top of page

The use of AI in Biomedical Literature to Identify Antimicrobial Resistance

Updated: Nov 8, 2023

Antimicrobial Resistance: Case study from the scientific literature

Antibiotic resistance is ,an inevitable biological process but as humans we accelerate it by squandering it. For example: data from the Centres for Disease Control and Prevention (CDC) indicates that roughly 30 percent of antibiotics used in hospitals are unnecessary or prescribed incorrectly [1]

And that’s just in healthcare! Antibiotics are also used in agriculture, aquaculture, and for growing fruit (yes fruit) but to name a few. Each time we use an antibiotic, we give bacteria billions of chances to “crack the codes” of the defenses we have constructed.

However, antibiotic resistance is not something new, it was predicted more than 70 years ago by Sir Alexander Fleming (who discovered one of the world’s first antibiotics; penicillin). Fleming predicted not only how useful antibacterial drugs would be, but how dangerous a world without them could be. In an interview shortly after winning the Nobel Prize in 1945 for discovering penicillin, Fleming said: "The thoughtless person playing with penicillin treatment is morally responsible for the death of the man who succumbs to infection with the penicillin-resistant organism." [2]

Antibiotics have not only saved patients’ lives, but they have also played a pivotal role in achieving major advances in medicine and surgery. The World Health Organisation (WHO) has declared that antimicrobial resistance is one of the top 10 global public health threats facing humanity and that misuse and overuse of antimicrobials are the main drivers in the development of drug-resistant pathogens [3].

An estimated 2 million people annually in the U.S. develop infections that are resistant to antibiotics and in some cases, these infections result in death. It has been estimated that by 2050, 10 million deaths worldwide could result from antibiotic resistance, making it deadlier than cancer[4].

Literature Screening Automation

There is so much information available on antibiotic resistance; however, the question is what to do with it and if the right technology were to be deployed, could we extrapolate this information to predict future resistance trends? Recent advances in artificial intelligence applied to biomedical text are opening exciting opportunities for improving pharmacovigilance activities currently burdened by the ever-growing volumes of real-world data.

Machine learning has emerged as one potential avenue by which to address antibiotic resistance. When we apply machine learning to pharmacovigilance data (such as biomedical literature) it could constitute an important part of the wider multi-disciplinary approaches used for resistance surveillance and warning.

By integrating biologit MLM-AI's machine learning and workflow automation functionality, we could access a large repository of biomedical literature and deliver high quality scientific results in one place, reducing the screening workload and helping identifying trends, signals, and key information quicker than any manual process.

biologit Natural Language Processing (NLP) techniques identify relevant sections of text by highlighting important aspects of the text for example, the technology can be fine-tuned to identify a specific cohort of patients using tags such as “resistance”, “paediatric”, “pregnant”, or “elderly”.

The WHO`s actions - AWaRe

The WHO Essential Medicines List (EML) breaks down antimicrobials into three categories known as “Access”, “Watch”, and “Reserve” or (AWaRe), based on their indication, availability, and awareness [5]. A global campaign by the WHO was launched in 2019 to increase the proportion of global consumption of antibiotics in the Access group to at least 60%, and to reduce the use of the antibiotics most at risk of resistance from the Watch and Reserve groups.

By applying the AWaRe principles to a real-world data set, Jean Marie Vianney Habarugira in the Netherlands [6] showed that the WHO AWaRe principle can help classify spontaneous reports and help to identify trends in antimicrobial resistance. Habarugira reviewed the local Lareb database coded with antimicrobial resistance (AMR) relevant MedDRA Preferred Terms and found that 73% of AMR relevant reports were received over a period of 20 years.

The report showed evidence of an increase in AMR for several antibiotics including tobramycin, colistin aztreonam and doxycycline. Applying the same methodology to biomedical literature databases can possibly lead to information regarding how and why different Watch and Reserve antibiotics are potentially overused or used off-label, depending on the indications and available formulations.

A Review of the Literature

To assess current trends in AMR biologit conducted a review of the literature using biologit MLM-AI’s database. The team at biologit decided to focus on one area of interest which is the Carbapenem-resistant Enterobacteriaceae (CRE). CRE poses a significant public health concern and contribute to the growing problem of antibiotic resistance, the CDC has listed CRE as an “urgent global threat” [7]. These drugs are considered the last line of defense or drugs of last resort for infections.

The search was run over a one-year period between January 2021 to December 2021. An initial search for “carbapenem” and related synonyms (automatically populated by the MLM-AI Platform) was conducted across global databases available out-of-the-box with biologit MLM-AI: PubMed, Crossref and the Directory of Open Access Journals (DOAJ). The total literature hits returned was 6,493 articles.

