Abstract
Handwriting is a basic and essential skill in school. Children who have difficulty with handwriting tend to feel frustrated about and unsuccessful in handwriting tasks, including assignments and exams, potentially leading to low academic achievement and poor self-esteem (Feder & Majnemer, 2007). The prevalence of handwriting difficulties ranges from 6% to 37% among school-age children using alphabetic languages (Karlsdottir & Stefansson, 2002; Overvelde & Hulstijn, 2011) and from 5% to 8% in those using Chinese (Tseng, 1993). Early identification and intervention for kindergarteners at risk for developing handwriting problems may help them successfully transition into elementary school. However, which factors can be used to screen at-risk kindergarteners is not clear.
One study investigated the consistency of letter handwriting performance from kindergarten to first grade (Marr & Cermak, 2003). Researchers found that only 42% of the children who were classified in the poorest handwriting group in kindergarten still remained in this group in first grade (Marr & Cermak, 2003). This result implies that kindergarteners’ handwriting performance may be not a good indicator of later handwriting issues. In addition, the handwriting evaluations administered in kindergarten are not practical in some countries. For example, on the basis of the curriculum guidelines set by Taiwan’s Ministry of Education (2017), the kindergarten curriculum does not include formal handwriting instruction; children learn to read and write Chinese phonetic symbols and characters after entering elementary school. Most public and few private kindergartens in Taiwan follow these guidelines (Liu, 2016); therefore, not all Taiwanese kindergarteners know how to write Chinese phonetic symbols and characters.
Sufficient developmental skills, including fine motor, visual–perceptual, and visual–motor integration (VMI) skills, are assumed to be prerequisites for proficient handwriting. A broad base of studies on both alphabetic (Cornhill & Case-Smith, 1996; Volman et al., 2006) and Chinese (Tseng & Chow, 2000) handwriting have supported the contribution of fine motor and VMI skills to handwriting legibility and speed in school-age children. Although the relationship between visual perception and alphabetic handwriting performance seems inconclusive (Feder et al., 2005; Volman et al., 2006), its contribution to Chinese handwriting performance has been proven across studies (Tse et al., 2014; Tseng & Chow, 2000). Whether these handwriting prerequisites can be indicators used to screen kindergarteners at risk for later handwriting difficulties remains unclear. Two studies of children using alphabetic handwriting demonstrated that fine motor (i.e., Nine-Hole Peg Test) and VMI (i.e., Beery–Buktenica Developmental Test of Visual–Motor Integration [Beery VMI]; Liu & Lu, 1999) scores in kindergarten are predictive of handwriting performance in first grade (Marr & Cermak, 2002; van Hartingsveldt et al., 2015). However, no studies have investigated this issue in children who use Chinese handwriting. Moreover, little is known about the predictive value of visual perception in kindergarten for either later alphabetic or Chinese handwriting performance.
This study had two objectives. The first was to determine the predictive value of handwriting prerequisites in kindergarten for poor Chinese handwriting in first grade. The second was to develop a nomogram model constructed of the proven risk factors to estimate kindergarteners’ probability of later poor Chinese handwriting and to explore the model’s predictive validity. A nomogram is a pictorial representation of a predictive model. It includes a scale for each variable that is based on the statistical model’s results; users can easily use the graphic interface to calculate the cumulative point score for all the variables and further estimate the probability of an event for a given person (Balachandran et al., 2015). Nomogram models have been widely used to predict outcomes in medical research (Balachandran et al., 2015), but they are rarely used to predict child development. To our knowledge, this study is the first to establish a nomogram model to predict the probability of later poor handwriting among kindergarteners.
Method
Design and Participants
This study used a 1-yr prospective longitudinal, observational design with two measurement points. Participants were administered fine motor, visual–perceptual, and VMI tests in kindergarten and handwriting assessments in first grade.
The study was conducted from April 2015 to July 2017. We contacted kindergarten principals in an urban area of Tainan, Taiwan, by phone and asked them to give our research proposal to their students’ parents on our behalf. Parents who wanted their children to participate in this study were advised to contact the research assistant by phone to obtain the detailed study procedure. The inclusion criteria for participation were (1) age 5–6 yr and (2) having native Chinese-speaking parents. Children with (1) a diagnosis of neuromuscular or developmental disabilities (e.g., intellectual disabilities), (2) significant neuromuscular injury (e.g., finger fracture), (3) visual or hearing impairment, or (4) any remediation intervention (e.g., interventions for handwriting) during the time of the study were excluded.
