Journal of Dental Research

http://jdr.sagepub.com/

Building Caries Risk Assessment Models for Children

X.-L. Gao, C.-Y.S. Hsu, Y. Xu, H.B. Hwarng, T. Loh and D. Koh
J DENT RES 2010 89: 637 originally published online 16 April 2010
DOI: 10.1177/0022034510364489

The online version of this article can be found at:

http://jdr.sagepub.com/content/89/6/637

Published by:

http://www.sagepublications.com

On behalf of:
International and American Associations for Dental Research

Additional services and information for Journal of Dental Research can be found at:

Email Alerts: http://jdr.sagepub.com/cgi/alerts

Subscriptions: http://jdr.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

>> Version of Record - May 14, 2010

OnlineFirst Version of Record - Apr 16, 2010

What is This?

Downloaded from jdr.sagepub.com at University of Alabama at Birmingham on January 7, 2013 For personal use only. No other uses without permission.

© 2010 International & American Associations for Dental Research

 RESEARCH REPORTS
Clinical

X.-L. Gao1, C.-Y.S. Hsu2*, Y. Xu3,

H.B. Hwarng4, T. Loh2, and D. Koh5
Building Caries Risk Assessment
1
Dental Public Health, Faculty of Dentistry, University of
Models for Children
Hong Kong; 2Department of Preventive Dentistry, Faculty of
Dentistry, National University of Singapore, 5 Lower Kent
Ridge Road, Republic of Singapore 119074; 3Department of
Information Systems, School of Computing, National
University of Singapore, and School of Management, Fudan
University; 4Department of Decision Sciences, School of
Business, National University of Singapore; and 5Department
of Epidemiology and Public Health, Yong Loo Lin School of
Medicine, National University of Singapore; *corresponding
author, pndhsus@nus.edu.sg

J Dent Res 89(6):637-643, 2010

Abstract
Despite the well-recognized importance of caries risk
assessment, practical models remain to be established.
This study was designed to develop biopsychosocial mod-
els for caries risk assessment in various settings. With a
questionnaire, an oral examination, and biological (sali-
vary, microbiological, and plaque pH) tests, a prospective
study was conducted among 1782 children aged 3-6 years,
with 1576 (88.4%) participants followed in 12 months.
Multiple risk factors, indicators, and protective factors
were identified. Various risk assessment models were
constructed by the random selection of 50% of the cases
and further validated in the remaining cases. For the pre-
diction of a “one-year caries increment”, screening mod-
els without biological tests achieved a sensitivity/
specificity of 82%/73%; with biological tests, full-blown
models achieved the sensitivity/specificity of 90%/90%.
For identification of a quarter of the children with high
caries burden (baseline dmft > 2), a community-screening
model requiring only a questionnaire reached a sensitiv-
ity/specificity of 82%/81%. These models are promising
tools for cost-effective caries control and evidence-based
treatment planning. Abbreviations: decayed, missing,
filled teeth in primary dentition (dmft); receiver operation
characteristics (ROC); relative risk (RR); confidence
interval (CI); National Institutes of Health (NIH); World
Health Organization (WHO); US Department of Health
and Human Services (US/DHHS); American Academy of
Pediatric Dentistry (AAPD).
KEY WORDS: caries risk assessment, biopsy-
chosocial model, preschoolers.
DOI: 10.1177/0022034510364489
Received November 14, 2008; Last revision December 14,
2009; Accepted January 19, 2010
A supplemental appendix to this article is published elec-
tronically only at http://jdr.sagepub.com/supplemental.
Introduction
D espite the great strides made in caries control in developed countries,
caries remains the most common chronic childhood disease and a major
financial burden on society (WHO, 2006). In recent years, caries in the pri-
mary dentition has been increasing in the USA (US/DHHS, 2007), UK (Pitts
et al., 2006), Canada (Speechley and Johnston, 1996), Australia (Armfield
and Spencer, 2004), Norway (Haugejorden and Birkeland, 2002), and the
Netherlands (Truin et al., 1993). These epidemiological signs point to an
urgent need for effective caries control among children.
A skewed distribution of caries has been observed in many developed
countries, with 25% of children bearing 75-80% of affected surfaces (Kaste
et al., 1996; Pitts et al., 2006). As concluded in a National Institutes of Health
(NIH) conference, caries prevention should be timely targeted at high-risk
individuals (NIH consensus panel, 2001). Since cumulative evidence has
linked early childhood caries with caries in the permanent dentition (Skeie
et al., 2006; Alm et al., 2007), early and accurate selection of high-risk pre-
schoolers through caries risk assessment for prevention and intervention is of
great importance for cost-effective caries control. In the clinical setting, caries
risk assessment is an essential element of diagnosis, constituting a basis for
evidence-based treatment planning (Featherstone et al., 2003).
Considerable efforts in past decades have identified multiple caries risk fac-
tors, indicators, and protective factors (Hunter, 1988; Graves et al., 1991; Harris
et al., 2004). Multifactorial modeling has been used to increase risk assessment
accuracy; however, few models have met the requirement for a useful model,
i.e., sensitivity + specificity > 160% (Stamm et al., 1988; Zero et al., 2001).
Several conceptual models have been suggested by professional organiza-
tions, such as the Caries-risk Assessment Tool proposed by the American
Academy of Pediatric Dentistry (AAPD, 2002) and the Caries Management
by Risk Assessment program advocated by the California Dental Association
(Featherstone et al., 2003). However, the accuracy of these models has not
been demonstrated. A computerized program, Cariogram, has been developed
to streamline the caries risk assessment process, with multiple weighed fac-
tors and interactions (Bratthall, 1996). Although Cariogram has been satisfac-
torily validated among Swedish schoolchildren (Petersson et al., 2002) and
the elderly (Petersson et al., 2003), its validity in preschoolers was unsatisfac-
tory (Holgerson et al., 2009).
Currently, the clinical and community practice of caries risk assessment
is hampered by the lack of a model with sufficient simplicity and proven

