Predicting motor outcome at preschool age for infants tested at 7, 30, 60, and 90 days after term age using the test of infant motor performance.

Early diagnosis of normal or abnormal motor performance is an important aspect of early intervention for infants who are at risk for developmental delays or disabilities due to biological or environmental risk factors. As the survival rate for infants who are born with low birth weights or with other prenatal medical complications has improved, the lack of accuracy of diagnostic and predictive measures continues to impede early identification of infants with motor delays. (1) Several factors are believed to contribute to the difficulty of accurate predictions. They include: (1) the dearth of information on the diagnostic efficiency of most diagnostic measures used with infants, (2) the spontaneous recovery during the first 2 years of life of suspicious or abnormal neuromotor signs observed in some of the infants born prematurely, (2-5) (3) the scarcity of prospective longitudinal studies that extend beyond the infancy and toddler periods, and (4) the lack of clarity on how the environment influences motor development. (6,7)

The accuracy of diagnosis depends on a test's diagnostic efficiency, which includes its sensitivity, specificity, and predictive values. (8) Sensitivity and specificity pertain to how well a test performs in the presence (or absence) of the condition of interest. A sensitive test is not likely to miss infants with atypical motor development (low false negative). A specific test is not likely to misclassify children with typical development as atypical (low false positive). Positive predictive values (PPVs) and negative predictive values (NPVs) provide information on the probability of having (or not having) the condition given the results of the test. Sensitivity and specificity are used to guide the selection of a measure, whereas positive and negative predictive values guide decisions about the meaning of the score once a test has been selected. Determination of the diagnostic efficiency values of a test requires comparisons of a new test with a "gold standard" or other diagnosis of the condition with known accuracy. (8) In the case of motor performance, determination of predictive values also requires observations over time (longitudinal studies) because motor skills change with maturation.

Of the numerous developmental tests that are available to identify infants who are likely to have motor delays or disabilities, only a few have known diagnostic efficiency. (9-14) The criterion measure or "gold standard" in most of these studies, however, was administered when the children were 24 months of age or less, a period during which transient abnormal neuromotor signs continue to resolve. (3,5,15) In one study that that predicted motor outcome at preschool age, (9) the target measure was administered at 6 months of age in one study. (9) In this study, the Bayley Infant Neurodevelopmental Screener was used at 6 months. In another study, (14) only 10% of the neonates diagnosed with neurological abnormalities had neurological disability at 4 years of age. Touwen's standardized neurological examination was used in this study. None of the motor developmental tests have been shown to predict motor outcomes at preschool age for infants tested at or before 3 months "after term, when decisions for referral for early intervention are most difficult.

The transient neuromotor signs during infancy, together with instability in the rate and patterns of motor development in infants, (2,16) contribute to misclassiffications. Using the original version of the Peabody Developmental Motor Scales (PDMS), for example, Darrah and associates (15) found that 76% of the infants in their study demonstrated instability in their gross motor scores during repeated testing at 9, 11, 13, 16, and 21 months of age. Other researchers have reported unstable Movement Assessment of Infants (MAI) scores in 47% of infants who were tested at 5, 10, and 18 months of age (16) and monthly intra-individual variations in the Test of Infant Motor Performance (TIMP) scores. (17) The instability in rate and patterns of motor development underscores the need for repeated assessments to improve the accuracy of diagnosis.

Susceptibility of infants to environmental influences is another factor that contributes to the difficulty of predicting outcome in preterm infants. (7,18,19) Although the nature of environmental influences is not well understood, (20) the findings with preterm infants suggest that the home environment and socioeconomic factors may modify or mediate between early and later motor development. (21) For example, the quality and quantity of the amount of stimulation in the home, as measured by the Home Observation for Measurement of the Environment (HOME), have been shown to correlate with children's scores on tests of motor skills. (21,22) Children in these studies were younger than 4 years of age. Therefore, the potential for the home environment to moderate or obscure the nature of the relationship between early motor and later motor performance cannot be overlooked. The purpose of this study was to examine the ability of the TIMP, when administered to infants at 7, 30, 60, and 90 days after term age, to diagnose and predict motor skills at preschool age.

The TIMP is the only comprehensive motor test designed to assess functional motor performance of infants 32 weeks postconceptional age to 16 weeks after term age. (23) The TIMP is a discriminative measure designed to identify infants, as early as the neonatal period, who are likely to benefit from early intervention. Evaluating the extent to which such infants are correctly identified and classified, therefore, is essential. The TIMP comprises 42 items grouped into 2 sections: the Observed and Elicited sections. The Observed section items are used to rate spontaneous movement, and the Elicited section items assess the infants' motor responses to placement in various positions and to visual or auditory stimulation. Some of the items for the TIMP were taken from neurological tests developed by Dubowitz and Dubowitz (24) and Brazelton (25) and from motor assessment procedures developed by Amiel-Tison and Grenier (26) and Cioni and Prechtl. (27) The TIMP developers created scoring descriptors for each of these items. (23)

Previous research on the TIMP suggests that it yields reliable and valid data for the purpose of discriminating motor performance of preterm and very young infants. (23,28) Campbell and associates (23) reported that TIMP scores are sensitive to changes in infants' motor performance due to maturation and medical complications. The TIMP scores were negatively correlated, for example, with the number of medical complications as measured by the newborn form of the Problem-Oriented Perinatal Risk Assessment System (POPRAS). (29) Concurrent validity of TIMP scores has been demonstrated at 3 months corrected age in comparison with scores on the Alberta Infant Motor Scale (AIMS). (30) Using a cutoff of 0.5 standard deviation below the mean, the TIMP identified 80% of the same infants as the AIMS cutoff score of the 10th percentile did at 3 months corrected age. The interrater and test-retest score stability on the TIMP over a 3-day period have been reported. (31) Satisfactory interrater reliability among experienced examiners was demonstrated with less than 5% misfitting items, and the test-retest reliability coefficient was r =.89. The 5% misfitting items were based on Rasch analysis using the Facets program (32) and represent the number of unexpected ratings by the testers, given the difficulty of the items and the infant's ability.

The results of a study by Campbell et al (17) indicated that the TIMP predicted motor outcome in infants at 12 months of age. The highest correlation coefficient (Pearson r) was observed between the TIMP scores at 90 days after term age and the AIMS scores at 12 months of age (r=.55, P=.01). (17) An earlier version of the TIMP (version 3.2), using a convenience sample of infants between 32 weeks gestational age and 4 months after term age, was found to correlate with the Bruininks-Oseretsky Test of Motor Proficiency (BOTMP) at school age (r =.36). (33) Version 3.2 of the TIMP consisted of 59 items. Although version 4.0 of the TIMP demonstrates attributes of a valid motor test for very young infants, no study to date has examined its ability to predict motor outcome at preschool or school age, when the diagnosis of motor disability is definite.

