NFHS‑4 · NFHS‑5 · NFHS‑6 Methodology & Indicator Definitions

1The survey‑weighted estimator

NFHS‑5 is a stratified, multi‑stage probability sample. To recover nationally and sub‑nationally representative quantities, every record is weighted by its design weight. DHS distributes the weight as an integer scaled by $10^{6}$; the analytic weight is

$$ w_i \;=\; \frac{V005_i}{1{,}000{,}000}\qquad\text{(women/children: }V005;\ \text{household: }HV005;\ \text{men: }MV005\text{)} $$

For a binary indicator $x_i\in\{0,1\}$ defined over an eligible population, the point estimate is the weighted proportion, expressed as a percentage:

$$ \hat{p} \;=\; \frac{\displaystyle\sum_{i\in\mathcal{E}} w_i\,x_i}{\displaystyle\sum_{i\in\mathcal{E}} w_i}\times 100 $$

where $\mathcal{E}$ is the set of records that satisfy the indicator's eligibility rule (the denominator). Records outside $\mathcal{E}$, and those with a missing value of $x$, are excluded — they are coded NaN and never enter either sum. The reported sample size $n=\lvert\mathcal{E}\rvert$ is the unweighted count and drives the reliability flag below.

Every indicator is produced at nine aggregation levels — national, state/UT, urban–rural, wealth quintile, women's education, and the state×residence, state×wealth, state×education cross‑tabs, plus district — by restricting $\mathcal{E}$ to the corresponding subgroup. Mortality and fertility (§14) are the sole exception: they are model‑based rates, not weighted means.

Design‑based inference Point estimates use the sampling weights only. For standard errors and confidence intervals that respect the design, declare the survey structure — PSU V021, strata V022, weight V005 — in a survey package (svyset in Stata, svydesign in R, samplics in Python). This dashboard reports point estimates; it does not display sampling errors.

2Reliability & small‑sample suppression

Sub‑national cells — particularly wealth or education breakdowns for small UTs, or rare‑event indicators whose eligible population is a slice of a slice — can rest on a handful of observations, where a single respondent moves the estimate by tens of points. Following the DHS reporting convention, every cell carries a reliability band derived from its unweighted denominator $n$:

Band	Rule	Interpretation	Dashboard treatment
ok	$n \ge 50$	Reliable	Shown normally
caution	$25 \le n < 50$	Interpret with caution	Muted, "small sample" flag
suppress	$n < 25$	Unreliable	Hidden / greyed as no‑data

Model‑based rates (§14) use a more conservative band keyed to birth/exposure counts (suppress below ~250, caution below ~500), because rates require substantially larger samples than proportions for comparable precision.

3Child anthropometry & nutritional status

Children under 5 with a valid anthropometric z‑score (WHO Child Growth Standards). Z‑scores are stored ×100; the plausible range is $[-600,600]$ (i.e. $\pm6$ SD); flagged values are excluded.

$$ \text{stunted}= \mathbb{1}[\,HAZ<-2\,],\quad \text{wasted}= \mathbb{1}[\,WHZ<-2\,],\quad \text{underweight}= \mathbb{1}[\,WAZ<-2\,] $$

Stunting / Wasting / Underweight / Overweight FS #81–85 · 35.5 / 19.3 / 32.1 / 3.4

Height/weight-for-age or -height beyond ±2 SD of the WHO child growth standard, children under 5.

Population: Children age <60 months with a valid z‑score for the relevant index.
Definition: Stunting hw70<−200; wasting hw72<−200; underweight hw71<−200; overweight hw72>+200; severe wasting hw72<−300.
Variables: hw70 HAZ · hw71 WAZ · hw72 WHZ · b19 age

Child anaemia (6–59 months)child_anaemia FS #92 · 67.1 approx

Haemoglobin below the WHO-defined anaemia threshold for their age, measured by a finger-prick blood test, among children aged 6–59 months.