637
Downloaded from jdr.sagepub.com at University of Alabama at Birmingham on January 7, 2013 For personal use only. No other uses without permission.

© 2010 International & American Associations for Dental Research

638
accuracy (Zero et al., 2001). Empirical estimation, which is not
reliable, remains the main way for clinicians to judge the caries risk
for formulating preventive and treatment plans (Zero et al., 2001).
The aim of this population-based prospective study was to
identify caries risk factors, indicators, and protective factors and
to develop and validate biopsychosocial caries risk assessment
models for preschoolers.
Materials & Methods
Participants
The participants in this study were recruited from the govern-
ment kindergartens in Singapore, with the ethical approval of
the National University of Singapore Institutional Review Board
and parents’ informed consent (Gao et al., 2009). Only one of
the 14 randomly selected kindergartens refused to participate.
All nursery and Kindergarten Grade-1 children aged 3-6 yrs
were approached, including those with compromised health
conditions and special needs.
Data Collection
Questionnaire Survey
We used a parent-administered questionnaire pre-tested in our
previous studies to collect information on (a) children’s demo-
graphic background (age, gender, race, and country of birth), (b)
socio-economic status (parents’ education attainment and hous-
ing condition), (c) children’s oral health practice (feeding histo-
ries, diet habits, oral hygiene measures, fluoride applications,
and dental attendance), (d) systemic diseases, and (e) parental
knowledge and attitudes on oral health.
Caries Examination and Oral Hygiene Evaluation
All participants were examined by the same examiner, who was
trained and calibrated against an experienced pediatric dentist
with representative cases. The participants were examined while
seated in a portable dental chair with a fiber-optic light. World
Health Organization procedures and diagnostic criteria (WHO,
1997) were followed for the caries examination. The tooth status
was mainly assessed by visual inspection, aided by tactile
inspection if necessary. A disposable probe, instead of a CPI
probe, was used prudently to avoid damaging sound enamel
surfaces. No radiographs were taken. A dmft score was calcu-
lated for each child. The oral hygiene status was evaluated based
on a modified Silness-Löe Plaque Index (Silness and Löe,
1964), with 6 index teeth (1E, 1B, 2D, 3E, 3B, and 4D). For
monitoring intra-examiner reliability, duplicate examinations
were carried out for 10% randomly selected children, with at
least 10 other children examined between duplicate examina-
tions for each selected child.
Biological Tests
The plaque pH at 6 sites (mesial surfaces of right-upper and left-
lower central incisors and first molars in each quadrant) was
measured with a microelectrode set (Beetrode®, World Precision
Instruments Inc., Sarasota, FL, USA). Stimulated saliva flow rate
was measured after the children chewed for 5 min on paraffin
Downloaded from jdr.sagepub.com at University of Alabama at Birmingham on January 7, 2013 For personal use only. No other uses without permission.
© 2010 International & American Associations for Dental Research
Gao et al. J Dent Res 89(6) 2010
pellets. Saliva buffering capacity and levels of mutans strepto-
cocci and Lactobacilli were assessed with Dentobuff® Strip,
Dentocult® SM Strip mutans, and Dentocult® LB kits (Orion
Diagnostica, Espoo, Finland), respectively. A trained dental nurse
