Multivariate regression analysis was carried out after multiple imputation?

[smartnikocj]

Neto, Ary S.*, Rodrigo O. Deliberato*, Alistair E. W. Johnson*, Lieuwe D. Bos*, Pedro Amorim, Silvio Moreto Pereira, Denise Carnieli Cazati et al. "Mechanical power of ventilation is associated with mortality in critically ill patients: an analysis of patients in two observational cohorts." Intensive care medicine 44, no. 11 (2018): 1914-1922.
Additional Methods described in the supplementary material:
To avoid bias introduced by missing data, and assuming that data were missing at random, the analysis of the primary outcome was replicated after multiple imputation. For this imputation the following variables were considered: age, gender, BMI, prognostic score (APACHE IV in eICU, OASIS in MIMIC–III and eICU, and SAPS II in MIMI–III), Elixhauser comorbidity score (only in MIMIC–III), SOFA, PaO2 / FiO2 ratio, pH, mean arterial pressure, SpO2, temperature, PaCO2, need of renal replacement therapy in the first 24 hours, need of vasoactive drugs in the first 24 hours, ICU and hospital length of stay, and hospital mortality. Multiple imputation was conducted using the method of predictive mean matching for continuous variable, logistic regression for categorical variables, and ten databases were created. The multivariable model was reproduced in the ten databases after multiple imputation and the results were pooled.

Thanks for your great jobs.

However, I am confused about the process of analysis and imputation.

Here is my thinking. When we query the data from the database, and after applying exclusion, we get the data with missing values.
After that, do we first fill the Nan value with multiple imputation and subsequent calculate baseline and multivariate analysis, finally compare to the MICE method for logit in ten databases.?

OR

we query the data from the database, and after applying exclusion, we get the data with missing values.
After that, we first calculate baseline and do the multivariate analysis with Nan value, and then fill the Nan value with multiple imputation, and subsequently do the multivariate analysis again?

Thanks!!!

[alistairewj]

Sorry, I can't meaningfully speak to this. I would contact Ary Neto who did this part of the analysis.

[smartnikocj]

Sorry, I can't meaningfully speak to this. I would contact Ary Neto who did this part of the analysis.

Thank you! Hope we can get the answer soon!