Population: Children 6–59 months with a valid haemoglobin reading.
Definition: Any anaemia = altitude‑adjusted Hb <11.0 g/dL, read from the DHS level variable hw57 ∈ {1,2,3} (severe/moderate/mild); 4 = not anaemic.
Variables: hw57 · b19

On this figure The weighted rate (68.0) runs 0.9pp above the published figure (67.1). This traces to how survey weights interact with this indicator, not a definitional error — treat it as accurate to ~1pp.

4Child immunization

Children age 12–23 months, with vaccination established from the card or the mother's recall — each antigen coded received if its status is in {1,2,3} (1 = date on card, 2 = mother's report, 3 = marked on card).

Fully vaccinated (basic)full_imm FS #57 · 76.4

Received all basic vaccines by 12–23 months: BCG (1 dose), DPT (3 doses), Polio (3 doses), and Measles (1 dose), per the national immunization schedule.

Population: Living children 12–23 months.
Definition: BCG and 3 doses penta/DPT and 3 doses polio (excl. birth dose) and measles‑containing vaccine, dose 1.
Variables: h2 BCG · h7 DPT3/penta · h4,h6,h8 polio 1/2/3 · h9,h9a MCV · b5 alive · b19

Three pitfalls this indicator corrects (1) Dead children. The KR file holds deceased children with blank vaccination fields; counting them as "not vaccinated" deflates every antigen ~3.5pp. The denominator is restricted to living children (b5==1).
(2) Polio birth dose. The fact‑sheet footnote defines polio‑3 as three doses excluding the dose given at birth (h0); the count uses routine doses h4,h6,h8 only.
(3) Measles / MR. India's measles‑rubella rollout records some first doses in the second‑dose slot, so MCV1 = h9∈{1,2,3,4} OR h9a. Without these corrections the naïve estimate is ~60%.

Antigen	Variable	Fact sheet
BCG	`h2 ∈ {1,2,3}`	95.2 (#59)
DPT/penta 3	`h7 ∈ {1,2,3}`	86.7 (#61)
Polio 3 (excl. birth)	`h4 & h6 & h8`	80.5 (#60)
MCV 1	`h9∈{1‑4} \| h9a`	87.9 (#62)

5Infant & young child feeding

Feeding indicators use the 24‑hour dietary recall, collected for the youngest child living with the mother. Food items are mapped to the eight WHO‑2021 IYCF food groups.

Early initiation of breastfeedingearly_bf FS #75 · 41.8

Put to the breast within 1 hour of birth, among children under 3 years.

Population: Last child born in the 3 years before the survey (bidx==1 & age<36 mo).
Definition: Put to the breast within one hour of birth: m34 ∈ {0,100} (0 = immediately, 100 = within the first hour). Never‑breastfed / missing remain in the denominator as 0.

Exclusive breastfeeding (<6 months)excl_bf FS #76 · 63.7

Fed only breast milk in the 24 hours before the survey — no water, formula, or other food — among infants under 6 months living with their mother.

Population: Youngest child <6 months, alive, living with mother.
Definition: Breastfed (v404==1) and given nothing else in the 24‑hour recall (no water, other milk, or foods across v409–v414*).

Adequate diet — minimum acceptable diet (6–23 mo)adequate_diet_6_23 FS #80 · 11.3

Received a sufficiently varied and frequent diet for their age, among children aged 6–23 months.

Definition: Two formulas by breastfeeding status (per fact‑sheet footnote):

$$ \text{MAD}=\begin{cases}\mathbb{1}[\,\text{food groups}\ge 4\,]\;\wedge\;\text{MMF}, & \text{breastfed}\\[4pt] \mathbb{1}[\,\text{milk feeds}\ge 2\,]\;\wedge\;\text{MMF}\;\wedge\;\mathbb{1}[\,\text{non‑milk food groups}\ge 4\,], & \text{non‑breastfed}\end{cases} $$

MMF = minimum meal frequency (breastfed: ≥2 solid feeds at 6–8 mo, ≥3 at 9–23 mo; non‑breastfed: ≥4 total feeds with ≥1 solid). Validated against the breastfed (11.1, #78) and non‑breastfed (12.7, #79) sub‑components.

