Recent Trends in Clinical Outcomes and Resource Utilization for Pulmonary Embolism in the United States: Findings From the Nationwide Inpatient Sample FREE TO VIEW

Pulmonary embolism (PE) is a leading cause of mortality and morbidity in hospitalized patients in the United States. Between 5% and 10% of hospital deaths are attributable to PE,^1–3 leading to an estimated 100,000 to 200,000 deaths annually in the United States from PE.^4–6 Because prophylaxis is clinically effective and cost-effective,^7–9 PE is the most common preventable cause of death in hospitalized patients.⁷

Contemporary studies estimate that annual health-care expenditures related to venous thromboembolism (VTE) and PE are in excess of $1.5 billion.¹⁰ At the individual patient level, hospital costs incurred by patients in whom VTE complications develop are double those for patients in whom these complications do not develop.¹¹ Several advisory groups^12–16 have sponsored initiatives to require all hospitalized patients to be assessed for VTE risk and to have appropriate thromboprophylaxis administered. Despite these efforts, the use of prophylaxis remains unacceptably low for several high-risk groups of patients. In particular, multiple studies^17,18 have demonstrated disparities between surgical and nonsurgical patients in terms of the use of appropriate prophylaxis. These findings highlight the necessity to continue to evaluate the clinical impact of PE on hospitalized surgical and nonsurgical patients and to determine the impact of nationwide initiatives to increase the use of VTE prophylaxis. The objectives of the present study were to determine the recent trends in clinical outcomes and resource utilization for patients hospitalized with a clinically recognized episode of acute PE in the United States.

Data used in this study were obtained from the Nationwide Inpatient Sample (NIS), from the Healthcare Cost and Utilization Project of the Agency for Healthcare Research and Quality. This database contains information abstracted from approximately 8 million patient hospitalizations per year and comprises a stratified sampling frame of 20% of all US hospital discharges. These data can be used to produce a weighted estimate of approximately 35 to 39 million patient hospitalizations per year. All patient identifiers have been removed from this database. The NIS represents the largest all-payer inpatient care database available and provides the unique opportunity to estimate nationwide trends for hospital admissions related to specific diseases and their associated clinical outcomes.^19,20 Data used for this analysis were adjusted national estimates based on the stratified sampling frame of discharges. The total number of weighted discharges per year reflected in the NIS database were as follows: 34,874,046 Data processing and statistical ana (1998); 35,467,673 (1999); 36,417,565 (2000); 37,187,641 (2001); 37,804,021 (2002); 38,220,659 (2003); 38,661,786 (2004); and 39,163,834 (2005). The NIS database was queried for an 8-year period from January 1, 1998, to December 31, 2005, for patients discharged with primary or secondary PE. These patients were defined according to the International Classification of Diseases, ninth revision (ICD-9), clinical modification codes that correspond to PE (415.11 to 415.19). The total cohort was further stratified according to surgical or nonsurgical hospital discharge status to permit comparison between subgroups. Surgical patients were identified using the ICD-9 clinical modification procedure codes 01 to 86.99, which pertain to major surgical procedures. Codes for minor procedures were excluded, using a method described previously.²¹

The total cohort, together with surgical and nonsurgical subgroups, was assessed for baseline characteristics, including age, gender, ethnicity, and type of hospital admission (eg, emergency, urgent, or elective). The analysis of patients' baseline characteristics was performed to determine whether certain subgroups of patients with PE were more high risk, therefore skewing the results of our comparisons of clinical outcomes and resource utilization. Recognized risk factors for PE were assessed, including malignancy, previous VTE, obesity (body mass index > 30 kg/m²), hormone-replacement therapy, congestive heart failure, prior stroke, coronary artery disease, nonambulatory status, smoking history, comorbid lupus, recent infection, recent hip or long-bone fracture, and clinically reported varicose veins.^12–16

The groups were assessed for specific clinical outcomes related to their current hospital admission, including in-hospital mortality, average length of stay, major bleeding (ie, hemorrhage leading to hemodynamic instability or requiring blood transfusion), and the incidence of heparin-induced thrombocytopenia. Relative resource utilization per hospital admission was estimated using mean hospital charges per PE-related hospital admission. Charges were normalized to 2005 US dollars adjusted to reflect the US consumer price index. Additional analyses were performed to determine whether surgical and nonsurgical patients were at differential risk for adverse outcomes or higher resource utilization. This study was conducted in full compliance with institutional review board policies for clinical research at the University of Massachusetts Medical School (Worcester, MA) and in compliance with rules for data use stipulated by the Healthcare Cost and Utilization Project of the Agency for Healthcare Research and Quality in granting the authors access to NIS data.^19,20

Data were analyzed and compared using analysis of variance for continuous data and χ² tests for proportions. Variables with a p ≤ 0.05 were considered statistically different. Variables that were significantly different between comparison groups underwent further post hoc testing with a Student-Newman-Keuls test. Data processing and statistical analyses were performed with statistical software (SAS, version 9.1; SAS Institute, Inc; Cary, NC).

