v3 · icmr-amrsn + ncdc nars-net trend tracker

National AMR trends, read by hand-checkable rule.

India runs two national AMR surveillance networks — ICMR's AMRSN and NCDC's NARS-Net — and both publish their results only as annual report PDFs. No API, no bulk download. This project turns the published national trend tables from both into structured data where every number carries the report edition and table it came from, so it can be traced back and checked against the source PDF. The two are carried as parallel series and never pooled.

Yashita Thakur · ORCID 0009-0004-7895-5250 · MIT licensed code

Independent, unofficial analysis. Not endorsed by or affiliated with ICMR or NCDC. Where this dataset and a published report differ, the published report is authoritative and the difference is a limitation of this extraction.
1,286AMRSN national rows
1,365AMRSN RC rows
558NARS-Net rows
2017–24years covered
3AMRSN editions
8NARS-Net editions
418tests, pytest
17 / 432genuine revisions found
scope

What this is — and isn't

This is

  • National trend data from both networks' own printed tables, 2017–2024: six organisms across three chapters from AMRSN's three editions, and E. coli and S. aureus from all eight NARS-Net editions.
  • Fully provenance-annotated: every value cites its network, report edition and table number.
  • Cross-checked three ways — internal arithmetic, narrative cross-check against report prose, and structural assertions.

This is not

  • Community prevalence — these are tertiary-care lab isolates, a hospital-skewed population.
  • Patient-level data, or a national burden/incidence estimate.
  • A pooled cross-network figure — the two are carried in separate files and never combined. Why, in full, below.
charts

The trends

E. coli national susceptibility trend, 2017–2024, ten antibiotics
Carbapenems (imipenem, meropenem, ertapenem) drifting down alongside everything else — none of the ten drugs in this panel is trending upward.
Klebsiella pneumoniae national susceptibility trend, 2017–2024, ten antibiotics
K. pneumoniae, same panel. Points with fewer than 30 isolates tested, or where the source does not print a percentage, are omitted rather than estimated.
Acinetobacter baumannii national susceptibility trend, 2017-2024
A. baumannii. Almost everything sits below 25%; meropenem reaches 9.0% by 2024, the 91.0% resistance quoted in the report's own summary. Minocycline is the outlier. Colistin is not plotted — the source reports it as intermediate susceptibility, not susceptibility.
Pseudomonas aeruginosa national susceptibility trend, 2017-2024
P. aeruginosa, an 11-drug panel that differs from A. baumannii's: gentamicin, tobramycin and ciprofloxacin in place of minocycline.
Staphylococcus aureus national susceptibility trend, 2017-2024
S. aureus, an entirely different drug-class mix. Vancomycin, teicoplanin and linezolid hold near 100%; cefoxitin and erythromycin fall away.
MRSA national susceptibility trend, 2017-2024
MRSA isolates only — how susceptible they are to each drug, not how common MRSA is. Cefoxitin sits at 0%, which is definitional: MRSA is methicillin/cefoxitin-resistant, so any other value would mean the wrong table had been read.
Chart showing a count revision between report editions for E. coli piperacillin-tazobactam 2022
Where editions report the same year differently. 17 genuine revisions across 432 comparable (organism, antibiotic, year) combinations. Shown here are the count revisions — changes to the numerator or denominator itself, which a percentage-only comparison would miss entirely. Those 432 comparisons exist only because every AMRSN edition reprints the whole series back to 2017 rather than just its own year — more to publish each time, and what makes a revision like this findable at all.
regional centres

By Regional Centre

The RC-wise tables are a single-year snapshot with no year axis, so cross-edition revision detection at RC level is empty by design, not an unfinished feature.

E. coli susceptibility by Regional Centre, 2024 edition, nine antibiotics
E. coli, 2024 edition, nine drugs (no cefazolin). RC15 was dropped from the 2024 table and breaks every line; RC1 and RC18 appear but are tested against fewer than 30 isolates each, so no point is drawn for them.
Klebsiella pneumoniae susceptibility by Regional Centre, 2024 edition, nine antibiotics
K. pneumoniae, same 2024 panel and the same RC15 break. Susceptibility swings hard between centres — meropenem alone runs from 1.4% to 65.8%.
Staphylococcus aureus susceptibility by Regional Centre, 2024 edition, eleven antibiotics
S. aureus, 2024 edition, eleven drugs. Three centres — RC1, RC15 and RC18 — are absent from the 2024 table and break their lines. Vancomycin, teicoplanin and tigecycline hold a flat 100% wherever they are reported; ciprofloxacin sits near 15%.
cross-network

