Secondary Logo

Journal Logo

Research Report

Discriminating Infants From Different Developmental Outcome Groups Using the Test of Infant Motor Performance (TIMP) Item Responses

Barbosa, Vanessa Maziero PhD, OTR/L; Campbell, Suzann K. PhD, PT, FAPTA; Berbaum, Michael PhD

Author Information
Pediatric Physical Therapy: April 2007 - Volume 19 - Issue 1 - p 28-39
doi: 10.1097/PEP.0b013e31802f65f9
  • Free


Only through reliable, early identification of cerebral palsy (CP) will it be possible to implement timely intervention and evaluate its efficacy, but improving the accuracy of identification remains a challenge. Researchers have assessed the diagnostic validity of various tests, but few have cited specific early neuromotor behaviors that are most predictive of CP.1–3 When specific behaviors were considered, a neuromaturational approach typically was taken. This included assessment of muscle tone, reflexes, and motor milestones.4,5 On the other hand, tests that include items to assess voluntary movement, such as the Movement Assessment of Infants (MAI)6 and the Alberta Infant Motor Scales (AIMS),7 usually have the best psychometric properties, including predictive validity for identifying delayed motor development, for children four months of age and older. Individual movement behaviors that are most predictive of CP at earlier ages are just beginning to be identified and, until recently, no quantitative assessment was available for assessing infants' functional motor performance at ages younger than four months.

The Test of Infant Motor Performance (TIMP) is an assessment of posture and movement for infants from 32 weeks' postconceptional age through four months' corrected age (CA).8 The test (Version 3) consists of two sections: 28 items for recording observations of spontaneous movements, such as head centering, reaching, and individual finger movements, and 31 elicited items that present functional movement problems for the infant to solve, such as placing the infant in supine position and presenting a visual stimulus to assess the infant's ability to maintain the head in midline position. Rasch analysis9 guided the development of the TIMP. Because it fits the Rasch psychometric model, the TIMP is a hierarchical scale of difficulty that assesses the fit between the child's response pattern and the item's difficulty levels.

Previous research on the TIMP8 showed that the test is sensitive to age-related changes in motor performance (r = 0.83) and medical complications.10 A TIMP cutoff of −0.5 SD at 90 days' CA in comparison with AIMS percentile ranks (PR) greater than and less than the 5th PR at age 12 months correctly identified 78% of the infants at 12 months' CA, with sensitivity at 92% and specificity at 76%.11 The TIMP total score has also been shown to be able to identify children with CP.12 Children with CP began presenting low TIMP scores as early as seven days' CA, and they tended to be consistently less than −0.5 SD in TIMP performance, ie, showing delayed functional motor performance relative to same-age peers.

Specific item responses were recently studied13 with the goal of identifying the differences in item performance of infants classified at one to one and one-half years of age as having CP, typical development (TD), or developmental delay (DD). Numerous items made specific contributions to identifying differences among the groups. Items that assessed the alignment of the head in midline, anterior/posterior head control in supported sitting, and lateral head righting, especially at the ages of two and three months' CA, appeared to be most helpful in characterizing differences between children with and without CP.

Although the previous study identified items that differed among the three outcome groups, it did not evaluate how fast these skills developed nor how item performances combined differently leading to specific developmental outcomes. Although developmental skills are gained at different rates in typically developing children,7 of particular interest is how rates of change in different skills combine or cluster differently in those with atypical developmental outcomes. Are abnormal rates of change on single skills as captured by each individual test item (ie, inability to maintain head in midline) representative of abnormal developmental patterns, or are the skills' rates of change grouped together (ie, unable to maintain head in midline plus inability to sustain legs flexed against gravity) in ways that prevent the child from developing as expected? According to the knowledge that typically developing children at early ages (ie, zero to four months) exhibit gains in motor skills about every two weeks as measured by the TIMP,8 we hypothesized that assessment of the rates of change at specific ages can further help in the diagnosis of developmental disability. The major question posed for this study was how do rates of change in performance on TIMP items cluster differently among children with different outcomes, especially a later diagnosis of CP? The goal of this study was to explore whether the rate of change in motor skills early in development can contribute to the early diagnosis of developmental motor problems.



