Paediatrica Indonesiana
p-ISSN 0030-9311; e-ISSN 2338-476X; Vol.65, No.5 (2025). p.416-21; DOI: https://doi.org/10.14238/pi65.5.2025.416-21

Original Article

Target height prediction in Indonesian children:
a population-based and clinically relevant model
Annang Giri Moelyo, Hanum Ferdian

Abstract

Background The traditional calculation of target height (TH)
overlooks two important factors: the assortative mating correlation, which reflects the tendency for people of similar height to
partner, and the parental-offspring correlation, which measures
the strength of the relationship between parents’ and children’s
heights. A more accurate model is needed for the Indonesian
population.
Objective To develop a target height (TH) prediction model
for Indonesian children and to compare its performance with
traditional formulas.
Methods This retrospective study used nationally representative data from the Indonesia Family Life Survey (IFLS). Adult
height data from the IFLS-5 (2014) for 2,506 subjects and the
corresponding parental height data from the IFLS-3 (2000)
were analyzed. We used a new model, namely estimated target
height (eTH), which combined the Hermanussen-Cole and
van Dommelen methods, to estimate each participant’s TH. This
new model was compared to traditional models which used Tanner
(TTH) and modified Tanner (mTH) formulas.
Results The new eTH model yielded the following formulas:
TH (boys)=0.36 × father’s height + 0.43 × mother’s height +
42.77; and TH (girls)=0.30 × father’s height + 0.36 × mother’s
height + 50.47. Correlations with the observed adult height were
highest for the new model (r=0.528 for boys; r=0.534 for girls),
compared to traditional models.
Conclusion This study provides a locally validated model
for TH estimation in Indonesian children that demonstrates
improved clinical applicability over traditional formulas. [Paediatr
Indones. 2025;65:416-21; DOI: https://doi.org/10.14238/
pi65.5.2025.416-21 ].
Keywords: target height; growth prediction;
Indonesia; Hermanussen–Cole method; child growth
modeling

T

arget height (TH) is a widely used clinical
indicator to evaluate whether a child’s
growth aligns with their genetic potential.
It also serves as a reference point for
determining the appropriateness of therapeutic
interventions in children with short stature.
Traditionally, TH has been calculated using the
mid-parental height method, most commonly via
the Tanner formula (TTH), in which a sex-specific
adjustment is applied to the average parental height.1,2
In Indonesia, a modified version of the TTH
(mTH) is often applied in clinical practice. In this
formula, a 6.5-cm addition for boys and a 6.5-cm
subtraction for girls is used, depending on the child’s
sex, along with a ±8.5 cm range for interpreting normal
growth expectations.3 However, increasing evidence
suggests that the mTH may underestimate actual adult
height, particularly in Asian populations.4-7
Foundational work by Hermanussen and Cole
introduced a revised model that incorporates two
key correlations: parental-offspring resemblance

Department of Child Health, Faculty of Medicine, Universitas Sebelas
Maret/Dr. Moewardi Hospital, Surakarta, Central Java, Indonesia.
Corresponding author: Annang Giri Moelyo. Department of Child
Health, Faculty of Medicine, Universitas Sebelas Maret/Dr. Moewardi
Hospital, Jalan Kolonel Sutarto No. 132, Jebres, Surakarta, Central Java,
Indonesia. Email: annanggm73@gmail.com.
Submitted MAY 11, 2025. Accepted November 12, 2025.

416 • Paediatr Indones, Vol. 65, No. 5, September 2025

Annang Giri Moelyo et al.: A population-based and clinically relevant model of target height prediction in Indonesian children

and assortative mating. This model expresses TH in
terms of standard deviation scores (SDS) and offers
improved predictive power.8 Despite the model’s
theoretical strength, the use of SDS is often seen
as impractical in routine clinical settings, where
estimates in centimeters are preferred. Hughes et
al.9 emphasized this limitation, which prompted the
development of conversion techniques.
Van Dommelen et al.10 proposed a stepwise
approach to convert TH SDS to centimetric values to
allow for broader clinical implementation. However,
this approach has not yet been evaluated or calibrated
for Indonesian children. The absence of localized TH
prediction models limits the precision of pediatric
growth assessments in the Indonesian context.
The aim of our study was, thus, to construct and
validate a TH prediction model for Indonesian children
by applying the Hermanussen-Cole (HC) method8 and
the van Dommelen conversion framework.10 We also
compared the predictive performance of the new
model against traditional TTH and mTH formulas
to evaluate its clinical relevance and accuracy in the
local population.1-3

