We conducted a prospective single-centre economic and processual evaluation study of an automated email alert system for orders of last-resort antibiotics, as defined by the German Federal Joint Committee (G-BA), at a large tertiary care university hospital (2210 beds) in Germany (Table 1). The key information regarding the infrastructure of the hospital, ID services and AMS are provided in Table 1.

At the time of initiation of the study, three G-BA-specified last-resort antibiotics (cefiderocol, ceftazidime/avibactam, ceftolozane/tazobactam) were included in the local hospital formulary.

This period was intended to determine the average duration of therapy with last-resort antibiotics in the absence of an alert system. For this purpose, we retrospectively analysed all prescriptions of the three last-resort antibiotics specified by the Federal Joint Committee (G-BA) from the date of market approval of each respective agent (ceftolozane/tazobactam: September 18th 2015; ceftazidime/avibactam: June 24th 2016; cefiderocol: April 24th 2020) until May 31st 2023. Only treatment courses without ID consultation during this pre-intervention period were included in the analysis. Treatment duration was extracted from the medical records and analysed descriptively (mean, median, interquartile range [IQR], minimum, maximum). The results of this analysis were used as baseline to enable the calculation of the mean costs of a course of a last-resort antibiotic therapy.

The wash-in period comprised the time during which the initial alert systems directed to the ID service was developed and piloted. On June 7th 2023, first automated email alerts were introduced and subsequently optimised. Since May 7th, 2024, the alert is fully implemented. No data from this period were included in the final analysis.

During the intervention period (beginning in June 1st 2024), the alert system for last-resort antibiotics was fully implemented (Fig. 1).

Orders for the three aforementioned last-resort antibiotics placed through the hospital pharmacy inventory and supply management system (SAP SE, Walldorf, Germany) required entry of the patient identification number, the name of the prescribing physician, and the respective hospital ward. Immediately upon order entry, the inventory management system automatically generated an email notification to the ID consultation team.

Upon receipt of the email notification, a senior ID specialist was required to assess the indication for the prescribed last-resort antibiotic. Based on this assessment, a recommendation was issued to the primary treating team and communicated without delay. Recommendations included continuation of therapy (with or without dose adjustment and with a defined duration or scheduled reassessment), de-escalation to an alternative non-last-resort antibiotic regimen, or discontinuation of therapy. All reviews and corresponding recommendations were prospectively documented in a REDCap database (21) for subsequent analysis.

Schematic overview of the intervention

To ensure a prompt initiation of therapy in critical cases, the pharmacy dispensed the last-resort antibiotic without delay. This process ensures that the therapy with a last-resort antibiotic, once prescribed by the treating physician, can be initiated immediately, even if the formal indication review by the ID specialist requires several hours. Furthermore, the process guarantees that every course of therapy involving a last-resort antibiotic is reviewed and documented by an ID specialist at the earliest possible time point.

Since a single package of the last-resort antibiotics typically covers three days of antibiotic treatment at standard dosing, a subsequent order by the treating physician is required if the continuation of the therapy is planned. This repeat order triggers a new alert to the ID specialist, thereby guaranteeing that every course of therapy is re-reviewed and documented by the ID service within a maximum of three days.

For the purpose of the present study, only the initial prescriptions and the subsequent implementation and outcomes of the corresponding ID recommendations were analysed; follow-up prescriptions were not included in the evaluation.

ID recommendations were communicated to the treating physicians using a multimodal approach tailored to the clinical complexity and urgency of the case. This included brief verbal consultations, digital documentation in the patient chart and detailed written reports in the medical record of the patient. Follow-up was ensured through the aforementioned repeat-order alert system.

The present analysis includes prescriptions for adult inpatients. Only indication reviews associated with initial prescriptions (“first-time prescription”) of one of the three by the G-BA specified last-resort antibiotics (cefiderocol, ceftazidime/avibactam, ceftolozane/tazobactam) were included; reviews of follow-up prescriptions were excluded from analysis.

To estimate potential economic benefits, we evaluated all indication reviews and subsequent recommendations of ID specialists carried out as a result of automated email alerts. For each case, expected medication costs under continued therapy (using the pre-intervention average duration of therapy) were compared to projected costs if the recommendation of the ID specialist (discontinuation, switch, or dose modification) was immediately implemented.

Cost calculations were performed based on the following parameters:

$$\begin }\left[ \text \right]: \hfill \\ }\left( }} \right) \times }\left[ \text \right] \hfill \\ \end$$

(1)

(References: validated RDD definitions according to German AM surveillance system (22, 23) and ABDA hospital acquisition prices as of May 1, 2025 (24) (see Supplement))

$$ \begin }\left[ \text \right]: \hfill \\ }\left[ \text \right] \times }\left[ } \right]\left( }} \right) \hfill \\ \end $$

(2)

For therapies that were continued with different agents or at an adjusted dose, the corresponding costs were calculated using the same methodology and the pre-intervention median treatment duration was assumed accordingly. The difference between expected costs (continued last-resort antibiotic treatment for full course of treatment) and projected costs (post-recommendation) represented the potential cost savings.

$$ \begin }\left[ \text \right] \hfill \\ }(} - })\left[ \text \right] \hfill \\ - \,}(} - })\left[ \text \right] \hfill \\ \end $$

(3)

Implementation adherence was retrospectively assessed by comparing the documented recommendation of the ID specialist to subsequent prescriptions by the primary treating team. A recommendation was classified as implemented if action was taken within 48 h after the email alert (or, for alerts received Friday/Saturday, by the following Monday).

The actual medication costs incurred were calculated using the same method as for potential savings. The actual cost savings were defined as the difference between the expected costs (continued last-resort antibiotic treatment for full course of treatment) and the actual costs after implementation of ID recommendations.

$$ \begin }\left[ \text \right] \hfill \\ }(})\left[ \text \right] \hfill \\ - \,}(} - })\left[ \text \right] \hfill \\ \end $$

(4)