Timely complementary feeding (6–8 mo)comp_feeding_6_8 FS #77 · 45.9

Started solid/semi-solid food on time, alongside breast milk, among children aged 6–8 months.

Definition: Breastfed and receiving solid/semi‑solid food, among children 6–8 months.

Definitional choice — dietary diversity Minimum dietary diversity (min_diet_diversity) has no NFHS‑5 fact‑sheet target. It is computed to the WHO‑2021 standard of ≥5 of 8 food groups (national value 23.6%). NFHS‑5's own "adequate diet" composite (above) instead uses the ≥4 food‑group threshold defined in its footnote, so the two are intentionally not identical.

6Childhood illness & care‑seeking

Denominators here are rare‑event slices (only children who were ill), so sub‑national cells lean heavily on the reliability flag.

ORS / Zinc for diarrhoeaors_diarrhoea · zinc_diarrhoea FS #70 / #71 · 60.6 / 30.5

Given oral rehydration salts (ORS) or zinc, among children with recent diarrhoea.

Population: Children with diarrhoea in the 2 weeks before the survey (h11 ∈ {1,2}).
Definition: Given ORS h13 ∈ {1,2}; given zinc h15e==1.

Care‑seeking for fever / ARIcareseek_ari_fever FS #74 · 69.0

Taken to a health provider, among children with fever or a breathing illness.

Population: Children with fever (h22==1) or symptoms of ARI — cough (h31∈{1,2}) with short, rapid breathing (h31b==1) that was chest‑related (h31c∈{1,3}).
Definition: Advice or treatment sought from a health facility or provider: the public and private medical sources in the h32* series. Excludes retail pharmacy, shop, unspecified "other", traditional healer, friend/relative, and no‑treatment.

Two conditions this indicator turns on. The chest‑related qualifier is what separates ARI from an ordinary cold; without it the denominator fills with children who seek care less often. And "health facility or provider" is narrower than "anywhere advice was sought" — a pharmacy or a shop is not a provider. Both rounds now apply the identical rule.

Residual. This definition reproduces NFHS‑4's published 73.2 exactly (73.21) and cuts the error against the 33 published NFHS‑5 state figures from RMSE 4.5pp (worst 11.9) to RMSE 1.5pp (worst 3.7) — but it lands NFHS‑5's national figure at 70.0 against a published 69.0. We leave the 1.0pp gap standing rather than close it by dropping legitimate facilities (urban health centres, NGO hospitals) from the numerator, which would fit the number at the cost of the definition.

7Maternal & newborn care

Antenatal and postnatal indicators refer to the woman's most recent live birth in the 5 years before the survey; delivery indicators are computed at the birth level over all live births in that window.

Institutional deliveryinst_delivery FS #50 · 88.6

Birth happened in a hospital or clinic rather than at home.

Population: All live births in the last 5 years (m15_1..m15_6 reshaped to birth level, age ≤59 mo).
Definition: Delivered in a health facility: m15 ∈ [20,89] (10–19 = home, 96 = other → 0).

Restricting to the most recent birth only would bias this ~1.7pp high, as recent births are more likely institutional.

Caesarean sectioncsection · _public · _private FS #54/56/55 · 21.5 / 14.3 / 47.4

Birth delivered surgically (C-section) — overall, and split by government and private facility.

Definition: m17==1 over all live births; the public split is conditioned on m15∈[20,29], the private split on m15∈[30,39].

Antenatal careanc4plus · anc_1st_tri FS #41 / #40 · 58.1 / 70.0

At least 4 antenatal check-ups, and a first check-up within the first 3 months of pregnancy, among mothers.

Definition: 4+ ANC visits: m14_1 ≥ 4. First‑trimester ANC: m13_1 ≤ 3 months. Both over all most‑recent births in the window.

Denominator rule "Don't know" / missing number of ANC visits are counted as not 4+ (i.e. 0), not dropped — dropping them would inflate the rate ~1pp. With this rule the national estimate is 58.4 (published 58.1).

IFA supplementationifa_100plus · ifa_180plus FS #43 / #44 · 44.1 / 26.0

Took iron-folic acid tablets for 100+ or 180+ days during pregnancy, among mothers.

