Study setting and description of routine AMS team practice

The Peter MacCallum Cancer Centre (Melbourne, Australia) is a quaternary 100-bed cancer center providing inpatient and outpatient hematology, medical oncology, surgical oncology and radiotherapy services.

The infectious diseases (ID) and AMS service are embedded throughout the hospital. The ID service provides a consult service at the request of treating teams and attend weekly interdisciplinary hematology ward rounds. The AMS team is responsible for local guidelines, audits and feedback, antimicrobial formulary restrictions and monitoring of antimicrobial consumption. The AMS team conduct twice weekly, pharmacist-led ward rounds providing antimicrobial PAF to the medical and surgical oncology teams. Ward rounds are also attended by a specialist nurse/nurse practitioner and a rotating ID physician. Patients who are receiving restricted IV antimicrobials (eg, piperacillin-tazobactam, meropenem), longer courses of IV therapy, have multiple comorbidities or are commenced on the institutional sepsis pathway are prioritized for AMS PAF. Following review, recommendations are documented in the electronic medical record (EMR). PAF recommendations are communicated to treating teams via an EMR secure chat function.

In October 2023, the AMS team implemented a hospital wide IV to PO switch guideline. Interdisciplinary education and promotion was conducted during this period. Following guideline implementation, the AMS team documented patient eligibility for IV to PO switch (Appendix 1) in the EMR following PAF.

Between 19/03/2024 and 23/08/2024, AMS ward rounds were rotated every four weeks between two inpatient streams, medical and surgical oncology. The intervention cohort comprised of inpatient admissions where AMS PAF occurred, and the comparator cohort comprised of inpatient admissions to the non-reviewed oncology team during the same rotational period and were managed by treating teams without AMS PAF (treatment team-directed care). For treatment team-directed care encounters, “potential review” was defined as the corresponding date on which AMS PAF would have occurred had that oncology stream been allocated to AMS PAF.

To assess impact of AMS PAF, a retrospective chart review was conducted to compare the patient cohorts who received AMS PAF to those who did not (treating team-directed care). Data was collected using an online database (REDCap®, Version 15.5.38, Vanderbilt University, Nashville, TN). Patients were included if they were prescribed an IV antimicrobial at the time of first AMS PAF or potential review. Patients were excluded if, at the time of first AMS PAF or potential review, the patient was being consulted by ID or if antimicrobials prescribed were for prophylaxis or for a non-infective indication.

Data

Baseline characteristics including age, gender, admitting specialty and Charlson Comorbidity Index were collected. Antimicrobial prescription data including the antimicrobial name, route, dose/frequency, start and end dates and indication (categorized by the study team into respiratory, intra-abdominal, urinary/genitourinary, fever or suspected infection without defined clinical syndrome and others) were collected. For antimicrobial courses given for an indefinite duration (eg, suppressive therapy), the date of IV to PO switch was assigned as the end date.

For the AMS PAF encounters, the types of recommendations made for each prescription were recorded (cease antimicrobial, IV to PO switch, seek expert advice, change spectrum, change dose, commence antimicrobial prophylaxis or duration advice) as well as acceptance of recommendations (within 24 h, 24–72 h or not accepted).

For treating team-directed care encounters, the date of potential AMS review(s) was collected (ie, the dates a patient was on an IV antimicrobial where an AMS round occurred in the corresponding group). The IV to PO switch criteria was assessed retrospectively by the study team AMS pharmacist to determine if the patient was eligible for switch on the days of a corresponding AMS round.

Outcomes

The primary outcome was the proportion of eligible IV antimicrobial prescriptions switched to oral therapy within 24 hours of meeting switch criteria (assessed at the time of first AMS review or potential review), excluding prescriptions where antibiotic therapy were ceased entirely.

Secondary outcomes included LOS, antimicrobial consumption, 90-day C. difficile infection and 30-day all-cause mortality, ICU admission, unplanned re-admission and antibiotic recommencement. The 30- and 90-day periods were measured from the date of the first AMS PAF or for those in the treatment team-directed care cohort, the first potential review.

