. Baseline characteristics were abstracted for each patient who met the inclusion criteria, including age at incident event, sex, calendar year of diagnosis, type of event (deep vein thrombosis, pulmonary embolism, or both), and category of the event (definite, probable, or possible). Data were entered directly into a relational database that included checks for ranges and consistency. During the first year, a sample of medical records was independently reviewed by all abstractors to provide a measure of interobserver and intraobserver variability to identify and correct problems in data collection, interpretation of definitions, and application of study criteria. Subsequently, we performed an extensive series of checks for consistency, proper sequences of dates, and evaluation of missing or incomplete data. All edits were reviewed with the nurse abstractor who originally collected the data. If necessary, medical records were reviewed again, and questions were resolved with a physician investigator.

Annual incidence rates (per 100000 population) were calculated using incident cases of deep vein thrombosis or pulmonary embolism as the numerator and age- and sex-specific estimates of the population of Olmsted County as the denominator. The population at risk was estimated using census data for 1960, 1970, 1980, and 1990, with linear interpolation for the intercensal years. Age- and sex-specific incidence rates were calculated overall and for each calendar year from 1966 through 1990. Confidence interval (CI) estimates were based on the Poisson distribution. The population of Olmsted County was 96% white in 1990 and, accordingly, the sex-specific and overall incidence rates were directly adjusted to the 1980 US white population. Poisson regression models were used to assess the relationship of crude incidence rates to year of diagnosis, age at diagnosis, and sex. Overall incidence of venous thromboembolism was modeled; separate models were constructed for the incidence of deep vein thrombosis, pulmonary embolism (with or without deep vein thrombosis), definite or probable venous thromboembolism, definite or probable venous thromboembolism excluding pulmonary embolism discovered after death categorized as not a cause of death, and possible venous thromboembolism. In all such models, the year of diagnosis was grouped into 3-year intervals (4-year interval for 1966-1969) because the data were too sparse for finer intervals. Age at diagnosis was grouped by 5-year intervals starting at age 15 years for the models analyzing overall incidence and by 10-year intervals for the analysis of the subgroups. The midpoint of each age group or year of diagnosis group was used as the independent variable for modeling linear (or higher order) relationships to incidence. We sought the simplest model first, preferring linear terms for the ordered variables (year at diagnosis and age at diagnosis) and only investigating second-degree terms if such a model fit well and if the main effects were significant at the P=.05 level.

After screening 9046 potential cases, we identified 2218 residents of Olmsted County who had first-time diagnoses of deep vein thrombosis or pulmonary embolism during their lifetimes between January 1, 1966, and December 31, 1990. The mean (±SD) age at onset was 61.7 ± 20.4 years, and 1244 patients (56%) were female. Of the 2218 patients, 938 (42%) had deep vein thrombosis, 969 (44%) had pulmonary embolism, 3 (0.14%) had chronic thromboembolic pulmonary hypertension, and 308 (14%) had evidence of both deep vein thrombosis and pulmonary embolism at the time of diagnosis.

The average annual incidence rate of venous thromboembolism in Olmsted County during the 25-year study period (age- and sex-adjusted to the 1980 US white population) was 117 per 100000 (95% CI, 112-122). The incidence of deep vein thrombosis alone was 48 per 100000 (95% CI, 45-51), and the incidence of pulmonary embolism (with or without associated deep vein thrombosis) was 69 per 100000 (95% CI, 65-73). Age- and sex-specific incidence rates for the entire 25-year study period and the most recent 5-year period are shown in Table 1. The incidence rates were somewhat higher in females during their childbearing years, and the incidence rates were generally higher in men older than 45 years. The overall age-adjusted incidence rate was 130 per 100000 per year (95% CI, 121-138) in males and 110 per 100000 (95% CI, 104-116) in females (male-female ratio, 1.2:1).

Incidence rates increased markedly with age for both males and females (Figure 1) and for both deep vein thrombosis and pulmonary embolism (Figure 2). Pulmonary embolism accounted for an increasing proportion of venous thromboembolism with increasing age for both sexes. Only 4 venous thromboembolism events occurred in Olmsted County residents younger than 15 years during the entire 25-year study period. After excluding these 4 patients, the age- and sex-adjusted venous thromboembolism incidence rate for Olmsted County residents aged 15 years or older was 149 per 100000 annually (95% CI, 143-155) (Table 2). There was a higher annual age-adjusted rate for males (165 per 100000; 95% CI, 154-175) than females (140 per 100000; 95% CI, 132-148). The annual incidence of deep vein thrombosis was 61 per 100000 (95% CI, 57-65), and the incidence of pulmonary embolism was 88 per 100000 (95% CI, 83-92).

