The EquityAMR project is a 4-year collaborative study (2021–2025) between Norway and India focused on the analysis of health equities and AMR in India. The project obtained permissions from the Norwegian Centre for Research Data (NSD) and the Health Ministry Screening Committee (HMSC), Government of India, before data collection. For enabling data collection to ensure informed consent, anonymization, and adherence to required guidelines, formal ethical approval was obtained from the HISP India Research Ethics Committee. Memorandum of Understanding and Non-Disclosure Agreements were signed between HISP India and the state hospitals to ensure permissions for data collection and the security and integrity of data.

This study employed a cross-sectional observational design as part of the larger Equity AMR initiative, a collaboration between Norwegian and Indian researchers inspired by systems thinking to analyze the interconnected relationships between surveillance, diagnostics, and prescriptions. The study was conducted at a community hospital (CH), a secondary care public health facility in North India. Data were collected from three primary sources: (i) antimicrobial procurement records from the community hospital between 2018 and 2022; (ii) antimicrobial susceptibility testing (AST) results of clinical isolates from urine, pus exudate, and stool samples collected from outpatients who underwent microbiology culture tests between January 2023 and December 2024; and (iii) antimicrobial prescription data (2023) extracted from patient prescription slips following their OPD encounters at the hospital. Each dataset was analyzed both individually and in combination to provide a holistic understanding of hospital-level drivers of AMR and to draw policy-relevant implications.

We collected data regarding inflow patterns of antimicrobials in the hospital, including from procurement to distribution within the facility. The data captured included the name of the healthcare community hospital, ordering time, generic names, dosage forms, strength, quantity purchased, expense, delivery time, etc. Data was then systematically collected from the manual stock ledgers in the stores.

For analysis, we compared the antimicrobials procured in terms of their compliance with the government-mandated Essential Drug List. We converted the compiled consumption data into defined daily doses (DDD) per 1,000 inhabitants per day, representing WHO’s standard measure (Misra et al., 2023). Based on 2011 census figures, we estimated the resident population for 2024. DDD per 1,000 inhabitants per day was used as a standardized metric, though in this hospital setting with a wider and partly undefined catchment, the values should be interpreted as approximate rather than precise. The DDD was calculated using the following formula:

For prescription data, randomly selected patients accessing antimicrobials from the hospital pharmacy after an OPD encounter were approached. A total of 1,151 prescription slips were captured. Simple random sampling was employed, selecting slips that included one or more antimicrobials or names of infections that would potentially require future antimicrobial treatment. Random sampling was carried out to ensure that the selected prescription slips accurately reflected the entire group of patients using antimicrobials, thus decreasing biased selection. This methodology provides impartial and broadly applicable results about prescribing antimicrobial trends. Approximately 100 patients visited the pharmacy daily, from which the researcher, a trained microbiologist, randomly selected ∼10 patients, and consent was taken to examine their slips and take an image of them using a camera phone while covering patient names. The data collected included demographics, prescription details (symptoms, diagnosis, drug names, dosage, duration), and legibility of prescriptions. Data was recorded in Excel and then transferred into a bespoke database for analysis and interpretation. All three datasets, procurement, prescriptions, and AST results, were obtained from the same community hospital, with records cross-verified against pharmacy and laboratory registers to ensure consistency and reliability.

The microbiology laboratory received 1,547 urine samples, 248 pus exudate samples, and 144 stool samples from patients attending the outpatient department (OPD) in the hospital. The samples were sent in for culture tests during the research period. However, blood samples were not tested as there is no facility for blood culture in this facility.

All clinical samples were processed in the Microbiology laboratory using conventional culture techniques. The urine samples were inoculated on Cysteine Lactose Electrolyte Deficient (CLED) agar medium. Inoculated agar plates were incubated initially aerobically at 37 °C for 24 h and finally for 48 h, and the plates were examined for pure growth. A growth of ≥ 10⁵ colony-forming units/ml was considered significant. Cultures with more than two colonies were considered contaminants, and such samples were discarded. Pus samples were processed for Gram staining and culturing. The samples were aseptically inoculated on blood agar (with 5% sheep blood) and MacConkey agar plates, incubated aerobically at 35 °C–37 °C for 24–48 h (Modgil et al., 2020). Stool samples were collected in sterile containers. They were cultured for the presence of bacterial species by standard procedures (Sharma et al., 2022). Briefly, samples were inoculated onto MacConkey agar, ampicillin blood agar, xylose lysine deoxycholate agar (XLD agar), thiosulfate-citrate-bile salts-sucrose agar (TCBS agar), alkaline peptone water (APW), and selenite F broth and incubated at 37 °C for 18–24 h. Pathogens were identified based on colony morphology, Gram staining, and standard biochemical tests such as indole, citrate, urease, triple sugar iron agar, and oxidase tests (Modgil et al., 2021).

