Empirical Antimicrobial Selection

ICMR AMRSN, 2024 Report

These are the figures the agent choices in this tool are read against. They come from the eighth annual report of the Indian Council of Medical Research Antimicrobial Resistance Surveillance and Research Network, covering January to December 2024, published November 2025. Percentages are susceptibility unless the row says resistance.

Where the local antibiogram beats this table

National figures are an average across 21 regional centres. Your own hospital's antibiogram is better evidence for your own patient, and both the ICMR guidance and the carbapenem-resistant organism pathway say so in as many words: empirical regimens should be built on the organisms found in that patient in the previous six months, the antimicrobial exposure in the previous thirty days, and the local antibiogram. Where your unit's figures differ from this table, follow your unit.

Understanding Empirical Therapy

1. What an Empirical Regimen Is For

Empirical therapy is a bet placed under time pressure, on the organisms a syndrome usually carries, before anybody knows which organism this patient has. It is deliberately broader than the infection in front of you, because the cost of missing the organism is measured in deaths and the cost of being too broad is measured in resistance. Both costs are real.

• The bet is settled at 48 to 72 hours by the culture. A regimen that is never narrowed was not empirical therapy, it was a decision nobody came back to.
• ICMR puts a limit on it that is worth quoting: empirical antibiotics should be reserved for selected critically ill patients, naming febrile neutropenia, severe sepsis and septic shock, community-acquired pneumonia, ventilator-associated pneumonia and necrotising fasciitis. Most fever is not one of those.

2. Two Sources Behind Every Line, and Why They Are Different Documents

The ICMR Treatment Guidelines for Antimicrobial Use in Common Syndromes, third edition 2022, is the Indian authority on which agent to choose. It very largely does not print doses.

• Across the respiratory chapter, all three treatment tables name drugs with no dose, no route and no frequency, and the intra-abdominal chapter prints none at all. The central nervous system chapter and the carbapenem-resistant organism pathway are the exceptions, and both print doses higher than the general adult dose.
• So this tool separates the two claims and labels them separately in every result: Agent source and Dose source. A recommendation whose agent comes from ICMR and whose dose comes from a product label is a normal and honest thing. A recommendation that quietly attributes both to ICMR is not.

3. The Penicillin Allergy Label, and What It Costs

The three categories in this tool are not decorative. A delayed maculopapular rash and an anaphylaxis carry different risks, and they leave different drugs available.

• Mild, delayed rash. Cephalosporins and carbapenems are given, with observation. Cross-reactivity follows the side chain rather than the beta-lactam ring: ampicillin and cephalexin share one, which is why the old figure of ten per cent cross-reactivity is wrong and too high for most pairs.
• Severe or IgE mediated. Every penicillin and cephalosporin is avoided. Aztreonam, a monobactam, has no meaningful cross-reactivity with penicillin and becomes the Gram-negative backbone, with the caution that it covers no Gram-positives and no anaerobes, so something else must carry those.
• Aztreonam and ceftazidime share a side chain. This is the one genuine aztreonam cross-reactivity, and it matters in a patient whose reaction was to ceftazidime.
• The label itself is treatable. Around nine in ten people carrying a penicillin allergy label are not allergic when tested. Delabelling is a service the patient gets for life, and the admission where the label caused a worse antibiotic is the right time to arrange it.

4. Reading the Resistance Data Without Over-Reading It

The AMRSN figures come from tertiary care laboratories, and the report says plainly that they should not be extrapolated to the community.

• A young woman with uncomplicated cystitis gets nitrofurantoin, not a carbapenem, and the ICMR guideline says so.
• Resistance data justifies escalation in the sick and the hospitalised. It does not justify treating everybody as though they were in an intensive care unit, and the surveillance report is the document that says so.

5. De-escalation Is the Part That Gets Skipped

De-escalation is more than narrowing one drug. ICMR sets out six moves, and states that de-escalation is safe in every patient group including febrile neutropenia and septic shock, and reduces both mortality and length of stay.

1. Stop the polymyxin and the glycopeptide if no carbapenem-resistant organism and no MRSA grew.
2. Remove double Gram-negative or double anaerobic cover.
3. Stop antibiotics altogether if a non-infectious cause is found.
4. Collapse combination therapy to a single agent.
5. Narrow the spectrum.
6. Change intravenous to oral.

• Two ideas that sound like de-escalation and are not: continuing a broad agent because the patient improved on it, and stopping one drug out of three while leaving the other two unexamined.
• The shape of the Indian guidance matters here, and in the febrile neutropenia regimens in particular. ICMR's escalation ladder is essentially the de-escalation arm chosen as the default. That is a defensible response to Indian resistance rates, and it is broader than most international first-line advice. So de-escalation here is not an optional tidying-up step at the end. It is the step the whole strategy rests on.

6. Duration, Which Is Shorter Than Most of Us Were Taught

The ICMR durations are worth learning as a block, because they are the numbers most often exceeded.

Write the stop date in the notes when you write the first dose. A course with a planned end is the single cheapest stewardship intervention available on a ward.

7. Where Indian Guidance and International Guidance Disagree, and Which This Tool Follows

Four divergences are worth knowing because a resident will meet both positions in the same week.

• Fluoroquinolones in respiratory infection. Western pneumonia guidelines endorse them. ICMR says they are best avoided in India, because tuberculosis presents as pneumonia, a quinolone partially treats it, and the diagnosis is then delayed and the organism made harder to treat. The same argument bars linezolid and aminoglycosides from routine respiratory empirical use. This tool follows ICMR.
• Antibiotics in exacerbations of chronic obstructive pulmonary disease. The joint Indian Chest Society and National College of Chest Physicians recommendation is that antibiotics be prescribed for all exacerbations, which is broader than the international position that gates them on sputum purulence. Its strongest-graded statement is against routine fluoroquinolone use.
• Enteric fever. Fluoroquinolone resistance above 95 per cent has removed ciprofloxacin from empirical use in India, while cotrimoxazole, which most of us stopped using, is 97.7 per cent susceptible again.
• Malaria. Where the ICMR syndromes guideline and the national vector borne disease programme differ on the falciparum regimen, the national programme governs. This tool does not treat malaria as malaria, and it points to the national policy instead. It carries one artesunate line, inside the acute encephalitis syndrome pathway, because cerebral malaria is one of the four treatable causes that pathway covers at once until the tests exclude them. Once malaria is the diagnosis, the national programme is the document and this tool is not.

8. What This Tool Does Not Cover

• It is an adult tool. It does not carry paediatric or neonatal regimens, and paediatric dosing is not a scaled adult dose.
• It does not cover tuberculosis, HIV related opportunistic infection, viral hepatitis, or antifungal therapy beyond the point at which empirical antifungal cover is added.
• Malaria goes no further here than the empirical artesunate in the acute encephalitis syndrome pathway. Confirmed malaria follows the national vector borne disease programme.
• It does not replace a culture, and it does not replace an infectious diseases opinion in a patient failing therapy or growing a carbapenem-resistant organism.
• Where the patient falls outside what a syndrome covers, the tool says so on screen rather than answering anyway.