The convenience sample consisted of infants assessed in previous TIMP research,10,11 born during the years 1996–1998, and recruited from the special care nurseries of three hospitals or from the community within the Chicago metropolitan area. Subject 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 and at each field-testing site. The original sample was 96 infants with a wide range of risk for poor developmental outcome, including 19 infants born low-risk full-term (FT) with no significant medical problems, 20 infants born prematurely (PT) weighing more than 1500 g and with no significant medical problems, 11 infants with very-low-birth weight (VLBW) born weighing less than 1500 g or before 32 weeks' gestational age (GA) but without chronic lung disease or brain insults, 27 infants with chronic lung disease (bronchopulmonary dysplasia), and 19 infants with documented central nervous system (CNS) insult. The overall sample included 37% white infants, 35% black, 26% Latino/a (primarily Mexican or Puerto Rican), 1% Asian, and 2% mixed race. There were 51 male and 45 female infants. The infants were classified at one to one and one half years of age as TD, DD, or as having CP. Table 1 provides a summary of information relating children's original risk group to their developmental outcome.

Original Risk Group Distribution By Developmental Outcome at One and One-Half Years of Age


Infants were tested weekly on the TIMP from age of recruitment shortly after birth (varying from 32 weeks' GA to approximately three weeks' CA based on age and health status) until 16 weeks of CA, resulting in anywhere from two to 22 assessments (median = 8). Infants also were tested on the AIMS7 at 12 months' CA. Eleven testers participated in this study. Methods of subject recruitment, testing, and determining rater reliability on the TIMP and AIMS have been previously reported.10,11 A follow-up with testers at each site was performed to obtain information on which infants had received a subsequent diagnosis of CP after the 12-month testing was complete, but there was no further testing of these infants in this study.

The criteria used to classify developmental outcome at one to one and one half years of age were a developmental pediatrician's clinical judgment and the child's performance on the AIMS7 at 12 months. For the purposes of this study, a score less than the 16th percentile rank (PR) on the AIMS at 12 months without a diagnosis of CP was used to define delayed development. Eight infants presented DD and no CP. The medical diagnosis of CP was based on the physicians' clinical judgment. Ten infants were diagnosed as having CP. Nine infants also had an AIMS score at 12 months below the fifth percentile; one infant's AIMS score was at the 49th percentile and was diagnosed as having mild CP at one and one-half years of age. Details on these infants' TIMP performance and services received are discussed elsewhere.7 Eleven infants did not have any outcome information available either on the AIMS or relative to the diagnosis of CP and they were eliminated from the analysis leaving a sample size of 85 infants. Children were followed up to four to five years of age as part of another TIMP-related longitudinal study14 confirming their CP diagnosis.

Individual items' responses were explored for the three groups of children according to their developmental outcome at one and one-half years of age (ie, TD, n = 67; DD, n = 8; or CP, n = 10). We attempted to identify unique patterns in TIMP item performance/rate of change that characterized each group and that might suggest developmental problems and discriminate among infants with different outcomes.

Because our interest was to assess the relationship between each item's rate of change and outcome group and because the developmental rate of each skill might vary at different points in time (ie, early vs later periods in development), regressions with cubic polynomials in terms of weeks of age at testing (independent variables) were run for each individual TIMP item score (dependent variable) and each individual child. Cubic polynomials were chosen because they capture nonlinear changes and more variability among children, thus allowing more discrimination among the groups.15 This procedure allowed identifying the items' trend lines over time for each TIMP item for each child separately (rates of change at different ages). Because previous TIMP research pointed to eight and 13 weeks (ie, 60 and 90 days' CA, respectively) as times when most differences in infants' performance appeared to occur,13 these ages were selected as specific time points to derive items' slope coefficients. Slope coefficients (tangent line to the polynomial curves) were derived to determine the instantaneous rate of change at both eight and 13 weeks CA. These coefficients represent a change trend of all the data each child has in each TIMP item. Next, a mean slope coefficient was calculated for each outcome group to assess the rate of change of each item per group. All of these procedures were preparatory steps to performing the analysis of interest. We do not present these intermediate results here; however, this information is available upon request.