Antibiotic susceptibility testing was performed with the disk diffusion method for the following antimicrobials: Amoxicillin + clavulanic acid, doxycycline, azithromycin, ampicillin, ciprofloxacin, amikacin, imipenem, levofloxacin, gentamicin, cefepime, piperacillin-tazobactam, ertapenem, cotrimoxazole, cefoxitin, ceftriaxone, etc., according to Clinical & Laboratory Standards Institute (CLSI) guidelines (Modgil et al., 2025). Quality control was performed using standard reference strains, including Escherichia. coli ATCC 25922 and Staphylococcus aureus ATCC 25923, to validate the potency of antibiotic disks, in line with CLSI guidelines.

Correlation analysis was conducted to explore potential associations between three variables: (i) antimicrobial procurement volumes (in DDDs), (ii) prescription frequencies of antimicrobials (percentage of total prescriptions), and (iii) AST-derived sensitivity percentages for selected organisms. Scatter plots were generated to visualize whether higher procurement or prescription corresponded with variations in sensitivity. These analyses were exploratory in nature and aimed to identify indicative trends and mismatches between antimicrobial use and resistance, rather than to infer causality.

We compiled test results from a total of 1,547 urine samples, 248 pus exudate samples, and 144 stool samples processed in 2023. E. coli was the most prevalent organism isolated from urine (47.1%) and stool (69.7%) samples, while Staphylococcus spp. was the most isolated (46.8%) organism from pus exudate samples. The prevalence of S. aureus in pus exudate samples and the presence of Salmonella spp. in stool samples was 55 (29.3%) and 24 (16.6%), respectively. “Other isolates” referred to organisms detected in smaller numbers, such as Proteus spp., Enterobacter, Enterococcus spp., and Candida spp. The sample types tested, and their isolation rates are presented in Figure 3 below.

Table 2 presents the data between the top procured antimicrobials in the year 2022 in DDD/gm, the most prescribed antimicrobials in OPDs, and sensitivity results in 2023–24. Notably, Ofloxacin + ornidazole was the most procured antimicrobial, followed by Azithromycin and Doxycycline. Among the top prescribed antimicrobials in OPD in 2023, Amoxycillin + Clavulanic Acid had the highest prescription rate (24%), followed by Cefixime (15%) and Azithromycin (11%). Nitrofurantoin exhibited high sensitivity against E. coli (82.26%) and Klebsiella (65.7%) in urine samples, while Ciprofloxacin showed relatively low sensitivity (6.72% and 27.2%) against the same bacteria.

Correlation analysis was performed to assess associations between antimicrobial procurement volumes, prescription frequencies, and AST-derived sensitivity percentages. Scatter plots were used to visualize whether higher use corresponded with changes in susceptibility, highlighting mismatches such as the frequent prescribing of Amoxicillin–clavulanic acid alongside high resistance in E. coli. These findings are exploratory and should be interpreted as indicative trends rather than causal relationships, underscoring the need for richer longitudinal datasets. It is worth noting that, given the limited laboratory resources and incomplete testing across all antimicrobials and organisms, the available AST data were restricted. Therefore, there is a need for more tests and continuous data collection over longer time frames. As can be seen in Table 3, the most tests were conducted for E. coli in urine, across 9 (all except for Amoxicillin, Metronidazole, and Cefixime-sulbactam) of the 12 procured antimicrobials. Pseudomonas spp. in urine was tested across 4 (Nitrofurantoin, Ciprofloxacin, Cefixime, Norfloxacin), Klebsiella in urine was tested across 7 (Doxycycline, Amoxyclav, Amoxicillin, Metronidazole, Nitrofurantoin, Ciprofloxacin, Cefixime, Norfloxacin, Ceftriaxone, Cefuroxime), while Staphylococcus in exudate was tested in 5 (Azithromycin, Doxycycline, Amoxyclav, Amoxicillin, Ciprofloxacin). This results in data sparsity and limits the extent to which the data can be statistically analyzed.

The primary analysis conducted was correlation analysis, over two conditions, to explore the nature of links between antimicrobial sensitivity and prescription:

Figure 6 presents a scatterplot of the prescribed antimicrobials (percentage) and sensitivity of all antimicrobials, with the 2023–24 data and split between 2023 and 2024.