Among patients in our cohort who were aged 15 years or older, 215 pulmonary embolism events discovered after death were categorized as "not a cause of death" (eg, pulmonary embolism not stated in either the autopsy report or the death certificate as an immediate, underlying, or contributory cause of death; Table 2). After excluding these events, the age- and sex-adjusted incidence of pulmonary embolism in Olmsted County residents aged 15 years or older was 73 per 100000 annually (95% CI, 68-77). We also stratified events of both deep vein thrombosis and pulmonary embolism based on the level of diagnostic certainty (eg, definite, probable, or possible events; Table 2). After excluding venous thromboembolism events with the lowest level of diagnostic certainty (possible events) and pulmonary embolism events discovered after death and categorized as not a cause of death, the age- and sex-adjusted incidence of deep vein thrombosis in Olmsted County residents aged 15 years or older was 28 per 100000 annually (95% CI, 26-31), and the incidence of pulmonary embolism was 42 per 100000 (95% CI, 39-46).

The age- and sex-adjusted annual incidence of venous thromboembolism declined during the 25-year study period (Figure 3). Overall, incidence rates were relatively high between 1966 and 1976, declined by approximately 35% between 1977 and 1979, and remained relatively stable during the period from 1980 through 1990. The decrease in incidence was most marked for pulmonary embolism (with or without deep vein thrombosis), while the incidence of deep vein thrombosis remained relatively constant. The overall age- and sex-specific incidence rate for the most recent 5-year period (1986-1990) was 96 per 100000 annually (95% CI, 87-106), and for Olmsted County residents aged 15 years or older the annual rate was 122 per 100000 (95% CI, 111-134).

Differences in the incidence of venous thromboembolism by age, sex, calendar year, and type of event (pulmonary embolism with or without deep vein thrombosis vs deep vein thrombosis alone) were examined using Poisson regression (Figure 4). The average annual incidence of pulmonary embolism changed over time (P<.001), with rates in more recent years being lower than those earlier in the study period for both sexes and all age groups. However, the changes by calendar year in the incidence of deep vein thrombosis differed by sex. For all age strata of males, the incidence of deep vein thrombosis did not change significantly during the study period. However, for females, the incidence of deep vein thrombosis by calendar year decreased among those aged younger than 55 years and increased among women aged older than 60 years.

When differences in incidence by calendar year were analyzed by level of diagnostic certainty (definite or probable vs possible), the decrease in the incidence of venous thromboembolism over time was predominantly due to a large decrease in venous thromboembolism categorized as possible (eg, venous thromboembolism diagnosed by clinical criteria alone) as shown in Figure 5. For both sexes and all age strata, the incidence of possible venous thromboembolism decreased, while the incidence of definite or probable venous thromboembolism increased slightly during the study period.

Olmsted County autopsy rates declined during the study period, from 50% to 60% of deaths in the 1960s to 30% to 35% in more recent years.⁹ The incidence of pulmonary embolism discovered after death also decreased during the study period, including events categorized as a cause of death as well as events categorized as not a cause of death (Figure 6). To determine if the decrease in pulmonary embolism incidence was due to decreased detection because of declining autopsy rates, we tested for a correlation between the incidence of pulmonary embolism discovered after death and the autopsy rate after adjusting for calendar year. The incidence of pulmonary embolism discovered after death and categorized as not a cause of death did not correlate significantly with autopsy rates after adjusting for calendar year (partial correlation coefficient, −0.003 [P=.99]). However, the partial correlation coefficient for pulmonary embolism discovered after death and categorized as a cause of death approached significance (partial correlation coefficient, 0.38 [P=.07]), suggesting possible underascertainment of important events of pulmonary embolism.

Our study demonstrates that venous thromboembolism is a major national health problem, with an overall age- and sex-adjusted incidence of more than 1 per 1000 annually. Based on these data, the incidence of venous thromboembolism is virtually equivalent to the incidence of stroke.¹⁰ Using our age- and sex-specific incidence rates for the most recent 3-year period, 1988 to 1990, projected to the 1990 US white population, we estimate that at least 201000 new cases of venous thromboembolism occur in this country annually, of which 107000 are deep vein thrombosis alone and 94000 are pulmonary embolism (with or without deep vein thrombosis). Additional cases would occur among the nonwhite population. However, we were unable to address venous thromboembolism incidence among minority populations due to the demographic structure of the Olmsted County population.

While our age- and sex-adjusted deep vein thrombosis incidence rate was slightly higher than the rate found in the only similar population-based study (48.3 vs 43.7 per 100000 annually), our annual age- and sex-adjusted pulmonary embolism incidence rate was more than 3-fold higher (68.9 vs 20.8 per 100000 annually).³ Even after excluding events discovered after death categorized as not a cause of death, our rate remained more than 2.5-fold higher (57.2 per 100000). We believe the differences in our findings are best explained by more complete case ascertainment. Our study identified an inception cohort from a well-defined geographic population⁷ and included the full spectrum of disease in all clinical settings (the community, nursing home, and hospital) in which venous thromboembolism may occur. In addition to our access to both outpatient and inpatient medical records, we used information from autopsy findings and death certificates to ensure essentially complete ascertainment of clinically recognized disease. Autopsy rates remained high in Olmsted County throughout the study period and even at the end of the 25-year period were approximately 3-fold higher than the rate in the general US population according to Nemetz et al⁹ and P. N. Nemetz, PhD (written communication, August 13, 1997). This was especially notable for adults aged 65 years or older, the ages in which most venous thromboembolism occurs. In 1989, the Olmsted County autopsy rate was 30% for individuals aged 65 to 74 years, 27% for those aged 75 to 84 years, and 24% for those aged 85 years or older; these rates were 4-, 6-, and 12-fold higher, respectively, than the average US autopsy rates in these age groups.

