Deaths attributable to bacterial AMR
An estimated 4.71 million deaths were associated with bacterial AMR, including those attributable to it. These figures are not additive.
Lancet · 2024 [2] ↗Antimicrobial stewardship for physicians and pharmacists
Antibiotic stewardship.
Less repetitive work.
Every prescription needs the right antibiotic, dose and duration. ABX Steward connects patient data, flags prescriptions for review and prepares consultation drafts with evidence, helping teams focus limited clinical time on decisions.
Create an account to reviewExplore the synthetic example below without signing in.
Resistance narrows treatment options and increases hospital and healthcare costs. WHO · 2026 [1] ↗
The case for sustainable stewardship
More care is needed. People and funding remain scarce.
Antimicrobial resistance (AMR) makes infections harder to treat. Antimicrobial stewardship (ASP) means ongoing review to provide needed treatment promptly and reduce unnecessary exposure. Appropriate use of today’s antibiotics is part of the response.
Projected health worker shortfall
The shortfall is concentrated in low and lower middle income countries. It covers the health workforce overall, not specifically infection specialists or ASP teams.
WHO [3] ↗Countries reporting dedicated domestic funding
For multisectoral national AMR action plans. This is not a hospital budget statistic, nor evidence that other countries have no AMR spending.
WHO · 2026 [4] ↗Invest in people. Reduce the repetitive work around them.
In Korea, staffing and funding constraints and infectious disease physician workload are reported barriers to expanding stewardship. JAMA · 2026 [8] ↗
ABX Steward addresses the repeated work of finding records, aligning timelines, checking evidence and documenting reviews. It cannot replace workforce investment or funding. Its purpose is to help existing expertise reach more reviews.
What ABX Steward helps prepare
From scattered records to a reviewable clinical draft.
Pharmacists can examine prescriptions and changes that need attention. Specialists review the clinical course and evidence. The hospital retains intervention and handoff records.
Less searching across records
Connect prescriptions, laboratory and culture results from Excel, ZIP or hospital APIs. Review new events alongside the existing admission history.
Orders · Lab trends · Culture and susceptibility
A clearer starting point for review
Flag prescriptions for review with explanations and recommendations. Compare source records and guidance, including applicability and missing information.
Monitoring rationale · Recommendations · Sources
Support for drafting and handoff
Prepare conditional treatment drafts covering drug, dose, interval and duration. Pharmacists and specialists review, revise and pass on their reasoning.
Consultation draft · Expert review · Intervention log
A workspace for connected reviewSynthetic example
+
Stewardship workspacePending reviewInpatient · Example 017
Orders and changes, together
Antibiotic ordersDay 1Day 2Day 3
Antibiotic AIntravenous
Prior course
Antibiotic BIntravenous
Current order
Aligned by time
Review prescription changes alongside test timing.
Serum creatinine1.6 mg/dL
Previous 0.9Day 1 · 0.9Day 2 · 1.1Day 3 · 1.6
Illustrative values, not results from a real patient.
Trace the basis of a suggestion
Antibiotic orders→Lab trends→Guidelines
Compare source records with clinical guidance.
Check which tests and orders support the suggestion, and whether the guidance applies to this patient.
Explore the orders, lab trends and evidence tabs. This is a synthetic illustration of the workflow.
The change we work toward
More of an expert’s time.
For more patients’ reviews.
Less preparation and documentation burden, with more attention for appropriate treatment.
Research vision · Not clinically validated
Beyond which antibiotic:
when should we intervene?
We aim to study an anticipatory stewardship agent that weighs evolving clinical evidence. Should treatment change now? Should we wait for more data? Which choice is safer? It would account for future treatment paths and uncertainty, with clinicians making the final decision.
What may happen next?
Study possible treatment trajectories as new clinical evidence arrives.
Research goal · temporal reasoning
Change now, or wait?
Compare flagging a treatment review, waiting for a result, or requesting more information.
Research goal · information value and intervention timing
Which option is safer?
Make uncertainty explicit, allow abstention, and keep authorized clinicians in control.
Research goal · safety-constrained decisions
Current service versus research vision
Today's pilot supports prescription monitoring, event-driven reassessment, evidence-linked drafts, and clinician review. Future-outcome prediction, counterfactual treatment comparison, intervention timing, and safe abstention are proposed research. They are not yet implemented or clinically validated.
Why invest in stewardship
Research supports the value of appropriate use.
The findings below come from external research and economic modelling. They do not measure ABX Steward’s performance.
Hospital antibiotic prescribing interventions
+15pp
Improvement in policy adherence
43% → 58% · 29 randomized trials
Cochrane review, 2017. Studies were mainly from North America and Europe, with varying interventions and effects. This is evidence for stewardship interventions, not for a particular software product.
Cochrane · 2017 [5] ↗Health resources consumed by resistant infections
28.9bn
Average annual health costs · USD PPP
34 OECD and EU/EEA countries in the model
OECD 2023 model: annual average through 2050, compared with eliminating resistant infections. This is neither a global total nor an amount this service can save.
OECD · 2023 [6] ↗Validate the product’s potential in your hospital.
ABX Steward has not yet established measured staffing, cost or resistance outcomes in clinical practice. The estimator below shows capacity and its monetary value if your assumptions hold.
What could change for your team
Estimate the time and capacity you could reclaim.
Adjust review volume and time per case for your hospital. Include the time needed for final review and corrections.