and an assistant completed the biological sample collection.
Follow-up of Caries Status
The caries examination was repeated after 12 mos with the same
criteria and procedures. The one-year caries increment (Δdmft)
was calculated. The information was also collected on dental
care received by the child in this 12-month period.
Statistical Analysis
Data were analyzed with the Statistical Package for the Social
Sciences (version 15). We used multiple stepwise logistic
regressions for identifying risk factors, indicators, and protec-
tive factors and for constructing risk assessment models. Data
from 50% randomly selected children were used for model
construction, with data from the remaining children for model
validation.
The independent variables (Appendix) fell into 5 categories,
including demographic, socio-economic, behavioral, clinical
(oral hygiene and baseline caries experience), and biological
variables. Multicollinearity was checked before modeling. For
various community and clinical applications, 5 models with dif-
ferent cost implications were constructed (Table 1). While pre-
diction models utilize all the potential factors to identify the
at-risk children, risk models focus on the modifiable etiological
factors conducive to tailoring preventive and treatment strate-
gies. Theoretically, risk models have greater external validity
and might be applicable in different populations (Bratthall and
Petersson, 2005).
The children’s caries risk was also assessed with Cariogram
(Table 2). The performances of Cariogram and models con-
structed in this study were evaluated by Receiver Operation
Characteristics (ROC) analysis.
Results
In the 13 participating kindergartens, 1782 children (889 males
and 893 females) were recruited, with a response rate of 86.3%.
The mean age was 4.8 yrs (age range, 3.6-5.7 yrs). The question-
naire was completed by the parents of 1754 (98.4%) children.
After 12 mos, 1576 (88.4%) children were followed up. During
this 12-month period, only 62 (3.5%) children visited dentists,
receiving restorations or extraction.
The intra-examiner reliability was high (Kappa = 0.958 for
caries examination; Intra-class Correlation Coefficient = 0.946
for oral hygiene evaluation). No significant difference was
found between lost-to-follow-up children and those followed
up, in their demographic and socio-economic profiles and base-
line caries status (all p > 0.05).
Caries Risk Factors, Indicators, and Protective Factors
At baseline, 40.3% of children were affected by caries. The
mean (SD) dmft was 1.57 (2.79), with 90% of the affected teeth
J Dent Res 89(6) 2010 Caries Risk Assessment Models for Children   639

Table 1. Various Models Constructed

Models Constructed**

Prediction Model Risk Model

Community-screening
Screening Full-blown Screening Full-blown Model
Outcome variable
1-year increment (∆dmft √ √ √ √
> 0 or = 0)
Baseline dmft > 2 or ≤ 2 √
Independent variables*
Features All features All features Selected features Selected features All features
Categories Demographic Demographic Demographic Demographic Demographic
Socio-economic Socio-economic Socio-economic Socio-economic Socio-economic
Behavioral Behavioral Behavioral Behavioral Behavioral
Clinical Clinical Clinical Clinical
Biological Biological
Source of information Questionnaire Questionnaire Questionnaire Questionnaire Questionnaire
Clinical examination Clinical examination Clinical examination Clinical examination
Biological tests Biological tests

* The independent variables are described in the Appendix.

