Discussion

These two institutional PPSs provide cross-sectional snapshots of HAI and AU burden at two survey dates a decade apart, rather than a longitudinal assessment of change. Both surveys identified a substantial HAI burden, similar observed hospital-level AU prevalence, high HAI prevalence in the ICU, and very frequent continuation of surgical prophylaxis beyond 1 day. The observed difference in HAI prevalence from 8.3% to 12.3% was not statistically significant and had a wide CI; it should therefore be regarded as a surveillance signal rather than evidence of a continuous 10-year increase or institutional deterioration. Repeated surveillance in Catalonia showed that HAI and antibiotic-use trajectories can be non-linear and need not move in parallel. ^{Reference Casas, Rodrigo and Rodríguez12} The 2022 estimate was close to the 12.1% reported nationally in Greece but higher than estimates from the 2022 German PPS and Lombardy. ^{Reference Palaiopanos, Krystallaki and Mellou3,Reference Aghdassi, Hansen and Peña Diaz13,Reference Antonelli, Ales and Chiecca14} Within the 2022 Greek national PPS, HAI prevalence was 14.3% in tertiary hospitals, 7.5% in secondary hospitals, and 3.5% in pediatric hospitals; the corresponding AU prevalences were 64.9%, 53.5%, and 39.5%. These figures illustrate the influence of hospital type and provide descriptive context rather than risk-adjusted benchmarks. ^{Reference Palaiopanos, Krystallaki and Mellou3} The wider ECDC PPS included 309,504 patients from 1,332 hospitals, but formal comparison with those hospitals was not possible using the available aggregate institutional data. ² The contribution of the present study is therefore the identification of institution-specific priorities rather than the estimation of a national or European trend.

The ICU had the highest HAI prevalence in both surveys, affecting 7/23 patients (30.4%) in 2012 and 6/21 (28.6%) in 2022. This pattern is consistent with regional PPS findings and with the established concentration of pneumonia and device-associated infection risk in critical care. ^{Reference Ioannou, Astrinaki and Vitsaxaki8,Reference Antonelli, Ales and Chiecca14,Reference Walter, Haller and Quinten15} Because device use, severity, prior antimicrobial exposure, and length of stay were unavailable, the present analysis cannot determine which factors explain the ICU prevalence or the descriptive differences in medical and surgical wards. The apparent rise in surgical AU deserves focused audit, but its nominal P value arose from an unadjusted secondary comparison among several specialty tests.

Pneumonia was the most frequently recorded HAI category at both survey dates, supporting continued emphasis on respiratory-infection prevention. By contrast, higher recorded proportions of urinary, bloodstream, and catheter-related infections and the sharp reduction in systemic infections cannot be interpreted as definitive temporal changes. Bloodstream- and catheter-related-infection categories were available under protocol version 4.3, but no such infections were recorded locally in 2012. Protocol revisions, improved microbiological ascertainment, changes in diagnostic and coding practices, and differences in patient case mix may have contributed. Prospective incidence surveillance with common definitions and device denominators would be required to distinguish true changes in acquisition risk from changes in detection and classification.

The different survey seasons further limit comparison. The 2012 survey was conducted during June–July, whereas the 2022 survey was conducted during November. Differences in respiratory-pathogen circulation, hospital activity, occupancy, and patient case mix could have influenced pneumonia and overall HAI prevalence. Because each year was represented by a single survey period, seasonal effects cannot be separated from other between-survey differences.

The interval between the surveys also encompassed the COVID-19 pandemic, which may have affected admission patterns, bed use, case mix, diagnostic testing, IPC practices, and antimicrobial prescribing. In addition, protocol version 6.1 included healthcare-associated COVID-19 categories that had no equivalent in 2012. No comparable intermediate PPS series was available, and the aggregate datasets did not contain standardized information on the timing, implementation, or effectiveness of intervening hospital policies or quality-improvement initiatives. Consequently, the study cannot determine whether the 2022 observations reflect pandemic-associated changes, subsequent stabilization, or the effect of a particular institutional intervention.

