Formulary restriction & preauthorization stewardship — reserve agents routed through ID/pharmacy review before dispense
Formulary restriction with preauthorization is one of the two "core" interventions the CDC and IDSA identify as most effective for hospital antibiotic stewardship programs (the other being prospective audit-and-feedback). A defined reserve list — typically carbapenems, polymyxins, newer β-lactam/β-lactamase-inhibitor combinations, and high-exposure regimens of daptomycin or linezolid — cannot be dispensed until an infectious-diseases (ID) physician or ID-trained pharmacist signs off. The moment a prescriber selects one of these agents in the EHR, the order is intercepted and diverted into an approval workflow rather than transmitted directly to pharmacy.
Reserve-tier lists are built around agents with a high resistance-selection footprint, narrow remaining spectrum against multidrug-resistant organisms, high acquisition cost, or significant toxicity requiring specialist titration. A representative list:
• Carbapenems — meropenem, imipenem-cilastatin, ertapenem: broad Gram-negative coverage, but overuse directly selects for carbapenem-resistant Enterobacterales (CRE) and Pseudomonas • Polymyxins — colistin, polymyxin B: nephrotoxic agents of last resort for extensively drug-resistant Gram-negatives • Novel β-lactam/β-lactamase-inhibitor combinations — ceftazidime-avibactam, meropenem-vaborbactam, imipenem-relebactam, cefiderocol: reserved specifically for confirmed or suspected carbapenem-resistant organisms so susceptibility is not squandered on empiric use • High-exposure daptomycin — doses above renal-adjusted 6–8 mg/kg for refractory MRSA/VRE bacteremia or endocarditis • Extended-duration linezolid — beyond 7–10 days, given cumulative myelosuppression and serotonergic-interaction risk • Select antifungals — liposomal amphotericin B, isavuconazole — often co-managed on the same restricted-order pathway
The list is reviewed at least annually by the pharmacy and therapeutics (P&T) committee, informed by the hospital antibiogram, local CRE/VRE prevalence, and drug-acquisition cost.
Most CPOE (computerized provider order entry) systems implement the hold as a hard interrupt: selecting a restricted agent triggers an interruptive alert that blocks the order from routing to pharmacy verification until an approval flag is set. This differs deliberately from a passive "soft stop" advisory alert, which studies show is overridden more than 90% of the time. Hard stops on true restriction lists show override rates below 5% because the pharmacy simply will not dispense without the flag.
Exception pathways exist for genuine emergencies — e.g., septic shock with a known carbapenem-resistant organism — permitting a single empiric dose to be released immediately while the formal review proceeds in parallel (see Stage 5, "curbside" allowance). This prevents restriction programs from ever being blamed for delaying resuscitative antibiotics in sepsis, which remains the top clinical objection to preauthorization models.
A restriction program lives or dies on the quality of the data a prescriber is forced to submit before a reviewer ever sees the case. A well-designed structured request form converts an ambiguous verbal or free-text order into a standardized packet: indication, microbiology, allergy history, and treatment trajectory — the same four data elements an ID consultant would ask for on a curbside call.
1. Indication and site of infection — pneumonia (HAP/VAP vs. CAP), intra-abdominal, bloodstream, urinary, bone/joint, skin/soft tissue — because spectrum and duration norms differ enormously by site.
2. Culture and susceptibility data — Gram stain, organism identification, and susceptibility panel if available; if empiric and pending, the prior 90-day culture history and local unit antibiogram substitute as the best available evidence.
3. Allergy and intolerance history — specifically distinguishing a true IgE-mediated penicillin allergy (hives, anaphylaxis) from a benign childhood rash, since roughly 90% of patients labeled "penicillin allergic" can in fact tolerate a β-lactam — a distinction that alone eliminates a large fraction of "restricted agent because allergy" requests.
4. Prior antibiotic exposure and failure — what was already tried, at what dose, for how long, and why it was deemed inadequate (clinical non-response, culture growth despite therapy, or a documented adverse effect).
Forms that omit any of these four elements are auto-returned to the ordering prescriber before entering the reviewer queue — this pre-filtering step alone removes 15–20% of requests without consuming reviewer time.