In the first grade, participants were classified into the normal or poor handwriting group on the basis of scores on a direct handwriting assessment, the Battery of Chinese Basic Literacy (BCBL; Hung et al., 2003), and a self-report handwriting questionnaire rated by class teachers, the Chinese Handwriting Evaluation Form–School version (CHEF–S; Chang & Yu, 2012). As suggested in the BCBL and CHEF–S test manuals, poor handwriting performance was defined as a score below the 20th percentile of the norm on any handwriting subtest of the BCBL or below the 16th percentile of the norm on any dimension of the CHEF–S, except for pencil grasp dimension (the score for this domain is not considered to be a good indicator of poor handwriting).
This study was reviewed and approved by the Ethics Committee of National Cheng Kung University Hospital (B-ER-104–119). Written informed consent was obtained before data collection from all parents, children age 7 yr or older, and teachers.
Measures
Test of Visual–Perceptual Skills, Third Edition
The Test of Visual–Perceptual Skills, Third Edition (TVPS–3; Martin, 2006), is a standardized motor-free test that consists of seven subtests: Visual Discrimination, Visual Memory, Visual–Spatial Relationships, Form Constancy, Visual Sequential Memory, Visual Figure–Ground, and Visual Closure. Each subtest has 16 questions that use black-and-white designs as stimuli. The child is asked to choose one answer from among four or five designs. The TVPS–3’s internal consistency (α = .75–.88 for the subtests, α = .96 for the total score composite) and test–retest reliability (r = .46–.81 for the subtests, r = .97 for the total score composite) are acceptable to excellent, and its content, criterion-related, and construct validities have been demonstrated to be good (Martin, 2006).
Bruininks–Oseretsky Test of Motor Proficiency, Second Edition
The Bruininks–Oseretsky Test of Motor Proficiency, Second Edition (Bruininks & Bruininks, 2005), was designed to measure gross and fine motor performance of people ages 4–21 yr. Two subtests of finger motor coordination were used in this study: Fine Motor Precision (FMP; e.g., filling in shapes), which has no time limitation, and Manual Dexterity (MD; e.g., transferring pennies), which is time limited. Both subtests have acceptable test–retest reliability (FMP, r = .75; MD, r = .63) and excellent interrater reliability (FMP, r = .86; MD, r = .99). Significant correlations have been reported between the FMP (r = .61) and the MD (r = .47) subtests and the Fine Motor Quotient of the Peabody Developmental Motor Scales–Second Edition (Bruininks & Bruininks, 2005).
Beery–Buktenica Developmental Test of Visual–Motor Integration–Chinese Version
The Beery VMI–Chinese Version is based on the fourth edition of the English version (Liu & Lu, 1999). It consists of a test booklet containing 24 geometric forms that are copied without a time limitation. Its reliabilities (split-half r = .77, interrater r = .96, intrarater r = .91) and validities (concurrent validity with the Bender Visual Motor Gestalt Test, r = −.65; discriminant validity, r = .86) in the Taiwanese population are good (Liu & Lu, 1999). In this study, one investigator (Hsiao) and one research assistant scored all of the participants’ test booklets. Their interrater reliability for the 22 test booklets was satisfactory (intraclass correlation coefficient = .97).
Battery of Chinese Basic Literacy
The BCBL is a standardized test developed to measure the Chinese reading and writing abilities of students in Grades 1–3 (Hung et al., 2003). We used two subtests (Dictation without a time limitation and Far-Point Copy with a 2-min limit), suggested in the manual as the short form of the battery, to assess handwriting performance. In both tests, each character has to be written within a grid (18 mm × 18 mm), and 1 point is given for each correctly written character. The internal consistency (Far-Point Copy, α = .51; Dictation, α = .80) and split-half reliability (Far-Point Copy, r = .84; Dictation, r = .81) of both subtests have been found to be acceptable for first graders (Hung et al., 2003).
Chinese Handwriting Evaluation Form–School Version
The CHEF–S has 25 items used to evaluate five dimensions (construction, accuracy, speed, pencil grasp, and directionality) of handwriting performance in first and second graders (Chang & Yu, 2012). Each child’s class teacher was asked to rate the child’s handwriting performance for each item on a 5-point Likert scale ranging from 1 (never matching) to 5 (always matching). The internal consistency (α = .70–.93), split-half reliability (r = .64–.98), and test–retest reliability (r = .79–.90) of the CHEF–S are acceptable to good. The CHEF–S demonstrated good construct, discriminative, and concurrent validities (Chang & Yu, 2012).
Procedure
In the second semester of kindergarten (April–August), all participants were first invited to come to a quiet assessment room in our department, where one investigator (Hsiao) and one trained occupational therapist administered the VMI, visual–perceptual, and fine motor assessments. The following year, participants in the second semester of first grade (May–July) came to the assessment room again and were administered the Dictation and Far-Point Copy subtests of the BCBL. In addition, because class teachers are the people who most frequently and directly observe a child’s handwriting performance in a variety of handwriting tasks in a natural environment, they were asked to fill out the CHEF–S and return it to the first author (Hwang) by mail within 2 wk.
