01-11-2018 | Author's Reply
Reply to the Letter to the Editor of S. Shahsavari et al. concerning “Predicting medical complications in spine surgery: evaluation of a novel online risk calculator” by M. F. Kasparek et al. (Eur Spine J: doi: 10.1007/s00586-018-5707-9)
Published in: European Spine Journal | Issue 11/2018
Login to get accessExcerpt
We thank for the interest in our paper [1]. Regarding the method of the current paper, we would like to clarify that the statistic was performed by a specialized biomedical statistician. Moreover, the manuscript was reviewed by a statistical reviewer, during the review process. Nevertheless, we would like to explain our statistics and the criticized issues in detail below.
1.
Firstly, there are two possible ways to compare estimated risks with the percentage of patients with complications. (1) Either to use the percentage of patients with complications as the expected mean of the estimated risks and compare these two using a one-sample t-test [2] or (2) taking the median of the estimated risks as “expected” percentages and compare the observed complication rate with it (using a Chi-squared test as its binary variable “complications yes/no”) [3]. In the current paper, we decided to apply the second approach because the estimated risk is somehow theoretical but the complication rate is observed in reality.
2.
Secondly, we categorized according to our study protocol patients into three risk groups to analyze if calculated risks in low-, medium- and high-risk patients are comparable to the actually observed complication rate. The categorization was based on clinical experience and was defined before risk calculation. Patients were not allocated according to quantiles. As mentioned above, the Pearson Chi-squared test was used for comparison between the median of the predicted risks and the actually observed complication rate in the beforehand-defined risk groups. Comparison of results among the groups was not made, and therefore, we did not have an independent variable.
3.
In our opinion, the new risk calculator is comparable to a diagnostic test, as it should indicate whether there is a risk of complication or not. As we used a completely different sample than the authors of the risk calculator, our study is a non-biased validation study [4].