Research Design, Biostatistics, and Literature Evaluation
Practice Case
Questions 2 and 3 pertain to the following case.
You are a critical care pharmacist in the medical ICU and wish to retrospectively study the safety and efficacy
of four-factor prothrombin complex concentrate (4F-PCC) for coagulopathy correction in patients with end-stage
liver disease. You decide on a primary outcome of in-hospital mortality and secondary outcomes of INR correc-
tion, viscoelastic test correction, and thrombotic events.
Which would create the highest risk of bias in this study?
chart review.
(frequentist)
Bayesian analysis allows for incorporation of prior knowledge into current study and calculates the
probability of benefit.
deviation, median, and interquartile range.
Interpretation of studies may be different based on whether frequentist or Bayesian analyses are selected
rejecting or accepting (failing to reject) a null hypothesis (H0)
Rejecting the null hypothesis results in accepting the alternative hypothesis (H1)
Superiority trial example (N Engl J Med 2023; 388:499-510).
H0 = no difference in mortality exists between a restrictive versus liberal fluid strategy for
sepsis-induced hypotension
ii.
H1 = a restrictive fluid strategy would lead to lower mortality than liberal fluid strategy for
sepsis-induced hypotension
iii.
The study found no significant difference in mortality between groups, thereby resulting in a
failure to reject the null hypothesis.
Type I error (alpha [ฮฑ] error): To reject the H0 when, in fact, it is true. Decisional threshold to reject/
not reject the H0 is conventionally set at ฮฑ = 0.05. The ฮฑ value represents the likelihood that a type
I error will be made. An ฮฑ set at 0.05 means that the H1 will erroneously be accepted 1 in 20 times.