Among the aims of this study, there was certainly a description of the Italian frame in relation to HAI prevalence and antimicrobial use. Interestingly, the area of Southern Italy showed a considerably lower HAI prevalence (even below 5%): however, this data come from many smaller-sized hospitals (Table 1), which may have played a confounding role since HAIs occur more frequently in bigger hospitals, with higher care complexity. Moreover, the reduced participation in the survey (Supplementary Table A1) and the relatively lower amount of blood tests annually performed (Table 1) in this area allow speculating that, probably, less attention is paid to HAI diagnosis and surveillance, with a tendency to under-notification, as already spotted in previous national surveys [41].
The main aim of this study was to identify variables with a potential role as determinants of HAI prevalence and proportion of antibiotics prescribed with no specified reason. A greater proportion of single-bed rooms seems to lead to a reduction in HAIs: this agrees with results obtained in high-risk settings, where a reduction in infections by methicillin-resistant Staphylococcus aureus (MRSA) was observed after introducing new single-bed rooms [42], even though some studies about direct contact infections (Clostridium difficile) failed to show significant improvement in incidence rates in single-bed compared to multiple-bed rooms [12].
Remarkably, HAI prevalence did not appear to be impacted by any other structural determinants, including indicators related to ABHR consumption, which represents a provenly effective measure in HAI prevention [43]. Nevertheless, it is important to consider possible issues of this indicator, mainly linked to problems in measuring healthcare workers’ actual compliance to hand hygiene practice, and in making comparisons between diverse contexts (with extremely different hand hygiene requirements, e.g. intensive-care units vs low-intensity wards) [44]. Furthermore, the effect of hand hygiene on HAI transmission has been shown to be non-linear [45], thus some relationships - albeit existent - might have remained undetected by the regression model.
Having one more antibiotic consultant in the IPC staff per 100 beds appears to reduce HAI prevalence by 33% and antimicrobials with no specified reason by 21%: hence, activities performed by these professionals seem to have a positive impact, reasonably due to their bridging role between scientific evidence and routine clinical practice [27].
Also, the presence of IPC nurses can lower both HAIs and antimicrobial agents with no specified reason (-8% and − 22%, respectively), although a fully significant result is obtained in the latter outcome only. This finding agrees with previous studies underlining the substantial role of IPC nurses [21–23] and the importance of fulfilling the international standard that would require the presence of at least one IPC nurse per 250 beds [46]: indeed, having this threshold reached - in almost all Italian macro-regions - by less than half of the included hospitals might represent a partial limitation of the model on this variable. Hence, training the model on a higher number of hospitals would be advisable, to include more healthcare facilities with higher values for this variable, and thus to provide more precise estimates for effect sizes.
Analysis of both outcomes’ determinants demonstrates that surveillance is paramount, but setting up a systematic IPC programme is all the more important: as already pointed out in previous studies, drafting an appropriate plan [28] is crucial to involve all stakeholders required for the creation of a suitable patient safety climate aimed at HAI prevention. Equally, producing adequate reports is essential to give precise feedback to all involved operators, providing them with a tool for evaluation and improvement of good practice implementation [6, 7].
A systematic IPC activity - of which the production of a plan and a report could represent a proxy indicator - is probably the ground for establishing appropriate surveillance for HAIs and antimicrobial use, which have proved to effectively reduce both the overall HAI incidence (for example in Germany [29], the Netherlands [30] or South Korea [31]), and the incidence of specific HAI subgroups, such as surgical site infections [47] and central line-associated bacteraemia [48].
The establishment of a patient safety environment within healthcare facilities leads to improved application of prevention measures [49], which retain a major role in HAI control (-84% according to our model, Table 2): as shown in the literature, applying simple bundles, for example, in monitoring ventilated patients [50], or adequate training for correct use of central lines [51], can contribute to reducing respective infections even by 66%. Unsurprisingly, the greatest results are obtained by combining several prevention measures, as suggested by multimodal prevention strategies, whose effects have been widely demonstrated, especially for MRSA, central line-related and lower respiratory tract infections [11].
Regarding antibiotic use, the model output suggests that the presence of a stewardship programme could potentially halve HAI prevalence in a hospital, and its effectiveness significantly improves with an increasing number of stewardship measures (Table 3): this confirms that antimicrobial stewardship can improve antibiotic use - and reduce healthcare costs - without jeopardizing the quality of assistance [52]. However, the presence of antibiotic stewardship seems to be positively correlated with a higher share of antimicrobial drugs prescribed with no reason specified on medical records. However, this apparent paradox disappears if a specific stewardship indicator – i.e., presence of routine post-prescription review – is considered: a positive impact of this element is recorded on both HAI prevalence (-27%) and antimicrobials prescribed with no specified reason (-32%).
The seeming contradiction is probably explained by the fact that “antimicrobial stewardship” can be considered an umbrella term [53], i.e. a cross-functional word encompassing all evidence-based strategies aimed at improving healthcare quality through optimization of antimicrobial use. Monitoring healthcare quality through wide-scope indicators undoubtedly represents a valuable resource, since it provides an overall view on many different management features, but must be interpreted very carefully, considering also more specific indicators [54].
Not coincidentally, routine execution of post-prescription review is often used as specific antibiotic stewardship indicator [34], first because this is an element effectively reducing antimicrobial prescriptions [35], and also because this intervention has proved to be successful both in modifying physicians’ behaviour towards antibiotic prescription [55] and in creating a relationship of confidence and co-operation among teams handling patients’ therapies [56], thus working as an actual drive for stewardship.
Among the strengths of this research, the study design must be mentioned: the analysis has been projected by using national-level data, and by reviewing what was a priori known about possible relationships among variables involved, to investigate the relationship between the potential determinants monitored by the European PPS and two of the main outcomes of the surveillance. This acquires further importance, as relating determinants to outcomes is key to performing appropriate healthcare evaluations.
Nevertheless, some limitations must be acknowledged: a first shortcoming comes from the fact that healthcare facilities participated in the surveillance on a voluntary basis, after responding to regional coordinators' requests. Hence, this convenience sampling may have suffered from the well-known selection biases and lack of representativeness [57], especially considering the great heterogeneity between healthcare services in different Italian regions and the reduced participation of hospitals from Southern Italy: for instance, the proportion of medium- and high-size hospitals in our sample is slightly higher than the true proportion among Italian hospitals.
The final 55-hospital subsample tried to provide a solution to the issue of the representativeness of data provided to the ECDC for the European prevalence computation [58]: however, unluckily, the model proposed in this study would be hardly reproducible on that subsample, since there would not be a sufficient sample size to obtain estimates with sufficient statistical power. Repeating this study on a larger group of hospitals, possibly from several countries, would help to obtain more robust (and less setting-dependent) estimates of the actual effect of various determinants on examined outcomes.
Another possible limitation lies in the fact that this analysis proposes a causal inference model based on data from a prevalence survey, and therefore attempts to detect cause-effect relationships (whose development needs a latency period) starting from point data. However, the conditions for building an aetiological hypothesis starting from available prevalence data are almost all apparently met, such as stable patient population and lack of effect of the investigated variables on the duration of HAIs once acquired [59]. Instead, our study is rooted in the assumption that the evaluated hospital IPC strategies are not directly affected by the HAI prevalence observed during the study (inverse causation). Although not testable, this assumption can be based on the low probability that the great majority of participating hospitals changed their IPC strategies in response to an increase in HAI prevalence close to the time of the PPS. Instead, it is reasonable to suppose that changes in the IPC strategies were distributed over the years and there was enough time for HAI prevalence to return to the expected steady state, given the strategies in place.