An exploratory participant-level reanalysis of repeated dietary recalls and spot urine osmolality.
Publication type: Research Note
Version: 1.0
Date: 29 September 2026
Institution: The Null Institute
Author: Ilya Nikitin, Ph.D., Founder and Director, The Null Institute
Status: Exploratory analysis; not peer reviewed
Abstract
A single 24-hour dietary recall is often treated as evidence that a person is a “low water consumer,” but day-to-day intake may be noisy. We reanalyzed NHANES 2011–2012 data for 4,134 adults aged 20 years or older who had two reliable dietary recalls, a positive two-day dietary weight, and measured spot urine osmolality. The primary exposure was total moisture from foods, beverages, and water on Day 1 and Day 2.
Intake rank showed moderate persistence. The weighted Pearson correlation between Day 1 and Day 2 total moisture was 0.478 and the unweighted Spearman correlation was 0.527. Among adults in the lowest survey-weighted quartile on Day 1, 50.7% were again in the lowest quartile on Day 2, compared with approximately 25% expected under independence. Of those in the bottom half on Day 1, 69.1% remained in the bottom half on Day 2.
Adults who were in the lowest quartile on both days had more concentrated spot urine than adults who were never in the lowest quartile: weighted median urine osmolality was 697 versus 619 mOsm/kg. Robustness checks using sex-specific quartiles and a fixed approximately 2 L/day threshold produced the same qualitative pattern.
These data support a narrow conclusion: repeated low total-moisture intake identifies a partly stable phenotype associated with more concentrated urine, while a one-day low-intake label contains substantial day-to-day noise. The analysis does not establish that low intake causes disease or that increasing water intake improves health.
Research question
Does a stable low-intake phenotype exist, or does a single 24-hour recall mostly capture day-to-day noise?
Data and analytic population
This exploratory participant-level reanalysis used NHANES 2011–2012.
- Primary intake:
DR1TMOISandDR2TMOIS, total moisture from foods, beverages, and water. - Secondary intake:
DR1_320ZandDR2_320Z, plain water. - Physiology:
URXOAV, spot urine osmolality (mOsm/kg). - Linkage:
SEQN.
Participants were included if they were age 20 or older, had reliable Day 1 and Day 2 dietary recalls (DR1DRSTZ=1, DR2DRSTZ=1), had a positive two-day dietary weight (WTDR2D), and had urine osmolality available. The analytic sample contained 4,134 adults.
Official CDC documentation:
- Day 1 Dietary Total Nutrient Intakes (DR1TOT_G)
- Day 2 Dietary Total Nutrient Intakes (DR2TOT_G)
- Urine Osmolality (UCOSMO_G)
Primary analysis: rank stability between Day 1 and Day 2
Day 1 was collected in person and Day 2 by telephone 3–10 days later. Because the absolute Day 2 intake distribution was lower, each day was divided into its own survey-weighted quartiles using WTDR2D.
Weighted quartile cut points
| Measure | Q1 | Median | Q3 |
|---|---|---|---|
| Day 1 total moisture | 2,119.9 g | 2,840.9 g | 3,863.8 g |
| Day 2 total moisture | 1,889.5 g | 2,554.3 g | 3,403.2 g |
Day-to-day association: weighted Pearson r = 0.478; unweighted Spearman rho = 0.527.
Weighted transition matrix
Rows are Day 1 quartiles and columns are Day 2 quartiles.
| Day 1 → Day 2 | Q1 | Q2 | Q3 | Q4 |
|---|---|---|---|---|
| Q1 | 50.7% | 29.3% | 14.9% | 5.1% |
| Q2 | 25.9% | 32.3% | 27.8% | 14.1% |
| Q3 | 14.7% | 23.7% | 34.1% | 27.5% |
| Q4 | 8.8% | 15.0% | 22.9% | 53.3% |
Among adults in the lowest quartile on Day 1, 50.7% were again in the lowest quartile on Day 2. If the two intake ranks were independent, the expected value would be approximately 25%. 69.1% of adults in the bottom half on Day 1 remained in the bottom half on Day 2. Highest-intake status was similarly persistent: 53.3% of Day 1 Q4 remained Q4 on Day 2.
The interpretation is asymmetric. Intake rank contains a real person-level component, but one 24-hour recall is still noisy: roughly half of people classified as lowest-quartile consumers on one day are not in that quartile a few days later.