The study population comprised 1,378,670 patients of whom 397,188 (28.8%) were categorized as surgical and 981,482 (71.2%) as nonsurgical. These data represent national estimates extrapolated from the sampling frame of 20% of US hospitals. The estimated number of patients with primary or secondary PE discharged from the hospital each year increased from 126,546 in 1998 to 229,637 in 2005 (Fig 1). Of these discharges in 1998, 72,221 (57%) were given a primary discharge diagnosis of PE and 54,325 (43%) a secondary discharge diagnosis of PE. By 2005, 137,451 (60%) were discharged with primary PE and 92,186 (40%) with secondary PE. Nonsurgical patients accounted for a greater proportion of hospital admissions throughout the study, with 70 to 72% of discharges occurring among nonsurgical patients (Fig 1).

Both overall and in the two subgroups, the proportion of white patients decreased, whereas the proportion of African-American patients remained nearly constant (Table 1). The rates of elective hospital admissions decreased, whereas hospital admission rates from the emergency department increased (p < 0.001 in all three groups) [Table 1].

In the overall cohort, PE hospital admissions related to malignancy, prior VTE, obesity, smoking history, and concurrent infection increased over the study period, whereas hospital admissions with existing hip and extremity fractures decreased (all p < 0.01) [Table 2]. Similar findings were observed in both the surgical and the nonsurgical subgroups (all p < 0.001) [Table 2].

Although the absolute number of deaths in patients hospitalized for PE increased, the in-hospital mortality rate decreased over the period of this study from 12.3% (15,591 of 126,546 patients) in 1998 to 8.2% (18,744 of 229,637 patients) in 2005 (Fig 2A). The mean length of stay decreased from 9.4 to 8.6 days (Fig 2B). The incidence of complications related to anticoagulation therapy (major bleeding and heparin-induced thrombocytopenia) remained stable or decreased (Fig 2C and D). The mean total charges associated with hospital discharges for PE increased from $25,293 to $43,740 (Fig 3).

Similar decreases in hospital death and length of hospitalization were observed in the surgical and nonsurgical subgroups, but both remained higher in the surgical population (Fig 2A and B). Increases in total costs also were observed and increased at a higher rate in the surgical group than in the nonsurgical group (Fig 3).

This contemporary study is the first of temporal trends (1998 to 2005) in the rates of clinically recognized PE and in-hospital mortality in patients hospitalized in US acute care hospitals. Key findings include a doubling in the number of hospitalized patients with a clinically recognized episode of acute PE combined with a decrease of two-thirds in hospital mortality. These findings were consistent in both surgical and nonsurgical subgroups. The overall trends in the rates of major complications related to anticoagulation therapy (bleeding and heparin-induced thrombocytopenia) remained stable or decreased. Consistent with national trends in length and cost of hospitalization during this period, the duration of hospitalization fell slightly, whereas the average costs associated with hospital admissions for PE increased by > 70%.

Although increases in the number of patients with PE were seen in both the surgical and the nonsurgical populations, the increase was far greater in medical patients. Although we have no data to support these hypotheses, increasing use of d-dimer and spiral CT scanning to diagnose PE²² and inadequate thromboprophylaxis in US hospitals during this time period may have contributed to the observed increase in diagnoses of PE.¹⁸

Of interest, although improvements in clinical outcomes and length of hospitalization were observed in both the surgical and the nonsurgical populations, these indicators of clinical performance remained notably higher in the surgical subgroup. Although these findings could indicate inferior treatment strategies for surgical patients with PE compared with nonsurgical patients, it is more likely that these disparities reflect the greater acuity of illness associated with surgical diseases. In addition, the inability to fully anticoagulate surgical patients in the short-term postoperative period may have been associated with the observed increase in the incidence of PE. Surgical patients demonstrated greater prevalences of congestive heart failure, prior stroke, and concurrent infection than nonsurgical patients. However, they experienced lower incidences of bleeding complications while being treated for PE. Thus, although mortality and length of stay are greater for surgical patients, these differences seem unlikely to be due to inadequate VTE prophylaxis strategies compared with those used in nonsurgical patients.

An estimate of the annual rate of PE treated in US hospitals was derived from the National Hospital Discharge Survey in 1987 by Gillum,²³ who reported a decrease from 197,000 episodes of PE in 1975 to 120,000 episodes in 1985. Other studies that attempted to determine the incidence of PE in the United States identified PE cases from hospitals in limited geographic areas, leading to uncertainty about the generalizabilty of these findings to the United States as a whole. Using 1986 data, Anderson et al²⁴ conducted the first US community-wide study of VTE and calculated the attack rate of PE and deep vein thrombosis in patients treated in acute care hospitals within the well-defined region of Worcester, MA, which they extrapolated to 99,000 patients treated for PE in US hospitals per year. Subsequently, Silverstein et al²⁵ published a landmark study from Olmstead County, MN, in which they identified individuals in whom PE and deep vein thrombosis developed in a 25-year period from 1966 to 1990, observing a 45% decrease in the attack rate of PE. Using 1999 hospital discharges, Spencer et al²⁶ studied patients in hospitals located in area of Worcester, MA, and extrapolated an estimated 80,000 patients treated for PE in US hospitals. This estimate is lower than the 1999 rate reported here; however, their findings were based on a limited geographic region of the United States, and they subjected ICD-9 hospital discharge codes for PE to direct validation in a review of hospital charts.

