Appending limited clinical data to an administrative database for assessing institutional quality outliers for AMI care: Does it matter?

A European benchmarking system to evaluate in-hospital mortality rates in acute coronary syndrome: The EURHOBOP project

“…Model discrimination at the patient level was similar between our models and those reported in previous studies [5][6][7][8][9][10]. Our AMI models showed a C-index of 0.77 and 0.82 for administrative and clinical models, respectively.…”

“…Patient-level variables included in our models were also retained in previous models with similar effects [5][6][7][8][9][10]. In addition, prior AMI and PCI models included more comorbidities, admission and procedural variables [5][6][7][8], which were not included in our models due to nonavailability, presence of missing values or non-compliance with the reference time. Regarding hospital-level variables, only the AMI model by Seghieri et al included one, catheterization laboratory [6], which did not contribute significantly to our AMI models, even if CCU/ICU was removed from the list of candidate variables.…”

“…In the last decade, several models have been developed to analyse hospital performance in ACS management based on in-hospital or 30day mortality [5][6][7][8][9][10]. Krumholz et al provided the first model to estimate AMI 30-day mortality for benchmarking purposes, using administrative data [5].…”

“…Krumholz et al provided the first model to estimate AMI 30-day mortality for benchmarking purposes, using administrative data [5]. Following similar methodology, other authors developed models to estimate AMI [6][7][8], ACS [9], and percutaneous coronary intervention (PCI) [10] mortality, mainly using administrative data.…”

“…For this comparison to be reliable, the data used to develop the models should be representative of European hospitals. However, the available models were mainly developed in the US [5,7,8,10] and most of them, including the only PCI model, were developed for patients older than 64 years [5,8,10]. In addition, cross-European performance models, which would be representative of European hospitals and would allow a common strategy for ACS quality of care improvement in Europe, are lacking.…”