The Table 3 presents the relationship between procured antimicrobials, prescriptions and sensitivity, discarding isolate and antimicrobial information. It is to be noted that the prescription data relates to 2023. As can be observed, we note low (positive) correlation between 2023 prescriptions and 2023–24 sensitivity. Despite this, we observe a weak negative correlation between 2023 prescriptions and 2023 sensitivity.

To explore this further, we conduct correlation analysis between prescriptions and sensitivity, grouped by microbial sensitivities. We conduct this analysis on the sensitivity data for 2023–24 and, in Figure 7, we present a scatterplot of sensitivity vs. prescription, grouped by the microbe and isolates.

Table 4 presents correlation analysis of the antimicrobial sensitivity, procurement and prescriptions. The correlation values between pairs of microbial sensitivities are ignored in our interpretations as they do not correspond to the same populations, and therefore is not presented in the table.

There are several key insights we derive out of the correlation analysis: inspecting the combined data, we observe a weak negative correlation (−0.069) between prescriptions and E. coli (urine) – as with the previous analysis, we note that the 2023 E. coli sensitivity (−0.13) is higher than 2024 (−0.025). In inspecting specific sensitivities, Staphylococcus spp. was determined to be weakly positively correlated (0.139) with prescriptions. The same was determined to be perfect correlation (1.0) with prescription for 2023. However, this is likely due to the sparsity of the data and only 2 of the 2 procured antimicrobials and therefore discard this as an observation. With more tests done in 2024, we observe a slightly positive correlation between Staphylococcus spp. and prescriptions. The number of tests for Pseudomonas spp. and klebsiella (urine) during 2024 were insufficient and combining 2023-24 data, both were determined as weakly negatively correlated with prescriptions, with −0.04 and −0.101 respectively. The negative correlations indicate that with an increase in the value of prescriptions, antimicrobial sensitivity reduces. However, given the limited and sparse nature of the data, longer term data collection is necessary to establish more generalizable findings.

The development of ASPs in Indian institutions encounters several challenges, such as inadequate facilities, financial constraints, irregular prescription processes, and a lack of expertise and education among healthcare providers. The data gap is also a significant barrier because ASPs rely on accurate and up-to-date information to identify trends in AMR, monitor prescription practices, and optimize antibiotic use. While observational, our analysis directly points to actionable stewardship steps, aligning procurement with resistance, updating formularies, and integrating AST data into prescribing practice. A study carried out at a tertiary care hospital in India evaluated the impact of specific ASP strategies and infection control practices in ICUs on decreasing overall antimicrobial usage and infection rates through targeted interventions (García-Rey et al., 2006). Our study demonstrates several implications for the development of ASP strategies. First, there is a need for regular review and adjustment of the hospital antimicrobial formulary to ensure it aligns with current resistance patterns and stewardship goals. Second, integrating procurement, prescription, and sensitivity testing data can enhance the effectiveness of ASPs by providing a more comprehensive understanding of antimicrobial use and resistance trends. Third, addressing the economic and cost-effectiveness aspects of antimicrobial selection by monitoring and modifying procurement decisions. By using resistance surveillance data in ASPs, hospitals can develop evidence-based prescribing guidelines and combat the spread of resistant organisms, ultimately enhancing patient safety. Furthermore, in settings with high background resistance, promoting a premature shift toward narrower-spectrum “Access” drugs may risk treatment failure. A feasible first step is to strengthen diagnostic and AST capacity to generate reliable local antibiograms guiding safe de-escalation. Integrating this data into formulary and procurement decisions, introducing phased de-escalation in select departments, and maintaining clinician feedback loops can gradually reduce dependence on broad-spectrum “Watch” drugs while minimizing clinical risk.

The findings from the two phases were summarized under three main categories: (1) Hospital procurement and AST patterns, (2) Hospital prescription practices and AST data, and (3) Development of ASPs through integrated data analysis.

To the best of our knowledge, this study is among the first from India to integrate antimicrobial procurement, prescription, and resistance data within a single institutional setting. Although constrained by its single-center design, outpatient focus, data sparsity in AST, and temporal disconnect between procurement and prescription/resistance datasets, it nevertheless demonstrates the feasibility of leveraging routine institutional data to identify mismatches and inform stewardship strategies. The absence of bloodstream isolate data highlights the need for strengthened laboratory capacity to capture the full resistance burden. The findings underscore the importance of aligning procurement with local resistance patterns, regularly updating formularies, and incorporating AST evidence into empirical prescribing. Future work should adopt standardized AST panels and include less-used drugs to reduce surveillance bias. Overall, the study highlights the critical need for synchronized, data-driven antimicrobial stewardship Programmes in which procurement, prescribing practices, and surveillance are continuously integrated to mitigate the escalating threat of AMR.