Physiological contrast: urine osmolality
| Two-day intake pattern | N | Median Uosm | Mean Uosm | Uosm ≥500 | Uosm ≥800 |
|---|---|---|---|---|---|
| Low quartile both days | 719 | 697 | 670 | 73.5% | 34.0% |
| Low only Day 1 | 550 | 655 | 628 | 67.6% | 26.0% |
| Low only Day 2 | 572 | 644 | 633 | 68.5% | 31.9% |
| Never in low quartile | 2,293 | 619 | 588 | 62.6% | 26.0% |
The group that remained in the lowest quartile on both days had more concentrated urine than the group that was never in the lowest quartile. Among participants classified as low on Day 1, a second low-intake day also identified a group with higher median urine osmolality: 697 versus 655 mOsm/kg. The distributions nevertheless overlap substantially. This result does not establish harm from the stable-low phenotype.
Two-day average
| Two-day average quartile | Median Uosm | Uosm ≥500 |
|---|---|---|
| Q1 | 673 | 71.3% |
| Q2 | 661 | 68.8% |
| Q3 | 601 | 60.5% |
| Q4 | 613 | 60.9% |
The clearest contrast was between the lower half and upper half. Above the median, progressively greater total moisture did not correspond to progressively lower spot urine osmolality in these data.
Robustness checks
Sex-specific lowest quartiles
- Day 1 low → Day 2 low: 51.7% in one sex and 51.3% in the other.
- Stable-low median Uosm: 701 mOsm/kg.
- Never-low median Uosm: 613 mOsm/kg.
Fixed approximately 2 L threshold
Using <2,000 g total moisture on each day:
- both days <2 L: median Uosm 695;
- Day 1 only <2 L: 656;
- Day 2 only <2 L: 647;
- neither day <2 L: 617.
The qualitative pattern was therefore not dependent on the quartile definition.
Secondary analysis: plain water
| Plain-water pattern | Weighted share | Median Uosm |
|---|---|---|
| Zero both days | 8.3% | 679 |
| Zero one day | 17.7% | 672 |
| Positive both days | 74.0% | 624 |
This points in the same direction, but plain-water intake is not equivalent to total water intake because water also comes from other beverages and food.
What the analysis supports
- A low-intake classification has meaningful persistence across two dietary recalls.
- A single recall also contains substantial day-to-day classification noise.
- Repeated low total-moisture intake is associated with more concentrated spot urine.
- Above the median of two-day average total moisture, the relation with spot urine osmolality is not monotonically dose-responsive in this dataset.
What the analysis does not support
This analysis does not show that stable low intake causes disease, that a specific intake target is optimal, or that increasing water intake would improve clinical outcomes.
- Exploratory calculation, not peer reviewed.
- Survey weights were used for point estimates, but full NHANES strata/PSU variance estimation was not performed.
- Urine osmolality is a single spot measurement.
- Day 1 dietary recall was in person, while Day 2 was by telephone 3–10 days later.
- Urine osmolality is temporally closer to Day 1 than Day 2.
- Self-reported 24-hour intake retains measurement error even with two days.
- Kidney concentrating capacity and other physiological factors can affect Uosm independently of intake.
The defensible conclusion is therefore limited: repeated low total-moisture intake is moderately stable and is associated with more concentrated urine, while a one-day “low water” label contains substantial day-to-day noise.
Reproducibility notes
Data availability: Download the machine-readable results (CSV).
Participant-level parsing used these machine-readable mirrors:
- Day 1 dietary:
YvonneJ18/Hypertension,cleaned_file/2011/cleaned_dietary_data.csv, blobe735c58a7d859574b5b6283a076a6d28bf0dcad8. - Day 2 dietary:
YvonneJ18/Hypertension,cleaned_file/2011/cleaned_dietary_data_2.csv, blob1290878442e07b40b4eb9722a26a6bf220167d04. - Demographics:
YvonneJ18/Hypertension,cleaned_file/2011/cleaned_demographic.csv, blob5f06f492b2f0f1b82c176f4da86220d0dbaee8df. - Urine osmolality:
protobi/nhanes-continuous,Laboratory/UCOSMO_G.csv.
The dietary mirrors contain a tiny machine value (5.397605...e-79) in places corresponding to SAS zero values. This does not affect the primary total-moisture analysis because total moisture is strictly positive in reliable recalls. It was treated as zero for the secondary plain-water analysis.
Machine-readable data
A 66-row supplementary CSV has been prepared as null-institute-nhanes-water-intake-stability-data.csv. It contains the reported cut points, the full 4×4 transition matrix, urine-osmolality summaries, robustness checks, and secondary plain-water results. The key values are also reproduced in the tables above, so the public research record does not depend on the attachment.
Research provenance
The analysis was conducted as part of The Null Institute’s exploratory research program and is published to make the underlying calculations inspectable rather than leaving them embedded only inside a narrative article. AI tools assisted data handling, checking, and drafting under human direction. Responsibility for the published interpretation remains with the named author.
Related interpretation
A reader-oriented discussion of these results appears in Mike’s Balance: How Much Water Should You Drink?