Causality in administrative datasets

Dear Sir, We read with great interest the original contribution by Ananthakrishnan et al. on the effect of infection-related hospitalizations in patients with inflammatory bowel diseases on mortality, hospital stay and hospital charges.1 The authors conclude fittingly that infections account for significant morbidity and mortality in these patients and identify several risk factors of infectious events. We congratulate the authors for addressing this important topic, and laud their efforts to identify important patient characteristics, namely age and Elixhauser co-morbidity score,2 that are strongly correlated with infections in hospitalized IBD patients. Nonetheless, we would like to highlight two points that merit further discussion. First, although the investigators attempted to identify patient characteristics associated with infection-related hospitalizations, they missed an opportunity to examine the effect of several key structural determinants. Specifically, the impact of demographic characteristics, including insurance3 and median zip code income4 as surrogates for socioeconomic status, was not examined. Moreover, it would also have been interesting to document the impact of provider volume on the risk of infection-related hospitalization. Finally, hospital attributes, such as hospital bed size and location (rural vs. urban) were not examined, even though these variables are readily available in the Nationwide Inpatient Sample. Second, as the authors justifiably point out, the administrative claims-based nature of the Nationwide Inpatient Sample precludes the authors from concluding about causation. As such, reasonable doubt can be raised with regard to several of their key findings. For example, the authors highlight that 15% of sepsis occurred in patients requiring total parenteral nutrition (TPN). Therefore, the authors conclude that minimizing the need for such indwelling catheters would reduce the risk of such catheter-related infections. Yet, it is entirely possible that patients presenting with severe sepsis (and other subsequent acute-phase events) required TPN at some point during hospitalization. The same could be said about hospital stay: does infection lead to a prolonged stay or does a prolonged stay put a patient at risk for infection? Therefore, due to a lack of a clear chronology of diagnostic codes in this dataset (unlike procedure codes, which can be determined using the variable “PRDAYn”), the findings of this study need to be interpreted with caution. Indeed, is it the chicken or the egg?

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.