Skip to main content

Table 4 Falsification analyses with identical model structure and adjustment variables as primary model a

From: Prehospital intravenous access and fluid resuscitation in severe sepsis: an observational cohort study

Pre-specified falsification analyses

Odds ratio (95% CI) for catheter alone

Odds ratio (95% CI) for catheter and fluid

Nonsensical outcomes with identical exposures and model b

  

Sum of age integers is an even number

0.75 (0.51, 1.10)

1.02 (0.82, 1.29)

Admission month is a prime number

1.11 (0.76, 1.60)

0.93 (0.60, 1.45)

Nonsensical exposures with identical model and outcome a

Odds ratio (95% CI)

Age is even number

0.92 (0.66, 1.27)

Admission month is a prime number

0.90 (0.67, 1.20)

  1. aIdentical models as above except intravenous catheter and fluid indicator variables removed and replaced with non-sensical exposures; ball estimates use imputed data after Rubin’s rules, including generalized estimating equations to account for within hospital clustering. Referent category for odds ratios in nonsensical outcomes analysis was patients who received no catheter and no fluid. We tested both nonsensical outcomes and nonsensical exposures in separate analyses. CI, confidence interval.
Back to article page

Critical Care

ISSN: 1364-8535