The fragility index is a
statistical metric used primarily in the
medical literature
Medical literature is the scientific literature of medicine: articles in journals and texts in books devoted to the field of medicine. Many references to the medical literature include the health care literature generally, including that of denti ...
to assess study results. It denotes the minimum number of
subjects whose status would have to change from a non-event to an event in order to convert a
statistically significant
In statistical hypothesis testing, a result has statistical significance when a result at least as "extreme" would be very infrequent if the null hypothesis were true. More precisely, a study's defined significance level, denoted by \alpha, is the ...
result of the study into a non-significant result.
The smaller the fragility index, the more fragile the result of the study.
__TOC__
Derivation
The result of
scientific studies
Scientific study is a creative action to increase knowledge by systematically collecting, interpreting, and evaluating data. According to the hypothetico-deductive paradigm, it should encompass:Popper, Karl (1959). The Logic of Scientific Discover ...
is said to be statistically significant if the calculated probability of a
type I error
Type I error, or a false positive, is the erroneous rejection of a true null hypothesis in statistical hypothesis testing. A type II error, or a false negative, is the erroneous failure in bringing about appropriate rejection of a false null hy ...
(assumption of a correlation that does not really exist) is below the arbitrarily chosen but generally accepted limit of 5%. Study results that meet these significance criteria can be published more easily and in more highly ranked journals (
publication bias
In published academic research, publication bias occurs when the outcome of an experiment or research study biases the decision to publish or otherwise distribute it. Publishing only results that show a Statistical significance, significant find ...
).
The sole consideration of the
significance level
In statistical hypothesis testing, a result has statistical significance when a result at least as "extreme" would be very infrequent if the null hypothesis were true. More precisely, a study's defined significance level, denoted by \alpha, is the ...
for the assessment of study results is increasingly considered questionable (see also Problems with the Interpretation of Statistical Significance). The fragility index describes the extent to which the attribution of statistical significance is subject to random influences. It indicates the number of patients or subjects whose results would have to change in order to alter the statistical interpretation of the study. The Fragility Index applies only to studies that examine
dichotomous
A dichotomy () is a partition of a set, partition of a whole (or a set) into two parts (subsets). In other words, this couple of parts must be
* jointly exhaustive: everything must belong to one part or the other, and
* mutually exclusive: nothi ...
outcomes (deceased/surviving, cured/not cured).
Importance
In a review of 56
intensive care
Intensive care medicine, usually called critical care medicine, is a medical specialty that deals with seriously or critically ill patients who have, are at risk of, or are recovering from conditions that may be life-threatening. It includes p ...
studies with a significant mortality outcome, a median Fragility Index of 2 was calculated, and in 40% of the studies it was 0 or 1. That is, in at least half of the studies examined, it would have been sufficient for 2 patients to have had a different outcome such that the result of the study would no longer have been considered significant. In 12.5% of the studies examined, the fragility index was smaller than the number of patients with no follow-up data available (''lost to follow-up''). Thus, it is quite possible that a more complete data collection would have nullified the significant result.
This problem of fragility is discussed especially in the intensive care literature. This is attributed, among other things, to the fact that the patient numbers in these studies are necessarily small.
However, this problem of fragility of study results has also been described in other areas of medicine.
References
{{reflist
Biostatistics
Medical statistics