This study examined: (1) the relationship between early TIMP scores and scores on the Peabody Developmental Motor Scales, 2nd edition (PDMS-2), at preschool age, (2) the age at which the scores on the TIMP provided the best diagnostic information, and (3) whether the quality of' the home environment moderated the relationship between early motor performance on the TIMP and motor skills at preschool age.

Method

Participants

Sixty-one children, ages 4 to 5 years, participated in the study. As infants, the children had participated in a longitudinal study of the TIMP and had been tested weekly between 33 weeks postconceptual age and 4 months after term age. The infants were recruited from 3 nurseries in the Chicago metropolitan area and from the community.

Description of the original sample. The original sample consisted of 90 infants who were stratified into 5 groups based on degrees of risk for developmental disability and by ethnicity (Caucasian, African American, and Hispanic American). (17) The 5 groups were: (1) infants with central nervous system insults such as interventricular hemorrhage, (2) infants with bronchopulmonary dysplasia, (3) those born weighing less than 1,500 g or born before 32 weeks, (4) those born prematurely but without considerable medical problems, and (5) infants born at term. The POPRAS (29) (a perinatal medical complications index) also was used to assess the infants' medical risk. The POPRAS assesses prenatal, perinatal, and neonatal medical risk factors using information from the infant's medical record. Detailed information about the initial sample characteristics and the testing procedures has been reported elsewhere. (17,28)

Recruitment of participants for the current study fat preschool age). Recruitment methods were approved by the Institutional Review Board for the Protection of the Rights of Human Subjects at the University of Illinois at Chicago. Parents or guardians of all the children who participated in the TIMP longitudinal study were contacted. Telephone listings and addresses were updated using information from directory assistance for parents who had moved. The following methods were used to contact parents who had since changed places of residence and telephone numbers. First, letters were sent notifying parents about the study. Second, we searched the Internet directory assistance using the parents' last names or addresses. Third, the parents' or children's social security numbers obtained from the longitudinal study database were used to obtain information on the current residential address.

Of the 90 infants who had participated in the longitudinal study, 73 were located, 2 had died, and the remaining 15 were lost to follow-up. Of the latter group, 4 children and their families were reported to have moved to other states. Of the 73 children who were located, 61 completed the study, 4 families declined to participate, and 8 could not be tested due to repeated schedule conflicts. The characteristics of the children who completed the study as well as the characteristics of those who were lost to follow-up are presented in Table 1. To determine if this sample was biased, we used a t test to compare the mean TIMP logit scores, POPRAS scores, birth weights, and gestational ages for the children who completed testing at preschool with the mean scores of those who were lost to follow-up. None of the mean differences were statistically significant. The mean TIMP logits for the children who completed testing at preschool age, although lower than those who were not available for testing, was not statistically different (t=-1.88; df=2,71; P=.06).

Characteristics of participants at preschool age. The children's mean age at the time of the PDMS-2 testing was 57 months (SD=4.8, range=47-65). Thirty-one children (58%) were male. Fifty-two percent of the children were Caucasian, 25% were African American, 17% were Hispanic American, and 2% identified themselves as "other." The majority of the parents were married (73%), and 83% of the mothers had completed high school.

Instruments

The PDMS-2 (34) was used to assess the children's motor developmental status, and the Early Childhood Home Observation for Measurement of the Environment inventory (EC-HOME) (35) was used to assess the quality of the children's home and caregiving environment at preschool age. The PDMS-2 is a norm-referenced motor test designed to assess motor development in children from birth to 71 months of age. (34) The PDMS was recently revised and normed (in 1997-1998) on 2,003 children in 46 states and one Canadian province. (34) The PDMS-2 has 6 subtests: reflexes, stationary, locomotion, object manipulation (Gross Motor Composite), grasping, and visual-motor integration (Fine Motor Composite). A Total Motor Composite score combines the Gross Motor Composite and Fine Motor Composite subtests. Five types of scores can be obtained from the PDMS-2: raw scores, age equivalents, percentiles, standard scores for the subtests, and quotients for the composites. According to the test developers, (34) the most reliable score is the quotient. The average Cronbach alpha coefficients for the 6 age groups (0-1, 12-23, 24-35, 36-47, 48-59, and 60-71 months) ranged from .89 to .97. The test-retest reliability and content and construct validity are reported in the test manual. (34) Interrater reliability for the preschool age children was not reported in the manual. Interrater reliability data for age groups 3 to 11 months (n=30) and 15 to 36 months (n=30) were obtained from a scored set of 60 completed protocols by 2 independent raters (the raters did not administer the test). Interrater reliability coefficients ranged from r=.97 to r=.99 for subtests and from r =.96 to r=.98 for composite scores." (34)

The PDMS-2 was chosen for several reasons. Like the TIMP, it is a comprehensive measure of motor development and function. The test items of the PDMS-2 provide detailed descriptions of motor skills and were developed based on research on the effects of interventions on motor skill development. The test provides separate composite scores for gross and fine motor skills. Like the TIMP, item analysis and item response theory modeling were used to select test items. The first edition of the PDMS (36) has been widely used by pediatric therapists to assess children's gross and fine motor as well as overall motor ability" levels. (37,38) The PDMS-2 and PDMS have correlation coefficients of r=.84 and r=.91 for gross motor and fine motor composite scores, respectively.

The EC-HOME (35) consists of 55 binary items that are organized into 8 subscales. The EC-HOME is scored using semistructured interviews and observations of the home and mother-child interactions during a home visit. The environmental characteristics assessed by the EC-HOME include: the amount of developmental, academic, and language stimulation in the home; the child's physical environment; the parent's responsiveness to the child; availability of play materials; discipline; and other home characteristics indicative of parental concern with achievement. The EC-HOME scales are widely used in research and to plan interventions (18) and have sound psychometric properties. (39) Item analysis, using point-biserial item correlations, demonstrated moderate to high coefficients (.28-.64). (35) Correlation coefficients between the EC-HOME and some of the indexes of socioeconomic status, such as maternal and paternal levels of education, were moderate (r=.47-.57). (35) A recent factor analysis of 870 children identified 6 factors in the EC-HOME, and these factors explained 73.1% of the variance in the responses of children tested. (39) Three of the EC-HOME subscales have been found to be related to children's achievement test scores (r=.51-.55) (35) and motor skills (standardized path coefficient of .24) (21,22)

Reliability

Three testers participated in data collection. Prior to data collection, intrarater and interrater reliability for the PDMS-2 and the EC-HOME were estimated with 6 preschool-age children of varying abilities who were not part of the study. The first author (THK) had used both measures in research and clinical practice. (37) The reliability training for the PDMS-2 consisted of an approximately 6-hour didactic session, observations of videotaped test administrations, and actual administration of the test items and scoring by the testers. Intraclass correlation coefficients (ICC [2,1]) ranged from .92 to .94 for the subtests and Total Motor Composite score and from .91 to .96 for interrater reliability. For the EC-HOME, following a 4-hour didactic session, testers observed a videotaped interview by one of the developers of the EC-HOME (Bradley). All the testers scored one videotaped interview and an interview conducted by one other tester. The ICCs ranged from .91 to .94.