Methods
In this study, we utilized data from the Indonesia
Family Life Survey (IFLS), which is a nationally
representative, longitudinal household survey
conducted by the RAND Corporation in collaboration
with Universitas Indonesia and Universitas Gadjah
Mada.11 The IFLS data are freely available from
the Institute of Demography, Faculty of Economics
and Business, Universitas Indonesia for IFLS-1 and
IFLS-2, which were conducted in 1993 and 1997,
respectively;12 and a from private Indonesian research
agency, Surveymeter, for IFLS-3, IFLS-4, and IFLS-5,
which were conducted in 2000, 2007, and 2014,
respectively. 13-15 The IFLS surveys and their
procedures have been reviewed and approved by an
institutional review board (IRB) in the United States
(RAND Human Subjects Protection Committee). In
Indonesia, the survey protocols have been cleared
by the IRB of Universitas Gadjah Mada for IFLS-3,
IFLS-4, and IFLS-5 and Universitas Indonesia for
IFLS-1 and IFLS-2.16 All methods were carried out in
accordance with relevant guidelines and regulations.

All participants in the IFLS survey had given written
informed consent prior to data collection.17 Stratified
multistage sampling design across 13 provinces was
employed in the IFLS to cover approximately 83%
of the Indonesian population. The dataset is publicly
available and was accessed in October 2019.11-15
Data from IFLS 5 (2014) were used to predict the
adult height of subjects aged 19-25 years, while the
corresponding parental height data, which were
collected during the participants’ childhoods, were
extracted retrospectively from IFLS 3 (2000). This
approach helped minimize potential bias due to stature
shrinkage in older adults. Data were processed and
analyzed from January to March 2025.
The final sample included 2,506 individuals
(1,195 males and 1,311 females) who had complete
records of their adult height and the heights of both
parents. Subjects were aged between 19 and 25
years at the time of the IFLS-5 measurement, which
reflected the attainment of their full adult height. The
paternal and maternal heights were drawn from their
IFLS-3 records to avoid age-related bias.
The TH was estimated using three different
approaches: the HC-van Dommelen method
(new model or eTH formula), the original TTH
formula, and the mTH formula. The new model
(eTH, estimated Target Height) incorporated both
the parent–offspring correlation (r=0.57) and the
assortative mating correlation (r=0.27) as follows:8
eTH SDS in centimeters =
TH SDS × r(P,O) × √[2/(1+r(P,P)] + error
[eTH=estimated target height; r(P,O)=parentoffspring
correlation;
r(P,P)=assortative
mating correlation; TH SDS=(TH - mean height)/
standard deviation]

Conversion from the TH SDS to a centimeterbased TH (eTH) was conducted by following the
procedure outlined by van Dommelen et al., 10
in which population-specific height means and
standard deviations were used. The new model (eTH)
calculated the actual mean height of subjects, hence
the possibility of a secular trend would be considered.
In comparison, TTH was calculated using the
formula (father’s height + mother’s height)/2 +

Paediatr Indones, Vol. 65 No. 5, September 2025 • 417

Annang Giri Moelyo et al.: A population-based and clinically relevant model of target height prediction in Indonesian children

6.5 cm in males and (father’s height + mother’s
height)/2 − 6.5 cm in females.1,2 In mTH, a sexspecific adjustment based on the observed final
height differences was applied: +6.15 cm for boys
and −6.15 cm for girls.3 Based on the two formulas
(TTH and mTH), an actual mean height of subjects
was not included, hence the secular trend could not
be considered.
Descriptive statistics were calculated for all
the relevant anthropometric variables. Pearson’s
correlation was used to examine the correlation
between actual adult height and each of the TH
estimates (eTH, TTH, and mTH) as well as individual
parental heights. Linear regression models were
applied to determine the variances in adult height
explained by each formula. The mean difference
between the predicted TH and actual adult height was
also calculated. Statistical analyses were performed
using STATA/SE 16.1 software (StataCorp LLC,
College Station, Texas, USA).

