Year: 2019
Type: peer-review
Abstract: Thank you for the opportunity to respond to this Letter to the Editor by Kaylor et al.1 As we described, this was a secondary analysis from a much larger study, and as such, the ability to control for a variety of potential limiting factors related to OR time (not a primary focus of the main study) was not possible after the fact. We do acknowledge these limitations in our article, although not as specifically as the authors of this letter suggest. The OR bed type and preventative strategies we list may in fact include the limitations described (ie, OR surface; OR patient position; and pre-, intra-, and postoperative pressure injury prevention interventions). We agree we could have listed these more specifically to enhance clarity. We are fully aware of the textbook criteria for model estimation; indeed, these textbook guidelines are highly useful when planning and designing a study. However, inference in smaller studies and observational studies can still be accomplished. To address instability in the parameter estimates when there are too few cases compared to the number of parameters, the model can be bootstrapped to examine the variability in the model parameters. To examine instability in the model predictions, cross-validation methods can be used to quantify how much predicted probabilities vary. Descriptive methods that examine the clustering of groups with respect to the variables in the model provide empirical evidence of the relations being modeled. Finally, comparisons of estimates between the univariate models and the multivariable models can provide information as to instability as a result of correlation in the model. These methods can provide evidence of stability—not only in small samples, but also in large ones that rely on textbook criteria for sample size. Only one of these methods was directly observable in our study, although we examined all of them. The stability and clustering of observations was reasonable and sufficient. Further, our model with three variables was graphable, and the clustering of responses was examinable (more variables would not be simultaneously observable in a graph). Empirically, and by using these methods to examine model instability, we feel our model and estimates were stable enough to report. Overall, the findings were surprising to our team as well; however, it is incumbent upon all of us to report both supporting and opposing outcomes of research, including the limitations of the study and the analysis. If not, erroneous assumptions may be made. Further, when carefully analyzed, small observational samples may still provide us with information and the opportunity to ask more questions.
Source: AORN Journal
DOI: 10.1002/aorn.12826
Language: English
Subfield: Occupational Therapy
Field: Health Professions
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2976788645