Procedure

Following telephone contact with the parents or guardians, an explanation of the study ensued using information from the consent form. Written parental consent to participate in the study was obtained from parents prior to testing. Testers were not aware of the children's TIMP scores or neonatal history. The TIMP scores were obtained after the PDMS-2 and EC-HOME were completed.

First, testers administered the PDMS-2 to the children according to instructions specified in the test manual. The test took approximately 1 to 1 1/2 hours to administer, depending on the degree of the child's cooperation. Next, testers completed the EC-HOME using interview and observation methods. The EC-HOME took approximately 45 minutes to administer. Testing was conducted in the children's homes with parents present. Although participants for this study were young children, only 2 children were not cooperative. This was handled, as instructed in the PDMS-2 manual, by postponing the completion of the session until another day or by administering items in a different order than that in the manual. The tests were scored according to the standardized instructions in the manual.

At the end of the session, the parent or guardian was asked to complete a questionnaire with items regarding demographics and the child's medical and developmental history. This information was used to determine whether the children had experienced episodes of medical problems, such as seizures, during the period between the TIMP and PDMS-2 testing. Conditions such as seizures would put them at risk for developmental delays or disabilities and would obscure the nature of the relationship between the 2 tests.

Data Analysis

Raw data obtained with the TIMP had been subjected to psychometric analysis in our previous study. (17) Using BIGSTEPS, a Rasch model of analysis, (40) raw ordinal scores were transformed into interval-level logit measures (logits). The mean was set at 50, and one logit was equal to 10 points. (41) The TIMP manual (42) presents the mean TIMP raw scores and logits for infants tested at various ages.

The Pearson product moment correlation coefficient was used to examine the relationship between the TIMP and the PDMS-2 scores. We selected the TIMP total raw scores and logit measures from tests performed 7, 30, 60, and 90 days after term age. The TIMP scores were correlated with scores on the Gross Motor Quotient (GMQ), Fine Motor Quotient (FMQ), and Total Motor Quotient (TMQ) of the PDMS-2.

Stepwise multiple regression analyses (43) were used, first to assess the contribution of TIMP scores at 7, 30, 60, or 90 days to the variance in the PDMS-2 scores at preschool age, and then to examine the additional contribution of the EC-HOME and POPRAS scores. (29) The POPRAS scores were included in the equation because previous research had indicated a negative relationship between a high degree of risk and motor performance. (30,33)

To determine the age at which the TIMP provided the best diagnostic values, we computed the sensitivity, specificity, PPVs, and NPVs for predicting performance on the PDMS-2. We examined various cutoff points on the

TIMP (eg, -0.25, -0.1), even though previous research on the predictive validity of data obtained with the TIMP and the AIMS indicated that -0.5 standard deviation below the mean created the best prediction with the AIMS at 12 months of age. (17) Two cutoff points were used for the PDMS-2:-2 and -1 standard deviations below the mean. The 2 cutoffs are used by various states to determine eligibility for children in need of services provided under the Individuals With Disabilities Education Amendments of 1997 (IDEA '97) mandate. (44)

Because fewer that 90 children were located, the 95% confidence interval (CI) was used to interpret the findings. According to Fletcher et al, (8) point estimates (eg, of sensitivity and specificity values) tend to be misleading when the sample is small. These authors recommended reporting the 95% CI for the range of values. We calculated the CIs for the sensitivity, specificity, PPVs, and NPVs using the formula proposed by Fletcher et al. (8) Lastly, the Pearson product moment correlation coefficient (partial correlation) between the TIMP and PDMS-2 scores, controlling for HOME scores, was used to determine whether the quality of the home environment affected the relationship between early motor development and the children's motor outcome at preschool age.

Results

The mean TIMP total raw scores and logits at each testing period are presented in Table 2. The mean TMQ at preschool age for the entire group was 88 (SD=22, range=38-115). The mean FMQ was 94 (SD= 22, range=46-36), and the mean GMQ was 85 (SD=20, range=41-111). Overall, these means were slightly below the PDMS-2 normative sample mean of 100. Seventeen children had scores below -1 standard deviation and 12 had scores below -2 standard deviations on the PDMS-2. The mean total EC-HOME score was 46 (SD=10, range=22-55), which was within the average range of the normative data provided in the test manual. (35) Fourteen (23%) of the children were reported to have received early intervention (therapy) after the TIMP testing was completed. Of these 14 children, 10 were diagnosed with cerebral palsy (CP), 1 was diagnosed with learning disabilities, 2 were diagnosed with attention-deficit disorders, and 1 was blind. No major illnesses were reported except that one child developed asthma at 2 years of age. Two of the children with CP had surgical interventions (adductor release and posterior dorsal rhizotomy, respectively).

The results of the Pearson product moment correlations are summarized in Tables 3 and 4. The correlation coefficients between the TIMP logit and PDMS-2 scores at preschool age were statistically significant at 30, 60, and 90 days after term age, but not at 7 days after term age. Coefficients ranged from .01 at 7 days to .69 at 90 days. The coefficients were the highest at 90 days. The TIMP score at this age explained 48% of the variance in the PDMS-2 scores.

Tables 5 through 7 summarize the results of the multiple regressions. The TIMP, EC-HOME, and medical risk (POPRAS) scores explained 24% of the variance in the PDMS-2 scores at 30 days and 23% of the variance at 60 days. At 90 days, these variables explained 52% of the variance in the PDMS-2 scores (an additional 4% of the variance of the PDMS-2 scores was explained by the EC-HOME and POPRAS scores).

The diagnostic efficiency values of the TIMP are presented in Tables 8 and 9. Using the cutoffs of -0.5 standard deviation on the TIMP and -2 standard deviations on the PDMS-2, the sensitivity increased from .33 at 30 days to .72 at 90 days. Specificity values were consistently high and ranged from .86 to .94. The PPVs ranged from .60 at 30 days to .75 at 90 days. The NPVs increased from .83 at 30 days to .91 at 90 days. The CIs for the sensitivity, specificity, PPVs, and NPVs also are reported in Tables 8 and 9. The age at which prediction of later motor development was most accurate was 90 days after term. The TIMP cutoff score of -0.5 standard deviation below the mean maximized the values and correctly classified 87% of the infants with or without motor developmental delay on the PDMS-2. Although the cutoff points of -0.25 and -0.1 were examined, their predictive values were low.