Definition: Iron‑folic‑acid taken for ≥100 / ≥180 days in pregnancy: m46_1 ≥ 100 / 180 (with <900 to drop the "don't know" code 998).

Neonatal‑tetanus protection & postnatal carennt_protection · pnc_mother_2days · pnc_newborn_2days FS #42 / #46 / #49 · 92.0 / 78.0 / 79.1

Birth protected against newborn tetanus by maternal vaccination, and a check-up for mother and newborn within 2 days of delivery.

Definition: NNT: ≥2 lifetime tetanus injections (m1_1 + m1a_1 ≥ 2). PNC: a check by health personnel (provider 11/12/13) within ≤2 days (timing code ≤202).

NFHS‑4PNC pathways. The pre‑discharge (m62) and post‑discharge (m50/51/52) checks are two complementary pathways, not a double‑count: m62 only applies to institutional births, the post‑discharge path carries home births, and no mother satisfies both. Their union is the correct estimate — dropping either path would erase real PNC. It lands a little above the fact sheet (~65 vs 62.4) because m62 has no provider sub‑variable in NFHS‑4 (NFHS‑5's m64 doesn't exist here), so institutional pre‑discharge checks can't be filtered to health personnel the way the fact sheet does — a data limitation, not a definition choice.

8Family planning

Contraception indicators are defined over currently‑married (in‑union) women 15–49 — the standard CPR denominator. Computing over all women is a frequent error that understates use by ~14pp and inverts the education gradient.

Modern contraceptive use (mCPR)modern_contra FS #29 · 56.5

Currently using a modern contraceptive method (pill, IUD, sterilization, condom, etc.), among currently married women.

Population: Currently‑married women 15–49 (v502==1) with valid method status.
Definition: Currently using a modern method: v313==3.

Unmet need for family planningunmet_need FS #36 · 9.4

Want to avoid or delay pregnancy but aren't using any contraception, among currently married women.

Definition: DHS‑derived v626a ∈ {1,2} (unmet need for spacing or limiting), among currently‑married women. The derived variable is used directly — the multi‑clause definition is never rebuilt by hand.

9Women's nutrition & anaemia

Women 15–49. BMI is stored ×100 (so 18.5 kg/m² → 1850); pregnant women and those with a very recent birth are excluded from BMI per the survey convention.

Thin / Overweight‑or‑obesethin_women · women_overweight FS #86 / #88 · 18.7 / 24.0

BMI below 18.5 (thin) or 25 and above (overweight/obese), among women.

Definition: Thin v445 < 1850; overweight/obese v445 ≥ 2500 (valid BMI 1200–6000).

Anaemia (women 15–49)anemia_women FS #95 · 57.0

Haemoglobin below the WHO anaemia threshold, among women.

Definition: Any anaemia v457 ∈ {1,2,3}, altitude‑ and smoking‑adjusted. NFHS uses capillary blood, so values are not comparable to venous‑blood surveys.

10Women's empowerment, literacy & digital access

All women 15–49 (menstrual hygiene: 15–24).

Indicator	Plain language	Key	Definition	FS
Literacy	Can read and write, among women.	`women_literate`	`v155∈{1,2}` (reads a sentence) OR `v133≥9`	71.5 (#14)
10+ yrs schooling	10 or more years of education, among women.	`women_edu10`	`v133 ≥ 10`	41.0 (#16)
Owns house/land	Owns a house or land, alone or jointly, among women.	`own_house_land`	`v745a` or `v745b ∈ {1,2,3}` (alone or jointly)	43.3 (#121)
Uses a bank account	Operates a bank account in their own name, among women.	`bank_account_self`	`v170 == 1`	78.6 (#122)
Uses a mobile phone	Uses a mobile phone of their own, among women.	`mobile_self`	`v169a == 1`	54.0 (#123)
Ever used internet	Has used the internet at least once, among women.	`internet_ever`	`v171a ∈ {1,2,3}`	33.3 (#18)
Menstrual hygiene (15–24)	Uses a hygienic method of period protection, among women aged 15–24.	`menstrual_hygiene`	any hygienic method `s260{b,c,d,e}==1`	77.3 (#124)
Married before 18 (20–24)	Married before turning 18, among women aged 20–24.	`child_marriage_women`	`v511 < 18` among women 20–24	23.3 (#20)
Begun childbearing (15–19)	Already a mother or currently pregnant, among women aged 15–19.	`teen_preg`	`v201>0` or currently pregnant `v213==1`	6.8 (#23)

11Household environment & assets

Household‑level indicators, weighted by HV005.