Our study further demonstrates that venous thromboembolism is predominantly a disease of older age; only 4 incident events occurred among patients younger than 15 years during the entire 25-year study period. These findings are consistent with previous reports.^3,6 However, our study identified several additional findings. In patients younger than 55 years, the incidence of venous thromboembolism was higher in females. This observation may relate to differential exposure to clinical risk factors by sex and age (eg, pregnancy, postpartum state, or oral contraceptive use among younger women). The increase in venous thromboembolism incidence with increasing age was somewhat greater among men than women, and pulmonary embolism (with or without deep vein thrombosis) accounted for an increasing proportion of events of venous thromboembolism with increasing age. These results might be explained by differential case finding by age because of a higher clinical index of suspicion for venous thromboembolism among older patients, a higher frequency of pulmonary embolism symptoms or signs with increasing age, or a greater exposure to venous thromboembolism risk factors with increasing age. Alternatively, age- or sex-related changes in peripheral vein vascular biology may promote both an increased propensity for local thrombosis and venous thrombus growth and embolization. Further studies addressing the effect of age and sex on the vascular biological characteristics of venous thrombosis are needed to resolve these issues.

We found important changes over time in the incidence of venous thromboembolism for the 25-year study period. Overall, venous thromboembolism incidence decreased for both males and females and across all age groups, with the lowest rates occurring in the last 15 years of the study period. There was a complex pattern in the incidence of deep vein thrombosis by calendar year, with no change in men, a decrease in young females, and an increase in older women. On the other hand, the incidence of pulmonary embolism decreased in both males and females across all age groups. However, when we tested for a correlation between autopsy rates and incidence rates of pulmonary embolism discovered after death by calendar year, events of pulmonary embolism categorized as a cause of death approached significance while events categorized as not a cause of death did not. This finding suggests that the decline in autopsy rates during the study period may have been associated with underascertainment of clinically important events of pulmonary embolism. Thus, we cannot exclude the possibility that the decreasing trend in the incidence of pulmonary embolism by calendar year was confounded by decreasing autopsy rates. Several factors may explain the complex pattern of changes in the incidence of deep vein thrombosis over time, including improvement in the management of pregnancy and the postpartum state, changes in physicians' threshold for the use of newer noninvasive diagnostic testing in patients with symptoms suggestive of venous thromboembolism, and changes in the burden of comorbid conditions that are risk factors for venous thromboembolism.

Our observed changes in incidence rates cannot be explained by a change in the demographics of the Olmsted County population at risk since all the incidence rates were age- and sex-adjusted to a standard population. Trends in migration also are unlikely to explain our results since we required confirmation of residency in Olmsted County at the time of onset of venous thromboembolism. Moreover, 94% of our entire cohort (95% of males and 93% of females) were Olmsted County residents for at least 1 year prior to the onset of deep vein thrombosis or pulmonary embolism. When cohort residency was assessed by age group, 78% of those aged 15 to 24 years, 87% of those aged 25 to 34 years, and 96% of those aged 35 years or older resided in Olmsted County for at least 1 year prior to the onset of venous thromboembolism. Taken together, these findings suggest that migration into Olmsted County for reasons related to health care is an unlikely explanation for the incidence trends observed in our study.

In summary, venous thromboembolism is a major national health problem, especially among the elderly. The incidence of venous thromboembolism increases markedly with advancing age, and pulmonary embolism, a potentially fatal complication of deep vein thrombosis, represents an increasing proportion of total venous thromboembolism events with advancing age. Although the incidence of pulmonary embolism may be decreasing, the incidence of deep vein thrombosis over time is unchanged among men and is actually increasing among older women. These findings have serious implications for the future. As the US population ages, the absolute number of events of venous thromboembolism likely will increase. A large proportion of these events will manifest as pulmonary embolism with its associated poor survival rates. Future studies should be directed toward the role of age and sex in the vascular biologic characteristics of venous thrombosis, more precise identification of individual patients at risk, and improved primary methods of prevention.

Accepted for publication July 9, 1997.

Funded by grants HL46974 and AR30582 from the US Public Health Service, National Institutes of Health, Bethesda, Md; and the Mayo Foundation, Rochester, Minn.

We thank C. Mary Beard, RN, for assistance in managing the study; Janet Ebersold, RN, Kay Traverse, RN, Mary Lou Noterman, RN, and Susan Stotz, RN, for untiring medical record review; Randall Stick for programming; Diana Rademacher and Christine Lohse for assistance with data analysis; and Darcy Jacobson for secretarial support.

Reprints: John A. Heit, MD, Hematology Research, Plummer 549, Mayo Clinic, 200 First St SW, Rochester, MN 55905 (e-mail: heitj@rcf.mayo.edu).