Why first_day_lab calculates the min and max but not average of lab items?

[zheng523]

I guess the reason is to identify the "worst" of all measured values during the first day of ICU stay. For example, when we try to evaluate the association between a particular variable (such as TyG) and the mortality of sepsis patients, we need to include some covariates. These covariates should be the worst value in the first day to reflect the severity of illness.
For some lab items, lower is worse, such as albumin and blood pressure (only for sepsis). For some lab items, higher is worst, such as temperature (only for sepsis) and creatine. However, for potassium, glucose and sodium, both higher and lower is worse. So my question is, which measure should be included as covariates for these lab items? Max, min or avg?

[zheng523]

Should I create a new variable - max deviation from normal range? For example, the normal range of potassium is 3.5-5.5 mmol/L. If potassium has 3 measures during the first day of ICU stay, 3.2, 4.0 and 6.8, their deviation are 0.3, 0, 1.3, respectively. The max deviation is 1.3.

[alistairewj]

Defining worst is hard to do generally though most values exhibit a J shaped curve in relation to mortality.

You might try looking at some existing severity of illness scores like SAPS/APS/OASIS, which try to score each individual value. They're only available for a subset of variables but you could take the value that gives the highest score.

With enough sample size though, you can include both! Which is what I would do, absent of any more context.