Indicator	Plain language	Key	Definition	FS
Electricity	Household is connected to the electricity grid.	`electricity`	`hv206 == 1`	96.8 (#7)
Improved drinking water	Household has access to a safe drinking water source (piped, tube well, etc.).	`imp_water`	`hv201` ∈ JMP improved set	95.9 (#8)
Improved sanitation	Household has a private, hygienic toilet, not shared with other households.	`imp_sanitation`	`hv205` ∈ improved set AND not shared (`hv225==0`)	70.2 (#9)
Clean cooking fuel	Household cooks with a clean fuel (like LPG) instead of wood, coal, or dung.	`clean_fuel`	`hv226 ∈ {1,2,3,4}` (electricity/LPG/gas/biogas)	58.6 (#10)
Iodized salt	Household's salt has adequate iodine.	`iodized_salt`	`hv234a == 1` among tested households	94.3 (#11)
Health insurance	Household has at least one member covered by health insurance.	`health_insurance`	any member covered by a scheme	41.0 (#12)

Two definitional points Sanitation must exclude shared facilities (hv225==0); the improved‑regardless‑of‑sharing figure is ~9pp higher and does not match the published indicator. Health insurance is the household coverage flag (any usual member covered), validated at 41.2% against the published 41.0%; note that coverage predates the full rollout of national schemes at survey time, so it understates current coverage.

NFHS‑4Improved water. The improved‑water set is hv201 ∈ {11,12,13,21,31,41,51,72} — piped variants, tube well / borehole, protected dug well, protected spring, rainwater, and community RO plant. Tanker truck, cart‑with‑tank, and bottled water are excluded, matching the fact sheet's own "improved" footnote.

12Adult biomarkers & non‑communicable disease (15+)

Person‑level (PR recode), de‑facto household members aged 15+, computed separately for women (_w) and men (_m). Anaemia is offered in two framings — see the note.

Indicator	Plain language	Definition	FS (W / M)
High blood sugar	Elevated blood sugar on a finger-prick test, among adults 15+.	random plasma glucose `shb74 > 140` mg/dL OR on glucose medication	13.5 / 15.6
Hypertension	High blood pressure, among adults 15+.	mean of 2nd & 3rd readings ≥140 systolic and/or ≥90 diastolic, OR on BP medication	21.3 / 24.0
Any tobacco use	Uses any form of tobacco, among adults 15+.	`sh25 == 1`	8.9 / 38.0
Any alcohol use	Consumes alcohol, among adults 15+.	`sh26 == 1`	1.3 / 18.8
Anaemia (15+, biomarker)	Low haemoglobin on a blood test, among adults 15+.	women `ha57 ∈ {1,2,3}`; men <13.0 g/dL — `hb57` (NFHS‑5), `hb56 < 130` (NFHS‑4)	57.0 / 25.0

Blood pressure uses the average of the 2nd and 3rd measurements (the 1st is discarded); out‑of‑range special codes are treated as missing.

NFHS‑4 Male anaemia. Computed from adjusted haemoglobin directly at the <13.0 g/dL threshold (hb56 < 130), rather than from NFHS‑4's hb57, which is coded to the women's <12.0 g/dL cut‑off. NFHS‑5's hb57 is already correctly sex‑specific and is used as‑is — both rounds measure male anaemia at the same <13.0 threshold.
Hypertension. Both rounds use the medication‑inclusive definition (≥140/90 or on BP medication) so the Compare view is like‑for‑like. This correctly counts successfully‑treated hypertension, but runs ~2pp above a measurement‑only fact‑sheet band — the two are not expected to tie exactly.