Data Analysis

Discriminant analysis was used to investigate how individual TIMP items' rates of change at eight and 13 weeks CA, alone or combined, behaved in classifying children with different outcomes (ie, TD, DD, and CP). Discriminant analysis, using each item's rate of change as independent variables and the developmental outcome groups as dependent variables, were run at eight and 13 weeks' CA. The goal of discriminant analysis15 is to provide group separation based on predefined outcome groups in terms of multivariate measurement profiles. Different methods of entering the independent variables in the discriminant analysis were explored. First, all TIMP item slopes were entered into the analysis simultaneously so that information from all items was combined to obtain the maximum discrimination among groups. Next, stepwise discriminant analysis was used in an attempt to reduce the discriminant function to a parsimonious minimum of relevant variables. Stepwise analyses were performed twice: first with the SPSS® statistical analysis program (SPSS Institute, Chicago, IL) default F value (3.84) to enter into the equation, second with the probability of F to enter into the equation set to be greater than 0.10 but smaller than 0.20, which allowed more items to enter into the discriminant equation while still being parsimonious in comparison to when all item slope variables were simultaneously entered into the equation.

The Wilks' Lambda statistic was used to determine whether the different TIMP items' slopes (predictor variables) could, as a set, differentiate among the infants with CP, DD, and TD. This is a test of the equality of mean scores on each of the predictors among the criterion groups (group centroids—points defined by a group's means on all items included). The standardized coefficient (ie, discriminant coefficient) represents the relative importance of the independent variables (ie, TIMP item slopes) with which each discriminant function is associated. The structure coefficient represents the correlation between the discriminant scores for each child and the slopes on each one of the TIMP items. In other words, structure coefficients show how much each item correlates with each child's discriminant score. Structure coefficients of 0.30 or greater are considered meaningful.15 The square of the structure coefficient is the proportion of variance in a particular variable explained by the discriminant function.


The results of all analyses at eight and 13 weeks are presented in Tables 2–4. Table 2 describes the results of the discriminant analysis. In this table, one can identify the items that correlated with each discriminant function for each separate set of analyses. This allows assessing the specific item's rate of change contribution in correctly classifying the infants according to their developmental outcome group. Table 3 describes the predictive values of all discriminant analysis runs for each outcome group separately. Table 4 presents a summary of the items' slopes (slope value and direction) per group at eight and 13 weeks for the slopes that contributed to the discriminant function using the stepwise analysis with probability of F values set to be >0.10 and <0.20.

Discriminant Functions' Summary per Individual Analyses: Differentiated Group, Percentage of Variance Explained, Contributing Items
Summary of Predictive Validity of the Discriminant Function Analysis
Items That Load in Each Discriminant Function at 8 and 13 Weeks Using the Stepwise With Probability of F for Variable to Enter into the Equation of 0.10 < F > 0.20

Discriminant Analysis of Items' Slopes at Eight and 13 Weeks

When all item slopes at eight weeks were entered simultaneously into the analysis, two discriminant functions were statistically significant indicating that the 59 TIMP items' slopes as a group discriminated among the children with different outcomes (Lambda = 0.001, equivalent to an F ratio of 171.491 with 118 degrees of freedom). Whereas function 1, which explained 68.6% of the between-group variance, was better able to identify children with TD, function 2, although explaining a smaller percentage of variance (31.4%), differentiated among all groups (Table 2). Using 59 items to predict classification of 85 infants, both functions together correctly classified 100% of the children (Table 3), but none of the independent variable's structure coefficients reached the 0.30 recommended cutoff15 to explain how much an item contributes to the function. Table 2 shows the variables that correlated best with each discriminant function.

Because greater parsimony is needed in practical work, we focused on the more economic approach of using the stepwise method of entering the variables in the discriminant analysis. In the default stepwise method of entering variables (ie, SPSS® statistical analysis program, F value = 3.84), two discriminant functions were identified. Function 1 explained 75.5% of the between group variance and function 2 explained 24.5%. In function 1 all groups were different from each other but, in function 2, only children with DD appeared distinct from the others. Overall, 63 (74.2%) of the 85 infants were correctly classified. Table 3 presents the predictive values for each outcome group separately. Only five TIMP item slopes statistically contributed to the discriminant functions (Table 2). In this analysis each item loading was greater than the 0.30 criteria used to identify important predictors for the discriminant function.