The second network

India has a second national AMR surveillance network: NCDC's NARS-Net, which feeds WHO GLASS and draws on a different set of hospitals — state government medical colleges, where AMRSN is dominated by central institutes, private and mission hospitals. No institution appears on both rosters. All eight NARS-Net editions, 2017–2024, are extracted here for the two organisms both networks report at species level: E. coli and S. aureus, 558 rows. NARS-Net prints every drug separately by specimen type rather than pooling them into one figure — more work to produce than a single table, and the only reason a specimen-matched comparison between the two networks is possible at all.

These two networks share no comparison value. AMRSN publishes % susceptible; NARS-Net publishes % resistant. Converting one to the other needs the intermediate fraction, and AMRSN publishes no % intermediate for either organism — so an AMRSN % resistant cannot be computed at all. The two are shown here as parallel series and are never joined on one value or one axis, not even with an inverted axis: a reader who misses the inversion gets the conclusion backwards. Every figure below is one network alone, with a single exception, noted where it appears.
E. coli NARS-Net percent resistant, blood and urine isolates, 2017-2024, seven antibiotics
E. coli, NARS-Net. High is bad here — this is % resistant, the opposite reading from every chart above. Blood and urine are the only specimen columns NARS-Net prints in all eight editions; the rest are either composites whose membership changes between editions or columns that begin in 2021. Only the drugs both networks report are drawn, and ceftazidime is left out because NARS-Net prints it in one edition only. Amikacin, meropenem and piperacillin-tazobactam start mid-chart because the panel widens from nine drugs to seventeen in 2021.
S. aureus NARS-Net percent resistant, blood isolates, 2017-2024, eight antibiotics
S. aureus, blood only — NARS-Net never surveils it from urine, and its pus-aspirate column exists only from 2021. Teicoplanin joins the panel in 2021; tetracycline runs for 2017 and 2018 alone. Linezolid sits near zero throughout. Vancomycin is not drawn: it is a printed row in the 2018 edition only, against 14 isolates, and the 2019 and 2020 editions explain the omission — no site reported vancomycin MICs by broth microdilution. That is a method-availability constraint, not a gap in interest.

Which network reports what

The panels overlap without matching. These two grids answer one question per cell — which network reports this organism, drug and year, on which metric, from which specimen basis — and nothing else. The companion file comparability_matrix.json carries no percentage and no count from either network, enforced at build time rather than intended, because a matrix carrying values would be exactly the join that must not exist.

Grid of 21 E. coli antibiotics by 8 years showing which network reports each cell, with a specimen-basis strip beneath
E. coli: 21 drugs × 8 years = 168 cells — 45 reported by both networks, 56 by NARS-Net only, 35 by AMRSN only, 32 by neither. Panel-level overlap is not cell-level overlap. Eight drugs are reported by both networks somewhere in the series, but only four in every year. Ceftazidime is the sharp case: it is in the AMRSN panel throughout and in a NARS-Net table in 2017 alone, so it counts as a shared drug on one of its eight cells. The strip beneath shows the 2021 break, where the pooled and PA+OSBF columns end and pus aspirate and OSBF begin.
Grid of 13 S. aureus antibiotics by 8 years showing which network reports each cell, with a specimen-basis strip beneath
S. aureus: 13 × 8 = 104 cells — 55 both, 16 NARS-Net only, 33 AMRSN only, none unreported by both. Vancomycin is this organism's sharp case, shared on one cell of eight (2018). Oxacillin is AMRSN-only throughout: it never appears as a row in any NARS-Net edition, which uses cefoxitin as the sole MRSA surrogate. The specimen bases are not equivalent either, and not in the same way for the two organisms — AMRSN's pooled column is captioned "all samples (except faeces and urine)" for E. coli but "all samples", with no exclusion stated, for S. aureus.

