An early resistance warning helped doctors change a woman’s antibiotic

智抗菌平台, also called the Intelligent Anti-Microbial System or i.A.M.S. [1]

The sources identify a machine-learning model for resistance prediction but do not establish its exact architecture. [1] [4]

The hospital account does not report a model release. [1] [4]

China Medical University Hospital, Taichung, Taiwan. [1]

The resistance-prediction component analyzed the bacterial spectrum. The broader platform supported antibiotic selection. [1]

The exact instrument model, software settings, and prediction threshold are not reported in the accessible patient account. [1] [4]

A routine MALDI-TOF mass spectrum informed a prediction of ceftazidime resistance and a suggestion to consider levofloxacin. [1]

The system supplied a prediction and treatment suggestion. Clinicians chose and administered the antibiotic. [1]

The laboratory cultured the organism and obtained its mass spectrum. The system predicted resistance, and the treating team changed the antibiotic. [1]

Laboratory staff prepared and identified the isolate. Physicians interpreted the warning and made the treatment decision. [1]

The platform analyzed a bacterial mass spectrum to predict resistance and suggest an alternative antibiotic. Clinicians selected and administered treatment. The hospital attributes earlier selection to the system, but the available account cannot isolate its effect on survival.

The reported contribution was an earlier resistance warning and antibiotic suggestion. The antibiotic and clinical care treated the infection. [1]

AIBD-0114-outcome-01 [1]

The hospital reported that an early resistance warning helped clinicians change the antibiotic before the woman recovered. [1]

Improved Physical Health

Recovery was reported by the treating hospital. The platform’s five-year mortality forecasts were intended benefits. [1]

People.

One 78-year-old woman with severe pneumonia, diabetes, hypertension, and chronic kidney disease. [1]

One reported recovery. The hospital’s figure of more than 100,000 platform uses is not a count of distinct beneficiaries. [1]

The reported clinical result concerns one patient at one hospital. Platform-wide usage does not count confirmed individual benefits. [1]

The woman went home in late July 2022. The sources do not report her subsequent health. [1]

Yes

The account does not report false predictions, adverse effects of the antibiotic change, or the full costs of the system. [1]

CMUH clinical account, NewsArticle 8151 [1]

A single-patient clinical account without a control. The hospital also reported separate, aggregate observations about its platform. [1]

The hospital described the laboratory warning, antibiotic change, improvement, extubation, and discharge. Public reports do not include the full medical record. [1]

One patient at China Medical University Hospital. The hospital separately reported more than 100,000 uses of its platform. [1]

Time to a resistance prediction and the patient’s subsequent clinical recovery. [1]

The hospital’s stated turnaround for conventional susceptibility testing. There was no controlled comparison of this patient’s recovery. [1]

The resistance prediction was available within an hour. Clinicians changed the antibiotic, and the patient later recovered and was discharged. [1]

The evidence supports the system’s role in selecting treatment earlier. It does not establish what would have happened without that warning. [1]

The clinical evidence comes from the treating hospital. News reports repeat its account. The original resistance-model letter could be identified but not read in full. [1]

The admission began in April 2022, pneumonia developed in late May, and discharge occurred in late July. The exact date of the AI warning is not established. [1]

The hospital reproduces a figure from its resistance-model letter. The reviewed public materials do not provide model code, weights, or patient-level data. [1]

Reported reductions in drug spending across the hospital’s platform. [3] [5]

The hospital reports 30–40% lower inpatient drug spending. A news report quotes annual savings of “10 million,” without stating a currency or calculation. [3] [5]

The annual figure does not state its currency or price year. The hospital report gives only percentages. [3] [5]

Editor’s note. New Taiwan dollars are a plausible reading because the report concerns a Taiwanese hospital’s domestic drug spending. The cited sentence does not name the currency, so no conversion has been applied.

