Chart 1. Description of the quality indicators of anthropometric data. Children 0-59 months of age. SISVAN, Brazil, 2008-2017.

Quality indicator	Breakdown of indicator	Data used
Completeness
Coverage of the target population, SUS users (%)	Number of children with nutritional status recorded in SISVAN divided by the number of children under 5 years old using the SUS, multiplied by 100. The population using the SUS was obtained by subtracting the total population (estimated by the IBGE) by the population covered by private health insurance (obtained from the ANS).	All children with nutritional status records in SISVAN. A single record per child each year (last follow-up date) was considered for the calculation of coverage.
Coverage of the total population, residents (%)	Number of children with nutritional status records in SISVAN divided by the total number of children under five years old (estimated by IBGE), multiplied by 100.	A single record per child each year (last follow-up date) was considered for the calculation of coverage.
Completeness of birth date (%)	Number of records with complete day, month, and year of birth divided by the total number of child records in SISVAN, multiplied by 100.	All nutritional status follow-up records in SISVAN.
Completeness of anthropometric measurements (%)	Number of records with completed weight and height measurements divided by the total number of child records in SISVAN, multiplied by 100.	All nutritional status follow-up records in SISVAN.
Sex ratio
Sex ratio	Number of boys divided by the number of girls. The expected ratio is that represented by the national population.	All nutritional status follow-up records in SISVAN.
Age distribution
Histograms of age (in years, months and month of birth)	Histograms were used to assess the pattern of age distribution in months, years, and month of birth. An approximately uniform distribution is expected. The age of the child was calculated from the date of birth and date of nutritional status monitoring.	All records with complete birth month and year.
Index of dissimilarity for age (in months)	The Myers “unblended” index was analyzed, according to the formula below, to identify the percentage of records deviating from a uniform distribution of age in months (0-59) $\frac{\sum _{i=1}^{60}|X_{is}-X_{ie}|}{2}$ in which Xis is the observed percentage and Xie is the expected percentage. The index ranges from 0 to 90, being 0 the optimum value.	All records with complete birth month and year.
Preference for height and weight digits
Histograms of terminal digits and whole numbers for height (cm) and weight (kg)	Histograms were used to assess the pattern of distribution of terminal digits and whole numbers of height and weight. An approximately uniform distribution is expected.	All records with weight and/or height information.
Index of dissimilarity for terminal digits of height (cm) and weight (kg)	The Myers “unblended” index was analyzed, according to the formula below, to identify the percentage of records deviating from a uniform distribution for terminal digits of height and weight $\frac{\sum _{i=1}^{10}|X_{is}-X_{ie}|}{2}$ in which Xis is the observed percentage and Xie is the expected percentage. The index ranges from 0 to 90, being 0 the optimum value.	All records with weight and/or height information.
Implausible z-score values
Implausible z-scores for each anthropometric index (%)	Number of records with implausible z-score values divided by the total number of child records in SISVAN, multiplied by 100. Using the WHO macro “STATA igrowup package” flagging system, implausible z-score values were detected according to plausibility criteria (WHO, 1995): H/A (-6, +6), W/H (-5, +5), W/A (-6, +5) and BMI/A (-5, +5). Percentages above 1% is indicative of poor data quality.	All records with date of birth, weight and/or height information.
Dispersion of z-score values
Standard deviation of plausible z-score values for each anthropometric index	Standard deviation was calculated using the following formula $\sqrt{\frac{\sum _{i=1}^n{(Y_i-\overline{Y})}^2}{n-1}}$ in which n is the total number of observations, Yi is each value in the database, and Yˉ is the mean of observations	All records with biologically plausible z-score values for the anthropometric index of interest.
Normality of z-score values
Distribution of plausible z-score values for each anthropometric index	Kernel density plots were used to examine the distribution pattern of the z-score values for each index.	All records with biologically plausible z-score values for the anthropometric index of interest.
Skewness of plausible z-score values for each anthropometric index	Skewness was calculated by the Fisher-Pearson coefficient: $\frac{\sum _{i=1}^n{(Y_i-\overline{Y})}^3/n}{s^3}$ in which Yˉ = mean, s = standard deviation (calculated with ‘n’ in the denominator instead of n-1) and n = sample size. It is generally accepted that a coefficient <-0.5 or >0.5 indicates skewness	All records with biologically plausible z-score values for the anthropometric index of interest.
Kurtosis of plausible z-score values for each anthropometric index	Kurtosis was calculated by the Fisher-Pearson coefficient: $\frac{\sum _{i=1}^n{(Y_i-\overline{Y})}^4/n}{s^4}$ in which Yˉ = mean, s = standard deviation (calculated with ‘n’ in the denominator instead of n-1) and n = sample size. In general, it is accepted that a coefficient <2 or >4 indicates kurtosis.	All records with biologically plausible z-score values for the anthropometric index of interest.

SISVAN: Food and Nutrition Surveillance System; SUS: Unified Health System; IBGE: Brazilian Institute of Geography and Statistics; ANS: National Supplementary Health Agency; H/A: Height/Age; W/H: Weight/Height; W/A: Weight/Age; BMI/A: Body Mass Index/Age.