Abstract
Background: Many hospitalized patients with PE die. A large registry study described a mortality rate of 17.4% in patients with PE and suggested 45% of these deaths were due to the PE. Data on death and PE is usually derived retrospectively from hospital databases without chart confirmation and to our knowledge no study has attempted to determine the accuracy of coding for PE deaths. Furthermore, it is unclear how often deaths caused by PE could have been prevented.
Methods: A retrospective chart review of PE cases hospitalized at a tertiary care center. Charts over an 8 year period ending in 2004 were reviewed if the hospital database record identified PE as a diagnosis by the ICD-10 coding system. Charts of those who died were independently reviewed by two thrombosis experts with discrepancies resolved by consensus or a third reviewer. Prior to chart review definitions were agreed upon. The coding as PE was considered correct (confirmed PE) if there was supportive imaging, an autopsy, or in the case of death without imaging or autopsy, the clinical scenario was such that PE could have occurred. The degree of certainty that PE contributed to the death was classified as certain (unexplained hypotension, hypoxia, cardiac arrest with no other explanation other than PE and autopsy confirmation or radiographic confirmation), highly probable (same as certain but no autopsy confirmation), probable (criteria for highly probable but another disease could have caused the death). We considered these cases to be death due to PE. Deaths were also classified as possible (other cause suspected based on clinical evidence but 100% certainty not available), or unlikely due to PE (all other cases). In cases defined as death due to PE we determined whether any further intervention could have prevented death.
Results: 612 cases were identified of whom 68 had radiographic or autopsy data that ruled out the diagnosis and in 46 the coding was clearly an error. 498 cases of PE were identified, 111 of whom died during hospitalization; the mortality rate in those the hospital coded as PE was 18% vs 22% of those with confirmed PE. Death due to PE was diagnosed in 70 patients (14% of patients with confirmed PE and 11% of all patients coded as PE). In the remaining 41 deaths, PE was possible in 24 and unlikely in 17. Disagreement was uncommon. There was no difference between the likelihood of death from PE in the group diagnosed by imaging and autopsy compared with the group where PE death was confirmed by an appropriate clinical scenario. 38 deaths due to PE may have been prevented with an additional intervention: prophylaxis (55%), earlier diagnosis (45%), inferior vena cava filter (IVCF) (32%), anticoagulation (18%), embolectomy 5%, thrombolytics (3%). The remaining deaths due to PE were not preventable since 15 patients were palliative and did not receive active treatment, 9 died before a diagnosis was made and in 8 another disease prevented treatment.
Conclusions: Using hospital database records is a reasonable means to evaluate PE mortality and our death due to PE rates are similar to those in registry publications. Surprisingly imaging and autopsy results do not increase the probability of reaching the conclusion that death is due to PE. Over half of preventable PE deaths may have been prevented by prophylaxis and one third with an IVCF.
Disclosure: No relevant conflicts of interest to declare.
Author notes
Corresponding author