Análisis de supervivencia en presencia de riesgos competitivos: estimadores de la probabilidad de suceso

Mortality study in patients at weaning from mechanical ventilation

OBJECTIVE

To explain mortality in the ICU and in hospital among patients subjected to invasive mechanical ventilation.

DESIGN

A prospective, 9-month observational cohort study was carried out.

SETTING

A Department of Intensive Care Medicine.

PATIENTS

Consecutive patients requiring invasive mechanical ventilation were followed-up on until hospital discharge or death.

INTERVENTIONS

None.

INTEREST VARIABLES

Date of admission, day of first spontaneous breathing test, length of mechanical ventilation, final extubation date, days in ICU, days in hospital or discharge from ICU, SAPS-3 score, WIND study classification, day of death, hospital discharge.

RESULTS

There were 266 patients: 40 in group 0 of the WIND classification (15%; 95% CI 11-20%); 166 in group 1 (62%; 95% CI 56-68%); 38 in group 2 (14%; 95% CI 11-19%); and 22 in group 3 (8%; 95% CI 6-12%. Logistic regression analysis showed group 3 to have the highest hospital mortality (group 3 vs. group 1; odds ratio 4.0; 95% CI 1.5-10.8; P=.007). However, Cox regression analysis showed no significant differences (hazard ratio group 3 vs. group 1, 1.6; 95% CI 0.7-3.4; P=ns).

CONCLUSION

In our study, considering exposure time, the probability of mortality was the same among the 3 different groups of patients with at least one spontaneous breathing test.

WAERS: An application for web-Assisted estimation of relative survival

Net cancer survival estimation is usually performed by computing relative survival (RS), which is defined as the ratio between observed and expected survival rates. The mortality of a reference population is required in order to compute the expected survival rate, which can be performed using a variety of statistical packages. A new Web interface to compute RS, called WAERS, has been developed by the Catalan Institute of Oncology. The reference population is first selected, and then the RS of a cohort is computed. A remote server is used for this purpose. A mock example serves to illustrate the use of the tool with a hypothetical cohort, for which RS is estimated based on three different reference populations (a province of Spain, an autonomous community (Region), and the entire Spanish population). At present, only mortality tables for different areas of Spain are available. Future improvements of this application will include mortality tables of Latin American and European Union countries, and stratified (control variable) analysis. This application can be also useful for cohort mortality studies and for registries of several diseases.

[Relative survival computation. Comparison of methods for estimating expected survival]

Relative survival is the most commonly used method to determine survival in patients diagnosed with cancer. This method takes into account estimation of expected survival in cancer patients based on the observed mortality in the geographical area to which they belong. The most frequently used methods for estimation of expected survival are the Ederer (I and II) and Hakulinen methods. Survival tables for the geographical areas stratified by age and calendar year are required for these calculations. The present article presents an example of how to perform these estimations and how to choose the most appropriate method for the type of analysis to be performed. This article shows that if the follow-up of the cohort is less than 10 years, any of these methods should give similar results. However, the Hakulinen method is preferred, since it accounts for heterogeneity due to potential withdrawals.

Cálculo automatizado de la supervivencia relativa vía web. El proyecto WAERS del Instituto Catalán de Oncología

The most commonly used measure to estimate cancer survival is relative survival, defined as the ratio between observed and expected survival. Expected survival is computed on the basis of the mortality of a reference population. Mortality tables for the general population are not always available and their calculation requires specific software. For that purpose, the Catalan Institute of Oncology developed WAERS (Web-Assisted Estimation of Relative Survival), a web-based application that estimates the relative survival for a cohort of patients. The user prepares data in a specific format and sends them to a remote server located at the Catalan Institute of Oncology. This server computes relative survival and returns a file with the results to the electronic address supplied by the user. By means of this application, hospital- and population-based Spanish cancer registries and registries of other diseases can estimate relative survival of their cohorts using their reference population (province or autonomous community). This application could also be useful for cohort mortality studies.