Monthly result under your assumptions
- Potential time released
- hours
- Capacity in full time equivalents
- FTE
- Staff time valued at the entered rate
- Staff time value minus additional cost
FTE is a time equivalent, not headcount to remove or cash savings. More time or higher costs can produce negative values.
Method and excluded effects
Monthly hours = reviews per day × review days × (current minutes − assumed future minutes) ÷ 60.
FTE = monthly hours released ÷ monthly hours per full time role.
Staff time value = monthly hours released × hourly staff cost.
Value after cost = staff time value − additional monthly operating cost.
All amounts are Korean won (KRW), with no currency conversion. Drug spending, length of stay, resistance and mortality effects are excluded. Whether released time becomes additional care must also be assessed.
What to measure before and after adoption
Record preparation, review, correction and handoff time for comparable patients and tasks. Use a concurrent comparator where possible and independent clinicians to assess recommendation quality, missed risks and false alerts. Measure review coverage, intervention acceptance and actual costs to replace assumptions with observations.
Building sustainable stewardship in Korea
Turn stewardship policy into daily clinical work.
Korea’s third national AMR plan (2026–2030) calls for phased ASP expansion and institutionalization in 2028. ABX Steward supports the work; it does not imply government certification or compliance with evaluation criteria. KDCA · 2026 [7] ↗
Start with a defined scope
Evaluate the workflow using supported Excel or ZIP files, or REST/JSON APIs. Code and timing mappings, transmission adapters and local guideline validation require hospital specific work.
Keep clinical judgment central
AI provides rationale, recommendations and conditional drafts. Pharmacists and physicians check the records and evidence and decide whether to start, change or stop treatment.
Agree on data handling first
Draft generation may send pseudonymized clinical information to an external AI API. It remains sensitive data; confirm institutional rules for access, storage and transfer first.
From each claim to its source
Evidence you can inspect.
Dates, populations and limits accompany the figures. Original sources open in a new tab. Reviewed 27 October 2026.
- WHO · 2026Antimicrobial resistance: health and health-system burden ↗
Fact sheet dated 16 July 2026. Describes treatment limitations, hospital and healthcare costs, and the need for appropriate use.
- GBD 2021 AMR Collaborators · The Lancet · 2024Global burden of bacterial AMR, 1990–2021 ↗
Modelled 2021 estimates for 204 countries and territories: 1.14m attributable deaths (95% uncertainty interval 1.00–1.28m), 4.71m associated deaths. Attribution compares resistant infection with a counterfactual susceptible infection.
- WHO · Undated · accessed 27 Oct 2026Health workforce: projected shortfall by 2030 ↗
Projected shortfall of about 11m health workers, mainly in low and lower middle income countries. This is not an estimate of infection specialist or ASP staffing shortages.
- WHO · A79/5 Add.2 · 2026Draft global AMR action plan 2026–2036: financing gaps ↗
WHO submission dated 13 May 2026, page 3, paragraph 10: 10% of countries reported dedicated domestic funding in 2024. A national reporting indicator, not a share of hospitals without budgets.
- Davey et al. · Cochrane · 2017Interventions to improve antibiotic prescribing practices for hospital inpatients ↗
Policy adherence +15 percentage points (95% CI +14–16), 29 RCTs and 23,394 participants. Mostly high income settings with heterogeneous interventions and effects. Not a trial of this product.
- OECD · 2023Embracing a One Health Framework to Fight AMR: health costs ↗
Modelled annual average through 2050 for 34 countries: USD PPP 28.9bn compared with eliminating resistant infections. Replacing them with susceptible infections gives about USD PPP 5.9bn. Not converted into KRW or product savings.
- KDCA · 2026Korea’s third national AMR action plan (2026–2030) ↗
Announced 25 February 2026. Phased expansion of the ASP pilot launched in November 2024, targeting institutionalization in 2028. A policy goal, not government certification or endorsement of the service.
- Lee et al. · JAMA Network Open · 2026National Pilot Program for Antimicrobial Stewardship in the Republic of Korea ↗
Special Communication, 29 January 2026. Describes staffing and funding constraints and infectious disease physician workload as barriers to ASP expansion in Korea. A program design and implementation report, not a software efficacy trial.
Before you begin
Common questions
Is the goal simply to prescribe fewer antibiotics?
No. Stewardship aims to provide necessary treatment promptly while reducing unnecessary or inappropriate use. Infection status, culture and susceptibility, organ function and clinical course must be considered together.
How much staffing or cost can it save?
There is no measured product saving rate yet. Explore a scenario with your volume, time and costs, then measure it in a pilot. Released time is capacity to review more patients or complex cases, not a promise of staff reductions.
Do the research results transfer directly to this service?
No. The cited stewardship studies evaluated varied staffing, education and feedback interventions. ABX Steward’s time, cost and clinical effects require separate evaluation in the adopting hospital.
Does AI change prescriptions automatically?
No. It provides monitoring explanations, recommendations and conditional treatment drafts. The responsible clinical team makes prescribing decisions.
Which patient data can be used?
Supported information includes prescription history, blood tests, cultures and susceptibility, drug levels, clinical context and CT or X-ray reports. The service does not interpret images or DICOM scans. Hospital integration requires local code and timing mappings.
Can uploaded information be sent outside the service?
Data organization and visualization do not require an external model call. AI drafting may send permitted pseudonymized clinical information to an external AI API. Confirm institutional approval and processing settings before using real patient data.
ABX Steward
Make limited time count.
Start with stewardship.
Explore how connected orders, results and evidence could support your team’s reviews.
Create an account to review