Sources & Methodology

What to know before relying on it

• IFCT 2017's copyright page does not allow its contents to be stored or reproduced electronically for a product without NIN's written permission. So our Indian food pages cite the IFCT record by code instead of printing its values; where USDA FoodData Central has analysed the same food (a reviewed, exact match), the page shows USDA's values, credited to USDA.
• USDA's analysis is of the same food, not of IFCT's Indian sample, so figures for a particular Indian variety may differ.
• History: while these pages still printed IFCT values, we found the dataset declares no units and stores energy in kilojoules, not kilocalories; 528 pages had overstated calories by 4.184× until that was caught and fixed. The values are no longer shown at all.
• IFCT lists distinct species under shared common names — two different 'Cat fish', two different 'Crab' — whose nutrition differs materially. We publish each separately under its scientific name rather than silently keeping one.

What to know before relying on it

• We publish the Foundation and SR Legacy datasets, which are laboratory analyses of whole foods, and exclude the 1.98 million branded supermarket products, whose values are manufacturer-declared rather than analysed.
• USDA files energy under a single nutrient name in both kilocalories and kilojoules. We match on the unit rather than the name, because reading whichever value came first is precisely how a 4.184x calorie error reached our Indian food pages.
• Descriptions get extremely granular -- dozens of near-identical beef cuts distinguished only by trim and grade. We exclude those, because a page per trim level is noise, not information.
• Where a food appears in both Foundation and SR Legacy we keep the Foundation record, which is the more recent analysis.
• Composition varies with variety, growing conditions, storage and preparation. These are typical values, not measurements of the item in your kitchen.
• These pages are separate from our Indian food pages rather than merged with them. The two sources name foods differently and rest on different analyses, and flattening them together would hide which one a number came from.

What to know before relying on it

• Pregnancy intakes appear only inside free-text notes, not as structured fields. We publish a percentage only where the note states an unambiguous absolute intake, which is folate (570 mcg) and calcium (1,200 mg).
• We deliberately publish no iron percentage. The dataset carries no pregnancy figure for iron; its 29 mg value is for a menstruating adult woman, and presenting that as a pregnancy target would be inventing a clinical number.
• Protein is excluded for the same reason: the source gives an increment (+9.5 g in the second trimester, +22 g in the third) that depends on body weight, not an absolute daily figure.
• Indian RDAs are deliberately higher than US values for some nutrients because Indian cereal-based diets have lower bioavailability. Every percentage we publish names the authority it was computed against, because a figure is not comparable across countries: folate in pregnancy is 600 mcg under US guidance and 570 mcg under India's, and the US does not raise calcium in pregnancy at all.

What to know before relying on it

• This tier was documented as ICMR data on 2026-08-10 and it is not. Its figures are US values -- vitamin A 900/700 mcg, vitamin D 15 mcg rising to 20 above 70 -- and the app reads it through parseUsTargets as the US source, while ICMR's separate tier backs the India view. Attributing US intakes to an Indian authority on the sources page was a mapping error made while closing a different gap.
• The records carry no source_url. ods.od.nih.gov refuses both automated clients and a browser from this environment, and the NCBI DRI summary tables need JavaScript, so no URL could be read and attached. The citation names the body but not the document.
• Only folate states an explicit pregnancy figure, so it is the only nutrient for which a pregnancy percentage is published. The rest are adult reference values and are labelled as such.

What to know before relying on it

• LactMed uses US generic names — acetaminophen rather than paracetamol — which may not match the name on an Indian prescription.
• We publish only records carrying actual safety content. The database also contains its own front matter (glossary, record format, disclaimers); eleven of these were being published as empty drug pages until we filtered them out.
• This is reference information about medicines in breast milk, never a recommendation. Dosing decisions belong to your doctor.

What to know before relying on it

• We publish a deliberately narrow subset. The full list of 2,439 products includes abortifacients, prescription hormones, antibiotics and central-nervous-system drugs, none of which belong in a browsable list on a pregnancy app.
• Broad keyword matching was tried and rejected because it pulled in look-alikes: matching 'calcium' surfaced acamprosate calcium, an alcohol-dependence treatment, and calcium polystyrene sulphonate, a potassium binder for kidney patients. Every published item matches an explicit whole-name rule.
• Prices are the published MRP and change over time. Confirm at the store.