Results
Table 1 provides a summary of baseline anthropometric
characteristics of subjects. Mean adult height was
165.4 (SD 6.2) cm for males and 152.9 (SD 5.2) cm
for females. The observed mean intergenerational
height difference between father and son was 4.1 cm,
while the difference between mother and daughter was
2.7 cm, which indicated a secular increase in height
across generations.
The eTH formulas derived from IFLS data were
as follows:
• Males: eTH in centimeters (cm) = 0.36 × father’s
height in cm + 0.43 × mother’s height in cm +
42.77

•

Females: eTH in cm = 0.30 × father’s height
in cm + 0.36 × mother’s height in cm + 50.47
Using this model, the 95% confidence interval
(95%CI) for the TH estimates corresponded to ±1.64
SDS, which was equivalent to approximately ±10.2
cm in males and ±8.5 cm in females.
Table 2 displays the correlation coefficients
between subjects’ adult heights, the TH estimates
(eTH, TTH, and mTH), and the parental heights.
The highest correlations were observed between the
final adult heights and the estimated THs derived
from the new formula (eTH). These correlations were
marginally higher than those observed using the TTH
and mTH formulas.
Compared to the TTH and mTH approach,
mean difference between estimated and observed final
heights was the smallest using the new eTH model
(Table 3). In contrast, adult height was overestimated
by 3 to 3.6 cm using the TTH and mTH formulas. We
then performed subgroup analysis to test the internal
consistency of eTH, with subjects split into those who
lived in urban and rural areas. The mean difference
in urban subjects was 0.30 (5.24) cm in males and
0.27 (4.35) cm in females. In rural subjects, the mean
difference was -1.15 (4.42) cm in males and -0.55
(5.24) cm in females. Linear regression models showed
that the new eTH model accounted for 27.8% of the
height variance in males and 28.5% in females, and
that these variances were slightly higher than those of
the TTH and mTH formulas (27.4% each).

Discussion
In this study, we present a novel, population-specific
approach to estimating TH in Indonesian children

Table 1. Anthropometric characteristics of the study population
Variables

Males
(n=1,195)

Females
(n=1,311)

Total
(N=2,506)

Median age (range), years

22 (19-25)

22 (19-25)

22 (19-25)

Mean final height (SD), cm

165.4 (6.2)

152.9 (5.2)

158.8 (8.4)

Mean father’s height (SD), cm

161.3 (6.2)

161.3 (6.2)

161.3 (6.2)

Mean mother’s height (SD), cm

150.2 (5.3)

150.3 (5.2)

150.2 (5.2)

Mean FH – MH (SD), cm

11.1 (7.5)

11.0 (7.4)

11.0 (7.4)

-

-

12.5 (0.2)

Mean male-female height gap
FH=father’s height; MH=mother’s height

418 • Paediatr Indones, Vol. 65, No. 5, September 2025

Annang Giri Moelyo et al.: A population-based and clinically relevant model of target height prediction in Indonesian children

based on the HC model,8 which was operationalized
through the van Dommelen conversion framework.10
The newly derived formula accounts for assortative
mating and parent-offspring height correlations to
enhance both biological plausibility and statistical
robustness. By expressing the results in centimeters,
the model bridges the gap between advanced
statistical modeling and routine clinical practice,
where centimeter-based metrics are more intuitive
for interpretation.
A key strength of this study was the use of
nationally representative data from the IFLS, which
provided a demographically diverse sample. The
resulting model, which we call estimated target height
(eTH), showed superior performance when compared
with conventional methods, such as the TTH formula
and its modified version.1-3 The new model (eTH
formula) yielded the strongest correlation with the
observed adult heights in both the males (r=0.528)
and females (r=0.534), thereby slightly outperforming
the TTH and mTH methods, with correlations
hovering around 0.52. These findings suggest the
model’s improved predictive power, particularly for
female subjects, which is a known limitation of earlier
Asian population models.7
Our results affirmed prior critiques of the
limitations of the TTH model, particularly in nonWestern settings. Previous studies in Korea and
India have revealed the consistent underestimation
of final adult height when using the TTH formula.6,7
Table 2. Coefficient correlation of the subjects’ final heights
with the parental heights and TH estimates
Correlation

Males

Females

Father’s height

0.379

0.300

Mother’s height

0.500

0.459

Estimated TH (eTH)

0.528

0.534

Tanner TH (TTH)

0.524

0.527

Modified TH (mTH)

