Concordance of AI-driven antimicrobial recommendations using rapid molecular versus phenotypic-optimized workflows for bloodstream infections: a prospective study in a resource-limited setting
This prospective study demonstrates that an integrated diagnostic-therapeutic program combining rapid molecular diagnostics with AI-driven initial guidance and subsequent phenotypic optimization significantly reduces time to appropriate antimicrobial therapy for bloodstream infections in resource-limited settings while maintaining high diagnostic and therapeutic concordance.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
When a patient develops a bloodstream infection, the clock starts ticking immediately. The bacteria are multiplying in the blood, and every hour that passes without the correct antibiotic increases the risk of death. For decades, the standard way to fight these infections has been a slow, deliberate process. Doctors take a blood sample, place it in a machine to see if anything grows, and then wait days for the bacteria to multiply enough to be identified and tested against different drugs. This waiting period, often lasting two to three days, forces doctors to guess which medicine to use first. If the guess is wrong, the patient suffers, and the bacteria may develop resistance to the drugs, making them harder to kill later.
To solve this, scientists have developed two powerful tools: rapid molecular tests and artificial intelligence. Molecular tests act like a high-speed scanner, looking for the unique genetic fingerprints of bacteria within hours rather than days. Artificial intelligence, in this context, is a computer system trained on vast amounts of medical data to act as a second opinion, suggesting the best drug based on the specific bacteria found and the patient's history. The big question for the medical world has been whether these fast tools can be trusted to guide treatment on their own, or if doctors must still wait for the slow, traditional tests to confirm the results. A new study from Peru explores whether combining these fast tools with the old, reliable ones creates a safer, faster path to recovery.
Researchers at a major hospital in Lima, Peru, set out to test a new, integrated approach for treating bloodstream infections. They followed 237 patients who had confirmed infections in their blood. For each patient, the laboratory ran two parallel tracks simultaneously. On the first track, they used rapid molecular tests to identify the bacteria and check for known resistance genes within a few hours of the blood sample turning positive. This information was immediately fed into an artificial intelligence system called OneChoice, which generated a specific antibiotic recommendation based on the genetic data and the patient's individual health factors. On the second, traditional track, the same blood samples were grown in culture dishes, a process that takes much longer. Once the bacteria grew large enough, they were identified using a different technology and tested against a wide range of antibiotics to see exactly which ones could kill them. This traditional data was then fed into a second, more advanced artificial intelligence system called OneChoice Fusion, which combined the new genetic results with the traditional lab results to create an optimized treatment plan.
The study found that the rapid molecular approach was dramatically faster. On average, the molecular tests provided results about 33 hours sooner than the traditional culture methods. This represents a reduction in waiting time of nearly two-thirds. More importantly, the researchers checked how often the fast, AI-driven recommendations matched the final, optimized recommendations. In more than 81 percent of the cases, the initial suggestion made by the fast system was exactly the same as the final suggestion made after the slow tests were complete. This means that for the vast majority of patients, the doctors could have started the correct treatment more than a day earlier without needing to change the plan later.
However, the study also revealed where the fast method has limits. In about 19 percent of the cases, the initial recommendation needed to be adjusted once the traditional test results arrived. The researchers discovered that the main reason for these adjustments was not a failure to identify the bacteria, but a failure to detect a specific type of resistance. The rapid molecular tests look for known genetic markers of drug resistance. If a bacterium has a resistance mechanism that is not yet mapped to a specific gene, or if the resistance comes from a complex chromosomal change rather than a simple gene, the fast test might miss it. In these specific instances, the traditional test was essential to catch the resistance and ensure the patient received the right medicine.
The findings suggest that the best strategy is not to choose between speed and accuracy, but to use both in a layered sequence. The rapid molecular test, guided by artificial intelligence, can get patients on the right treatment path almost immediately, saving critical time. Then, the traditional test acts as a safety net, arriving later to confirm the plan or make necessary fine-tuning. In this study, this combined approach meant that 100 percent of patients eventually received the perfect, personalized therapy, but the majority of them received appropriate care 33 hours earlier than would have been possible with traditional methods alone.
This work is particularly significant in settings where drug-resistant bacteria are common, such as in parts of Peru where resistance to standard antibiotics is high. In these environments, guessing the wrong antibiotic can be fatal. By using the fast test to narrow down the options quickly and the slow test to verify the details, hospitals can reduce the time patients spend on the wrong medication. The study did not measure whether this faster approach directly saved lives, as that would require a different type of long-term trial, but the data strongly suggests that reducing the time to the correct antibiotic is a vital step toward improving survival rates. The research confirms that while rapid tests are incredibly powerful, they work best when paired with the thoroughness of traditional methods, creating a system that is both swift and precise.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.