The diagnostic efficiency values of the TIMP using the cutoff of -1 standard deviation below the mean on the PDMS-2 are presented in Table 9. The values for specificity and PPV were higher compared with those obtained using the cutoff of -2 standard deviations. The sensitivity and NPV, however, were lower, indicating a high rate of false negatives.

The correlation coefficient between the EC-HOME scores and the PDMS-2 TMQ was .31 (P=.02). Table 4 presents the results of the partial correlation between the TIMP scores and PDMS-2 quotients. The coefficients were still statistically significant when the effects of the EC-HOME were removed, indicating that the quality of the home environment did not modify the relationship between TIMP and PDMS-2 scores.

Discussion

In earlier studies, we observed that the TIMP scores were correlated with scores on other motor tests such as the AIMS at 3, 6, and 12 months of age (17) and BOTMP at school age. (33) In the current study, the TIMP scores at 30, 60, and 90 days after term were correlated with the PDMS-2 total motor scores at age 4 to 5 years. In addition, an association was observed between the TIMP scores and the Gross Motor Composite and the Fine Motor Composite subtests of the PDMS-2. This relation ship was observed as early as 30 days after term and during subsequent testing up to 3 months after term. The results of the regression analysis suggest that the TIMP score at 3 months is likely to correctly identify an infant with or without motor delays compared with the scores at 1 or 2 months of age.

Information derived from correlations, although valuable in determining the relationships between early and later development, is limited in discriminating between children with and without developmental delay. To establish a diagnosis or predict outcome, information obtained from sensitivity, specificity, and predictive values of a test is needed. (8) At 90 days after term age, the TIMP cutoff score of 0.5 standard deviation below the mean provided the highest and the best prediction of motor performance at 4 to 5 years of age. At this age, the TIMP correctly identified 72% of the infants with poor motor performance and more than 90% of those who had no motor delays or disabilities at preschool age. Of the children identified by the TIMP as having atypical motor development at 90 days, 75% demonstrated poor motor performance at 4 to 5 years of age, as defined by PDMS-2 scores below -2 standard deviations. Of those children identified by the TIMP as having typical motor development, 91% exhibited typical motor performance.

Predictive values, and to some extent sensitivity and specificity values, are influenced by the prevalence of a condition in the population being tested. These values will be high depending on the proportion of individuals with or without the condition of interest and consequently need to be examined in the context of published findings on the prevalence of poor motor outcome in preterm infants. In our study, 12 children (19%) scored below -2 standard deviations on the PDMS-2 at preschool age. Of these children, 10 were diagnosed with CP, 1 was diagnosed with attention-deficit disorder, and 1 was legally blind. Eighty-one percent scored within the average range on the Fine Motor Composite and Gross Motor Composite subtests of the PDMS-2. The PDMS-2 misclassified 2 children who were diagnosed with mild CP and learning disabilities, bringing the total of children with disabilities to 14. Therefore, the prevalence of 19% to 23% observed in this study is within the 15% to 35% range of poor motor outcomes often reported in the literature. (45-48) For example, Paul et al (47) observed that, of the infants who were at risk for developmental delays and were admitted to the neonatal intensive care unit, 85% had a normal outcome and 15% had an adverse outcome at 3 years of age. Children in the study by Paul and associates were closer in age to our sample compared with the other studies.

Comparison of the TIMP With Other Tests

Analysis of the diagnostic values of the TIMP suggests that the test compares favorably with other developmental and neuromotor tests used with infants younger than 4 months of age. The tests include the MAI, (49) PDMS, (36) and Neurological Assessment of the Preterm and Full-Term Newborn Infant. (24) The sensitivity and specificity values reported for these tests ranged from 72% to 81% and 71% to 93%, respectively. (10) The TIMP offers an advantage over the other tests because of its ability to predict poor outcome at preschool age (PPV=75%). For example, the PPVs at 18 months of age were 39% for the AIMS, 58% for the MAI, and 31% for the PDMS. (10) The diagnostic values of these tests for preschool- or schoolage children are not known.

The results of the TIMP at 3 months of age are consistent with the General Movement Assessment (GM) method. The GM is an assessment tool that has been used to diagnose and predict neuromotor outcome in infants younger than 4 months of age. (50) The predictive validity for the GM was found to be good in identifying infants with CP. Cioni et al (51) reported the highest sensitivity (94%) and specificity (82%) for the GM method at 3 months. Additionally, it was found that repeated videotaping and analysis of the videotapes increased the predictive validity of data obtained with this approach. (52) The criterion measure was administered when the children were 2 years of age. The GM method is predicated upon the notion of age-related properties of the nervous system and the need for a neurological assessment to diagnose the system's impairments. (27,53) The unique feature of the GM method is the qualitative aspect of the movement patterns. According to Prechtl, (50) in the event of nervous system dysfunction, the movements assessed by the GM lose their complexity, while very, little change occurs in the quantity of spontaneous movements. The GM method does not offer age-related developmental scores that therapists need to determine children's eligibility for early intervention services. The GM method's fidgety and writhing movements, however, are included in the Observed section of the TIMP, but are only scored as present or absent. Their inclusion provides clinicians with an opportunity to observe these and other qualitative aspects of the infant's spontaneous movements during assessment.

Our findings underscore the need for the criterion measure to be administered past the toddler stage. Approximately 9 of the infants whose performance on the PDMS-2 was below -2 standard deviations at preschool age had average scores on the TIMP at 7 and 30 days after term. Research findings show that children who are premature and have low birth weight, who may not show deficits as infants, may fall behind peers as standardized test items become more challenging and developmental deficits become more evident. Goyen and Lui (54) in a sample (N=58) of premature infants with low birth weight (< 1,000 g), observed that infants tested at 12 months of age had no motor deficits but that 81.1% scored below average range on the PDMS Gross Motor Composite subtest and 63.8% scored below average range on the Fine Motor Composite subtest at age 5 years. Children classified as having false negative results had deficits that only became evident as they grew older and higher motor skills were expected. Similarly, Hadders-Algra (55) observed that forms of neurological dysfunction, such as dyskinesia, tended to be more pronounced as children grew older.

Ability to Diagnose

False positive and false negative results merit special attention. False positive test results represent those children who are likely to be erroneously referred for early intervention and generate undue concern by parents. False negative results, however, imply that some children with motor problems may not receive the necessary interventions. The TIMP at 90 days after term misclassified 6 of the 61 subjects. Of these children, 3 were classified as having normal motor performance on the TIMP but not on the PDMS-2 (false negative). Two of the 3 children were diagnosed with CP, and 1 child was diagnosed with attention and behavior problems. The latter child demonstrated difficulty in following test instruction, and this, rather than his motor abilities, contributed to his low PDMS score. An examination of the information gathered from the parents of these children revealed no remarkable medical history after 4 months of age when the TIMP testing was completed. Their performance on the TIMP at 7, 30, 60, and 90 days was unremarkable (Tab. 10). All 3 children, however, were born at less than 29 weeks gestational age and had many medical complications during the neonatal period (high POPRAS scores).

