ORIGINAL INVESTIGATION Venous Thromboembolism According to Age The Impact of an Aging Population Paul D. Stein, MD; Russell D. Hull, MBBS, MSc; Fadi Kayali, MD; William A. Ghali, MD, MPH; Andrew K. Alshab, MD, MPH; Ronald E. Olson, PhD Background: With the aging of the US population, there the use of diagnostic tests over 21 years were markedly is concern that the rate of venous thromboembolism will higher in elderly than in younger patients (PϽ.001). Al- increase, thereby increasing the health burden. In this though the rate of diagnosed DVT in elderly patients strik- study we sought to determine trends in the diagnosis of ingly increased over the past decade (PϽ.001), that of deep venous thrombosis (DVT) and pulmonary embo- PE has been relatively constant. There was a proportion- lism (PE) in the elderly as well as the use of diagnostic ately greater use of venous ultrasonography, ventilation- tests. perfusion lung scanning, and pulmonary angiography in elderly than in younger patients. Methods: Data from the National Hospital Discharge Survey were used. These data are abstracted each year Conclusions: Extensive use of diagnostic tests in el- from a sample of records of patients discharged from non- derly patients in the past decade has resulted in an in- federal short-stay hospitals in the entire United States. creased diagnostic rate for DVT but not PE. The reason Main outcome measures were trends in rates of diagno- for this disparity is uncertain but may reflect early diag- sis of DVT and PE as well as trends in the use of diag- nosis and treatment of DVT. With the aging of the popu- nostic tests between 1979 and 1999. lation, DVT will increase the health burden. Results: The rates of diagnosis of DVT and PE and of Arch Intern Med. 2004;164:2260-2265 HE POPULATIONS OF THE tion of patients with DVT and/or PE in the Western industrialized United States, which avoids the bias that can countries are aging, and it occur with regional, often less diverse, is predicted that the el- samples. The size of the NHDS database and derly population will in- the broad representation it affords make it Tcrease from 26.6 million in 2003 to 36.0 well suited to assess trends in the rate of ve- million by 2020 in the United States—a nous thromboembolic disease in the United 35% increase.1 The risk of venous throm- States over 2 recent decades. boembolism is strongly associated with In view of the paucity of empirical na- age.2-4 Furthermore, deep venous throm- tional literature for venous thromboembo- bosis (DVT), which leads to pulmonary lism in the elderly and the scope provided Author Affiliations: embolism (PE), is the third most com- by the NHDS database, we analyzed (1) 21- Department of Research, mon cardiovascular disease after myocar- year diagnostic trends for DVT and PE; (2) St Joseph Mercy Oakland 5 Hospital, Pontiac, Mich, dial infarction and stroke. Yet, neither the diagnostic process involved; and (3) (Drs Stein, Kayali, and Alshab) trends in the rates of diagnosis of DVT and changes in the proportion of elderly per- Department of Internal PE in the elderly nor trends in the rates sons with PE and DVT in the US popula- Medicine, Wayne State of use of diagnostic tests in the elderly have tion between 1979 and 1999. By providing University, Detroit, Mich been assessed using a survey of the entire a representative hospital survey that per- (Dr Stein); Departments of United States. mits accurate estimates of rates of PE and Medicine (Drs Hull and Ghali) To assess the public health burden of DVT in the population, the NHDS is and Community Health DVT and PE in the elderly, we evaluated the uniquely suited to evaluate trends over time Services (Dr Ghali), University database of the National Hospital Dis- in the rates of diagnosis of DVT and PE of Calgary, Calgary, Alberta; and 6,7 Department of Grants, charge Survey (NHDS) , one of the larg- among the elderly. We present an epide- Contracts, and Sponsored est known databases of records of patients miological study of hospitalized elderly per- Research, Oakland University, hospitalized with PE and/or DVT. This sur- sons in the United States assessing the bur- Rochester, Minn (Dr Olson). vey allows for a methodologically rigorous den of venous thromboembolism over the Financial Disclosure: None. sample, distributed over the entire popula- recent 20-year interval. (REPRINTED) ARCH INTERN MED/ VOL 164, NOV 8, 2004 WWW.ARCHINTERNMED.COM 2260 ©2004 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 10/01/2021 METHODS STATISTICAL ANALYSIS The rates of diagnosis of DVT and PE were calculated by di- DATA SOURCES viding the number of individuals having the conditions in a given period by the sum of the yearly census estimates