Stepwise analysis with the probability of F to enter into the equations set to be >0.10 and <0.20, a more relaxed criterion to enter variables in the discriminant analysis, resulted in 10 variables accounting for the variability of the discriminant functions. Figure 1 shows how successful the discriminant functions are in distinguishing group membership. The group centroids indicate how the groups are separated. Function 1 discriminated the three outcome groups from each other. Discriminant function 2 discriminated the infants with CP from the infants with TD and with DD. Although the specific contribution of function 1 decreased in comparison with the default stepwise analyses, this analysis not only increased the number of predictors in the equation from five to 10, but it also improved the predictive values of the discriminant functions for each outcome group (Table 3). Overall, 74 (87%) out of the 85 infants were correctly classified by these functions.

Fig. 1.
Fig. 1.:
Discriminant analysis of items' slopes at eight weeks; stepwise method – probability of F to enter variables >0.10 and <0.20. Centroids of outcome groups: DD, developmental delay; CP, cerebral palsy; TD, typical development.

The same procedures were repeated at 13 weeks. On the basis of the entire sample, both functions together correctly classified 100% of the children (Table 3). In this analysis none of the TIMP items' slopes clearly stood out in its ability to classify the infants into different groups. None of the items' slopes structure coefficients reached the 0.30 suggested cutoff15 to explain how much an item contributes to the function. Table 2 shows which items correlated with each function. However, again, a more parsimonious approach to the data is more informative regarding which items contribute to correctly identifying the infants by outcome.

In the default stepwise method (ie, F = 3.84), only one function was identified with only one related item (observed item 17, ie, isolated right ankle movements). On the basis of the entire sample, this function misclassified 26 (31%) infants. Specific predictive values per outcome group are presented in Table 3.

Stepwise analysis with the probability of F to enter into the equations set to be >0.10 and <0.20 resulted in eight items' slopes entering into the analysis. Figure 2 shows that function 1 discriminates better those children with DD and function 2 does a better job of classifying infants with CP. The overall correct classification in this analysis (69%) was the same as in the previous one (ie, default stepwise at 13 weeks). However, a more relaxed criterion to enter variables in the discriminant analysis not only identifies seven more TIMP items that contributed to the classification of the infants, but it also improved the correct classification of the infants with DD (100% correctly classified) and with CP (60% correctly classified).

Fig. 2.
Fig. 2.:
Discriminant analysis of items' slopes at 13 weeks; stepwise method – probability of F to enter variables >0.10 and <0.20. Centroids of outcome groups: DD, developmental delay; CP, cerebral palsy; TD, typical development.

Individual Items' Contributions to Correct Classification of Infants into Different Groups

Analyzing the specific items that correlated with each discriminant function helps to identify the most useful items in discriminating among infants with different outcomes. This section is focused on the stepwise analyses, because these were the only analyses in which the TIMP items met the 0.30 criterion to explain the variance of the discriminant function.15Table 2 shows the items that correlated with each function in the different sets of analyses.

TIMP Items Slope.

The plots of the group centroids (Figs. 1 and 2) show that different functions are more likely to discriminate a specific outcome group, thus correlation of an item with a discriminant function shows that the item tends to be more related to the outcome group that the function is discriminating. This correlation can be either positive or negative, and refers to the relationship between the item rate of change (not the direction of change) and the discriminant function score (Table 4, column 2). The positive correlation value of an item slope and a discriminant function means that the direction of the relationship between the item rate of change and the function is the same, ie, higher item rate of change leads to higher discriminant function values and lower item rate of change leads to lower discriminant function values. A negative correlation represents an inverse relationship between the item rate of change and the discriminant function. However, changes can be either positive meaning gain in ability, or negative meaning loss in some skills previously observed (Table 4, columns 3, 4, and 5). A zero rate of change is also possible indicating a plateau in the ability being measured. In children with TD plateaus are expected because they reach the maximal scores on easier items as development proceeds.

Studying the relationship between items' slopes and discriminant functions allows one to learn about the contribution of individual item's rate of change in differentiating among the outcome groups. The mean slopes of items by group were calculated separately for each group at the two different ages to interpret each item's contribution to the discriminant functions in each discriminant analysis. Table 4 presents a summary of the items' slopes per group that contributed to the discriminant function in the stepwise analyses with probability of F values set to be >0.10 and <0.20 at both ages studied. This procedure allowed examining and interpreting the direction of items' slopes correlation with each discriminant function for each outcome group separately.