The primary strength of this study is the power afforded to the analysis by the NIS database, which includes a large, representative sample of inpatients with acute PE. The availability of data over an 8-year period allows for a robust and informative study about recent national trends in clinical outcomes and resource utilization for patients with PE. In addition, the database includes all-payer information from 20% of inpatient hospital admissions throughout the United States; therefore, deficiencies common to previous studies, including extrapolation to US-wide estimates from regional populations and data sources for patients hospitalized before 2000, are avoided.

Despite these strengths, several important limitations pertain to this study. The NIS database comprises information extracted from hospital discharge summaries, which are subject to coding errors. However, studies using medical record audit have demonstrated that NIS data are coded with adequate sensitivity and specificity.^19,20 Additionally, a satisfactory sensitivity of ICD-9 codes has been demonstrated²⁷ for identifying patients with objectively confirmed PE. Detailed data are not available for the diagnostic tools used to confirm the diagnosis of PE or the hospital treatments administered (eg, type and findings of diagnostic testing, type and duration of PE prophylaxis or treatment). Another limitation concerns our method of estimating health-care resource utilization through the analysis of mean charges. Typically, charges are significantly higher than actual costs, but cost data were not available for these patients. Despite these limitations, charge data can provide a gross index of resource utilization for these PE-related hospital admissions over the period of the study. Previous estimates of resource utilization were based on extrapolations from much more limited data sets, and no information has been available about trends in resource utilization over consecutive years. Physicians' increasing awareness of PE during the period of this survey, with a corresponding increased utilization of increasingly accessible and sensitive diagnostic tests, may be a confounding variable. The observed increase in PE-related hospital admissions may reflect an increase in the detection of minor and asymptomatic PE, which could account for the lower in-hospital mortality and shorter length of hospitalization observed. Although the detection of these minor PEs may diminish the impact of the apparent improvements in clinical outcomes for patients with PE over time, the observed increase in hospital admissions for PE highlights a concerning rise in the clinically apparent prevalence of a largely preventable and potentially lethal disease.

A final limitation is our inability to separate PE that developed during hospitalization from PE that developed prior to hospital admission. Because the data extracted for the NIS database are derived from hospital discharge summary diagnostic codes, which do not make this differentiation, it was not possible to make correlations between the temporal onset of PE and the primary diagnosis at hospital admission. This limitation is highlighted by the finding of Spencer et al²⁸ that only approximately 25% of VTE episodes are hospital acquired. The results of this current study must be interpreted in light of these limitations.

In conclusion, we have provided robust data regarding recent nationwide trends for clinical outcomes and health-care resource utilization for patients with acute PE. Despite steadily increasing hospital admissions for PE over the past 8 years, in-hospital mortality and length of hospitalization have decreased consistently but with an increasing cost for health-care resources. PE remains a major risk for hospitalized patients in the United States. Our findings indicate that important improvements have been made over the past 8 years, possibly due to physicians' increased awareness of PE, more aggressive diagnostic testing, and greater use of thromboprophylaxis in high-risk patients. The large proportion of PE identified in patients hospitalized for nonsurgical illness, including clinically recognized and fatal PE, suggests that an opportunity exists to focus quality improvement efforts in US hospitals to further improve patient outcomes. The overall increase in hospital admissions for PE highlights a substantial rise in the prevalence of this largely preventable and potentially lethal disease, and emphasizes the need to continue aggressive surveillance, prophylaxis, and treatment.

Author contributions: Drs. Park, Messina, Ciocca, and Anderson contributed to the study design. Drs. Park, Dargon, Huang, and Anderson performed the data collection/processing. Drs. Park and Huang contributed to statistical analysis. All authors contributed to the manuscript preparation, and Drs. Park, Messina, Ciocca, Huang, and Anderson contributed to the editorial review.

Financial/nonfinancial disclosures: Dr. Anderson has received grants from Sanofi-Aventis, The Medicines Company, and Ortho McNeill Jansen. He also has received consulting and speaker fees from GlaxoSmithKline and Sanofi-Aventis. Drs. Park, Messina, Dargon, and Ciocca and Ms. Huang have reported to the ACCP that no significant conflicts of interest exist with any companies or organizations whose products or services may be discussed in this article.

Other contributions: We thank Sophie Rushton-Smith, PhD, Medical Writer, Center for Outcomes Research, University of Massachusetts Medical School, for her editorial support in the preparation of this manuscript.