Association of a Nurse-Led Antimicrobial Stewardship Intervention with Inappropriate Antibiotic Prescribing in a Secondary-Care Hospital in Mexico: A Quasi-Experimental Before-and-After Study

1. Introduction

Antimicrobial resistance is one of the leading threats to global public health because of its consequences for mortality, patient safety, and the effectiveness of available treatments. In 2021, resistant bacterial infections were associated with approximately 4.71 million deaths worldwide, including 1.14 million deaths directly attributable to antimicrobial resistance [1]. Inappropriate antibiotic prescribing contributes to this burden by increasing selective pressure on microorganisms and exposing patients to unnecessary or excessively prolonged treatment. Improving antibiotic use has therefore become an international priority for research and health-service strengthening [2].

In Mexico, several studies have identified substantial opportunities to improve inpatient antibiotic prescribing. A multicenter survey in secondary-care hospitals of the Instituto Mexicano del Seguro Social reported an antimicrobial-use prevalence of 61%, guideline adherence of 53.9%, and predominance of antibiotics classified in the Watch category of the WHO AWaRe system [3]. Other national studies have reported a high proportion of empirical treatment, frequent intravenous administration, and shortcomings in documented reassessment of prescribed regimens [4]. Although institutional antimicrobial stewardship initiatives have been developed, barriers remain concerning implementation, specialized resources, and operational continuity, particularly in secondary-care hospitals [5].

Antimicrobial stewardship programs aim to optimize treatment selection, dose, route, and duration while preserving clinical effectiveness. Prospective audit and feedback to prescribers are among the recommended strategies, allowing opportunities for improvement to be identified during treatment without delaying initiation when clinically indicated [6,7]. Nurses may contribute meaningfully through their continuous involvement in medication administration, monitoring of clinical progress, and communication with other professionals. Existing research has identified opportunities to involve nursing staff in treatment review, identification of antibiotic-related problems, and communication of recommendations, although such involvement depends on training, professional competencies, and institutional support [8,9,10].

Despite these findings, evidence on structured nurse-led antimicrobial stewardship interventions in Mexican secondary-care hospitals remains limited. Evaluating such strategies under routine clinical conditions may help improve prescribing quality and organization of inpatient care. This study aimed to assess the association between a nurse-led intervention and inappropriate antibiotic prescribing among hospitalized patients at a secondary-care hospital in Apodaca, Mexico, using a quasi-experimental before-and-after design.

2. Materials and Methods

A prospective, analytical, quasi-experimental before-and-after study without a control group was conducted in a public secondary-care hospital in Apodaca, Mexico. The study comprised 27 weekly reporting periods in 2026, including 14 pre-intervention and 13 post-intervention periods. The appropriateness of systemic antibiotic prescriptions was compared before and after implementation of a nurse-led antimicrobial stewardship strategy.

Therapeutic prescriptions for systemic antibiotics administered to hospitalized patients with suspected or confirmed infections were included when sufficient clinical information was available to assess appropriateness. Exclusively prophylactic prescriptions, topical antimicrobials, outpatient treatments, records lacking information necessary for assessment, and duplicate records of the same prescription were excluded. The minimum sample size was calculated for a comparison of two independent proportions, assuming inappropriate-prescribing rates of 52% and 37%, 95% confidence, 90% power, and 1:1 allocation. The target was 229 prescriptions per period. The final analysis included 458 prescriptions from 390 patients, comprising 229 prescriptions and 195 distinct patients per phase. The primary unit of analysis was the prescription, and prescriptions from the same patient were linked through a common identifier.

The intervention consisted of structured audit and feedback to the medical team [11]. During the pre-intervention phase, prescriptions were evaluated without systematically implementing this strategy. Before the intervention began, the nurses responsible for prescription audits received standardized training based on the study protocol and the predefined clinical appropriateness criteria. The training addressed the assessment domains used in the audit, including antibiotic indication, agent selection and spectrum, dose, dosing interval, route of administration, renal adjustment, therapeutic duplication, treatment duration, microbiological findings, and potential treatment modifications. All auditors used the same structured assessment form and operational definitions, with criteria applied consistently to prescriptions in both study periods. In the post-intervention phase, nurses used this framework to identify opportunities for de-escalation, intravenous-to-oral conversion, and discontinuation when clinically appropriate. Recommendations were communicated to the responsible physician, and full acceptance, partial acceptance, or rejection was documented. Therapeutic decisions remained the responsibility of the treating medical team.