0.524

0.527

TH=target height

For instance, a study found significant discrepancies
between the observed and Tanner-predicted adult
heights in Korean adolescents,6 while another study
reported similar findings in Asian Indian populations,
particularly among girls.7 In contrast, our Indonesianspecific model demonstrated more balanced accuracy
across sexes, with narrower 95% confidence intervals
(±10.2 cm for boys, ±8.5 cm for girls) than those
reported in Australian and Dutch studies.9,10 Our new
model showed greater accuracy in males than females
(Table 3), with a discrepancy between estimated
and observed adulg height of only -0.02 cm in males,
compared to -1.17 cm in females. A more pronounced
secular trend in males may explain this difference.
Notably, our model achieved similar performance
in explaining the variance in both sexes: 27.8% in
males and 28.5% in females. In earlier studies, this
variance often differed markedly by sex; for example,
in an Indian study, the TH explained 29.7% of the
variance in females, but only 16.7% in males.7 The
improved balance in our findings likely reflects the
benefits of incorporating the correlations into the
model, as originally advocated by Hermanussen and
Cole.8
The observed intergenerational height increases
of 4.1 cm in boys and 2.7 cm in girls underscore
an ongoing secular trend and align with previous
findings in the Javanese population. 18 Similar
secular gains have been observed across various
low- and middle-income countries, which suggests
environmental improvements in nutrition, healthcare,
and education. However, such trends may confound
static TH prediction models and potentially lead to
the misclassification of normal growth acceleration
as pathological. While our model does not explicitly
adjust for secular trends, its retrospective design (using
childhood-era parental data) helps mitigate this bias,
as advised in prior literature.9
This study offers the first validated TH prediction
model for the Indonesian pediatric population that

Table 3. Comparison of TH formulas: accuracy and explained variance
Formula

Mean difference (SD), cm
Males

Females

eTH (new)

−0.02 (0.15)

Tanner TH (TTH)

3.00 (5.41)

Modified TH (mTH)

3.36 (5.41)

Explained variance, %
Males

Females

−1.17 (0.12)

27.8

28.5

3.65 (4.70)

27.4

27.7

3.30 (4.70)

27.4

27.7

Paediatr Indones, Vol. 65 No. 5, September 2025 • 419

Annang Giri Moelyo et al.: A population-based and clinically relevant model of target height prediction in Indonesian children

incorporates population-specific references and
genetic theory. Unlike previous models derived from
Western or East Asian data, our formula reflects local
anthropometric distributions and growth dynamics,
which improves its clinical relevance.
The precision and reliability of our method hold
particular utility in evaluating children with growth
concerns. Compared to the TTH model, which
consistently overestimated height by approximately
3 cm in our sample, the eTH method yielded a nearzero mean prediction error, especially in males, and
lower SD. This enhances its potential for the early
identification of pathological growth deviations.
By applying this model to the IFLS cohort,
our study provides a replicable framework for
future updates as secular trends evolve. This
flexibility is critical for ensuring long-term clinical
applicability in a rapidly changing epidemiological
and socioeconomic landscape. The present model
offers clear advantages in both clinical interpretation
and methodological rigor. Its derivation from a large,
nationally representative sample enhances external
validity. The reduced mean error, high correlation with
final height, and operational simplicity strengthen the
case for its adoption in routine pediatric evaluations
in Indonesia.
This study had some limitations. Although we
attempted to reduce bias by using earlier height data
from IFLS-3, residual shrinkage-related inaccuracies
may have persisted. Additionally, the dataset reflects
conditions up to 2014 and thus may not have captured
more recent shifts in child growth norms. The timing
and duration of puberty, which are known to influence
final height, were not included in the current model.5
Incorporating these variables in future iterations of the
model could further improve its predictive accuracy.
Although eTH was the more accurate formula,
using absolute values in cm in the final calculation
may reduce the accuracy, as this cannot account for
secular trends. Nonetheless, the consistency of the
results within a large, representative national dataset
enhances the model’s internal validity.
We found that the correlation between male
adult height and mother’s height was almost as high
as that between male adult height and target height
derived from both parents; further explanation is
needed for this finding. A previous study has proposed
a formula for estimation of adult height when the

420 • Paediatr Indones, Vol. 65, No. 5, September 2025

father’s height is unknown.10 Further studies are
needed to estimate target height based on mother’s
height only. In our study, the correlation of assortative
mating and parent-offspring height were based on
Dutch data. Future studies should investigate such
associations in the Indonesian population.
Our results present a refined and populationspecific model for TH prediction in Indonesian
children and offer improved clinical utility and
predictive accuracy over traditional formulas. By
incorporating genetic correlations through the HC
method and using population-relevant conversion
steps, the resulting equations better reflect actual
adult height outcomes across sexes. These findings
affirm the value of recalibrating global pediatric
tools using local anthropometric data. The new
TH estimation model has the potential to enhance
clinical decision-making in pediatric endocrinology
across Indonesia and could serve as a template for
similar adaptations in other low- and middle-income
countries.

Conflict of interest
None declared.
Funding acknowledgment
The authors received no specific grants from any funding agency
in the public, commercial, or not-for-profit sectors.

References
1.

2.

3.

