*Limitations*

We are aware that a major limitation is introduced by the study design: retrospective, based on administrative data, and with no possibility to assess whether AKI was cause or complication of hospitalization. We observed that day-of-week of hospital admission has a significant impact on outcome, but we cannot extrapolate from the administrative database some important items, such as cause of admission and death, intensive care level or hospitals' facilities, device use, type of treatment, and impact of clinical or biochemical parameters. It is known that administrative databases, born to be used for other reasons (i.e., reimbursement), lack specific clinical information and may cause possible misclassification of outcomes, thereby generating confounding factors [35]. Moreover, we did not identify AKI on the basis of international Kidney Disease Improving Global Outcomes (KDIGO) guidelines [36], nor di fferentiate patients on the basis of the cause of AKI and the treatment setting, with the exception of dialysis treatment.

We previously stated that medical and nursing understa ffing, shortage of diagnostic or procedural services, and the presence of inexperienced residents could be related to WE e ffect [9]. Unfortunately, administrative databases do not allow us investigate these aspects, being conceived for financial reasons.

Some years ago, concerns about WE e ffect were raised due to three main potential limitations of administrative databases: (1) coding mistakes, (2) insu fficient consideration of comorbidity, and (3) failure to consider the severity or acuity of patients [37]. According to several authors, the performance of ICD-9-CM for diagnosis of acute renal failure showed poor sensitivity, and high specificity, while positive and negative predictive value could di ffer [38–41]. However, Grams et al. underlined that sensitivity was significantly higher when the selected individuals were aged ≥ 65 years; moreover, AKI diagnosed by administrative data detected more severe disease and higher IHM mortality [41]. Due to this reason, we decided to focus on patients aged > 65 years. Finally, we also have to underline some strengths of our study: (1) a high number of records derived from a national database, (2) the long period of time analyzed, and (3) the utilization of a hard outcome indicator, such as IHM.