What to know before relying on it

• This is the source that closed a gap we had previously declared open. LactMed answers the breastfeeding question and not the pregnancy one, and our only other herb data was ethnobotanical folk records that flag carrots and garlic as abortifacient. NCCIH reviews evidence and states something specific about pregnancy on all 56 monographs.
• We reproduce their safety section whole rather than summarising it, and add no verdict of our own. NCCIH's own statements differ sharply between herbs — ginger "may be safe", ashwagandha "should be avoided", fenugreek "not safe ... linked to increased risks of birth defects", chamomile "little is known" — and paraphrasing would flatten exactly the distinction that matters.
• "Little is known" appears often and is not a safety clearance: it means the studies do not exist, which is true of most herbs because trials in pregnancy are rarely run.
• NCCIH states on every page that it is not copyrighted and is in the public domain, and we publish only records where that statement is present.
• Captured by fetching rather than vendored, so it is a point-in-time snapshot. A length cap during capture once left a fragment of markup at the end of one safety quote; the build now rejects any record containing markup rather than repairing it, because a mangled safety statement is worse than a missing page.
• Evidence is about the herb, not about the bottle in your cupboard. Dose and form change the answer, and supplements are not regulated for strength or purity the way medicines are.

What to know before relying on it

• We publish a filtered subset. Topics outside this app's remit — cancers, natural-disaster preparedness, transplantation, male anatomical health — are excluded so the site stays within what it can cover credibly.
• Male infertility is kept, because it matters to readers trying to conceive, even though neighbouring topics in the same source group are excluded.
• This is US-authored guidance. Where Indian clinical practice differs, follow your own doctor.

What to know before relying on it

• Extracted from each PDF's text layer rather than read off the rendered page, then cross-checked: all six signs appear verbatim in both the parent card and the parent fact sheet, and tests assert that against the extracted text.
• This covers children generally, not newborns specifically, and it is about sepsis rather than every reason a baby might need care. We say so on the page instead of letting it stand in for a newborn triage list.
• We have found no US public-domain, parent-facing list of newborn-specific warning signs we can reproduce faithfully. We publish none rather than assemble one from mismatched sources or invent thresholds to fill the gap.
• CDC's phrasing "one or more of the following" is kept, because a list read as needing several signs at once means something different from what CDC wrote.

What to know before relying on it

• Every sign, bullet and threshold on that page is CDC's own wording. We added no sign, dropped none, did not reorder them by severity, and did not reword a threshold — on this page in particular, our judgement has no place.
• CDC nests some bullets under a lead-in, such as "You may think about hurting yourself because you:". We preserve that nesting; flattening it leaves three bare feelings and states something CDC did not.
• CDC scopes these to the full year after delivery, not the six weeks much pregnancy content assumes. We keep its scope.
• CDC says plainly that the list does not cover every symptom, and that you should call if something just isn't right. We quote that rather than paraphrasing it, because it matters more than the list.
• This is guidance for the mother. We have not found an equivalent US public-domain list of newborn warning signs in a form we can reproduce faithfully, so we publish none rather than assemble one.

What to know before relying on it

• We publish as routine only the doses CDC marks as recommended for all children. Doses for high-risk groups, and those left to shared decision-making, are shown separately and labelled, because hiding them would misrepresent the schedule and merging them would misrepresent your child.
• This replaced the WHO registry as our US source. WHO records what a country reports offering and cannot distinguish routine childhood vaccines from travel or occupational ones -- its 2023 US rows list Ebola, anthrax, yellow fever, cholera, typhoid, dengue and adult shingles alongside DTaP.
• The schedule is published as an HTML table built with merged cells. Read naively, cell position puts MMR at 2 months rather than 12-15 months. We reconstruct the table's true grid and pin the published ages in tests, because that error looks entirely plausible on the page.
• Where CDC states a dose as "1 or 2 doses annually" or "See Notes" we reproduce that wording rather than inventing a dose number for it.
• We do not reproduce the catch-up schedule or the detailed notes, which carry conditions we are not the right place to interpret. Both are linked on every page.
• cdc.gov refuses automated clients, so the schedule is captured through a browser and the extraction is committed as a fixed input. It will not silently drift, but it does need re-capturing when CDC publishes a new schedule year.

What to know before relying on it

• This is WHO's registry of what countries report offering, not a national programme's own schedule document. We previously described it as India's Universal Immunisation Programme, which overclaimed what the source says.
• The dataset spans 1995 to 2025 and we show only the most recent year. Merging years had been putting superseded guidance beside current guidance on the same page — India's polio entry blended 1995 and 2025 dosing.
• It carries no field separating routine childhood vaccines from travel, occupational or risk-group ones. We exclude doses explicitly recorded against risk groups, but that filter is imperfect.
• For this reason we publish no US vaccine pages from it. The 2023 US rows include Ebola, anthrax, yellow fever, cholera, typhoid, dengue and adult shingles vaccines, none flagged as non-routine, and presenting those as a child's schedule would mislead.
• Your own schedule comes from your paediatrician and your child's immunisation card.