The other 3 children were classified as having motor delays on the TIMP but not on the PDMS-2 (false positive). Of these children, 2 scored -1 standard deviation below the mean on the PDMS-2 (TMQ=82 and 73). The 2 children had a diagnosis of "learning disability" and of "mild cerebral palsy," respectively (Tab. 11). These findings underscore the errors in classification that are associated with decisions regarding cutoff scores (56) and misclassifications that may be associated with the criterion measure (reference standard). For example, Glascoe (57) investigated whether children tested on a screening tool and classified as having false positive results had scores similar those of children who were classified as having true negative results. The children, ages 7 months to 8 years (N=512), were tested in areas of intelligence, language, and academic achievement.

Younger children were tested on diagnostic tests that assessed all developmental domains (including motor development). Children with false positive scores performed lower on diagnostic measures of adaptive behavior, intelligence, and academic achievement than children classified as having true negative results. Seventy percent of children with false positive results scored below the 25th percentile on one or more diagnostic measures compared with 29% of children with true negative results. In our study, the 2 children with CP and learning disability were misclassified by the PDMS-2 cutoff of -2 standard deviations and correctly diagnosed by the TIMP cutoff score of -0.5 standard deviation at 60 and 90 days of age. Cutoff scores, although useful for clinical decision making, impose an artificial dichotomy, suggesting that a certain number of misclassifications should be expected with any good test.

Sometimes false positive or false negative results can be explained by intervening factors. Analyses of the information from the parent questionnaire revealed that all except one of the children whose PDMS-2 scores were -2 standard deviations below the mean were receiving physical therapy, occupational therapy, or speech therapy. For many of these children, therapy was initiated as early as 6 months of age. Only 2 of the children with PDMS-2 scores greater than -2 standard deviations received therapy. Therefore, it is unlikely that therapy contributed to the false positive and negative findings of these children. Three children received surgical interventions (posterior dorsal rhizotomy, adductor release, and hernia repair).

In our study, the quality of the home environment, as measured by the EC-HOME, did not affect the relationship between the TIMP and PDMS-2 scores. We found a positive correlation between the EC-HOME and PDMS-2 scores, but not the TIMP and EC-HOME scores. The results of the partial correlation between the TIMP and PDMS-2 scores, controlling for the EC-HOME, and multiple regression suggest that the quality of the home, assessed at preschool, did not influence the relationship between early motor performance and later motor outcome. Neither could the false positive or false negative findings be explained by the quality of the home environment. In a previous study that demonstrated a mediating effect of the home environment, the HOME was administered before 36 months of age. (22) Because we administered the EC-HOME at 4 to 5 years of age rather than during infancy in our study, no assumptions can be made about whether the home did or did not relate to motor development. To make such assumptions, the home environment must be assessed during infancy.

Clinical Implications

Tests are unlikely to be both highly sensitive and specific. The trade-offs between these 2 indicators depend on the risk of misclassification that clinicians are willing to accept. A sensitive test is particularly valuable for ruling out the presence of a condition, and a specific test is particularly valuable for confirming a diagnosis. (8) A highly specific test also is important when the cost (physical, financial, or emotional) of false positive results is high. For therapists who work in developmental follow-up programs, a decision to erroneously refer an infant for early intervention services on the basis of test results may be costly in terms of financial and emotional burdens to families. On the contrary, the decision not to refer may result in a delay in initiating rehabilitation services or ineligibility for early intervention services under Part C of the Individuals With Disabilities Education Amendments Act of 1997. (44) Therefore, the implications of these findings to practice pertain to decisions about test selection and interpretation of test results.

Regarding test selection, using the TIMP at 3 months and a cutoff of -0.5 standard deviation will result in 72% of infants who are likely to demonstrate poor motor performance or CP at 4 to 5 years of age being correctly identified and referred. Rarely would an infant whose TIMP score at 3 months is average or above average be misdiagnosed (specificity=91%). In terms of prognosis or prediction, using the TIMP cutoff of -0.5 standard deviation, 75% of infants who are classified as having delayed development on the TIMP at 3 months are likely to be identified as having poor motor outcome or the PDMS-2 at preschool age. Eighty-five percent of these infants are likely to be diagnosed with CP or learning disabilities. Just as important, 91% of those infants whose scores are above -0.5 standard deviation are likely to have a good motor outcome at 4 to 5 years of age. Tables 8 and 9 present the diagnostic information for younger months. Overall, the accuracy of correctly diagnosing and predicting the outcome of an infant's motor performance at 3 months using the TIMP is 87%. The cutoff scores for various ages are reported in the test manual. (42)

False negative results suggest that clinicians should use their clinical judgment and other factors such as birth weight, gestational age, and neonatal history when making decisions about whether or not to refer infants for early intervention services or to initiate the services. The children who were misclassified as having false negative results in this study were born at less than 29 weeks gestational age and had many medical complications during the neonatal period.

The low correlation and diagnostic values at 7 and 30 days suggest that many infants who are likely to experience motor problems will be missed using the TIMP when tested as young as 7 or 30 days. This is a concern because the TIMP was intended to assess motor skills during the preterm and early infancy periods. It should be noted, however, that the diagnostic ability of the TIMP for infants at and below term age was not the focns of this study; therefore, professionals should exercise cantion when using the TIMP to predict later motor outcome with preterm infants up to 7 days after term.

Our findings also highlight the importance of repeated assessments when examining or screening infants for the presence of motor delays or risk of motor disability. Based on our results, an infant who scores below average on the TIMP at 60 and 90 days is most likely to exhibit poor motor performance at preschool age. These findings underscore the need for developmental surveillance for infants with average or low average TIMP scores, but who also were born at less than 29 weeks gestational age, and have experienced multiple medical complications during the perinatal and postnatal periods.

Conclusions

This study examined the ability of the TIMP to predict motor development of children at preschool age. The TIMP score at 90 days after term identified the majority of infants who experienced developmental motor delays or disabilities at preschool age and most of those with typical development (85%). The TIMP's diagnostic efficiency values compare favorably with data obtained with other tests of neurological integrity and can be used to predict motor outcome with infants at younger ages compared with other developmental tests. The number of false negatives observed in this study suggests the need for clinicians to consider factors such as gestational age at birth and neonatal medical history of infants when interpreting early negative TIMP test scores and to monitor the infants closely over time.