The demographic data and school handwriting curriculum information were obtained, respectively, from the parents and the kindergarten and first-grade teachers using a self-report questionnaire designed by the investigators.
Statistical Analysis
Participant characteristics are presented in the form of descriptive and inferential statistics. The handwriting prerequisites (e.g., fine motor) and demographic variables (e.g., age, sex) potentially related to handwriting performance (Feder et al., 2005; van Hartingsveldt et al., 2015) were analyzed, and univariate and multivariate logistic analyses were used to identify significant predictors of poor handwriting. A receiver operating characteristic (ROC) curve for the logistic regression model was used to assess the diagnostic accuracy of the probability of poor handwriting. The area under the ROC curve (AUC) was considered to be a measure of discriminative validity. Finally, we created a predictive nomogram of poor handwriting constructed of the proven risk factors. All statistics and figures were prepared using R software (Version 3.5.1; R Foundation for Statistical Computing, Vienna, Austria). Each test of the model parameters was two-sided, and statistical significance was set at p < .05.
Results
A total of 170 kindergarteners were enrolled. Fourteen children were excluded as a result of loss of contact in first grade or because they received a diagnosis of developmental disability, rehabilitative interventions, or medicine for inattention between the two assessments. Data for 156 children studying in 54 elementary schools were analyzed, and 37 children (23.7% of sample) met the criteria for poor handwriting. The children’s characteristics and standard scores on the kindergarten tests are provided in Table 1.
Child Characteristics and Standard Scores on Tests Administered in Kindergarten (N = 156)
Note. Beery VMI = Beery–Buktenica Development Test of Visual–Motor Integration–Chinese version; FMP = Fine Motor Precision subtest of the Bruininks–Oseretsky Test of Motor Proficiency, Second Edition (BOTS–2); M = mean; MD = Manual Dexterity subtest of the BOTS–2; SD = standard deviation; SES = socioeconomic status; TVPS–3 = Test of Visual–Perceptual Skills, Third Edition.
The univariate regression analyses identified four significant predictors: child’s age at entry to elementary school and scores on the FMP and MD subtests and the TVPS–3 (Table 2). Previous research has indicated that around 10 cases or controls are required per parameter to reliably estimate regression coefficients (Harrell, 2001). On the basis of this guideline, we determined that a multivariate logistic analysis with fewer than four variables would be more appropriate for this study’s small sample size (n = 37 in the poor handwriting group). Thus, considering similar properties measured by the FMP and MD subtests, as well as a moderate correlation between both tests (r = .35, p < .001, for the present sample and r = .44 for the normative sample; Bruininks & Bruininks, 2005), we chose the MD subtest (a stronger variable correlated with handwriting scores) for the multivariate logistic analysis instead of the FMP subtest. The results showed that child’s age (odds ratio [OR] = .24, p = .039) and MD (OR = .86, p = .015) and TVPS–3 (OR = .96, p = .036) scores independently predicted poor handwriting. We then used these three variables to construct a nomogram model to predict the occurrence of poor handwriting.
Predictors for Poor Handwriting in First Grade
Note. A dash indicates that the factor was not included. Beery VMI = Beery–Buktenica Development Test of Visual–Motor Integration–Chinese version; CI = confidence interval; FMP = Fine Motor Precision subtest; MD = Manual Dexterity subtest; OR = odds ratio; SES = socioeconomic status; TVPS–3 = Test of Visual–Perceptual Skills, Third Edition.
FMP score was not included in the multivariate regression because of its significant correlation with the MD subtest and weaker prediction for poor handwriting than the MD subtest.
p < .05. **p < .01.
The AUC of the logistic regression model was .75 (95% confidence interval [.66, .84]), indicating that the ability of this nomogram model to distinguish between kindergarteners predicted and not predicted to have poor handwriting in first grade was acceptable (Figure 1). In the nomogram point system, the MD subtest was the greatest contributor to predictions of the probability of poor handwriting, followed by the TVPS–3 and age. For example, if a child was 78 mo old (corresponding to 24 points) and scored 10 on the MD subtest (90 points) and 85 on the TVPS–3 (75 points), the child received a total of 189 points (approximately a 70% probability of poor handwriting), as calculated by the proposed nomogram (Figure 2).
ROC curve of the nomogram model.
Nomogram model for predicting poor Chinese handwriting in first grade.
We set the optimal cutoff point with a sensitivity of at least 80% to avoid false-negative results. Under this requirement, a cutoff point of 119 (a 20% probability of poor handwriting) in the nomogram (see Figure 2) with a sensitivity of .811 (true positive) and a specificity of .546 (i.e., false positive [1 − specificity] = .454) was most optimal because it had the lowest false-positive fraction among the cutoff points, with a sensitivity greater than .80 (see Figure 1).