The primary outcome was the proportion of inappropriate prescriptions, defined as the presence of at least one clinically relevant criterion for inappropriateness. Assessment considered the infectious diagnosis, disease severity, comorbidities, renal function, clinical course, and microbiological findings when available. Appropriateness of indication, selection and spectrum, dose, interval, route, renal adjustment, duplication, combination therapy, treatment duration, and treatment modification was evaluated. Criteria for de-escalation, oral conversion, and discontinuation were applied only when clinical conditions warranted consideration of these changes. The absence of cultures or susceptibility testing was not automatically classified as inappropriate prescribing. The same criteria were used in both periods, drawing on institutional recommendations, local protocols, applicable clinical practice guidelines, and relevant scientific evidence.

To promote uniform application of the criteria, prescription assessments were guided by a common structured form, predefined decision criteria, and the same institutional recommendations, local protocols, and applicable clinical practice guidelines throughout the study. The protocol provided for a second independent assessment of uncertain prescriptions, referral to a third evaluator with antimicrobial expertise when disagreements persisted, and additional quality-control reviews. However, the final dataset did not allow verification of the implementation of these individual review steps or calculation of inter-rater agreement. Therefore, consistency was supported procedurally through standardized training and assessment materials, but formal inter-rater reliability cannot be claimed.

Age, comorbidities, sepsis, admitting service, infectious focus, and treatment characteristics were recorded. Secondary outcomes included documentation of the indication, culture requests, specific appropriateness criteria, treatment duration, length of stay, clinical resolution, treatment failure, transfer to a higher-level facility, in-hospital mortality, 30-day readmission, and adverse events. Implementation indicators included nursing review coverage, opportunities for improvement, recommendations communicated, and acceptance by physicians. Antimicrobial utilization was described using treatment duration per prescription; days of therapy per 1000 patient-days were not calculated because a validated hospital-level denominator was unavailable [12].

Data were obtained from medical charts, physician orders, nursing records, medication administration documentation, and laboratory and microbiological findings, using coded identifiers. The primary outcome was analyzed at the prescription level (229 records per phase). Clinical outcomes were consolidated at the patient level, retaining one observation per individual and phase (195 patients per phase).

Categorical variables were summarized as frequencies and percentages, and quantitative variables as means and standard deviations or medians and interquartile ranges, according to distribution. Pearson chi-square or Fisher exact tests were used for categorical variables and Student t or Mann–Whitney U tests for quantitative variables, as appropriate. The main association was examined using the absolute difference in proportions and relative risk. A multivariable logistic regression model was fitted with inappropriate prescribing as the outcome and study period as the main explanatory variable, adjusting for age, sepsis, number of comorbidities, hospital service, and infectious focus. Patient-clustered robust standard errors accounted for within-patient correlation, and adjusted odds ratios (aORs) with 95% confidence intervals (CIs) were calculated. Clinical outcomes were compared by patient using Fisher exact tests for categorical variables and the Mann–Whitney U test for length of stay. Secondary comparisons were exploratory and were not adjusted for multiplicity. All tests were two-sided, with p < 0.05 considered statistically significant. Analyses were conducted using IBM SPSS Statistics, version 24.0 (IBM Corp., Armonk, NY, USA).

The hospital bioethics committee authorized initiation of the study in March 2026. The Ethics and Research Committee of Family Medicine Unit No. 32, Instituto Mexicano del Seguro Social, subsequently issued its favorable determination on 30 September 2026 (CONBIOETICA-19-CEI-008-20180807; institutional protocol R-2026-1909-123). Written informed consent was obtained before inclusion, and records were processed using coded identifiers to protect confidentiality.

3. Results

A total of 458 systemic antibiotic prescriptions from 390 hospitalized patients were analyzed, with 229 prescriptions and 195 distinct patients in each period. Mean age was 56.82 ± 16.13 years before and 58.25 ± 15.55 years after the intervention (p = 0.502). Mean number of comorbidities was 1.10 ± 0.90 and 0.94 ± 0.86, respectively (p = 0.038). Differences were also observed in heart failure and a history of antibiotic allergy. Internal Medicine accounted for the largest proportion of prescriptions in both phases, followed by General Surgery. Pneumonia was the most common infectious diagnosis before the intervention (31.44%), whereas urinary tract infections predominated afterward (31.88%). Baseline clinical and service characteristics are presented in Table 1.