With the help of AI predictions available on the platform, our team was able to significantly reduce the “background noise” and focus on articles of interest. Automatic article de-duplication eliminated 1,524 articles from the search straight away.

The remaining results were automatically tagged allowing for the rapid filtering of data i.e., aggregate report information, adverse events, and special situations.

CRE in the Pediatric Population

When the team searched for “resistance” data a total of 826 articles were found, constituting a 54% reduction in review time and cost. From this set, limited yet interesting information was present on the pediatric patient cohort.

CRE is not widely discussed in the general literature and there is limited information available on pediatric use: only 5 articles pertaining to the pediatric population were retrieved at the time of this review. We discuss those findings below:

Mzimela [8] conducted a retrospective study from 2015 to 2016 to evaluate the prevalence and clinical outcomes of neonates with CRE and found that CRE infections are on the increase in children and neonates and that neonates developed CRE much earlier than previously reported.

Seesahai [9] presented a case report of a carbapenem-producing E. coli-positive mother who delivered premature twins delivered at 24 weeks and the care practices initiated to contain CP-CRE including antibiotic treatment for the neonates, isolation, the use of person protective equipment (PPE) and in-depth cleaning of the used room(s). The author also highlighted that Public Health Ontario has been tracking the rising incidence of colonization and infection of CPE through a mandatory reporting system from May 2018.

In their November 2019 report, over a period of one year, 315 cases were reported.

In the article by Romandini [10] a study by the Worldwide Antibiotic Resistance and Prescribing in European Children (ARPEC) group obtained detailed information on antibiotic use in hospitalized neonates and children across 226 hospitals in 41 countries (of which 6499 inpatients received at least one antimicrobial). The study showed a high level of use of broad-spectrum antibiotics in certain regions, possibly explained by high incidence of carbapenem-resistant organisms. The study also found a striking regional difference in antibiotic prescribing was observed among hospitalized children (older than 28 days). A high proportion of African, Australian, Western European, and Northern European children continued to receive older narrow-spectrum antibiotics, mainly benzylpenicillin, sulfamethoxazole/trimethoprim, amoxicillin, and gentamicin. In Eastern and Southern Europe, Asia, North and Latin America, children received more broad-spectrum antibiotics, mainly third-generation cephalosporins and carbapenems.

Zejuan [11] presented the first published report of ceftazidime/avibactam (CAZ-AVI) for Carbapenem-resistant Klebsiella pneumoniae (CR-KP) osteomyelitis in a child. The 1.5-year-old male revealed osteomyelitis one month post operative for cardiopulmonary bypass. Blood culture showed a bacterial infection of K. pneumoniae. Drug susceptibility testing of the isolated CR-KP strain showed susceptibility to CAZ-AVI and a second surgery was conducted. Within 48 hours of administration there was no evidence of infection. CRE induced osteomyelitis has yet to be listed as an approved indication for the administration of CAZ-AVI and optimal treatment dosages and durations for osteomyelitis have yet to be investigated.

Yang [12] described an audit carried out in a maternity and child health care hospital in China in 2014. The results revealed that there was a sharp increase of carbapenem resistance rates in K. pneumoniae. Resistance rates of CRKp jumped swiftly from 0.6% in 2013 to 16.9% in 2014. It was found that inpatients in the neonatal wards had no corresponding clinical symptoms, albeit they had the highest rate of carbapenem-resistant Klebsiella pneumoniae (CRKp) pathogen isolation. This suggests that the CRKp in the neonatal patients are likely colonizing in the body, but not responsible for the infection. It appeared that prescription of carbapenem antibiotics was not well restricted, and there was a lack of robust justification for using carbapenem via collective discussion with experts during those years. Based on the analysis, the swift increase in the CRKp was a result of irregular use of imipenem in the previous years and the poor hospital environmental hygiene.


The urgency for new, safer, and more efficient drugs is greater than ever. The capability in applying AI algorithms to improve antibiotic resistance provides deeper insight into patterns, and emerging trends and expands pathways for antibiotic development.

This review emphasized the challenges in prescribing antibiotics in the pediatric population. The pharmacokinetics of many drugs are different in children compared to adults. The pharmacokinetic processes of absorption, distribution, metabolism and excretion undergo changes due to growth and development. Finding the correct doses for children is complicated by a lack of pharmacokinetic studies. Children’s doses cannot always be extrapolated directly from adult studies. Many pediatric doses are based on the child’s age or weight. These may need adjustment depending on the child and the clinical response. A simple infection in this patient population can prove to be fatal and therefore the correct course and dose of treatment are vital.