of the US popu- Data from the NHDS were used for this study.6 The NHDS da- lation in the same period and reporting the number per 100000 tabase is described in detail elsewhere.7-25 The survey design, population. The numerator was obtained from the NHDS. The sampling, and estimation procedures were planned to pro- denominator was obtained from estimates based on the US cen- duce calendar-year estimates. Trained medical personnel coded sus, as described. Yearly rates of diagnosis of DVT and PE were diagnoses and procedures using the International Classifica- calculated for the 21-year period of observation. Descriptive sta- tion of Diseases, Ninth Revision, Clinical Modification (ICD-9- tistics were used to graphically display trends over time in rates CM).26 A minimum of 1 and a maximum of 7 diagnostic codes of diagnosis of DVT and PE and rates of use of diagnostic tests. were assigned for each sample abstract. If an abstract included These data show estimated rates of diagnosis or rates of use of surgical and/or diagnostic procedures, a maximum of 4 pro- diagnostic tests based on sampling rather than on a census of cedure codes was assigned. cases or procedures. Linear regression analyses were performed to calculate the 28 NHDS SAMPLING SCHEME slopes of selected segments of the curves describing the data. Pearson correlation analyses were performed for the same lin- ear segments to assess the extent of dispersion of points around The NHDS is based on a national probability sample of pa- 28 tients discharged from short-stay hospitals—exclusive of fed- the regression lines. More complex equations were used to de- eral, military, and Department of Veterans Affairs hospitals— scribe the curves that related rate ratios for DVT and PE to age. located in the 50 States and the District of Columbia. The Linear regression analyses were performed using InStat soft- sampling plan, performed in 3 stages, is described else- ware, version 3.0 (GraphPad Software, San Diego, Calif), and where.6,7,27 nonlinear analyses were performed using SPSS software, ver- sion 11.0 (SPSS Inc, Chicago, Ill). Differences between groups and differences in the rates of ESTIMATION PROCEDURES use of diagnostic tests performed over time were assessed us- ing t tests when 2 groups were compared and analysis of vari- Estimates of patients with DVT and PE and the total number ance when multiple groups were compared.28 Differences of rates of diagnostic tests performed in the entire United States for DVT were assessed by ␹2 test. An analysis of covariance was done and PE were derived from the number of sampled patients with with sex and race (white and black races only) as covariates DVT and/or PE and the number of diagnostic tests performed using SPSS software, version 11.0. The adjusted values based in sampled patients using a multistage estimation procedure. on the covariates for DVT per 100000 population per year were This procedure, which produces essentially unbiased national submitted to a linear regression analysis. estimates, is described elsewhere.7,27 RESULTS IDENTIFICATION OF DVT CASES All available diagnostic code fields were screened for specific TRENDS IN THE RATE OF DVT DIAGNOSIS codes to identify patients with DVT and/or PE. Since 1979, the ICD-9-CM has been used for classifying diagnoses and proce- Trends for the 21-year study period show that the rate dures in the NHDS. Although the ICD-9-CM is modified an- of diagnosis for DVT in elderly patients (Ն70 years) was nually, the diagnostic codes for “PE and infarction” and “phle- constant from 1979 to 1990 but increased markedly from bitis and thrombophlebitis of deep vessels of lower extremities” 1990 to 1999. The rate of DVT diagnoses was higher for have changed little. elderly patients (Ն70 years) than for younger patients The specific ICD-9-CM codes that we used for identifica- (20-69 years old) (rate ratio, 4.72; 95% confidence in- tion of patients with PE were 415.1, 634.6, 635.6, 636.6, 637.6, Ͻ Figure 1 638.6, and 673.2. The codes used for identification of patients terval, 4.30-5.14; P .001) ( A). The rate of DVT with DVT were 451.1, 451.2, 451.8, 451.9, 453.2, 453.8, 453.9, diagnoses among elderly patients increased from 454 per 671.3, 671.4, and 671.9. Five-digit codes, such as 451.11 (in- 100000 population in 1990 to 655 per 100000 popula- cluded under the code 451.1), were not listed because they were tion in 1999 (Figure 1A). included under the corresponding 4-digit codes. The ICD- The 21-year trends for diagnosing DVT according to 9-CM codes used for diagnostic tests were the following: 88.77 age distribution (by 10-year increments) are shown in for diagnostic ultrasound examination of the peripheral vas- Figure 1B.