Two anaemia measures on this page The person‑level (15+, biomarker) anaemia here and the woman‑level (15–49) anaemia in §9 are distinct indicator keys and are never merged. They agree closely nationally but diverge by state because they use different weights and denominators — that divergence is real and is itself worth exploring in the correlation view.

13Men's indicators (15–49)

Men's recode (MR), restricted to 15–49 for fact‑sheet parity. Men's BMI is not carried in MR and is merged from the person recode (hb40) on the household‑member key, weighted by mv005.

Indicator	Plain language	Key	Definition	FS
Literacy	Can read and write, among men.	`men_literate`	`mv155∈{1,2}` OR `mv133≥9`	84.4 (#15)
10+ yrs schooling	10 or more years of education, among men.	`men_edu10`	`mv133 ≥ 10`	50.2 (#17)
Thin (BMI<18.5)	BMI below 18.5, among men.	`men_thin`	`hb40 < 1850`	16.2 (#87)
Overweight (BMI≥25)	BMI of 25 or above, among men.	`men_overweight`	`hb40 ≥ 2500`	22.9 (#89)
Ever used internet	Has used the internet at least once, among men.	`men_internet`	`mv171a ∈ {1,2,3}`	57.1 (#19)
Comprehensive HIV knowledge	Has accurate, complete knowledge of how HIV spreads and is prevented, among men.	`men_hiv_knowledge`	5‑condition composite (2 prevention facts, healthy‑can‑have, reject 2 myths)	30.7 (#116)

14Mortality & fertility model‑based rate

These are not weighted proportions. They are synthetic‑cohort rates following the DHS Guide to Statistics, computed from the full birth history (BR) with woman‑level exposure (IR). Available at national / state / residence / wealth / education only — never by district, where they would be unstable.

Childhood mortality

Deaths per 1,000 live births — in the first 28 days (neonatal), before age 1 (infant), and before age 5 (under-five).

Over a reference window of the 60 months before interview, deaths and child‑months of exposure are accumulated in eight age segments $s\in\{[0,1),[1,3),[3,6),[6,12),[12,24),[24,36),[36,48),[48,60)\}$ months. Each segment yields a conditional death probability

$$ q_s \;=\; \frac{\sum_i w_i\,d_{i,s}}{\sum_i w_i\,e_{i,s}} $$

where $d_{i,s}$ flags a death of child $i$ inside segment $s$ and within the window, and $e_{i,s}$ is the fraction of the segment the child actually lived in‑window. Segment probabilities are chained into cumulative mortality:

$$ \text{NNMR}=1000\,q_{[0,1)},\qquad {}_{1}q_{0}=1000\Big(1-\!\!\prod_{s:\,<12\text{mo}}\!\!(1-q_s)\Big),\qquad {}_{5}q_{0}=1000\Big(1-\prod_{\text{all }s}(1-q_s)\Big) $$

Targets — NNMR 24.9 (#25), IMR ${}_1q_0$ 35.2 (#26), U5MR ${}_5q_0$ 41.9 (#27), per 1,000 live births.

Fertility

The average number of children per woman (TFR), births per 1,000 women aged 15–19, and the number of girls born per 1,000 boys.

Over the 36 months before interview, age‑specific fertility rates divide births in the window by woman‑years of exposure in each five‑year age group $g$ (exposure drawn from all women, including the childless, in the IR file):

$$ \text{ASFR}_g=\frac{\sum w\cdot(\text{births to women aged }g)}{\sum w\cdot(\text{woman‑years aged }g)},\qquad \text{TFR}=5\sum_{g} \text{ASFR}_g $$

Targets — TFR 2.0 (#22), adolescent fertility ASFR(15–19) 43 per 1,000 (#24), sex ratio at birth 929 females per 1,000 males (#4).