The highest-performing programs embed the justification form directly in the CPOE order set as a "smart form" that auto-populates from existing chart data: most-recent culture results pulled from the lab interface, allergy list pulled from the allergy module, and current antimicrobial regimen pulled from the medication administration record. This reduces prescriber burden from an estimated 12 minutes (free-text paging and manual chart summary) to under 4 minutes, and is the single largest driver of prescriber satisfaction scores with restriction programs — dissatisfaction with the paperwork burden, not the restriction itself, is the most common reason clinicians cite for wanting to bypass stewardship review.
Once a complete packet reaches the queue, an on-call ID physician or an ID-trained clinical pharmacist with prescribing collaboration privileges evaluates it against restriction criteria, the unit-specific antibiogram, dosing/renal adjustment, and drug-interaction risk. Most mature programs commit to a turnaround-time (TAT) service-level target of 4 hours for routine requests and near-immediate response for flagged emergent cases.
The reviewer cross-checks five things in sequence:
• Restriction criteria match — does the documented indication fall within the pre-approved criteria for this agent (e.g., ceftazidime-avibactam restricted to confirmed/suspected carbapenem-resistant Gram-negative infection)? • Local antibiogram concordance — is the empiric choice consistent with unit-level susceptibility data (CLSI M39-based cumulative antibiogram), or would a narrower agent achieve equivalent coverage given local resistance rates? • De-escalation opportunity — if culture data are already back, can the order be safely narrowed to a non-restricted agent rather than approved as-is? • Dosing and renal/hepatic adjustment — particularly critical for colistin (loading + maintenance nomograms) and vancomycin/daptomycin (AUC-guided dosing) • Drug interaction and toxicity screening — e.g., linezolid with serotonergic agents, daptomycin with statins (CPK monitoring)
A documented rationale is required regardless of outcome — approvals as well as denials are logged with the specific criterion met, which becomes the raw data for the quarterly stewardship report.
Coverage models vary by hospital size. Academic centers typically staff a dedicated ID stewardship pharmacist during business hours with an ID physician on-call for complex or after-hours cases; community hospitals more often rely on a remote tele-stewardship contract or a rotating on-call ID physician covering multiple facilities. After-hours and weekend coverage is the most common point of TAT failure — surveyed programs report median TAT of 1.8 hours during business hours versus 6.4 hours overnight/weekend, which is precisely why the curbside empiric-dose allowance (Stage 5) exists as a safety valve.
Every reviewed request resolves into one of three outcomes. Straight approval releases the order to pharmacy unchanged. Denial blocks the restricted agent outright. The third and clinically most instructive path — a suggested alternative — occurs when the reviewer denies the specific agent requested but simultaneously proposes a narrower or equally effective substitute, closing the loop in the same interaction rather than leaving the prescriber to re-submit blind.
When a request does not meet restriction criteria — most commonly because culture data support a narrower-spectrum choice, or the documented "allergy" does not preclude a first-line β-lactam — the reviewer does not simply deny and stop. High-performing programs require that roughly a quarter to a third of all denials be paired with a concrete substitute recommendation delivered through the same order channel: e.g., "deny meropenem; recommend ceftriaxone 2g q24h given culture-confirmed ceftriaxone-susceptible E. coli" or "deny vancomycin; penicillin allergy label does not preclude cefazolin — recommend cefazolin 2g q8h."
This pathway is where the largest measurable de-escalation gains occur: it converts a restriction event from a pure "no" into an active stewardship intervention, and audit data consistently show it is accepted by the ordering prescriber in the large majority of cases without further escalation or appeal.
A denial is not final. Every program maintains an appeal pathway: the prescriber can request a peer-to-peer discussion with the reviewing ID physician, escalate to the ASP medical director, or — in a true clinical emergency — invoke the one-time emergency-override provision that releases a single dose while documentation catches up. Override rates on genuine hard-stop restrictions are low (typically under 5%) precisely because the justification and alternative-suggestion steps upstream resolve the great majority of disagreements before they reach the override stage.