Inappropriate prescribing decreased from 51.97% (119/229) before the intervention to 37.12% (85/229) afterward, an absolute difference of −14.85 percentage points (95% CI, −23.61 to −5.75; p = 0.0014). The relative risk was 0.714, corresponding to an observed relative reduction of 28.6%. After adjustment for age, sepsis, number of comorbidities, hospital service, and infectious focus, the post-intervention period was associated with lower odds of inappropriate prescribing (aOR = 0.544; 95% CI, 0.368–0.804; p = 0.0023). Table 2 presents the primary findings.

During the post-intervention period, nurses reviewed 219 of 229 prescriptions (95.63%). Opportunities for improvement were identified in 93 of the 219 reviewed prescriptions (42.47%), and 88 of these 93 opportunities (94.62%) were communicated to physicians. Among communicated recommendations, 64 were fully accepted (72.73%), six were partially accepted (6.82%), and 18 were rejected (20.45%). Full or partial acceptance reached 79.55% (70/88). The most frequent opportunities concerned treatment duration (22.58%), dose adjustment (17.20%), intravenous-to-oral conversion (16.13%), and antimicrobial de-escalation (15.05%). Implementation indicators are shown in Table 3.

Documentation of the antibiotic indication increased from 83.41% to 91.70% (p = 0.007), and the proportion of prescriptions with a microbiological culture requested increased from 48.91% to 68.12% (p < 0.001). Among specific criteria for inappropriate prescribing, inappropriate duration declined from 13.10% to 6.55% (p = 0.019). The mean number of inappropriateness criteria per prescription declined from 0.73 ± 0.81 to 0.55 ± 0.79 (p = 0.005). No statistically significant differences were identified in individual criteria for selection, spectrum, dose, interval, route, de-escalation, or timely discontinuation.

Mean antibiotic treatment duration decreased from 7.05 ± 2.71 to 6.18 ± 2.26 days per prescription (p < 0.001), a difference of −0.87 days. Empirical treatments comprised 93.01% of prescriptions before and 88.21% after the intervention (p = 0.109). Under the AWaRe classification, Watch antibiotics predominated in both periods, although their share declined from 75.11% to 68.12%, while Access antibiotics increased from 22.71% to 31.44%. The overall distribution across categories differed statistically (p = 0.036), although Reserve-category counts were small.

For clinical outcomes, records were consolidated to 195 unique patients per period. Mean hospital length of stay was 10.06 ± 3.09 days before and 9.12 ± 2.69 days after the intervention, a difference of −0.94 days (p = 0.0033). Treatment failure occurred in 19 patients (9.74%) before and 17 (8.72%) after the intervention (p = 0.861), and in-hospital mortality occurred in five (2.56%) and two patients (1.03%), respectively (p = 0.449). Thirty-day readmission occurred in 17 (8.72%) and 18 (9.23%) patients (p = 1.000). No statistically significant differences were observed in clinical resolution, transfer to a higher-level facility, antibiotic-related adverse events, or Clostridioides difficile infection. These analyses used unique patients and are summarized with other secondary indicators in Table 4.

4. Discussion

The nurse-led intervention was associated with a lower frequency of inappropriate antibiotic prescribing in a Mexican secondary-care hospital. The proportion declined from 51.97% to 37.12%, an absolute difference of 14.85 percentage points. The association remained after adjustment for clinical and service characteristics (aOR = 0.544; 95% CI, 0.368–0.804). These findings suggest that structured nursing participation may contribute to improving hospital prescribing practices. However, the absence of a control group prevents the observed changes from being attributed exclusively to the intervention.

The direction of these findings is consistent with those of Kaur et al. [13], who assessed a nurse-led antimicrobial stewardship strategy in a hospital emergency unit. Guideline adherence increased from 82.4% to 91.0%, and culture collection before treatment increased from 23.0% to 40.3%. Although both studies observed improvements in care processes, differences in populations and indicators prevent direct comparison of effect sizes. Doshi et al. [14] conducted a cluster-randomized trial in 16 district hospitals in Vietnam and reported a −6.3-percentage-point difference in inappropriate prescribing versus control hospitals (95% CI, −10.9 to −1.7). That design affords stronger causal inference and places our findings in the broader context of nonspecialist hospitals.

