The findings of this study demonstrate that the incidence of pulmonary embolism diagnosis is increasing with the increasing use of CT scans, and specifically with the increasing use of spiral CT scans. Previous studies have been limited by small sample size and relatively homogeneous populations without wide applicability to the general population, such as the Worcester Deep Venous Thrombosis Study and the Olmstead County study.2,12
Because PCH4 reporting is mandatory, and virtually all patients with pulmonary embolism are hospitalized, the PCH4 database overcomes such limitations, capturing most, if not all, such diagnoses.
This study for the first time demonstrates an increase in population-based incidence rates of pulmonary embolism since the advent of spiral CT, consistent with the community hospital study showing a 1 per 1000 increase in pulmonary embolism admissions with spiral CT.13
The observed increase in pulmonary embolism we observed was accompanied by a 3.6-fold increase in spiral CT use and a 2.3-fold increase in enoxaparin use, during the period from 1997 to 2001, but not by any increase in D-dimer assays.
Those diagnosed with pulmonary embolism later in the time period (ie, 2000 and after) were “less sick” than those diagnosed earlier, as evidenced by the decrease in the 2 highest and an increase in the 3 lowest Atlas Severity of Illness scores. The in-hospital mortality rates also decreased from 1999 and before, compared with 2000 and after. We suspect that wider use of spiral CT during this period may have led to earlier pulmonary embolism diagnosis than with conventional methods, resulting in lower acuity of illness. This has significant public health implications, because earlier scanning may lead to better health outcomes.
The higher incidence of pulmonary embolism with advancing age and African American race is similar to that in previous studies,2,14,15
but the higher incidence in female patients was unexpected and in contrast with past reports.2,12
Although the gender difference was not large, it may have been related to the size of our database and may reflect the fact that PHC4 does not collect data from smaller federal hospitals, including Veterans Administration hospitals. Despite this, there has been no decrease in the proportion of female admissions in the Veterans Administration system database, either nationally (4.36% in fiscal year 2005 and 4.47% in fiscal year 2006) or locally in Pennsylvania (3.72% in fiscal year 2005 and 3.77% in fiscal year 2006), nor do these figures exceed the estimated 5% of veterans who are female (Michael Fine, MD, personal communication, June 2007). Given the widespread use of hormonal agents and other gender-specific risks in women, however, our findings suggest important areas for further research.
Decreasing pulmonary embolism mortality in the UPMC sample could be predicted by LMWH and warfarin use, and by D-dimer testing (P < .05) in logistic regression analysis, but only LMWH use significantly increased during this period (1997–2002). Although decreasing pulmonary embolism mortality may also be related to earlier diagnosis, spiral CT scan was not a significant predictor of decreasing mortality, perhaps related to the smaller subset for whom these data were available.
We recognize certain additional limitations to our dataset. First, because this analysis excludes Pennsylvania residents hospitalized in other states, non-Pennsylvania residents, and those managed as outpatients, the incidence of pulmonary embolism in our population may have been underestimated. The overwhelming majority of patients with pulmonary embolism are treated in the inpatient setting.16,17
The only “outpatient” pulmonary embolism treatment study of more than 40 subjects was not purely an outpatient study because it included those hospitalized for up to 3 days.18
Thus, the criteria for safe outpatient anticoagulation in patients with newly diagnosed pulmonary embolism are not yet established and an important area of ongoing research.19,20
The incidence rates we found were similar to or greater than those of previously published population-based studies,11,15
suggesting it is unlikely that a significant proportion of patients are missing because of these factors. Moreover, outpatient pulmonary embolism cases should have little to no effect on the change in incidence rates during a 5-year period.
A second limitation was the use of revenue codes for CT scans, because there are no specific codes for spiral scans. However, a subset analysis of spiral CT scan data on the UPMC discharges supports this conclusion, as do the findings of Prologo and colleagues,21
who found a significant increase in spiral CT in patients with suspected pulmonary embolism at their academic medical center between 1997 and 1998 and 2002 and 2003. Moreover, this is corroborated by the lower incidence of pulmonary embolism in clinical studies before the use of spiral scans,2
as well as the high sensitivity and specificity of spiral CT scans for pulmonary embolism.22,23
These data justify our assumption that the increased use of CT scans in our 5-year study was the result of an increase in spiral procedures evaluating pulmonary embolism.
It is possible, although not proven, that the increasing incidence of pulmonary embolism corresponds to the diagnosis of smaller, possibly very small, pulmonary emboli (eg, isolated subsegmental pulmonary emboli), the significance of which is not established.24
If so, then diagnosing an increasing number of smaller, possibly irrelevant, pulmonary emboli could be deleterious if these were treated with anticoagulant therapy with its inherent bleeding risk. Because neither bleeding symptoms nor anticoagulation use is available in the PHC4 database, it is not possible to determine this risk, although should significant bleeding have occurred and led to a decrease in hemoglobin or blood pressure, the severity of illness scores would be unlikely to have improved.