Antimicrobial consumption was measured using length of therapy (LOT), days of therapy (DOT) and inpatient DOT/patient days. LOT was defined as the total number of calendar days during which a patient received any antimicrobial therapy regardless of the number of agents or their frequency. DOT was defined as the total number of calendar days on which a patient received at least one dose of an antimicrobial agent (with each antimicrobial counted separately), regardless of the number of doses administered per day. LOT and DOT included the number of days antibiotics were prescribed during the entire hospital admission and the intended duration following hospital discharge. Inpatient DOT/patient days was defined as the DOT accrued during inpatient admission only, divided by the number of days admitted to mitigate the impact of differences in LOS between cohorts.

Time-and-motion study

A time-and-motion study was performed alongside the retrospective review; the aim was to quantify the nursing time taken to prepare and administer IV and PO antimicrobials. Nurses on the medical and surgical oncology wards were selected using convenience sampling and provided verbal consent to be observed from 8am to 11am from Wednesday to Friday over a four-week period (14/08/2024–06/09/2024).

Single observers measured the time taken to administer IV and PO medicines. This included the time taken from the point the nurse entered the medication room to find the antimicrobial to the time point where the medicine was marked as given in the clinical record. For IV antimicrobial infusions, the time the infusion ran was not included.

Statistical analyses

Categorical variables were summarized using frequency and percentage. Continuous variables were summarized using mean, standard deviation (SD) and range; and/or median, interquartile range (IQR) and range as appropriate. Time-to-event outcomes were summarized and visualized using Kaplan-Meier survivor functions and survival and failure curves.

The primary outcome was analyzed using Cox proportional hazards regression. Effect sizes were expressed as hazard ratios (HRs) with 95% confidence intervals (CIs). Hazard proportionality was assessed using scaled Schoenfeld residuals. Univariate and multivariate models were fitted, with multivariate analyses adjusting for age, sex, specialty type (medical vs surgical oncology), cancer type (gastrointestinal, head and neck, lung, sarcoma and other) and comorbidities (defined by Charlson Comorbidity Index variables). Variable selection was based on clinical relevance and univariate model performance. Models were assessed for collinearity and interactions.

Secondary outcomes were analyzed using regression models appropriate to the outcome type. Antimicrobial consumption (LOT, DOT and inpatient DOT/patient days) were analyzed using negative binomial regression with follow-up time included as an offset where appropriate. Binary clinical outcomes were analyzed using logistic regression. Effect sizes were expressed as risk ratios (RR) with associated 95% CIs. The same covariates previously described were included in the analysis as adjusting confounder covariates.

A stabilized inverse probability of treatment weight (IPTW) analysis was performed as a sensitivity analysis to further account for baseline differences between the AMS PAF and treatment team-directed care cohorts. Propensity scores were estimated using a multivariate logistic regression model including the covariates listed previously. Stabilized inverse probability weights were calculated to estimate the average treatment effect and covariate balance before and after weighting was assessed using standardized mean differences (SMDs), with an absolute SMD < 0.10 considered indicative of adequate balance. IPTW analyses were considered sensitivity analyses, with results compared with the primary multivariate-adjusted analyses to assess the robustness of the findings.

Estimated cumulative nursing administration time and mean nursing administration time per inpatient encounter were derived by applying the observed mean IV and PO antimicrobial administration times from the time-and-motion study to the estimated numbers of IV and PO antimicrobial administrations in each cohort. Uncertainty around these estimates was quantified using non-parametric bootstrapping. IV (n = 20) and PO (n = 31) administration times were resampled independently with replacement over 100,000 iterations. For each bootstrap sample, mean IV and PO administration times were recalculated and applied to the observed administration counts to estimate cumulative nursing time and mean nursing time per encounter. Point estimates are presented with percentile-based 95% CIs.

For all analyses, P < .05 was considered significant. All analyses were undertaken in Stata version 18 (StataCorp. 2023. Stata Statistical Software: Release 18. College Station, TX: StataCorp LLC.).

Ethics approval was granted by the Peter MacCallum Cancer Centre Human Research Ethics Committee (HREC/108201/PMCC).