In the post-intervention period, nurses reviewed 95.63% of prescriptions and communicated recommendations for 88 of 93 identified opportunities for improvement. Full or partial acceptance was 79.55%, although favorable responses do not establish that all proposed changes were carried out. Ergen et al. [15] also evaluated an audit-and-feedback strategy, delivered by pharmacists in intensive care. Despite differences between the models, systematic review and communication with prescribers were shared components. Participation by several professions may support implementation, provided their clinical competencies and responsibilities are respected.

Training is an important requirement for sustaining these activities. Hendy et al. [16] reported improvements in nurses’ knowledge, perceptions, and antimicrobial stewardship practices after an educational intervention and identified workload-related barriers. Although the present study did not measure changes in individual competencies, the achieved review coverage suggests that prescription reviews could be integrated into routine care during the study period. Long-term sustainability, however, was not assessed.

One notable improvement was the decline in inappropriate treatment duration from 13.10% to 6.55%, accompanied by a mean reduction of 0.87 antibiotic days per prescription. Livorsi et al. [17] reported improved optimal prescribing at hospital discharge following an audit-and-feedback intervention, without a significant reduction in overall antibiotic use. Together, these observations underscore the distinction between prescribing appropriateness and the volume of antibiotics used. This study documented changes in individual treatment duration but could not calculate standardized hospital antibiotic-use rates per patient-day.

Opportunities for improvement primarily involved duration, dose adjustment, intravenous-to-oral conversion, and de-escalation. Wong et al. [18] identified organizational and interprofessional factors shaping nurses’ participation in stewardship, while Zhao et al. [19] described competencies relating to antimicrobial monitoring and clinical communication. In our study, nurses issued recommendations, but treatment decisions remained with physicians. The lack of systematic verification of implemented changes limits identification of the specific components responsible for the observed improvement.

Treatment duration must be interpreted in light of the diagnosis and clinical course. Harvey et al. [20] described criteria for safe transition from intravenous to oral antibiotics, while Lee et al. [21] reviewed evidence supporting shorter courses for selected infections. Thus, reducing treatment days alone does not necessarily indicate better prescribing. Microbiological culture requests also increased, from 48.91% to 68.12%, but this increase does not prove that every request was appropriate. As Fabre et al. [22] noted, diagnostic stewardship requires attention to the indication, timing, and interpretation of microbiological tests.

Among clinical outcomes, the analysis of 195 patients per phase showed a shorter mean hospital stay after intervention (10.06 versus 9.12 days; p = 0.0033). No statistically significant differences were found in mortality, treatment failure, readmission, or antibiotic-related adverse events. These findings neither establish better clinical prognosis attributable to the intervention nor demonstrate clinical equivalence between periods. In a meta-analysis, Baur et al. [23] reported that some stewardship programs reduced infections with resistant organisms and Clostridioides difficile. The present study did not assess changes in bacterial resistance and recorded only one C. difficile infection in each period; consequently, conclusions about these outcomes are not warranted.

The main contribution of this study is its description of structured nursing participation in antimicrobial stewardship at a Mexican secondary-care hospital. Harun et al. [24] identified training, resources, interprofessional collaboration, and institutional support as factors relevant to implementation in low- and middle-income settings. The present findings indicate favorable initial operational integration but do not establish sustainability, cost-effectiveness, or reproducibility elsewhere.

Strengths include prospective data collection, evaluation of an intervention embedded in hospital care, and multivariable analysis with patient-clustered robust standard errors. Clinical outcomes were analyzed after consolidating records by patient, avoiding duplication caused by multiple prescriptions. Limitations include the uncontrolled before-and-after design, potential temporal changes unrelated to the intervention, and differences in clinical characteristics between phases. The single-center setting limits generalizability. Independent inter-rater agreement for appropriateness classification was not documented, and no validated denominator was available to calculate standardized antimicrobial use. Implementation of accepted recommendations was not systematically verified, and rare events limited precision in secondary clinical analyses. These comparisons should be regarded as exploratory.

5. Conclusions

The nurse-led antimicrobial stewardship intervention was associated with a lower proportion of inappropriate antibiotic prescriptions and shorter treatment duration. High prescription-review coverage and recommendation acceptance support its initial operational feasibility. Multicenter studies with comparator groups, tracking of actual treatment modifications, and standardized antimicrobial-use measures are needed to establish effectiveness and sustainability.