A novel, integrated approach for understanding and investigating Healthcare Associated Infections: A risk factors constellation analysis.

Introduction: Healthcare-associated infections (HAIs) and antimicrobial resistance (AMR) are major public health threats in upper- and lower-middle-income countries. Electronic health records (EHRs) are an invaluable source of data for achieving different goals, including the early detection of HAIs and AMR clusters within healthcare settings; evaluation of attributable incidence, mortality, and disability-adjusted life years (DALYs); and implementation of governance policies. In Italy, the burden of HAIs is estimated to be 702.53 DALYs per 100,000 population, which has the same magnitude as the burden of ischemic heart disease. However, data in EHRs are usually not homogeneous, not properly linked and engineered, or not easily compared with other data. Moreover, without a proper epidemiological approach, the relevant information may not be detected. In this retrospective observational study, we established and engineered a new management system on the basis of the integration of microbiology laboratory data from the university hospital "Policlinico Tor Vergata" (PTV) in Italy with hospital discharge forms (HDFs) and clinical record data. All data are currently available in separate EHRs. We propose an original approach for monitoring alert microorganisms and for consequently estimating HAIs for the entire period of 2018.

Methods: Data extraction was performed by analyzing HDFs in the databases of the Hospital Information System. Data were compiled using the AREAS-ADT information system and ICD-9-CM codes. Quantitative and qualitative variables and diagnostic-related groups were produced by processing the resulting integrated databases. The results of research requests for HAI microorganisms and AMR profiles sent by the departments of PTV from 01/01/2018 to 31/12/2018 and the date of collection were extracted from the database of the Complex Operational Unit of Microbiology and then integrated.

Results: We were able to provide a complete and richly detailed profile of the estimated HAIs and to correlate them with the information contained in the HDFs and those available from the microbiology laboratory. We also identified the infection profile of the investigated hospital and estimated the distribution of coinfections by two or more microorganisms of concern. Our data were consistent with those in the literature, particularly the increase in mortality, length of stay, and risk of death associated with infections with Staphylococcus spp, Pseudomonas aeruginosa, Klebsiella pneumoniae, Clostridioides difficile, Candida spp., and Acinetobacter baumannii. Even though less than 10% of the detected HAIs showed at least one infection caused by an antimicrobial resistant bacterium, the contribution of AMR to the overall risk of increased mortality was extremely high.

Conclusions: The increasing availability of health data stored in EHRs represents a unique opportunity for the accurate identification of any factor that contributes to the diffusion of HAIs and AMR and for the prompt implementation of effective corrective measures. That said, artificial intelligence might be the future of health data analysis because it may allow for the early identification of patients who are more exposed to the risk of HAIs and for a more efficient monitoring of HAI sources and outbreaks. However, challenges concerning codification, integration, and standardization of health data recording and analysis still need to be addressed.