The higher proportion of microbiologically documented HAI episodes in 2022 may facilitate targeted therapy and local epidemiologic assessment, but it does not by itself prove improved diagnostic performance. Testing availability, specimen collection, infection-site distribution, and data completeness may also have changed. The non-susceptibility findings similarly require caution. High protocol-specific carbapenem non-susceptibility markers among A. baumannii and other Gram-negative isolates are plausible in the Greek tertiary-care context and are consistent with longitudinal surveillance from western Greece and broader ICU evidence. ^{Reference Spiliopoulou, Giannopoulou and Assimakopoulos16–Reference Bassetti, Poulakou, Ruppe, Bouza, Van Hal and Brink18} However, the denominators were small, patients may have contributed multiple isolates, and the 2 surveys used non-equivalent susceptibility classifications. A deduplicated, isolate-level time series using consistent breakpoints is needed before inferring a temporal change in antimicrobial resistance.

Similar AU prevalence at the two survey dates should not be equated with appropriate prescribing. More than half of patients received an antimicrobial in both surveys, and the overall patient-level prevalence concealed important differences in specialty, indication, and duration. The most actionable finding was prolonged surgical prophylaxis: 51/58 prescriptions (87.9%) in 2012 and 47/48 (97.9%) in 2022 exceeded 1 day. For comparison, 1,253/1,656 surgical-prophylaxis prescriptions (75.7%) exceeded 1 day in the 2022 Greek national PPS. ^{Reference Palaiopanos, Krystallaki and Mellou3} Although this comparison was not adjusted for procedural case mix, the magnitude of prolonged use at both local survey dates identifies surgical prophylaxis as a high-priority stewardship target.

Current surgical-site-infection prevention guidance recommends discontinuation after incision closure for most procedures because continuation does not add preventive benefit and increases antimicrobial exposure. ^{Reference Calderwood, Anderson and Bratzler19} Nevertheless, the aggregate PPS data did not include procedure type, timing of the first dose, intraoperative redosing, allergy status, or documented exceptions. The appropriateness of each individual prescription could therefore not be assessed. A focused stewardship bundle should include procedure-specific guidance, documentation of indication and intended stop time, electronic or pharmacist-supported stop mechanisms, prospective audit and feedback, and department-level reporting. Audits should measure the proportion of prophylaxis prescriptions discontinued within 24 hours and report performance by surgical department and procedure type. A Greek tertiary-hospital program has shown that locally monitored stewardship can improve prescribing indicators, ^{Reference Spernovasilis, Kritsotakis and Mathioudaki20} and systematic stewardship interventions can reduce resistant organisms and Clostridioides difficile infection. ^{Reference Baur, Gladstone and Burkert21}

This study has strengths: both surveys were conducted at the same institution, covered eligible inpatients in participating wards, used the ECDC framework, and assessed complementary epidemiologic and stewardship indicators within the same institution. Its limitations are equally important. PPSs estimate prevalence rather than incidence and preferentially capture conditions associated with longer stays. The protocol versions differed; only aggregate outputs were available; case mix, device exposure, and illness severity could not be adjusted; non-patient outcomes could not account for clustering; multiple secondary comparisons were unadjusted; and isolate counts were small with nonharmonized susceptibility interpretation. The surveys were also conducted in different seasons and were separated by the COVID-19 pandemic. Furthermore, the available data did not permit evaluation of intervening IPC, AMS, or other quality-improvement policies. Because this was a single tertiary referral hospital with modest specialty and isolate counts, the findings should not be generalized to institutions with different case mix, referral roles, resources, or surveillance practices. The findings are therefore most useful for institutional priority setting, not causal inference. Repeated surveillance should now be linked to measurable interventions, particularly reduction of prophylaxis beyond 24 hours and prospective monitoring of respiratory and device-associated HAIs.