Self-brought antibiotics and their WHO AWaRe composition at the Hajj: dataset and analysis code for a two-season photo-verified pilot study

De-identified participant-level (n=60) and item-level (n=190) data, the data dictionary, the analysis code and the full analysis log from a cross-sectional pilot study of the antibiotics that pilgrims physically carry to the Hajj, conducted over the 1445 AH (June 2024) and 1446 AH (June 2025) seasons. Fifty pilgrims made their personal medicine bag available for a complete photo-verified inventory; 48 systemic antibacterials (ATC group J01 and oral metronidazole, ATC P01AB01) were coded to the WHO ATC/DDD Index and classified under the WHO AWaRe 2023 framework. Package photographs are not included, as consent covered study use only. README.md documents the file structure, the known data-quality features and the limitations on reuse: this is a convenience sample of 60 pilgrims, suitable for methodological work and for sizing a larger study, and not a prevalence estimate for the Hajj population. Ethics: the study was approved by The Biomedical Research Ethics Committee, Umm Al-Qura University, Makkah, Saudi Arabia, on 1 March 2024, before fieldwork began in either season. A single approval covers the whole protocol and is registered under approval numbers HAPO-02-K-012-2024-09-2818 and HAPO-02-K-012-2025-05-2712 for the Hajj 1445 AH and 1446 AH seasons respectively. Written informed consent was obtained from every participant, with separate verbal consent before the medicine bag was opened and before each package was photographed. Version 2 corrects an omission in the analysis log. The Mann-Whitney comparison of antibiotic knowledge scores between carriers and non-carriers was computed but not written to results/analysis_output.txt, because a guard in analyse.py skips that line when a knowledge score is stored as text rather than as a number. The line has been restored in its correct position. The three data files, the analysis code, the README and the licence are byte-identical to version 1, and no reported value changed. Version 3 corrects two country-level codes in participants.csv. The variable country_income_group, which is assigned at data cleaning from the country of nationality and is not participant-reported, had coded India as upper-middle income (3) and Türkiye as high income (4); under the World Bank classification for fiscal year 2026 (and fiscal year 2025) India is lower-middle income (2) and Türkiye is upper-middle income (3). Eight participant records are affected (HAJ-1445-0016, -0017, -0018, -0028; HAJ-1446-0014, -0015, -0029, -0030); no other field and no medicine record changed. The analysis log (results/analysis_output.txt) was regenerated with the unchanged statistical code: the Table 1 income-group counts, the Table 2 carriage estimates by income group, the exploratory income comparison and three Benjamini-Hochberg adjusted values changed; no headline estimate and neither of the two comparisons surviving multiplicity correction changed. One line of code/analysis.py changed, the footnote of Figure 2, which now reads "Most subgroups contain fewer than 30 inventoried bags". The README documents the correction in full. Version 4 completes the pre-specified double coding. In versions 1 to 3, 2 of 48 J01 items had been double-coded against a 10% target. A second author, blind to the first coding, has now coded 5 of 48 items (the two double-coded in the field plus three drawn with random.seed(20260907)) from the package photographs alone. The new folder double_coding/ contains the second coder's sheet, the comparison script (compute_kappa.py) and its output: agreement on ATC code 5/5 (Cohen's kappa 1.00) and on AWaRe category 4/5 (kappa 0.55). The single disagreement, amoxicillin + clavulanic acid (HAJ-1445-0009 M03) coded Watch by the second coder, was resolved against the WHO AWaRe classification, which lists it as Access; the first coder's value in medicines.csv is unchanged. Data change: double_coded set to 1 for the three newly double-coded items (HAJ-1445-0013 M02, HAJ-1445-0016 M02, HAJ-1446-0001 M01). No other field changed; participants.csv, the data dictionary and the code are identical to version 3. The analysis log was regenerated: the only line that changed is the data-quality audit line, now "J01 items double-coded: 5/48 (10.4%)". Version 5 corrects one ATC code and adds sensitivity analyses. The four oral metronidazole items (500 mg tablets) had been coded J01XD01, which the WHO ATC/DDD Index reserves for parenteral metronidazole; they are now coded P01AB01. They remain in the analytic set of systemic antibacterials, because the WHO AWaRe classification lists metronidazole as an Access antibiotic, so no reported number changes. code/analysis.py now reads the deposited CSV files directly, applies the pre-specified 20-item rule to the one stratum that reaches it (weakly enforced prescription regulation, 33 items: Access 18/33, exact binomial p = 0.059 against 0.70), and corrects the Figure 1 labels: the 70% Access target comes from the 2024 UN General Assembly political declaration, and one stratum does exceed 20 items. A new script, code/sensitivity_v5.py, with its output in results/sensitivity_v5_output.txt, adds participant-cluster and campaign-cluster bootstrap intervals, the Access share across regulatory categories, stratified analyses of sex and previous Hajj experience, a breakdown of incomplete courses, and the Access share weighted by defined daily doses remaining. Files are restricted and are available on reasonable request through the Zenodo access-request function. README.md gives full details of every change.