** Two types of models, prediction models and risk models, were explored, involving all features and selected features (age, gender, and etio-
logical factors for caries), respectively (Bratthall and Petersson, 2005). For each of the two types (prediction models and risk models), both
full-blown and screening models were built, with and without information from biological tests, respectively. These four models were all “any-
risk” models, with “any caries increment in 1 year (∆dmft > 0 or = 0)” as the dichotomous outcome variable. In addition, for screening the
high-risk children in the community setting without a clinic visit, a community-screening model was explored, with all features excluding the
clinical and biological categories. This model is a “high-risk” model for identifying about 25% of children with high caries burden (Beck
et al., 1992) (baseline dmft > 2 in this population) through a questionnaire.

as decayed teeth (dt) (Gao et al., 2009). In 1 yr, 43.7% of chil-
dren developed new caries (∆dmft > 0), including 43.3% of
children with ∆dt > 0. The mean (SD) increase of dmft in 1 yr
was 0.93 (1.42).
Multiple caries risk factors, indicators, and protective factors
have been identified (Table 2). The chance of “1-year caries
increment” (∆dmft > 0) increased with “age”, “Malay race”,
“prolonged breastfeeding”, “bedtime feeding with breast/milk/
formula/juice/sweet”, “frequent sweet intakes”, “bedtime sweet
intakes”, “no health problems”, “age that parents regarded as
appropriate for starting dental check”, “past caries experience”,
“Plaque Index”, “mutans streptococci”, and “Lactobacilli” and
decreased with “education attainment of father”, “never lived in
non-fluoridated community”, “use of other fluorides (not from
toothpaste)”, “no annual check because teeth did not bother the
child”, and “plaque pH” (all p < 0.05).
The chance of “high caries burden at baseline (dmft > 2)”
increased with “age”, “Malay race”, “using other fluorides”,
“parents’ unawareness of the detrimental effect of bedtime milk
bottle”, and “number of child’s decayed teeth estimated by par-
ents”, and decreased with “parents’ belief that ‘tooth worm’ is
the main reason of caries” (all p < 0.05).
Caries Risk Assessment Models and Their Performance
Various caries risk assessment models have been developed
(Table 2); the performances of these models and Cariogram are
compared (Table 2, Fig.).
For prediction models, the sensitivity/specificity of the
screening model was 82%/73%; the full-blown model reached a
sensitivity/specificity of 90%/90%. For risk models, the screen-
ing model showed a sensitivity/specificity of 81%/62%; the
full-blown model achieved a sensitivity/specificity of 83/92%.
The community-screening model was able to identify the 25%
of children with the highest baseline caries burden, with a sen-
sitivity/specificity of 82%/81%, through 6 simple questions.
Cariogram demonstrated a sensitivity/specificity of 71%/66%
and a moderate Area under ROC Curve (0.731). All performance
measures of our models were higher than those of Cariogram,
except for our screening risk model, which had significantly
higher sensitivity (p < 0.05), but similar performance in other
measures (p > 0.05), as compared with Cariogram (Table 2).
Discussion
As a population-based prospective study, this project was
designed to establish accurate biopsychosocial models for pre-
dicting caries. Due to the constraints of the field settings and the
potential health risks related to radiation exposure, no radio-
graphs were taken. This might result in an underestimation of
caries prevalence and incidence. For a complete profile of the
children’s caries experience, dmft, instead of dt, was used as the
outcome measure. A tooth was scored as missing only if trauma
was excluded as the reason and the child was at an age when
normal exfoliation would not be a sufficient explanation for the
absence of the tooth.

Downloaded from jdr.sagepub.com at University of Alabama at Birmingham on January 7, 2013 For personal use only. No other uses without permission.

© 2010 International & American Associations for Dental Research

 Table 2. Caries Risk Assessment Models and Their Performance

640
Odds Ratio (95% Confidence Interval)*

Prediction Models Risk Models

Community-screening
Factors/Indicators (X) Screening Full-blown Screening Full-blown Model Cariogram**