Table 1.
Characteristics of Infants Who Were Tested and Not Tested at
Preschool Age (a)

                    Birth       GA at
                    Weight      Birth   POPRAS   TIMP
                    (g)         (wk)    Score    Logits

Tested (n=61)
  X                 1,802.28    31.14   109.66   47.10
  SD                1,183.61     5.40    70.41    7.44
  Range             578-4,192   24-41   1-225    27-70
Not tested (n=29)
  X                 1,796.06    32.37   106.11   43.53
  SD                1,113.93     6.03    67.04    6.34
  Range             567-4,240   23-40   2-216    32-55

(a) GA=gestational age, POPRAS=Problem-Oriented Perinatal Risk
Assessment System, TIMP=Test of Infant Motor Performance.

Table 2.
Test of Infant Motor Performance (TIMP) Scores at
7, 30, 60, and 90 Days After Term

Measure           7 d      30 d     60 d     90 d

TIMP logit
  X              46.84    51.34     57.82    61.86
  SD              4.20     4.52      6.58     8.42
  Range          35-55    35-59     41-71    44-83
TIMP raw score
  X              71.14    86.75    105.65   114.55
  SD             18.06    15.73     20.20    25.03
  Range          34-108   38-113   48-143   65-159

Table 3.
Pearson Product Moment Correlation Coefficients Between Test
of Infant Motor Performance (TIMP) Scores at Various Ages
and Peabody Developmental Motor Scales, 2nd Edition (PDMS-2)
Scores (a)

                                     PDMS-2

                   TMQ            FMQ            GMQ

TIMP logit: 7 d    .01 (P=.74)    .04 (P=.77)    .06 (P=.66)
TIMP logit: 30 d   .48 (P=.001)   .47 (P=.001)   .45 (P=.002)
TIMP logit: 60 d   .45 (P=.001)   .44 (P=.001)   .44 (P=.001)
TIMP logit: 90 d   .69 (P=.001)   .67 (P=.001)   .66 (P=.001)

(a) TMQ=Total Motor Quotient, FMQ=Fine Motor Quotient,
GMQ=Gross Motor Quotient.

Table 4.
Pearson Product Moment Partial Correlation Coefficients Between
Test of Infant Motor Performance (TIMP) Scores at Various Ages
and Peabody Developmental Motor Scales, 2nd Edition (PDMS-2)
Scores, Controlling for the Early Childhood Home Observation
for Measurement of the Environment Inventory (EC-HOME) Score

                                   PDMS-2 (a)

                   TMQ            FMQ            GMQ

TIMP logit: 30 d   .44 (P=.003)   .41 (P=.006)   .43 (P=.004)
TIMP logit: 60 d   .42 (P=.003)   .41 (P=.005)   .42 (P=.003)
TIMP logit: 90 d   .68 (P=.001)   .67 (P=.001)   .65 (P=.001)

(a) TMQ=Total Motor Quotient, FMQ=Fine Motor Quotient,
GMQ=Gross Motor Quotient.

Table 5.
Results of Multiple Regression Equations for Prediction of Peabody
Developmental Motor Scales, 2nd Edition (PDMS-2) Scores From the
Infants' Scores on the Test of Infant Motor Performance (TIMP) at
30 Days, Early Childhood Home Observation for Measurement of the
Environment Inventory (EC-HOME), and Problem-Oriented Perinatal
Risk Assessment System (POPRAS)

                             Standardized
                             Regression
Predictor Variable           Coefficient (Beta)   t       P

R=.55, P=.003, adjusted       0.41                 2.79   .01
  [R.sup.2]=.24 TIMP logit
POPRAS                       -0.26                -1.73   .09
EC-HOME                      -0.06                -0.39   .70

Table 6.
Results of Multiple Regression Equations for Prediction of Peabody
Developmental Motor Scales, 2nd Edition (PDMS-2) Scores From the
Infants' Scores on the Test of Infant Motor Performance (TIMP) at
60 Days, Early Childhood Home Observation for Measurement of the
Environment Inventory (EC-HOME), and Problem-Oriented Perinatal
Risk Assessment System (POPRAS)

                             Standardized
                             Regression
Predictor Variable           Coefficient (Beta)   t       P

R=.52, P=.002, adjusted       0.35                 2.31   .02
  [R.sup.2]=.24 TIMP logit
POPRAS                       -0.12                -0.70   .48
EC-HOME                       0.23                 1.63   .11

Table 7.
Results of Multiple Regression Equations for Prediction of Peabody
Developmental Motor Scales, 2nd Edition (PDMS-2) Scores From the
Infants' Scores on the Test of Infant Motor Performance (TIMP) at
90 Days, Early Childhood Home Observation for Measurement of the
Environment Inventory (EC-HOME), and Problem-Oriented Perinatal
Risk Assessment System (POPRAS)

                             Standardized
                             Regression
Predictor Variable           Coefficient (Beta)   t       P

R=.74, P=.001, adjusted      0.76                 5.67    .001
  [R.sup.2]=.52 TIMP logit
POPRAS                       0.18                 1.25    .21
EC-HOME                      0.29                 2.52    .01

Table 8.
Summary of Sensitivity, Specificity, and Positive and
Negative Predictive Values Based on the Test of Infant
Motor Performance (TIMP) Cutoff Point of -0.5 Standard
Deviation and Peabody Developmental Motor Scales, 2nd
Edition (PDMS-2) Score of -2 Standard Deviations (95%
Confidence Interval in Parentheses)

Age    Sensitivity     Specificity

30 d   .33 (.19-.47)   .94 (.87-1.0)
60 d   .50 (.35-.65)   .86 (.76-.96)
90 d   .72 (.59-.83)   .91 (.83-.99)

       Positive        Negative
       Predictive      Predictive      %
Age    Values          Values          Correct

30 d   .60 (.45-.75)   .83 (.72-.94)   80%
60 d   .55 (.40-.70)   .84 (.73-.95)   79%
90 d   .75 (.63-.88)   .91 (.83-.99)   87%

Table 9.
Summary of Sensitivity, Specificity, and Positive and
Negative Predictive Values Based on the Test of Infant
Motor Performance (TIMP) Cutoff Point of -0.5 Standard
Deviation and Peabody Developmental Motor Scales, 2nd
Edition (PDMS-2) Score of -1 Standard Deviation (95%
Confidence Interval in Parentheses)

Age    Sensitivity     Specificity

30 d   .29 (.15-.43)   .92 (.84-1.0)
60 d   .44 (.30-.58)   .97 (.92-1.0)
90 d   .62 (.48-.76)   .97 (.92-1.0)

       Positive        Negative
       Predictive      Predictive      %
Age    Values          Values          Correct

30 d   .80 (.68-.92)   .72 (.59-.85)   73%
60 d   .89 (.80-.98)   .74 (.61-.87)   77%
90 d   .92 (.84-1.0)   .82 (.71-.93)   85%

Table 10.
Summary of the Test of Infant Motor Performance (TIMP) and Peabody
Developmental Motor Scales, 2nd Edition (PDMS-2) Scores of Infants
Who Were Misclassified by the TIMP (False Negatives) (a)

      TIMP                       PDMS-2
ID
No.   7 d   30 d   60 d   90 d   TMQ   GMQ   FMQ   Diagnosis

 50   A     A      A      A      64    61    76    SD
 72   A     N/A    A      A      38    41    46    QP
112   A     LA     A      A      70    74    70    ADD

(a) A=average, LA=low average, N/A-not tested, SD=spastic diplegia,
QP=quadriplegia, ADD=attention-deficit disorders, TMQ=Total Motor
Quotient, FMQ=Fine Motor Quotient, GMQ=Gross Motor Quotient.