The one axis they may share

Isolates tested per year for both networks, E. coli and S. aureus, on a log count axis
Counts, not percentages — which is why this is the single figure here that puts both networks on one axis. An isolate tested is the same unit on both sides in a way the two resistance metrics are not. Each line is the largest printed denominator in that year's panel: neither network publishes "isolates tested" for an organism, only isolates tested against each drug, and those differ widely inside one year. The trajectories diverge — NARS-Net grows steadily while AMRSN's E. coli denominators fall sharply in 2023, from 14,728 tested in 2022 to 7,559, recovering to 11,679 in 2024. That dip is in the source and is not a revision: the 2023 and 2024 editions print every 2023 cell identically. What is comparable here is the shape over time, not the size — the two cover different specimen populations.
method

Three ways this PDF fights back

Regex over pdftotext output silently produces plausible-looking wrong numbers on these reports. Extraction goes through pdfplumber's table detection instead, and still had to defend against three specific failure modes found in the actual PDFs:

found in the 2023 edition

Column indices lie

The header row carries an extra leading cell, so Year-2017 sits at index 4 while its data sits at index 3. Fix: cells are matched to years by x-coordinate overlap, never by column position.

found in the 2022 edition

Digits get clipped

A ruled box narrower than the digits inside it makes 5170 / 14729 come back as 170 / 1472 — off by an order of magnitude, while the percentage still looks fine. Fix: cell contents are assembled from whole words by centre point, so no digit can be cut off mid-cell.

found across editions

Wrapped labels create phantom rows

"Piperacillin-/tazobactam" wrapping to two lines makes the table library report a nested extra row, which then overwrites the next antibiotic's percentage. Fix: row bands wholly contained inside another band are discarded.

Full writeup, including how table locations are found across editions that renumber their own chapters, is in README.md.

Reproduce any number by hand

  1. Read source_url and source_report_year in the row — that's the report PDF.
  2. Read source_table — e.g. Table 2.6.
  3. Open that PDF, find the caption matching that table.
  4. Find the antibiotic row and the year column.
  5. The printed cell should match susceptible_n, tested_n, and susceptible_pct exactly.

Row schema

{
  "organism": "Klebsiella pneumoniae",
  "antibiotic": "meropenem",
  "year": 2024,
  "susceptible_n": 4276, "tested_n": 12189,
  "susceptible_pct": 35.1, // as printed — null where the source does not print a percentage
  "computed_pct": 35.08, // derived, labelled as such
  "source_report_year": 2024, "source_table": "Table 2.7",
  "flags": ""
}
project history

How this was built

  • v1E. coli and K. pneumoniae, 2017–2024, 3 editions.
  • v1.1Six organisms across three chapters — A. baumannii, P. aeruginosa (non-fermenters) and S. aureus/MRSA (staphylococci) added.
  • v2Regional Centre breakdowns for the three organisms with an RC-wise table — flagging editions whose RC panel changed rather than averaging across it as if it were stable.
  • v3 — currentNCDC NARS-Net carried as a parallel series — all eight editions, 2017–2024, E. coli and S. aureus. The two networks are published side by side and never joined: they share no comparison value.
cite / connect

Using this

$ cite --project→see CITATION.cff — cite this project and, for each value used, the network it came from (ICMR-AMRSN or NCDC NARS-Net) with its report edition and table
$ license --code→MIT
$ license --data→see DATA_LICENSE.md — fetched at run time, never redistributed here
$ read --research→docs/narsnet_v3_research.md — what each of the eight NARS-Net editions prints, edition by edition, and what can be checked against what
$ read --sources→docs/narsnet_investigation_artifacts.md — the eight editions with pinned SHA-256 hashes, and the 2024 edition’s full 54-site roster
$ connect --github→github.com/yashita-thakur
$ connect --orcid→orcid.org/0009-0004-7895-5250
sources