Age (mos) 1.045 (1.017-1.073) 1.060 (1.009-1.113) 1.058 (1.029-1.087) 1.060 (1.009-1.114) 1.042 (1.015-1.072) Related diseases
Malay race 1.837 (1.174-2.876) 2.053 (1.269-3.308) Diet frequency
Father’s education level 0.645 (0.542-0.767) 0.606 (0.436-0.841) Diet content
Months of breastfeeding 1.037 (1.012-1.063) 1.067 (1.016-1.120) 1.033 (1.010-1.057) 1.027 (1.008-1.087) Plaque amount
Bedtime feeding 1.484 (1.106-1.928) Fluoride applications
Frequency of between-meal sweets 1.368 (1.102-1.698) 1.338 (1.007-1.685) Lactobacilli level
Bedtime sweets 1.332 (1.006-1.682) Mutans streptococci
level
Never lived in non-fluoridated 0.676 (0.576-0.796) Saliva flow rate
community Saliva buffering
capacity
Using fluorides (other than fluoride in 0.420 (0.202-0.876) 2.633 (1.188-5.796)
toothpaste)
No annual check because teeth did 0.475 (0.256-0.781)
not bother the child
Age regarded by parents as 1.300 (1.071-1.639)
appropriate for dental check
Parent’s belief of ‘tooth worm’ as 0.103 (0.012-0.837)
reason for caries ***

© 2010 International & American Associations for Dental Research

Parents do not know bedtime milk 1.999 (1.228-3.326)
bottle is bad for teeth
Child’s number of decayed teeth 12.835 (8.908-18.673)
estimated by parent
No health problems 2.869 (1.737-4.736) 2.669 (1.198-5.945)
Past (baseline) caries 7.316 (5.135-10.423) 3.948 (1.933-8.060)
Plaque Index 5.050 (3.263-7.814) 8.902 (3.822-20.726) 9.061 (6.548-12.461) 7.367 (4.102-12.987)

Downloaded from jdr.sagepub.com at University of Alabama at Birmingham on January 7, 2013 For personal use only. No other uses without permission.
Level of mutans streptococci 2.700 (2.025-3.599) 2.555 (2.016-3.335)
Level of Lactobacilli 2.272 (1.587-3.471) 2.123 (1.481-3.304)
Average pH 0.010 (0.004-0.024) 0.016 (0.008-0.031)

(continued)
 Table 2. (continued)

Odds Ratio (95% Confidence Interval)*

Prediction Models Risk Models

Community-screening
Factors/Indicators (X) Screening Full-blown Screening Full-blown Model Cariogram**

Performance ****
Area under ROC curve 0.845II 0.961I 0.765III 0.945I 0.885II 0.731III
Cut-off point of predicted “possibility 0.328 0.445 0.342 0.551 0.158 0.376
of disease”
Sensitivity (%) 82.3II 90.4I 81.2II 83.1I, II 82.2II 70.5III
Specificity (%) 73.1III 90.0I 62.2IV 91.9I 81.2II 65.8IV
II I III I II
Sensitivity + Specificity (%) 155 180 143 175 163 136III
Accuracy (%) 77.1II 90.2I 70.5III 88.1I 81.4II 67.9III

* Only significant caries risk factors/indicators are listed in this Table, with the Odds Ratios (95% Confidence Intervals) indicating the associations between caries and these factors/indicators.
** Caries risk assessment with Cariogram was done by following the Cariogram user’s manual (Bratthall et al., 2003). “Group” was set as “standard” and the “Country/Area” was set

© 2010 International & American Associations for Dental Research

as “low-risk”, considering the nationwide water fluoridation in Singapore since 1958 (personal communication with Dr. Douglas Bratthall). To be more prudent, we have also tried other
setting combinations and confirmed that the current setting generates the best prediction.
*** With the survey question as “what do you think is the main reason of tooth decay?”, parents who chose ‘tooth worm’ other than the other 4 answers (heatiness, ineffective toothbrush-
ing, sugar, and bacteria) were defined as holding the belief of “tooth worm” as the reason for caries. “Heatiness” is an Asian belief emphasizing the heatiness resulted from the Yin-Yang
imbalance.
**** In a ROC curve, the true-positive rate (sensitivity) was plotted against the false-positive rate (1-specificity) for different possible diagnostic thresholds (Fig.). An ideal cut-off point of
“predicted possibility of caries” with the highest “sensitivity + specificity” was identified for each model. “Area under ROC curve” served as a summary measure indicating the overall
model performance in discriminating between low-risk and high-risk children.
I-IV
ranking: There were significant differences in the specific performance measures among models with different ranks. The Rank-I indicates the highest performance.