Discussion
Our results reveal that Taiwanese children who were younger at the time of school entry and performed poorly on the MD and TVPS–3 tests in kindergarten had increased risk of poor Chinese handwriting in first grade. Consistent with van Hartingsveldt et al.’s (2015) findings for alphabetic handwriting, our findings support the predictive value of fine motor skills in kindergarten for later handwriting performance. However, in contrast to van Hartingsveldt et al.’s finding that kindergarten VMI significantly predicted copying speed in first grade, we did not find Beery VMI–Chinese Version scores in kindergarten to be related to the risk of developing poor Chinese handwriting in first grade. This distinction might have resulted from the different methodologies (e.g., the measures and definitions of handwriting outcomes) and written languages used in the two studies.
We found that visual perception in kindergarten was also an early predictor of poor Chinese handwriting in first grade. This finding was in line with previous findings with older children using Chinese handwriting (Tse et al., 2014; Tseng & Chow, 2000). In contrast, in alphabetic handwriting, a significant correlation of visual perception with handwriting seems to occur only among first graders (Feder et al., 2005; Volman et al., 2006). These findings may lead to the conclusion that the importance of visual perception in handwriting performance is mainly observed in beginning writers or in children writing in characters that involve complex construction, such as Chinese characters.
We found that the odds of poor Chinese handwriting were lower among children who were older at the time of school entry. This finding was congruent with the results of previous research indicating that older students outperformed their younger peers in terms of academic attainment (Puhani & Weber, 2007). One reason for this is maturational differences when children start school (Bedard & Dhuey, 2006), which could also explain our findings. Younger children may have less mature prerequisite abilities (e.g., attention span, finger muscle endurance) for learning Chinese handwriting than their older classmates; thus, they cannot perform as well as their older classmates. In contrast to our results, van Hartingsveldt et al. (2015) did not find an age effect on first-grade alphabetic handwriting performance. Differences in the complexity of character construction between Chinese and alphabetic languages may be a possible explanation for this inconsistency.
Cutoff points for use in screening at-risk kindergarteners can be set from this nomogram model. Using the cutoff point of 119 suggested in this study, approximately 81% of children with poor Chinese handwriting in first grade could be identified in kindergarten; however, 45.4% of children without handwriting problems in first grade were false positive (i.e., 1 − specificity). Although a rather significant proportion of children without poor handwriting may mistakenly be classified into the at-risk group, more attention or early interventional programs may also be beneficial for their later handwriting skills; thus, the unexpected cost of false-positive results seems acceptable. Other cutoff points may also be optimal, depending on the availability and cost of educational or medical resources in other countries that use Chinese handwriting.
Limitations and Future Studies
This study had several limitations. First, the proposed nomogram is based on a small sample living in one area in Tainan, Taiwan. Thus, our results should be generalized with caution to children living in other geographic areas or using different language systems. Second, classroom teachers of 26 children (16.7% of the sample) did not return the completed CHEF–S. The grouping for these children was based only on their performance on the BCBL. Therefore, the number of children with poor handwriting may potentially have been underestimated.
The preliminary predictive value of our nomogram for later Chinese handwriting difficulties in first grade was demonstrated. Increasing sample size and heterogeneity (e.g., students living in other geographic areas) and adding more significant predictors (e.g., kindergarteners’ handwriting performance) to the nomogram may enhance its predictive validity and generalizability. In addition, future studies could further examine the nomogram’s long-term predictive value for poor Chinese handwriting beyond the first grade.
Implications for Occupational Therapy Practice
The results of this study have the following implications for occupational therapy practice:
Kindergarten fine motor and visual–perceptual abilities are important predictors of Chinese handwriting difficulties in first grade.
Although evidence has shown a significant correlation between Beery VMI–Chinese version score and handwriting performance in kindergarteners, further evaluation of the use of the Beery VMI for kindergarten screening for handwriting difficulties is warranted.
A predictive nomogram constructed of significant kindergarten predictors (e.g., fine motor skills) seems feasible for use as a screening tool to identify Taiwanese kindergarteners at risk for later Chinese handwriting problems for further referral and interventions.
A similar predictive nomogram can be developed for kindergarteners in different countries using Chinese or alphabetic language systems.
Conclusion
This study demonstrates that a nomogram built by combining predictions of school entrance age and scores on fine motor and visual perception tests could be used as a screening tool to early identify kindergarteners at risk for poor Chinese handwriting in the first grade.
Footnotes
Acknowledgment
This project was partially supported by the Ministry of Science and Technology (Grant MOST 104-2410-H-006-068) to Yea-Shwu Hwang.