15Known limitations

Item	Status	Detail
Child anaemia	±0.9pp	Weighted 68.0 vs published 67.1; a weighting nuance, not a definition error. Accurate to ~1pp.
Care‑seeking, fever/ARI (NFHS‑5)	+1.0pp	Weighted 70.0 vs published 69.0. The same definition reproduces NFHS‑4's published 73.2 exactly and matches the 33 published NFHS‑5 state figures to RMSE 1.5pp — so the residual sits in NFHS‑5's own national tabulation. Not closed by dropping legitimate facilities from the numerator. See §6.
Sex ratio at birth (NFHS‑4)	≈ −5	Survey‑weighted 914 vs published 919; the unweighted ratio is 921.7. Same weighting nuance as child anaemia — `b4` is clean (male/female only) and excluding twins changes nothing. The principled weighted estimate is kept.
NFHS‑4 targets	note	NFHS‑4 target figures have been cross‑checked against the published report at national and subgroup level; any indicator without an explicit note elsewhere on this page has been validated to the same tolerance as NFHS‑5.
Sampling errors	not shown	Point estimates only. Confidence intervals require a design‑based survey package (PSU/strata/weight).
District‑level rates	excluded	Mortality/fertility are not estimated at district level — sample sizes are too small for stable rates.
Small‑UT breakdowns	suppressed	Wealth/education cells for small UTs (e.g. Lakshadweep) often fall below n=25 and are hidden; national wealth quintiles mean affluent UTs may have no "poorest" cell at all.
Capillary‑blood anaemia	note	NFHS measures haemoglobin in capillary blood; values should not be compared with venous‑blood surveys.

Every numeric target above (FS #n) refers to the official NFHS‑5 India Fact Sheet. A national estimate is accepted when it falls within ±0.5pp of its target (±1.0 for mortality rates, ±0.1 for TFR); the one documented exception is child anaemia.

16NFHS‑6 (2023‑24) — what's different

Every other round on this page is computed from the unit‑level microdata. NFHS‑6 is not. Its recodes have not been released, so every NFHS‑6 figure here is read from the official fact sheets — one for India and one for each of 35 states/UTs. That single fact changes what the round can and cannot show, and it is worth being blunt about it.

Provisional IIPS publish the NFHS‑6 fact‑sheet figures as provisional. They may be revised when the full National Report and the microdata are released. Treat NFHS‑6 numbers here as indicative, not final.

State level only

The fact sheets print each indicator at national, state/UT, and urban–rural level — and nothing else. So for NFHS‑6 there are no district estimates, no wealth quintiles and no education breakdowns, and there can't be until the microdata is out. The dashboard doesn't fake them: the District Spread, Equity Lens, Education Gradient and Z‑score Distribution sections simply stand down for this round, and the map and correlation views fall back to state level.

Manipur is absent. NFHS‑6 was not conducted there, so the round covers 35 states/UTs, not 36. Manipur is not missing data — it has no NFHS‑6 survey at all.

About 40 of our 68 indicators

Only the indicators the fact sheets actually print can be carried across. Roughly 40 of the dashboard's 68 indicators are available for NFHS‑6; the rest are not in the sheets yet. Nothing is estimated, imputed or carried forward from NFHS‑5 to fill a gap — an indicator either has an NFHS‑6 number or it does not appear for that round.

What's absent, and why

Not in NFHS‑6 here	Reason (per the Ministry of Health & Family Welfare)
Anaemia (women, men, children)	Dropped from this round pending a methodological switch from capillary to venous blood sampling. It is to be reported separately through an ICMR survey, so there is no comparable NFHS‑6 anaemia figure.
Mortality & fertility rates — IMR, U5MR, NNMR, sex ratio at birth	Not printed in the fact sheets. (TFR is printed, and is carried.) The full birth‑history rates need the microdata.
Clean cooking fuel	Moved out of NFHS reporting to Swachh Survekshan Grameen.
Several child‑health and family‑planning indicators	Deferred to the full NFHS‑6 National Report; the fact sheets carry only a subset.

How the numbers get here

The India fact sheet has a real text layer, so its values are read exactly. The 35 state sheets do not — their tables were flattened to vector outlines — so those have to be read by machine, and any single reading of them can be wrong.