In the article by Zejuan, CRE induced osteomyelitis has yet to be listed as an approved indication for the administration of CAZ-AVI and optimal treatment dosages and durations for osteomyelitis have yet to be investigated. This demonstrates the lack of clinical trials and data for the pediatric population and a lot of the time prescribing of antibiotics is done “off label” as a result. Romandini found a regional difference in antibiotic treatment which only highlights the need for stricter control over antibiotic use as resistance is rapidly growing.

It is also worth mentioning the articles by Yang and Seesahai both showed a clear correlation between hospital environmental hygiene and the spread of antibiotic resistance between in-patients. The better the preventative measures i.e., isolation, use of PPE and cleaning, the better the outcome. However, there is still a lot to learn as Mzimela found that CRE infections are on the increase in children and neonates and that neonates developed CRE much earlier than previously reported.


This review highlights the need for more stringent control over the use of antimicrobial resistance and proves pharmacovigilance data can signal use-related issues which can be used to provide a bigger picture to prescribers who have a choice to make during each consultation.

Although there are some limitations to this review the findings were in line with the current emerging trends. By using the MLM-AI Platform the biologit team were able to significantly reduce the review time leading to faster identification of information. Implemented timely and properly, the biologit MLM-AI Platform can provide tangible business outcomes to an organization allowing for the rapid identification and review for emerging safety information.

About biologit MLM-AI

biologit MLM-AI is a complete literature screening platform built for pharmacovigilance teams. Its flexible workflow, unified scientific database, and unique AI productivity features deliver fast, inexpensive, and fully traceable results for any screening needs.

biologit MLM-AI



2. Sillankorva, S., Pereira, M.O. and Henriques, M. (2019). Editorial: Antibiotic Alternatives and Combinational Therapies for Bacterial Infections. Frontiers in Microbiology, 9. doi:10.3389/fmicb.2018.03359.

3. World Health Organization (2019). Antimicrobial resistance. [online] Available at:

4. Thermo Fisher Scientific. (n.d.). Procalcitonin (PCT) | Antibiotic Resistance. [online] Available at:

5. (n.d.). WHO releases the 2019 AWaRe Classification Antibiotics. [online] Available at:

6. Habarugira, J.M.V., Härmark, L. and Figueras, A. (2021). Pharmacovigilance Data as a Trigger to Identify Antimicrobial Resistance and Inappropriate Use of Antibiotics: A Study Using Reports from The Netherlands Pharmacovigilance Centre. Antibiotics, [online] 10(12), p.1512. doi:10.3390/antibiotics10121512.

7. CDC (2019). Antibiotic resistance threats in the United States, 2019. ANTIBIOTIC RESISTANCE THREATS IN THE UNITED STATES. [online] doi:10.15620/cdc:82532.

8. Mzimela, B.W., Nkwanyana, N.M. and Singh, R. (2021). Clinical outcome of neonates with Carbapenem-resistant Enterobacteriaceae infections at the King Edward VIII Hospital’s neonatal unit, Durban, South Africa. Southern African Journal of Infectious Diseases, [online] 36(1), p.223. doi:10.4102/sajid.v36i1.223.

9. Seesahai, J., Church, P.T., Asztalos, E., Eng-Chong, M., Arbus, J. and Banihani, R. (2021). Neonates with Maternal Colonization of Carbapenemase-Producing, Carbapenem-Resistant Enterobacteriaceae: A Mini-Review and a Suggested Guide for Preventing Neonatal Infection. Children, [online] 8(5), p.399. doi:10.3390/children8050399.

10. Romandini, A., Pani, A., Schenardi, P.A., Pattarino, G.A.C., De Giacomo, C. and Scaglione, F. (2021). Antibiotic Resistance in Pediatric Infections: Global Emerging Threats, Predicting the Near Future. Antibiotics, 10(4), p.393.

11. Ji, Z., Sun, K., Li, Z., Cheng, W. and Yang, J. (2021). Carbapenem-Resistant Klebsiella pneumoniae Osteomyelitis Treated with Ceftazidime-Avibactam in an Infant: A Case Report. Infection and Drug Resistance, Volume 14, pp.3109–3113. doi:10.2147/idr.s320056.

12. Yang, Y., Liu, J., Muhammad, M., Liu, H., Min, Z., Lu, J., Zhang, L. and Chai, Z. (2021). Factors behind the prevalence of carbapenem-resistant Klebsiella pneumoniae in pediatric wards. Medicine, [online] 100(36), p.e27186. doi:10.1097/MD.0000000000027186.


bottom of page