Downloaded from jdr.sagepub.com at University of Alabama at Birmingham on January 7, 2013 For personal use only. No other uses without permission.
  641
642
Figure. ROC curves of caries risk assessment models. There were
significant differences in the Area under ROC Curve among models
with different ranks (I-III). Rank I indicates highest model performance,
with Rank III indicating lowest performance. Due to different disease
outcomes, the ROC curve of the community-screening model is not
included in this Fig.
The children were followed over a 12-month period, inade-
quate for adolescents and adults, but sufficient for preschoolers
with faster caries progression (Shwartz et al., 1984). A sizeable
incidence rate (44%) has allowed us to develop various multi-
factorial models successfully. Dental treatments delivered to the
participants during the 12-month period might confound the
results; however, the related information revealed only 62
(3.5%) children as receiving restoration or extraction. Since no
regular oral health education program was delivered to pre-
schoolers during this period, dental intervention in the kinder-
garten setting was deemed to be negligible.
Caries Risk Factors, Indicators, and Protective Factors
This study has identified multiple biopsychosocial risk factors,
with “breastfeeding” among them. However, further analysis
indicated that, compared with children “never breastfed”, those
“breastfed within 12 months” did not demonstrate a higher car-
ies risk, with a relative risk (RR) of 1.03 and 95% confidence
interval (CI) of 0.84-1.23. Breastfeeding for “1-2 years” and
“beyond 2 years” elevated the caries risk, with a RR (95% CI)
of 1.42 (1.07-1.66) and 1.64 (1.27-1.92), respectively. This
result seems to indicate a “healthy window for breastfeeding”.
Downloaded from jdr.sagepub.com at University of Alabama at Birmingham on January 7, 2013 For personal use only. No other uses without permission.
© 2010 International & American Associations for Dental Research
Gao et al. J Dent Res 89(6) 2010
“Using other fluorides” indicated a reduced chance of “1-year
caries incidence” (∆dmft > 0), but an elevated chance of “high
caries burden at baseline” (dmft > 2). At the cross-sectional
level, “use of other fluorides”, most likely prescribed by den-
tists, could be a result, instead of a cause, of high caries burden
at baseline. Longitudinally, the application of additional fluoride
products served as an effective measure to control caries incre-
ment in susceptible children (odds ratio = 0.420; 95% CI =
0.202-0.876).
“Tooth worm” is a traditional conception of caries pathogen
and is more prevalent in low-socio-economic groups. Interestingly,
children whose parents held this concept had a lower caries risk,
possibly due to the “fear of tooth worm” being a strong motiva-
tor for children and parents to strive for good oral health. With
creative modifications, this concept, or its derivatives, might be
useful in oral health education.
Surprisingly, children with “no health problems” (reported
by parents) tended to develop more caries. This may be possible
since “health problems” may lead to a better parental awareness
of and attention to their child’s oral health (Rohde and Northrup,
1988).
There was a significant increase in caries risk associated with
plaque acidity, which is a cumulative reflection of multiple
microbiological (amount and virulence of cariogenic bacteria),
dietary (cariogenicity of food intake), and salivary factors
(saliva flow and buffer capacity), and their interactions
(Kleinberg et al., 1982). Our analysis comparing the predictive
values of single risk factors and indicators showed no significant
difference in the “sensitivity + specificity” between “plaque pH”
(158%) and “past caries” (153%), which has been regarded as
the strongest caries predictor (Zero et al., 2001). Nevertheless,
“plaque pH” was more sensitive than “past caries” (82% vs.
70%, p < 0.05), with no significant difference in their specificity
(76% vs. 83%, p > 0.05).
Caries Risk Assessment Models
A useful risk assessment program should be one with high
simplicity, sensitivity, and specificity (Stamm et al., 1988).
However, with the trade-off between simplicity and accuracy, it