What to know before relying on it

• It is a screening instrument, not a diagnostic one. A score above the cut-off does not mean a person has depression, and a score below it does not rule difficulties out. Only a clinician can diagnose, and the tool says so on the results screen.
• The scale was validated in specific populations and its cut-off is not equally sensitive in every language, culture or care setting. We report the band and encourage a clinical conversation rather than presenting the number as a verdict.
• We shipped all ten items with no attribution at all until 2026-08-10. The instrument's terms permit reproduction only with the authors, title and source cited, so this was a licensing failure as well as a credibility one.

What to know before relying on it

• Until 2026-08-10 this screen showed four Indian numbers to every reader in the world, and told them to call 112 — India's emergency number — inside the self-harm escalation. A reader in the US was handed numbers that do not connect from the US on the one screen where that causes direct harm.
• Four numbers were dropped rather than corrected: Vandrevala Foundation, NIMHANS and Snehi are real services, but their sites did not expose a number we could read and verify, and an unverifiable crisis number is worse than none.
• India's 112 emergency number is not published here. It is correct, but 112.gov.in was unreachable and the MHA page returns 404, so it fails the rule that every number trace to a page we actually read. Indian readers get Tele-MANAS, a 24/7 government line, plus the emergency-room instruction.
• Helplines change. Each entry carries the date it was last checked, and re-checking them is a standing task rather than a one-off.

What to know before relying on it

• The 13 food-safety records were re-sourced on 2026-08-10: each now names the specific document its guidance came from and carries the URL and the date that page was read, so a reader can check one claim rather than a set. Fish advice cites the FDA/EPA advisory; the rest cite the relevant MedlinePlus topic. Every URL was fetched before being attached.
• The 18 symptom records were re-sourced the same way on 2026-08-10. Thirteen cite the Office on Women's Health page on body changes and discomforts, one cites its labour page for Braxton Hicks, and the four that can signal an emergency -- headache, trouble breathing, vaginal bleeding and reduced fetal movement -- cite CDC's urgent maternal warning signs, because that is the document that speaks to them. They previously carried one string naming ACOG, NHS and Mayo Clinic with no document, two of which are not freely readable.
• One conflict is resolved in favour of the primary source rather than averaged: FDA lists albacore tuna as a Good Choice at one 4 oz serving a week, while MedlinePlus says 6 oz. We publish the FDA figure, because the FDA/EPA advisory is the source of record for fish and mercury.
• These are authored summaries of public guidance, not extracted quotations. The upstream lint that gated them (lint_health_curated.py) accepts a disclaimer in place of a citation and treats a missing one as a warning rather than a failure, so 'cited' there is a weaker guarantee than it is elsewhere in this registry.
• The mercury advice was checked against the FDA/EPA fish table on 2026-08-10 and was incomplete: it omitted marlin and orange roughy, both on FDA's Choices to Avoid list, and said 'tilefish' where FDA avoids only Gulf of Mexico tilefish — Atlantic tilefish is a Good Choice. Corrected, and pinned by test.
• Symptom pages describe when something is typical and when to call someone. They are not a triage tool and cannot replace one; every page says so.

What to know before relying on it

• Infant feeding and child activities were re-sourced per fact on 2026-08-10. Feeding stages cite the WHO fact sheet on infant and young child feeding; the honey and choking rows cite MedlinePlus infant nutrition; activities cite CDC's Learn the Signs. Act Early., whose 2-months-to-5-years range matches this tier's ages. Each URL was fetched and read before being attached.
• Three tiers still name a body without a document, because no URL could be verified from this environment: sleep ('AASM 2016, Paruthi et al., J Clin Sleep Med' — the most specific citation of the group, but PubMed returns a bot challenge here), child safety, and human development 5-18.
• The human development 5-18 tier was withdrawn on 2026-08-10 and is no longer published. It cited 'CDC Learn the Signs. Act Early.' for stages from 12 to 18, and Act Early covers 2 months through 5 years -- the document says nothing about adolescents. Rather than leave five entries standing on a citation that does not support them, we stopped publishing them, the same call made on the 621 Duke ethnobotanical herb records. The data still exists upstream and can come back once it is sourced to something that covers those ages.
• As with the pregnancy tiers, the citation is per tier rather than per fact, and these are authored summaries rather than quotations.
• The safety and development tiers are small (3 and 5 entries). They are starting points, not a complete guide to either subject, and should not be read as one.