Table 11.
Summary of the Test of Infant Motor Performance (TIMP) and Peabody
Developmental Motor Scales, 2nd Edition (PDMS-2) Scores of Infants
Who Were Misclassified by the TIMP (False Positives) (a)

      TIMP                       PDMS-2
ID
No.   7 d   30 d   60 d   90 d   TMQ   GMQ   FMQ   Diagnosis

 51   LA    A      BA     BA     82    81    88    LD
 81   N/A   N/A    A      BA     92    91    94    No diagnosis
109   N/A   LA     FBA    LA     73    68    85    "Mild CP"

(a) A=average, LA=low average, BA=below average, FBA=far below
average, N/A-not tested, LD=learning disability, CP=cerebral
palsy, TMQ=Total Motor Quotient, FMQ=Fine Motor Quotient,
GMQ=Gross Motor Quotient.

References

(1) Meisels SJ, Shonkoff JP. Early childhood intervention: a continuing evolution. In: Meisels SJ, Shonkoff JP, eds. Handbook of Early Childhood Intervention. 2nd ed. New York, NY: Cambridge Press; 2000:3-31.

(2) Coolman RB, Bennett FC, Sells CJ, et al. Neuromotor development of graduates of the neonatal intensive care unit: patterns encountered in the first two years of life. J Dev Behav Pediatr. 1986;6:327-333.

(3) Piper MC, Mazer B, Silver KM, Ramsay M. Resolution of neurological symptoms in high-risk infants during the first two years of life. Dev Med Child Neurol. 1988;30:26-35.

(4) Wildin SR, Anderson AE, Woodside M, et al. Prediction of 12-month neurodevelopmental outcome from a 6-month neurologic examination in premature infants. Clin Pediatr (Phila). 1995;5:290-299.

(5) de Vries AM, Groot L. Transient dystonias revisited: a comparative study of preterm and term children at 2 1/2 years of age. Dev Med Child Neurol. 2002;44:415-421.

(6) Leonard CH, Clyman RI, Piecuch RE, et al. Effect of medical and social risk factors on outcome of prematurity and very low birth weight. J Pediatr. 1990;116:620-626.

(7) Wildin SR, Smith KE, Anderson AE, et al. Prediction of developmental patterns through 40 months from 6- and 12-month neurologic examinations in very low birth weight infants. J Dev Behav Pediatr. 1997;18:215-221.

(8) Fletcher RH, Fletcher SW, Wagner EH. Clinical Epidemiology: The Essentials. Baltimore, Md: Williams & Wilkins; 1996.

(9) Aylward GP, Verhulst SJ. Predictive utility of the Bayley Infant Neurodevelopmental Screener (BINS) risk status classifications: clinical interpretation and application. Dev Med Child Neurol. 2000;42: 25-31.

(10) Darrah J, Piper MC, Watt MJ. Assessment of gross motor skills of at-risk infants: predictive validity of the Alberta Infant Motor Scale. Dev Med Child Neurol. 1998;40:485-491.

(11) Magaziner V, Molteno C, Lachman P, et al. A neuromotor screening test for high risk infants in a hospital or community setting. Pediatric Physical Therapy. 1997;9:66-72.

(12) Morgan AM, Aldag JC. Early identification of cerebral palsy using a profile of abnormal motor patterns. Pediatrics. 1996;98:692-697.

(15) Allen MC, Alexander GR. Using motor milestones as a multistep process to screen preterm infants for cerebral palsy. Dev Med Child Neurol. 1997;39:12-16.

(14) Touwen BCL, Lok-Meijer TY, Huisjes HJ, Olinga AA. The recovery rate of neurologically deviant newborns. Early Hum Dev. 1982;7: 131-148.

(15) Darrah J, Hodge M, Magill-Evans J, Kembhavi G. Stability of serial assessment of motor and communication abilities in typically developing infants: implications for screening. Early Hum Dev. 2003;72:97-110.

(16) Erickson C, Allert C, Carlberg EB, et al. Stability of longitudinal motor development in very low birthweight infants from 5 months to 5.5 years. Acta Paediatr. 2003;92:197-203.

(17) Campbell SK, Kolobe THA, Wright BD, Linacre JM. Validity of the Test of Infant Motor Performance for prediction of 6-, 9-, and 12-month scores on the Alberta Infant Motor Scale. Dev Med Child Neurol. 2002;44:263-272.

(18) Bradley RH. Children's home environments, health, behavior, and intervention efforts: a review using the HOME inventory as a marker measure. Genet Soc Gen Psychol Monogr. 1993;19:437-490.

(19) Levy-Shiff R. Einat G, Mogilner MB, et al. Biological and environmental correlates of developmental outcome of prematurely born infants in early adolescence. J Pediatr Psychol. 1994;19:63-78.

(20) Helms JE. Why is there no study of cultural equivalence in standardized cognitive ability testing? Am Psychol. 1992;47:1083-1101.

(21) Zahr LK. Predictors of development in premature infants from low-income families: African Americans and Hispanics. J Perinatol. 1999;19:284-289.

(22) Garret P, Ferron J, Ngandu N, et al. A structural model for the developmental status of young children. J Marriage Fam. 1994;56: 147-163.

(23) Campbell SK, Kolobe THA, Osten ET, et al. Evidence for the construct validity of the Test of Infant Motor Performance. Phys Ther. 1995;75:585-596.

(24) Dubowitz L, Dubowitz V. The Neurological Assessment of the Preterm and Full-Term Newborn Infant. Philadelphia, Pa: JB Lippincott; 1981.

(25) Brazelton TB, Neonatal Behavioral Assessment Scale. 2nd ed. Philadelphia, Pa: JB Lippincott Co; 1984. Clinics in Developmental Medicine.

(26) Amiel-Tison C, Grenier A. Neurologic Assessment During the First Year of Life. New York, NY: Oxford University Press; 1986.

(27) Cioni G, Prechd HFR. Preterm and early postterm motor behavior in low risk premature infants. Early Hum Dev. 1990;23:159-191.