The hospital compared the period after rollout in 2022 with the corresponding period in 2021. A separate news account gives an annual figure. [3] [5]

The sustainability report presents the spending reduction as observed. The annual figure’s calculation and period of observation are unclear. [3] [5]

The hospital describes lower inpatient drug spending. It does not allocate the saving between the insurer, hospital, and patients. [3] [5]

The figures concern drug spending. They do not subtract development, equipment, staffing, or operating costs. [3] [5]

Implementation and operating costs are not reported alongside the savings. [3] [5]

The hospital report uses the corresponding period of 2021. It does not describe adjustment for changes in patients or treatment. [3] [5]

The report states a percentage reduction without publishing the underlying spending amounts, denominators, or calculation. [3] [5]

The sources do not calculate net return or a payback period. [3] [5]

China Medical University Hospital developed and deployed the platform. Its laboratory and treating clinicians cared for the patient. [1]

Taichung, Taiwan, where the platform was developed and the patient was treated. [1]

The reviewed clinical sources identify the organizations involved, but do not establish their complete ownership or control arrangements. [1]

The service depended on a microbiology laboratory, MALDI-TOF mass spectrometry, susceptibility records, clinical staff, and an available alternative antibiotic. [1]

The reviewed sources do not document international transfers of patient data or quantify cross-border financial flows. [1]

Not assessed

The laboratory needed a suitable isolate and mass spectrum, and clinicians needed a timely warning and an effective alternative antibiotic. [1]

The reviewed sources do not report an independent repetition of this clinical result. [1]

The hospital’s 2022 report describes lower inpatient drug spending across the platform. It provides no later follow-up on this woman. [5]

The platform was in clinical use at the hospital in 2022. The patient account does not establish public or commercial access to that version. [1]

China Medical University Hospital sought faster resistance predictions and earlier selection of antibiotics. [1] [5]

Hospital leaders described delayed susceptibility results and the consequences of inappropriate treatment. [1] [5]

The hospital introduced the platform in 2022 and reported more than 100,000 uses by October. Its publications differ on whether rollout began in January or February. [1] [5]

The hospital described faster laboratory results and reduced drug spending. Its five-year mortality and medication-error figures were forecasts. [1] [5]

5 retained sources.

Original citations and the passages that support each finding. [1] [2] [3] [4] [5]

The hospital account, report, and news articles are in Traditional Chinese. The model letter’s metadata is in English. Summaries were translated from the Chinese sources. [1] [2] [3] [4] [5]

The hospital is the original clinical source. CNA and Liberty Times add reported details from its clinicians, without independent clinical verification. Their evidence relationship is therefore the deployer’s testimony. [1] [2] [3]

The original studies and their supporting materials are linked to the findings that they support. [1] [2] [3] [4] [5]

Access and version details are listed for each reviewed source. [1] [2] [3] [4] [5]

The platform predicted antibiotic resistance before clinicians changed treatment, after which the patient recovered. [1]

The hospital supports the clinical account. Its report provides separate figures about the wider platform. [1] [2] [4] [5]

A firsthand hospital account, supplemented by reporting from its press conference. The clinical outcome has not been independently verified in the reviewed materials. [1]

The exact deployed model, release, configuration, and patient-level treatment record remain unavailable. The source does not quantify this woman’s financial benefit. [1]

The benefit described here, for the beneficiary groups shown.

Depth 4 for the reported resolution of a serious condition.

Depth 4 for the reported resolution of a serious condition.

The depth describes the reported result. Evidence strength is assessed separately.

AIBD’s five-level depth rubric, reviewed on 25 October 2026.

AIBD, 25 October 2026.

Depth 5 would imply stronger evidence that the intervention preserved a life. Platform-wide forecasts do not increase this patient’s grade.

The original clinical evidence comes from the treating hospital, including its testimony reported by news outlets. The outline is appropriate.

A treating hospital’s account of one patient, with separate information about the wider platform. [1]

Not assessed

Healthcare - clinical microbiology and antibiotic therapy. Hospital laboratory AI (antibiotic-resistance prediction from MALDI-TOF spectra).

Help clinicians select an effective antibiotic sooner by predicting resistance from routine mass spectra.

The woman went home in late July 2022. The hospital presented the case on 26 October and published its account on 31 October. [1] [2] [5]

Clinical use at China Medical University Hospital in 2022. [1]

SHA-256: 60e0410e9896…

oecd_aim: 2022-10-25-1879.

Not assessed

Not assessed

6 revisions. Latest: 25 October 2026.