In a sample of 356 hospitalized patients, 25% were diagnosed with acute kidney injury and had on average a 14.1 longer length of stay than those patients without AKI. This difference was found to be statistically significant. A logistic regression model showed an odds ratio (odds of AKI to odds of not having AKI) of 2.418 for an increase of one day of stay. The logistic model with length of stay as the independent variable has relatively good predictive capability for acute kidney injury. Bear in mind that the data in this case was synthetic and it is not known the extent to which it represents real patient data. As always, it is a good idea to review statistical results with subject matter experts.

While the logistic regression model established a statistically significant relationship between AKI and length of stay, it should not be concluded that length of stay causes AKI. Causality cannot be substantiated solely by an empirical association. Temporal precedence of the causal factor, elimination of association with other causal factors, an understanding of the physiological mechanisms causing AKI, and replication of the association are needed to establish causality. Again, consultation with subject matter experts is recommended.

Next analysis steps would be to examine the bivariate relationship between AKI and other variables given in the data set such as co-morbidities (coronary artery disease, diabetes), age, and length of stay. Following this multivariate relationships should be pursued.