may be impractical for both to be achieved simultaneously.
Nevertheless, by combining a simple and low-cost screening
model and a full-blown model with high accuracy, various needs
in the community and clinical settings may be met. With a ques-
tionnaire and a clinical examination, our screening prediction
models can pick up the potential high-risk children in the clinic
for further assessment with our full-blown model. The explora-
tion of the community-screening model is a unique approach of
this study. Through a six-item simple questionnaire, 25% of
children with high caries burden (baseline dmft > 2) could be
identified with both sensitivity and specificity above 80%. This
model would be practical and useful for screening these children
for early treatment and intervention, especially in communities
where regular dental screening for preschoolers is uncommon
and/or costly.
Cariogram, even with biological tests, showed a sensitivity/
specificity (71%/66%) and a performance lower than that of our
screening models. The insufficient accuracy of Cariogram in
predicting early childhood caries was also revealed recently in
J Dent Res 89(6) 2010 Caries Risk Assessment Models for Children   643
Sweden, with a sensitivity/specificity of 46%/88% (Holgerson
et al., 2009). Compared with our findings, the relatively lower
sensitivity and higher specificity of Cariogram in this Swedish
study might be due to the strict threshold value for defining
“positive case” (i.e., chance of avoiding caries of 20% or below),
while the optimal cut-off point identified in our study was 37.6%.
Both studies reported similar “sensitivity + specificity” (134%
vs. 136%; p > 0.05) of Cariogram. This indicates that, although
Cariogram has been validated in Scandinavian adolescents and
the elderly, it may have limited accuracy among preschoolers.
In conclusion, this study has identified pertinent caries risk
factors, indicators, and protective factors and has established
caries risk assessment models for a wide range of uses in com-
munity and clinical settings for the early detection of high-caries-
risk groups and evidence-based treatment planning.
Acknowledgments
We appreciate the help from the staff in the participating kinder-
gartens and the financial support of the Singapore Ministry of
Education Academic Research Fund, R222-000-021-112 and
R222-000-022-112.
References
Alm A, Wendt LK, Koch G, Birkhed D (2007). Prevalence of approximal
caries in posterior teeth in 15-year-old Swedish teenagers in relation to
their caries experience at 3 years of age. Caries Res 41:392-398.
American Academy of Pediatric Dentistry (AAPD) (2002). Policy statement
on the use of a caries-risk assessment tool. http://www.aapd.org/pdf/
policycariesriskassessmenttool.pdf (accessed on Feb. 2, 2010).
Armfield JM, Spencer AJ (2004). Changes in South Australian children’s
caries experience: is caries re-surfacing? Aust Dent J 49:98-100.
Beck JD, Weintraub JA, Disney JA, Graves RC, Stamm JW, Kaste LM,
et al. (1992). University of North Carolina Caries Risk Assessment
Study: comparisons of high risk prediction, any risk prediction, and any
risk etiologic models. Community Dent Oral Epidemiol 20:313-321.
Bratthall D (1996). Dental caries: intervened—interrupted—interpreted.
Concluding remarks and cariography. Eur J Oral Sci 104(Pt 2):486-491.
Bratthall D, Petersson GH (2005). Cariogram—a multifactorial risk assess-
ment model for a multifactorial disease. Community Dent Oral
Epidemiol 33:256-264.
Bratthall D, Petersson GH, Stjernswärd JR (2003). Cariogram manual.
Cariogram Internet version 2.01. http://www.mah.se/fakulteter-och-
omraden/odontologiska-fakulteter/Avdelning-och-kansli/Cariologi/
Cariogram/
Gao XL, Hsu CY, Loh T, Koh D, Hwarng HB, Xu YJ (2009). Dental caries
prevalence and distribution among preschoolers in Singapore.
Community Dent Health 26:12-17.
Graves RC, Abernathy JR, Disney JA, Stamm JW, Bohannan HM (1991).
University of North Carolina caries risk assessment study. III. Multiple
factors in caries prevalence. J Public Health Dent 51:134-143.
Featherstone JD, Adair SM, Anderson MH, Berkowitz RJ, Bird WF, Crall
JJ, et al. (2003). Caries management by risk assessment: consensus
statement. J CA Dent Assoc 31:257-269.