(28) Campbell SK, Hedeker D. Validity of the Test of Infant Motor Performance for discriminating among infants with varying risk for poor motor outcome. J Pediatr. 2001;139:546-551.

(29) Davidson EC, Hobel CJ. POPRAS: A Guide to Using the Prenatal, Intrapartum, Postpartum Record. Torrance, Calif: South Bay Regional Perinatal Project Professional Staff Association; 1978.

(30) Campbell SK, Kolobe THA. Concurrent Validity of the Test of Infant Motor Performance with the Alberta Infant Motor Scale. Pediatric Physical Therapy. 2000;12:2-9.

31 Campbell SK. Test-retest reliability of the Test of Infant Motor Performance. Pediatric Physical Therapy. 1999;11:60-66.

(32) Linacre JM. FACETS: Computer Program for Many-Faceted Rasch Measurement. Chicago, Ill: Mesa Press; 1988.

(33) Flegel J, Kolobe THA. Predictive Validity of the Test of Infant Motor Performance as measured by the Bruininks-Oseretsky Test of Motor Proficiency at school age. Phys Ther. 2002;82:762-771.

(34) Folio MR, Fewell RR. Peabody Developmental Motor Scales. 2nd ed. Austin, Tex: Pro-Ed Inc; 2000.

(35) Caldwell BM, Bradley RH. Home Observation for Measurement of the Environment. Arkansas: Ark: Center for Research on Teaching and Learning at the University of Arkansas; 1984.

(36) Folio MR, Fewell RR. Peabody Developmental Motor Scales and Activity Cards Manual. Allen, Tex: DLM Teaching Resources; 1983.

(37) Kolobe TH, Palisano RJ, Stratford PW. Comparison of two outcome measures for infants with cerebral palsy and infants with motor delays. Phys Ther. 1998;78:1062-1072.

(38) Palisano RJ, Kolobe TH, Haley SM, et al. Validity of the Peabody Developmental Gross Motor Scale as an evaluative measure of infants receiving physical therapy. Phys Ther. 1995;75:939-948.

(39) Bradley RH, Mundfrom DJ, Whiteside L, et al. A factor analytic study of the infant toddler and early childhood versions of the HOME inventory administered to white, black, and Hispanic American parents of children born preterm. Child Dev. 1994;65:880-888.

(40) Wright BD, Linacre GN, BIGSTEPS: A Ranch-Model Computer Program. Chicago, Ill: MESA Press; 2000.

(41) Wright BD, Linacre JM. A User's Guide to BIGSTEPS. Chicago, Ill: MESA Press; 1996.

(42) Campbell SK. The Test of Infant Motor Performance: Test User's Manual Version 1.4. Chicago, Ill: Infant Motor Performance Scales; 2001.

(43) Daniel WW. Biostatistics: A Foundation for Analysis in the Health Sciences. New York, NY: John Wiley & Sons Inc; 1999.

(44) Individuals With Disabilities Education Act Amendments of 1997. 1997: 37-157. Available at: www.ed.gov/policy/speced/leg/idea/idea.pdf.

(45) Pinto-Martin JA, Whitaker AH, Feldman JF, et al. Relation of cranial ultrasound abnormalities in low-birthweight infants to cognitive performance at ages 2, 6, and 9 years. Dev Med Child Neurol. 1999;41: 826-833.

(46) Stevenson DK, Wright LL, Lemons JA, et al. Very low birth weight outcomes of the National Institute of Child Health and Human Development Neonatal Research Network, January 1993 through December 1994. Am J Obstet Gynecol. 1998;179(6 pt 1):1632-1639.

(47) Paul VK, Radhika S, Deorari AK, Singh M. Neurodevelopmental outcome of "at risk" nursery graduates. Indian J Pediatr. 1998;65: 857-862.

(48) Paneth N. The causes of cerebral palsy. Clinical Investigative Medicine. 1993;16:95-102.

(49) Chandler LS, Andrews MS, Swanson MW. Movement Assessment of Infants: A Manual. Washington: DC: Rolling Bay; 1980.

(50) Prechtl HF. General movement assessment as a method of developmental neurology: new paradigms and their consequences. The 1999 Ronnie McKeith Lecture. Dev Med Child Neurol. 2001;43:836-842.

(51) Cioni G, Prechfl HFR, Ferrari F, et al. Which better predicts later outcome in fullterm infants: quality of general movements or neurological examination? Early Hum Dev. 1997;50-71-85.

(52) Cioni G, Ferrari F, Einspieler C, et al. Comparison between observation of spontaneous movements and neurologic examination in preterm infants. J Pediatr. 1997;130:704-711.

(53) Prechtl HFR. State of the art of a new functional assessment of the young nervous system: an early predictor of cerebral palsy. Early Hum Dev. 1997;50:1-11.

(54) Goyen TA, Lui K. Longitudinal motor development of "apparently normal" high-risk infants at 18 months, 3 and 5 years. Early Hum Dev. 2002;70:103-115.

(55) Hadders-Algra M. Two distinct forms of minor neurological dysfunction: perspectives emerging from a review of data of the Groningen Perinatal Project. Dev Med Child Neurol. 2002;44:561-571.

(56) Dewyer CA. Cut scores and testing: statistics, judgment, truth, and error. Psychol Assess. 1996;8:360-362.

(57) Glascoe FP. Are overreferrals on developmental screening tests really a problem? Arch Pediatr Adolesc Med. 2001;155:54-59.

THA Kolobe, PT, PhD, is Associate Professor, Department of Rehabilitation Science, University of Oklahoma Health Sciences Center, 801 NE 13th St, Oklahoma City, OK 73104 (USA) ([e-mail]). Address all correspondence to Dr Kolobe.

M Bulanda, PT, MS, PCS, is Clinical Assistant Professor, Department of Physical Therapy, University of Illinois at Chicago, Chicago, Ill.

L Susman, PT. is a doctoral student, Department of Physical Therapy, University of Illinois at Chicago.

Dr Kolobe provided concept/idea/research design, project management, and subjects. Dr Kolobe and Ms Bulanda provided writing, fund procurement, and institutional liaisons. All authors provided data collection, and Dr Kolobe and Ms Susman provided data analysis. The authors acknowledge Suzann K Campbell for provision of subjects, fund procurement, and data for the Test of Infant Motor Performance (funded by a grant from the National Institutes of Health-National Center for Medical Rehabilitation Research) and Leticia Ishi for clerical support.

This study was approved by the Institutional Review Board for the Protection of the Rights of Human Subjects of the University of Illinois at Chicago.

This study was supported, in part, by a grant from the American Physical Therapy Association Section on Pediatrics.

This research was presented, in part, at the Combined Sections Meeting of the American Physical Therapy Association; February 12-16, 2003; Tampa, Fla.

This article was received October 14, 2003, and was accepted May 28, 2004.