A restriction program that delays antibiotics in a genuinely septic patient causes measurable harm — each hour of delay in appropriate antimicrobial therapy for septic shock is associated with an incremental increase in mortality. Mature programs therefore track turnaround time as a primary safety metric with the same rigor as approval rate, and build in an explicit empiric "curbside" allowance so review latency never becomes a barrier to timely first-dose antibiotics.
Median turnaround time flatters a program — a queue that clears most requests in 90 minutes can still leave a meaningful tail of patients waiting 8+ hours overnight. Programs that only report the median can mask exactly the failure mode most likely to cause harm: the after-hours or weekend request that sits unreviewed. Reporting the 90th (or 95th) percentile alongside the median, stratified by time-of-day and day-of-week, is now considered best practice for stewardship dashboards and is the metric most frequently cited in P&T committee reviews of program performance.
To decouple "time to formal approval" from "time to first dose," essentially all functioning restriction programs permit a single empiric dose of the restricted agent to be released immediately upon order entry in flagged high-acuity scenarios — severe sepsis, septic shock, neutropenic fever — with formal ID/pharmacy review proceeding in parallel rather than as a precondition. This single dose is time-limited (typically 24 hours) and automatically re-triggers the hold before a second dose can be dispensed without a completed review.
Retrospective audits show roughly 70% of curbside empiric doses are subsequently confirmed as appropriate on formal review; the remaining 30% are de-escalated or stopped once culture data return — meaning the safety valve does exactly what it is designed to do: protect time-sensitive patients without materially eroding stewardship oversight.
The 2016 IDSA/SHEA stewardship guideline explicitly endorses combining preauthorization with a documented emergency-override or empiric-allowance mechanism — programs that lack this valve show measurably higher rates of clinician workaround behavior (verbal orders bypassing CPOE, "splitting" a broader agent under a different indication) that undermine the restriction list entirely.
The entire architecture of preauthorization exists to move one population-level number: the incidence of carbapenem-resistant and other multidrug-resistant organisms in the facility. Quarterly stewardship reports pair restricted-agent utilization — days of therapy (DOT) per 1,000 patient-days — against CRE incidence density, closing the loop between an individual approval decision and hospital-wide resistance trends.
Antimicrobial consumption is tracked using WHO AWaRe-classified days of therapy per 1,000 patient-days, benchmarked against NHSN Standardized Antimicrobial Administration Ratios (SAARs). Facilities that implement preauthorization for carbapenems typically see carbapenem DOT fall 20–35% within the first year, driven by three simultaneous effects documented in this simulation: fewer restricted agents are started in the first place (Stage 1 gatekeeping), a meaningful share are redirected to narrower alternatives (Stage 4), and empiric courses are more aggressively de-escalated once culture data return (Stage 5 curbside conversion).
Because carbapenem exposure is the single strongest modifiable risk factor for carbapenem-resistant Enterobacterales colonization and infection at the patient level, sustained consumption reduction at the population level is mechanistically expected — and is what multi-year surveillance consistently shows — to translate into falling CRE incidence density, typically lagging the consumption decline by 12–24 months as colonization pressure within the facility gradually declines.
Preauthorization rarely runs in isolation — it is nearly always deployed alongside contact precautions, active CRE surveillance cultures on high-risk units, environmental cleaning audits, and hand-hygiene programs, making it difficult to attribute resistance reduction to stewardship alone in any single-site study. However, systematic reviews (Cochrane 2017 update on hospital ASP interventions; CDC Core Elements evidence summaries) consistently rank preauthorization plus prospective audit-and-feedback as the intervention pairing with the largest and most reproducible effect on both antimicrobial consumption and downstream resistance markers across dozens of before-after and interrupted-time-series studies, even after adjusting for concurrent infection-control bundles.
A widely cited interrupted time-series from a US academic medical center found that carbapenem-restriction preauthorization, sustained over 4 years, was associated with a 44% relative reduction in carbapenem-resistant Enterobacterales incidence density — with the resistance curve beginning to bend downward roughly 18 months after the consumption curve first fell, consistent with the expected biological lag between reduced selection pressure and measurable population-level de-colonization.