So no single reading is trusted. Four independent extractions of the same pages are compared cell by cell, and each value is decided by majority: two extractions landing on the identical wrong number is vanishingly unlikely, so a value that two or more of them agree on is treated as confirmed. 99.5% of the NFHS‑6 figures used here are confirmed by two or more independent readings; the handful that are not are reviewed by hand. Where the four disagree irreconcilably, or where one reading offers a number that the others read as a printed blank, no value is published rather than a guessed one.

A second, independent check runs on top of the vote. Every fact‑sheet row also prints its own NFHS‑5 total — and we already hold NFHS‑5 exactly, from microdata. So one of the four columns on every page is, in effect, a checksum: each reading is scored against the NFHS‑5 values we computed ourselves, and rows that disagree are flagged regardless of how the vote went. This is what catches a misaligned row, which is otherwise invisible — a transcription can be perfectly legible and still be reading the wrong line.

NFHS‑6 values are nonetheless transcribed, not computed — the reproducibility guarantee that applies to rounds 4 and 5 on this page does not extend to round 6 until its microdata is released.

17Comparing rounds when the boundaries moved

India's administrative map was redrawn between the rounds. A comparison is only meaningful if a name means the same territory on both sides of it — so where a territory's boundaries changed, its earlier figures are recomputed from the microdata rather than patched together from published numbers.

Why published estimates cannot simply be pooled

The tempting shortcut is to build a territory out of the published estimates of its parts — a state from its districts, or a merged UT from the two UTs it replaced — weighting each part by its sample size. That is wrong, and measurably so. A DHS estimate is weighted by $w = \text{V005}/10^{6}$, and those weights restore the population that a stratified, deliberately unequal sample was drawn from. Sample size is not population.

Leh and Kargil are the clearest case: together they are 8.9% of Jammu & Kashmir's households by sample but only 1.8% by weight — small districts are oversampled precisely so they can be reported at all. Reconstructing every NFHS‑4 state from its own districts by sample‑size weighting and comparing against the true state value gives a mean absolute error of 1.1pp, and up to 15.7pp. That is larger than most of the changes this dashboard is trying to measure.

So the two territories that changed are rebuilt at source, from the unit records, with the ordinary survey weights — at national, state, urban–rural, wealth and education level alike, using the same indicator definitions as every other row:

Territory	NFHS‑4 figure used in Compare
Ladakh (UT from 2019)	Recomputed from the NFHS‑4 records of Leh + Kargil, then districts of J&K. Includes mortality and fertility, which have no district estimates to pool.
Jammu & Kashmir	Recomputed with Leh and Kargil removed, so it covers the same 20 districts NFHS‑5 and NFHS‑6 report. Without this, NFHS‑4's J&K would still contain the territory it is being compared against.
Dadra & NH and D&D (merged 2020)	Recomputed as a single territory from the records of both former UTs.

Removing Leh & Kargil moves J&K by 0.2pp on average (largest: sex ratio at birth, 2 points). The correction is small — but it is the difference between comparing like with like and not, and it costs nothing to do properly.

Single-round views These rebuilt territories exist only in Compare. The NFHS‑4 dashboard still shows the 22‑district Jammu & Kashmir, and no Ladakh — because that is the country that was actually surveyed in 2015‑16, and a 2015‑16 map should not be back‑dated with borders that did not yet exist.

Districts

District comparisons match on state + normalised district name, resolving renames and spelling drift, which reconciles 637 of NFHS‑4's 640 districts. The three that remain — Warangal, Barddhaman and Jaintia Hills — were each split in two, and a split is one‑to‑many: neither half has a 2015‑16 baseline of its own, and copying the parent's value onto both halves would manufacture one. They are left out of district comparisons. NFHS‑5 districts with no NFHS‑4 counterpart are shown as new, not as change.

18Updates

A running log of material changes to the data and the dashboard. Newest first.

July 2026 — Dashboard live on theinferenceproject.com.

Band	Rule	Interpretation	Dashboard treatment
ok	\(n \ge 50\)	Reliable	Shown normally
caution	\(25 \le n < 50\)	Interpret with caution	Muted, "small sample" flag
suppress	\(n < 25\)	Unreliable	Hidden / greyed as no‑data