A Propensity Score Matched Study of the Positive Impact of Infectious Diseases Consultation on Antimicrobial Appropriateness in Hospitalized Patients with Antimicrobial Stewardship Oversight

Study design and setting.We conducted a cross-sectional study using data collected as part of the AS program at the University of Maryland Medical Center (UMMC). UMMC is a 750-bed tertiary care center in downtown Baltimore City with active solid organ and hematologic stem cell transplant and oncology programs and eight specialized intensive care units (ICUs). There are 32 infectious diseases physicians that practice within six inpatient adult infectious diseases (ID) consult services (general, ICU, transplant, surgical, oncology, and trauma), as well as a primary ID service. Monthly subspecialty service meetings are held to standardize practice. At UMMC, at any time, there are approximately 500 active antimicrobial orders daily. Per hospital policy, an IDC is required within 24 h of ordering a restricted antimicrobial. Antimicrobials included in the restricted category were based on breadth of spectrum, cost, and toxicity potential of the drug. Mandatory IDC upon starting antimicrobial therapy (except prophylaxis) is required for patients in the oncology ward, medical ICU, and trauma services, regardless of antimicrobial restriction classification. The above policies were in place during the entire study period.

A panel of infectious diseases specialists developed UMMC antimicrobial guidelines based on institutional adaptations (formulary, antibiogram, patient population, etc.) of locally published best-practice guidelines. UMMC antimicrobial guidelines were published originally in 2015 and updated annually or more frequently as needed. The guidelines are posted on the hospital intranet site and are available for download on smartphones. In addition, providers have access to the online Sanford Guide to Antimicrobial Therapy (26).

The UMMC AS program consists of 2 infectious diseases physicians (combined equivalent of 0.3 full-time) and 3 infectious diseases pharmacists (combined equivalent of 1.5 full-time). A prospective audit and feedback (PAF) process was introduced in July 2017 and is the mainstay of the program. PAF is carried out with the goal of reviewing all active antimicrobial orders within 3 to 5 days of starting therapy for patients without IDC and 4 to 6 days for patients with IDC. An AS PAF at 3 days was chosen to replicate the CDC-endorsed “antibiotic timeout” as microbiological and clinical information is available that can be intervened at https://www.cdc.gov/antibiotic-use/core-elements/hospital.html. An additional discretionary day for AS review was given to IDC in order to ensure that IDC communication and documentation were carried out. A list of new antimicrobial orders is obtained daily from the electronic medical record EPIC (Verona, WI). Each new selected prescription is evaluated for its appropriateness against UMMC antimicrobial guidelines by an AS team member who communicates as needed with the provider. The appropriateness of each new antimicrobial, indication, and interventions, and whether recommendations were accepted, are documented in the electronic medical record.

Study period, population, and data collection.The current analysis used data collected for the AS PAF activity from October 2017 to March 2019. The first 3 months of the AS PAF activity (i.e., when the antimicrobial review process was first piloted) were excluded from the current analysis. Antimicrobials eligible for inclusion into the study included all unique prescriptions (i) for adults age 18 years and older admitted to the hospital, and (ii) that were active for ≥3 days duration. Antimicrobials were excluded if they (i) had any missing data (e.g., indication), (ii) were delivered by nonenteral or nonintravenous route (e.g., topical), or (iii) were for prophylaxis.

Study measures, outcomes, and analysis.IDC was defined by the presence of an IDC order or an IDC progress note at the time of PAF review. IDC may have been ordered at any point from admission to the time of AS review. IDC is ordered by the primary service or is automatically ordered with a restricted antimicrobial and in designated required locations (oncology ward, medical ICU, and trauma).

Our primary outcome measure was antimicrobial appropriateness, which was operationalized as a binary variable (yes, no). Antimicrobial therapy was considered appropriate if (i) there was an indication for antimicrobial therapy (i.e., antimicrobial therapy was necessary, as determined by the AS reviewer), and (ii) the correct antimicrobial was used based on UMMC antimicrobial guidelines. However, therapy was also deemed to be appropriate if antimicrobial prescription was discordant with guidelines but justified by expert AS review (e.g., meropenem in a patient with sepsis and a history of an extended-spectrum, beta lactamase-producing, Gram-negative). Guideline-discordant antimicrobial prescriptions that were unjustified by expert AS review were considered inappropriate.

Other variables that could influence antimicrobial prescribing were collected and classified in the following five categories: (i) patient related (sex, age, multidrug-resistant organism [MDRO] colonization, location), (ii) antibiotic related (type, route of administration), (iii) infection related (indication, culture positivity), (iv) prescriber related (primary service), and (v) reviewer related (days till antibiotic review) (see Supplemental Table S1 at https://www.researchgate.net/publication/341459328 for full descriptions).

Statistical analysis.Data analysis was performed using SAS v9.4 (SAS Institute; Cary, NC). We quantified the unadjusted association of IDC and antibiotic appropriateness using logistic regression. Our adjusted multivariable model of the association between IDC and antibiotic appropriateness included variables, which, in bivariate analyses, were related both to our outcome (appropriateness) and exposure (IDC). The bivariate analyses were conducted using Pearson’s χ² statistic or Fisher’s exact test for categorical variables and Mann-Whitney U test or Student's t test for continuous variables, as appropriate.

All variables were evaluated for potential confounding. In addition, stratification by primary service was carried out to further assess the impact of the association of IDC and antimicrobial appropriateness in different settings in the hospital. Medical and surgical specialties were analyzed in additional to subspecialties (e.g., vascular, cardiology).

To decrease the probability of confounding by indication in our nonrandomized study, all of our analyses, those in the entire population and the stratified analyses, were conducted using data in which prescriptions written with IDC consult present versus those that were written without IDC consult present were 1:1 propensity score matched. Separate matching was carried out for the overall sample and within each primary service stratum. Variables included in the propensity score were those variables found in the bivariate analysis to be significantly related to receiving IDC (with P ≤ 0.05) or deemed to be clinically relevant. The nonstratified data included the following covariates in the propensity score model: restricted antimicrobial, culture positivity, intravenous route, MDRO, days until antimicrobial start, days until antimicrobial review, community-acquired pneumonia, genitourinary infection, endocarditis, osteomyelitis, oncology ward, and surgical specialty. We present crude unadjusted analyses run using all study data, results run using the propensity score matched data that have no further adjustment, and results run using the propensity score matched data adjusted for variables that remained imbalanced in the propensity score matched data, which included oncology ward, culture positivity, community-acquired pneumonia, and surgical specialties. Stratified analyses included variables in a propensity matched score that were found in the bivariate analysis to be significant (with P ≤ 0.05) with respect to surgical and medical specialty stratum (see Table S2 at https://www.researchgate.net/publication/341459328), and there was no additional adjustment.

Antimicrobial appropriateness subanalysis.Due to a possible overestimation of antimicrobial inappropriateness with strict guideline concordance, the proportion of antimicrobial reviews that were guideline discordant but justified on clinical grounds was investigated. All antimicrobials that were appropriate were further categorized as either “guideline concordant” or “guideline discordant but justified” at the time of AS review. Reasons for justification of appropriateness (i.e., allergy, antibiotic failure), despite discordance with guidelines, were documented. The proportion of “guideline discordant but justified” antimicrobial prescriptions among all appropriate antimicrobials was compared among the IDC and no-IDC groups.