Tanner JM, Goldstein H, Whitehouse RH. Standards for
children’s height at ages 2-9 years allowing for height of
parents. Arch Dis Child. 1970;45:755-62. DOI: https://doi.
org/10.1136/adc.45.244.755
Tanner JM. Normal growth and techniques of growth
assessment. Clin Endocrinol Metab. 1986;15;411-51. DOI:
https://doi.org/10.1016/s0300-595x(86)80005-6
Fredriks AM, van Buuren S, Burgmeijer RJF, Meulmeester JF,
Beuker RJ, Brugman E, et al. Continuing positive secular
growth change in the Netherlands 1955-1997. Pediatr Res.
2000;47;316-23. DOI: https://doi.org/10.1203/00006450200003000-00006

Annang Giri Moelyo et al.: A population-based and clinically relevant model of target height prediction in Indonesian children
4.

Zeevi D, Ben Yehuda A, Nathan D, Zangen D, Kruglyak L.
Accurate prediction of children’s target height from their
mid-parental height. Children. 2024;11;916. DOI: https://
doi.org/10.3390/children11080916
5. Luo ZC, Albertsson-Wikland K, Karlberg J. Target height
as predicted by parental heights in a population-based
study. Pediatr Res. 1998;44;563-71. DOI: https://doi.
org/10.1203/00006450-199810000-00016
6. Kim S, Yoo JH, Chueh HW. Tanner’s target height formula
underestimates final adult height in Korean adolescents and
young adults: reassessment based on the Korean National
Health and Nutrition Examination Survey 2010-2019.
BMJ Paediatr Open. 2024;8:e002653. DOI: https://doi.
org/10.1136/bmjpo-2024-002653
7. Atluri S, Bharathidasan K, Sarathi V. Tanner’s target height
formula underestimates final height in Asian Indians - A
cross-sectional observational study. Indian J Endocrinol
Metab. 2018; 22:441-4. DOI: https://doi.org/10.4103/ijem.
ijem_92_18
8. Hermanussen M, Cole J. The calculation of target height
reconsidered. Horm Res. 2003; 59:180-3. DOI: https://doi.
org/10.1159/000069321
9. Hughes IP, Davies PS. Proposed new target height equations
for use in Australian growth clinics. J Paediatr Child Health.
2008; 44:613-7. DOI: https://doi.org/10.1111/j.14401754.2008.01397.x
10. van Dommelen P, Schönbeck Y, van Buuren S. A simple
calculation of the target height. Arch Dis Child. 2012; 97:182.
DOI: https://doi.org/10.1136/archdischild-2011-301095
11. RAND Social and Economic Well-being. IFLS Data and

Documentation. [cited 2025 Jan 12]. Available from: https://
www.rand.org/well-being/social-and-behavioral-policy/data/
FLS/IFLS/download.html
12. RAND Social and Economic Well-being. Community-Facility
Surveys: IFLS1 and IFLS2. [cited 2025 Jan 12]. Available
from: https://www.rand.org/well-being/social-and-behavioralpolicy/data/FLS/IFLS/cf.html
13. Strauss J, Beegle K, Sikoki B, Dwiyanto A, Herawati Y,
Witoelar F. The third wave of the Indonesia Family Life
Survey (IFLS): overview and field report. 2004;1. WR-144/1NIA/NICHD.
14. Strauss J, Witoelar F, Sikoki B, Wattie AM. The fourth wave
of the Indonesian family life survey (IFLS4): overview and
field report. 2009;1. WR-675/1-NIA/NICHD.
15. Strauss J, Witoelar F, Sikoki B. The fifth wave of the Indonesia
family life survey (IFLS5): Overview and Field Report.
2016;1. WR-1143/1-NIA/NICHD.
16. RAND Social and Economic Well-being. IFLS Data Updates,
Data Notes, Tips, and FAQs. [cited 2025 Jan 13]. Available:
https://www.rand.org/well-being/social-and-behavioralpolicy/data/FLS/IFLS/datanotes.html#ethical
17. Strauss J, Witoelar F, Sikoki B. Household Survey
Questionnaire for the Indonesia Family Life Survey, Wave
5. [cited 2025 Jan 13]. Available from: https://www.rand.
org/content/dam/rand/pubs/working_papers/WR1100/
WR1143z3/RAND_WR1143z3.pdf
18. Moelyo AG, Sitaresmi MN, Julia M. Secular trends in
Javanese adult height: The roles of environment and
educational attainment. BMC Public Health. 2022; 22:712.
DOI: https://doi.org/10.1186/s12889-022-13144-6

Paediatr Indones, Vol. 65 No. 5, September 2025 • 421