A response to a letter to the editor titled “Nomogram for predicting mechanical ventilation need among acutely intoxicated patients with impaired consciousness: Correspondence”

Dear Editor,

We read the letter to the editor titled “Nomogram for predicting mechanical ventilation need among acutely intoxicated patients with impaired consciousness: Correspondence”, ¹ which comments on our article titled “Development and validation of a nomogram for predicting mechanical ventilation need among acutely intoxicated patients with impaired consciousness” ²

Subjects and methods

We thank HD and VW, for their interest in our research and engagement with its content. First, we want to clarify an important point regarding the study population. Our study was not limited to patients with acute alcohol intoxication, as mentioned in the correspondence. ¹ In fact, it included all acutely intoxicated patients who presented with impaired consciousness, regardless of the type of poison. Therefore, the causative agents included multiple poison categories as illustrated in Table 1, page 4 of our original study, not just alcohol. ² This distinction is essential to accurately define the scope and applicability of the developed nomogram.

Concerns were raised about potential selection bias. ¹ We used appropriate statistical methods to address the limitations of convenience sampling, including the risk of selection bias. We included all patients who met the eligibility criteria and avoided selective enrollment to mitigate the risk of bias. To further improve methodological rigor, we included patients much larger than the minimum required sample size for the derivation cohort to enhance the study’s population representativeness. Additionally, sufficient study power was confirmed through post hoc power analysis. ²

The correspondents also suggest using a longitudinal approach to monitor the nomogram’s performance and improve its reliability. ¹ We agree that this approach has several advantages. However, its application is limited in acute toxicological emergencies, where nomograms primarily predict short-term outcomes such as mortality, need for intensive care unit admission, need for mechanical ventilation, or acute adverse events.^3–7

Results and Discussion

Our study developed and validated a nomogram using four predictors: mean arterial blood pressure (MAP), partial arterial oxygen pressure (PaO2), pH level, and the glucose/potassium ratio. ² In their comments on our article, the correspondents raised two concerns regarding selecting these parameters. They questioned whether these parameters are consistently available in all clinical situations and suggested including demographic characteristics or substance-specific features to improve the model’s accuracy. ¹ As mentioned in the discussion section, these parameters are routinely measured for all poisoned patients upon admission, even in smaller healthcare facilities. ² The convenience and objectivity of these parameters were mentioned among the study’s strengths, making the nomogram applicable in various clinical settings. In the discussion, we also explained that we selected these four predictors based on statistical and clinical considerations. The relevance of each parameter in predicting the need for mechanical ventilation was clarified. ²

Substance-specific features or demographic characteristics were suggested to be included in the developed nomogram. ¹ Indeed, the discriminative power of our developed nomogram is excellent, so including substance-specific features or demographic characteristics is unlikely to improve the model’s performance further. Simultaneously, precisely identifying the agent responsible for disturbed sensorium will significantly limit the applicability of the proposed model. The developed nomogram is intended for use in intoxicated patients with disturbed consciousness, regardless of the causative agents, since most of these patients cannot provide adequate medical histories. ² Regarding demographic data, especially age and gender, they can serve as useful predictors in nomograms within clinical toxicology. ⁸ Therefore, we concur with the correspondents that these factors could be considered for future predictive models, provided they are supported by statistical and clinical justification.

Recommendations

Interestingly, the correspondents offered valuable recommendations that align with our views. ¹ They emphasized the importance of conducting future multicenter studies, which agrees with our suggestions at the end of the discussion section. ² They also recommended integrating this nomogram into clinical decision support systems to facilitate its practical use during emergencies. ¹ In this context, ZS, one of the current study authors, contributed to a systematic review that compiled all nomograms in clinical toxicology into a comprehensive article to enhance their utility in toxicological emergencies. ⁸ We are in the era of artificial intelligence, so we fully support the idea that developing a dynamic model capable of utilizing real-time data from electronic health records could significantly improve the accuracy and speed of clinical decision-making, especially in assessing the need for mechanical ventilation. Future studies should explore these promising modalities.