Elsevier

Thrombosis Research

Volume 97, Issue 5, 1 March 2000, Pages 267-273
Thrombosis Research

Regular article
Multivariate Analysis-Based Prediction Rule for Pulmonary Embolism

https://doi.org/10.1016/S0049-3848(99)00180-2Get rights and content

Abstract

The diagnosis of pulmonary embolism (PE)is still an unresolved problem. The aim of this prospective observational study was to derive and validate a prediction rule (PEscore) by which PE can be diagnosed by easily obtainable and rapidly available investigations. Included were consecutive patients with a clinical suspicion of PE admitted to a community hospital. Risk factors and clinical and instrumental investigations were registered. PE was diagnosed by angiography, scintigraphy, or autopsy. In 168 patients, PE was either diagnosed (angiography, n=28; autopsy, n=18) or excluded (angiography, n=12; scintigraphy, n=99; autopsy, n=11). Based on the results of clinical and instrumental findings, a PEscore was derived by a multiple regression analysis, calculated as: [0.29×proven leg vein thrombosis (0=no, 1=yes)] + [0.25×ECG right heart strain (0=no, 1=yes)] + [0.22×neck vein distension (0=no, 1=yes)] + [0.20×dyspnoea (0=no, 1=yes)] + [0.13×suspicious chest X-ray (0=no, 1=yes)] − [0.17 (constant)]. The PEscore was tested further in 139 subsequent cases. In these patients, the PEscore was 0.65±0.17 (diagnosed PE, n=47) and 0.18±0.17 (excluded PE, n=92), respectively (p=0.0001). Depending on a given PEscore, the level of probability of PE can be assessed. Calculation of the PEscore can be helpful in clinical decisions when PE is suspected.

Section snippets

Materials and Methods

Consecutively included were patients admitted to a medical department during 39 months whose initial presentation (history, clinical examination, ECG, blood tests) was sufficiently suggestive for PE that therapeutic heparinisation was initiated (Figure 1). Among the included patients, risk factors (malignancy, previous thrombosis or PE, trauma or surgery within the last 3 months, immobilization>3 days within the last 14 days, oral contraceptives) and symptoms (dyspnoea, chest pain, hemoptysis,

Part 1

Over 19 months, 216 patients (104 males and 112 females) were included, with a mean age of 65 years (range 20–87). Twenty-nine patients (13%) died within 10 days after the suspected pulmonary embolism.

Pulmonary embolism was diagnosed in 46 patients (21%) (PE+); 18 of them by autopsy and 28 of them by pulmonary angiography. PE was excluded in 122 patients (56%) (PE−); 99 of them by a negative perfusion scan, 11 of them by autopsy, and 12 of them by pulmonary angiography. PE was uncertain in 48

Discussion

In part 1 of the study, a prediction rule for PE was derived. This prediction rule contains clinical signs and symptoms (dyspnoea, neck vein distension), results of ECG (right heart strain), chest X-ray, and investigations of the legs. The validation of the prediction rule in different patients of part 2 of the study confirmed the rule's ability to distinguish between PE+ and PE− patients.

To diagnose PE expert systems and decisions, rules have been developed, but a prediction rule has not been

Acknowledgements

This work was supported by a grant from the Medizinisch-wissenschaftlicher Fonds des Bürgermeisters der Bundeshauptstadt Wien, and through the advice of M.T. Schlern.

References (22)

  • P. Pruszczyk et al.

    Noninvasive diagnosis of suspected severe pulmonary embolism. Transesophageal echocardiography vs spiral CT

    Chest

    (1997)
  • Cited by (24)

    • Immobilization and the risk of venous thromboembolism. A meta-analysis on epidemiological studies

      2009, Thrombosis Research
      Citation Excerpt :

      cohort studies among patients with suspicion of VTE within hospitalized patients (n = 7) [17–23] cohort studies among patients with suspicion of VTE within ambulatory patients (n = 21) [24–43]. In this group, a validation and a derivation analysis were reported in the same article [30].

    • Imputation of missing values is superior to complete case analysis and the missing-indicator method in multivariable diagnostic research: A clinical example

      2006, Journal of Clinical Epidemiology
      Citation Excerpt :

      We analyzed whether the following predictors independently contributed to the prediction of PE presence or absence: age (years), gender (male/female), duration of symptoms (number of days), period confined to bed (number of days), respiratory rate (breaths/min), cardiac rate (beats/min), Quetelet index (kg/m2), arterial oxygen pressure (mmHg), arterial carbon–oxygen pressure (mmHg), the presence (yes/no) of leg paresis, leg pain, a family history of deep venous thrombosis (DVT)/PE, crepitations, fever (temperature above >38.5°C), dyspnea, pleura rub, wheezing, palpitations, collapse with or without unconsciousness, surgery in past 3 months, malignancy, signs of DVT, and of previous history of PE or DVT, leg ultrasound of lower limbs (positive/negative for DVT), and the chest x-ray result (positive/negative for PE). The a priori selection of these predictors was based on previous diagnostic studies [17–19]. There were no missing data for the presence of PE (yes/no; diagnostic outcome variable), whereas there were missing data on various tests or predictors.

    View all citing articles on Scopus
    View full text