Harris R, Nicoll AD, Adair PM, Pine CM (2004). Risk factors for dental
caries in young children: a systematic review of the literature.
Community Dent Health 21(1 Suppl):71S-85S.
Downloaded from jdr.sagepub.com at University of Alabama at Birmingham on January 7, 2013 For personal use only. No other uses without permission.
© 2010 International & American Associations for Dental Research
Haugejorden O, Birkeland JM (2002). Evidence for reversal of the caries
decline among Norwegian children. Int J Paediatr Dent 12:306-315.
Holgerson PL, Twetman S, Stecksèn-Blicks C (2009). Validation of an age-
modified caries risk assessment program (Cariogram) in preschool
children. Acta Odontol Scand 67:106-112.
Hunter PB (1988). Risk factors in dental caries. Int Dent J 38:211-217.
Kaste LM, Selwitz RH, Oldakowski RJ, Brunelle JA, Winn DM, Brown LJ
(1996). Coronal caries in the primary and permanent dentition of chil-
dren and adolescents 1 ± 17 years of age: United States, 1988 ± 1991.
J Dent Res 75(Spec Iss):631-641.
Kleinberg I, Jenkins GN, Chatterjee R, Wijeyeweera L (1982). The anti-
mony pH electrode and its role in the assessment and interpretation of
dental plaque pH. J Dent Res 61:1139-1147.
National Institutes of Health (NIH) consensus panel (2001). National
Institutes of Health consensus development conference statement.
Presented at the Consensus Development Conference on Diagnosis and
Management of Dental Caries Throughout Life, March 26-28, 2001.
Natcher Conference Center, National Institutes of Health, Bethesda,
MD, USA. http://www.nidcr.nih.gov/NR/rdonlyres/778F8DE9-D5C4-
4C71-940E-07C2E0C9E348/0/Final_CDC_Statement.pdf (accessed on
Feb. 2, 2010).
Petersson G, Twetman S, Bratthall D (2002). Evaluation of a computer
program for caries risk assessment in schoolchildren. Caries Res
36:327-340.
Petersson G, Fure S, Bratthall D (2003). Evaluation of a computer-based
caries risk assessment program in an elderly group of individuals. Acta
Odontol Scand 61:164-171.
Pitts NB, Chestnutt IG, Evans D, White D, Chadwick B, Steele JG (2006).
The dentinal caries experience of children in the United Kingdom,
2003. Br Dent J 200:313-320.
Rohde JE, Northrup RS (1988). Feeding, feedback and sustenance of pri-
mary health care. Indian J Pediatr 55(1 Suppl):110S-123S.
Shwartz M, Grondahl HG, Pliskin JS, Boffa J (1984). A longitudinal analysis
from bite-wing radiographs of the rate of progression of approximal cari-
ous lesions through human dental enamel. Arch Oral Biol 29:529-536.
Silness J, Löe H (1964). Periodontal disease in pregnancy. II. Correlation
between oral hygiene and periodontal condition. Acta Odontol Scand
22:121-135.
Skeie MS, Raadal M, Strand GV, Espelid I (2006). The relationship between
caries in the primary dentition at 5 years of age and permanent dentition
at 10 years of age—a longitudinal study. Int J Paediatr Dent 16:152-160.
Speechley M, Johnston DW (1996). Some evidence from Ontario, Canada,
of a reversal in the dental caries decline. Caries Res 30:423-427.
Stamm J, Disney J, Graves R, Bohannan H, Abernathy J (1988). The
University of North Carolina Caries Risk Assessment Study. I: Rationale
and content. J Public Health Dent 48:225-232.
Truin GJ, van’t Hof MA, Kalsbeek H, Frencken JE, König KG (1993).
Secular trends of caries prevalence in 6- and 12-year-old Dutch chil-
dren. Community Dent Oral Epidemiol 21:249-252.
US Department of Health and Human Services (US/DHHS) (2007). Trends
in oral health status: United States, 1988-1994 and 1999-2004. Centers
for Disease Control and Prevention, National Center for Health
Statistics, Hyattsville, MD: USDHHS.
World Health Organization (WHO) (1997). Oral health surveys—Basic
methods. 4th version. Geneva: WHO.
World Health Organization (WHO) (2006). Oral Health Country/Area
Profile Programme (CAPP). Geneva: WHO. http://www.whocollab
.od.mah.se/index.html (accessed on Feb. 2, 2010).
Zero D, Fontana M, Lennon AM (2001). Clinical applications and outcomes
of using indicators of risk in caries management. Presented at the
Consensus Development Conference on Diagnosis and Management of
Dental Caries Throughout Life, March 26-28, 2001. Natcher Conference
Center, National Institutes of Health, Bethesda, MD, USA.