What to know before relying on it

• Attribution lives in the data as growth:source = 'WHO Child Growth Standards' rather than in a citation field, and carries no edition year or retrieval date.
• The 0-5 standards describe how healthy children grow under recommended conditions; the 5-19 set is a reference describing a surveyed population. They are not interchangeable, and a percentile is a position in a distribution rather than a verdict on a child.

What to know before relying on it

• Every figure is labelled with the survey round it came from. NFHS-4 is used only where NFHS-5 published no value for that state.
• These are population statistics, not a prediction about any individual pregnancy.

What to know before relying on it

• Only about 3% of the national directory carries any specialty tagging, so this is a starting point rather than a complete list. A hospital near you offering maternity care may simply not be listed.
• The directory retains legacy entries filed under pre-2014 state names — the same Hyderabad hospitals appear under both Andhra Pradesh and Telangana. We resolve each district to its current state and de-duplicate.
• An older 1,048-hospital directory was evaluated and rejected: it predated the 2014 state reorganisation entirely and returned nothing for Telangana.
• Always call ahead. We do not verify that any listing is still accurate or still offers these services.

What to know before relying on it

• Carries no specialty information, so it supports counts and an explanation of the system's tiers, not a search for a particular service.
• Population figures per tier are Indian Public Health Standards norms describing how the system is planned, not a guarantee about any individual centre.

What to know before relying on it

• This is the least-sourced content on the site, and the honest reason it appears here is that it would otherwise be the one dataset this page did not account for. It was compiled before we adopted the sourcing standard the rest of the site is held to, and its individual entries are not traceable to a named authority the way a nutrient value or a vaccine dose is.
• A minority of names carry an external citation — a verified Monier-Williams dictionary etymology, an epic character profile, or both. The remaining names carry a conventional meaning that is widely attested but not individually attributed. (Re-auditing this citation count against the current corpus size is open work.)
• Name meanings are genuinely contested. The same name is often given different glosses by different sources, and a single "correct" meaning frequently does not exist — treat these as the common reading, not a ruling.
• Rashi and nakshatra are traditional phonetic associations, recorded because families use them when choosing a name. Every page says on its face that this is cultural tradition rather than a rule, and nothing here is a claim about a child.
• Re-sourcing this corpus against a citable lexicon is open work, tracked in docs/data-sources-needed.md.

What to know before relying on it

• This is a record of what plants have traditionally been used for, gathered from folk medicine worldwide. It is not a safety assessment, and we do not publish it as one.
• Read as guidance it produces alarming nonsense: it flags okra, garlic, carrots, coriander, parsley and sesame as abortifacient or emmenagogue, because somewhere a tradition used them that way. Its fields are folk_use, has_activity and contains_chemical — no dose, no plant part, no evidence grade. Several entries are marked veterinary and one activity is recorded with a literal question mark.
• We considered publishing a page per plant — 621 pages against high-intent queries — and rejected it. Pages telling pregnant readers that carrots are abortifacient would have ranked, and would have discredited every properly sourced figure on this site.
• What we publish instead is the LactMed evidence for the same plants, at /guide/herbs. For herbs in pregnancy specifically we have found no equivalent evidence-based, openly licensed source, so we publish none.

What to know before relying on it

• A name appears only where the dictionary definition genuinely overlaps the meaning we already publish. An earlier unfiltered pass matched roughly half its names wrongly — 'Jai' to an unrelated plant, 'Kanika' to a king's name — so we require semantic agreement before publishing.
• Most popular modern names have no classical entry at all, which is accurate rather than a gap: they are modern coinages or non-Sanskrit in origin.
• An 1899 dictionary reflects the scholarship of its time.

What to know before relying on it

• Many characters share a name across the epics. We show one notable bearer, not a claim about the name's only meaning.
• We deliberately do not use the Purāṇic Encyclopaedia that covers similar ground: it remains in copyright, and checking it surfaced real errors including a 'Sita' entry that is not Rama's wife.

What to know before relying on it

• These are retellings in modern language, not translations. Violent endings are softened for a young audience.
• Two stories carry a reframed moral, noted on each, because the source's own stated lesson — keep your employer's problem half-solved, or do not warn others of a crime — is not one worth passing to a child.
• Six of the sixteen fables in the source books are not retold at all: three contain adult content, two are plots consisting only of deaths, and one teaches a lesson we would not want a child to take.