References

  1. Indian Council of Medical Research. Annual report: Antimicrobial Resistance Research and Surveillance Network, January 2022 to December 2022. 6th ed. New Delhi: Indian Council of Medical Research; 2022. Available from: https://www.icmr.gov.in/icmrobject/custom_data/pdf/resource-guidelines/AMRSN_Annual_Report_2022.pdf
  2. Indian Council of Medical Research. Annual report: Antimicrobial Resistance Research and Surveillance Network, January 2023 to December 2023. 7th ed. New Delhi: Indian Council of Medical Research; 2023. Available from: https://www.icmr.gov.in/icmrobject/uploads/Documents/1725536060_annual_report_2023.pdf
  3. Indian Council of Medical Research. Annual report: Antimicrobial Resistance Research and Surveillance Network, January 2024 to December 2024. 8th ed. New Delhi: Indian Council of Medical Research; 2024. Available from: https://www.icmr.gov.in/icmrobject/uploads/Report/1763981012_icmramrsnannualreport2024.pdf
  4. National Centre for Disease Control. National Antimicrobial Resistance Surveillance Network (NARS-Net) annual report: reporting period January-December 2017 [Internet]. New Delhi: National Centre for Disease Control, Directorate General of Health Services, Ministry of Health and Family Welfare, Government of India; [date unknown] [cited 2026 Sep 1]. Available from: https://ncdc.mohfw.gov.in/uploads/pdf/amr39.pdf
  5. National Centre for Disease Control. National Antimicrobial Resistance Surveillance Network (NARS-Net) annual report: reporting period January-December 2018 [Internet]. New Delhi: National Centre for Disease Control, Directorate General of Health Services, Ministry of Health and Family Welfare, Government of India; [date unknown] [cited 2026 Sep 1]. Available from: https://ncdc.mohfw.gov.in/uploads/pdf/amr38.pdf
  6. National Centre for Disease Control. National Antimicrobial Resistance Surveillance Network (NARS-Net) annual report: reporting period January-December 2019 [Internet]. New Delhi: National Centre for Disease Control, Directorate General of Health Services, Ministry of Health and Family Welfare, Government of India; [date unknown] [cited 2026 Sep 1]. Available from: https://ncdc.mohfw.gov.in/uploads/pdf/amr37.pdf
  7. National Centre for Disease Control. National Antimicrobial Resistance Surveillance Network (NARS-Net) annual report: reporting period January-December 2020 [Internet]. New Delhi: National Centre for Disease Control, Directorate General of Health Services, Ministry of Health and Family Welfare, Government of India; [date unknown] [cited 2026 Sep 1]. Available from: https://ncdc.mohfw.gov.in/uploads/pdf/amr36.pdf
  8. National Centre for Disease Control. National Antimicrobial Resistance Surveillance Network (NARS-Net) annual report: reporting period January-December 2021 [Internet]. New Delhi: National Centre for Disease Control, Directorate General of Health Services, Ministry of Health and Family Welfare, Government of India; 2022 [cited 2026 Sep 1]. Available from: https://ncdc.mohfw.gov.in/uploads/pdf/amr35.pdf
  9. National Centre for Disease Control. National Antimicrobial Resistance Surveillance Network (NARS-Net) annual report: reporting period January-December 2022 [Internet]. New Delhi: National Centre for Disease Control, Directorate General of Health Services, Ministry of Health and Family Welfare, Government of India; 2023 [cited 2026 Sep 1]. Available from: https://ncdc.mohfw.gov.in/uploads/pdf/amr34.pdf
  10. National Centre for Disease Control. National Antimicrobial Resistance Surveillance Network (NARS-Net) annual report: reporting period January-December 2023 [Internet]. New Delhi: National Centre for Disease Control, Directorate General of Health Services, Ministry of Health and Family Welfare, Government of India; [date unknown] [cited 2026 Sep 1]. Available from: https://ncdc.mohfw.gov.in/uploads/pdf/amr32.pdf
  11. National Centre for Disease Control. National Antimicrobial Resistance Surveillance Network (NARS-Net) annual report: reporting period January-December 2024 [Internet]. New Delhi: National Centre for Disease Control, Directorate General of Health Services, Ministry of Health and Family Welfare, Government of India; [date unknown] [cited 2026 Sep 1]. Available from: https://ncdc.mohfw.gov.in/uploads/pdf/amr30.pdf
  12. Kaur J, Dhama AS, Buttolia H, Kaur J, Walia K, Ohri V, Kumar V, Lynn AM, Srivastava A, Singh H. ICMR's Antimicrobial Resistance Surveillance system (i-AMRSS): a promising tool for global antimicrobial resistance surveillance. JAC Antimicrob Resist. 2021;3(1):dlab023. doi:10.1093/jacamr/dlab023. Available from: https://doi.org/10.1093/jacamr/dlab023
Scope and maintenance. Data is current through the ICMR AMRSN 8th edition (2024) and through all eight NCDC NARS-Net editions, 2017–2024. Built and completed independently before the author began medical school, and not maintained against editions published since — for anything published after 2024, check the ICMR AMRSN site or the NCDC reports page directly.