Eight-Week Slopes.

At eight weeks' CA, the item individual finger movement correlated positively with discriminant function 1 which, despite separating the three groups of infants from each other, seemed to discriminate better children with DD (Fig. 1). For this item, children with TD had a negative slope, and children with CP and with DD had a positive slope, which was steeper for children with DD (Table 4). In this case, fewer changes in children's abilities discriminated more children with TD and more changes were more discriminative of children with DD. The larger positive slope for children with DD revealed that improvement in the ability to perform individual finger movement was more predictive for them than for children with TD or with CP. Changes in the infants' ability to turn the head in prone also were positively correlated with discriminant function 1. Although all children's scores had positive slopes, more rapid improvement in this skill at eight weeks' CA was observed among infants with DD.

The only item slope that was negatively correlated with function 1 at eight weeks of CA was the hip/knee flexion item. This negative correlation means that higher rates of changes are associated with lower discriminant function values, which in this particular case, is more related to TD (Fig. 1). However, because this item slope was originally negative for children with DD and with CP, the interaction of the slope coefficient with the direction of the item loading in the discriminant function resulted in higher discriminant scores for children with DD and with CP, separating them even further from the infants with TD. This resulted in a higher effect of the negative slopes for children with DD and with CP, showing that losses in the ability to flex hips and knees at eight weeks CA were predictive for them. Children with TD did not change much in their ability to flex hip and knees at this age, a skill they acquired earlier and they maintain at this age (Table 4).

Function 2 at eight weeks was better in discriminating infants with CP (Fig. 1) who, in this function, had the lowest discriminant scores. Two items were positively correlated with this function, meaning that lower items' rates of change are more discriminative of children with CP than of the other children. Thus, children with CP had more difficulty in developing reaching skills than children with TD or DD and they also did not improve their ability to turn the head in supine. Among the three groups, children with CP were changing less and also losing the ability to turn their head in supine. The small but negative changes were predictive of CP as an outcome. Rates of change in the items that measure children's ability to bring their right hand to the mouth, to demonstrate left fingering of objects or a surface, to demonstrate head control using extensor muscles, to maintain head in midline with visual stimulation, or to rotate the head to the left following visual stimulation were all negatively correlated with the discriminant function 2, meaning that lower discriminant scores correlated with higher rates of change. Originally, the directions of change of these items for children with CP were all positive except for their ability to demonstrate head control with extensor muscles. The direction of changes for children with TD and DD were also all positive except for their ability to bring their right hand to their mouth, however, the magnitude of changes for children with CP was smaller than for children with DD and TD (Table 4). The interaction of these items' slope coefficients with the negative discriminant function coefficient made the score of children with CP on the discriminant function decrease and thus it maximized the difference between infants with CP and those without CP. Children with CP had difficulty with these items and presented slower gains on them than children without CP. The only exception was that the mean gain of children with CP at eight weeks on the ability to finger objects or surfaces was improving at the fastest rate.

13-Week Slopes.

At 13 weeks, the discriminant analysis of TIMP slopes resulted in two discriminating functions. Function 1 discriminated mostly infants with DD, who had higher values on it (Fig. 2). Infants' changes in the ability to demonstrate right fingering of objects or surfaces was positively correlated with this function, meaning that fewer changes in children's skill discriminated more children with CP. Children with DD were gaining most in these skills. The infants' changes in their ability to bring the left hand to the mouth, to present the defensive reaction with arms response and to stand were negatively correlated with this function. In these three items, whereas children with TD and with CP were gaining skills, children with DD were losing skills. The magnitude of changes also varied among the infants. Children with TD were the ones changing less in their ability to bring their hand to the mouth and children with DD were the ones changing most on this item. In contrast children with DD were changing less on defensive reactions and in their ability to stand, followed by children with CP and then children with TD (Table 4). The magnitudes and directions of developmental changes interacted with the item loading in the discriminant function resulting in further separation of the infants with DD from the others.

Function 2, which was better in discriminating infants with CP, had only one item positively correlated with it, right isolated ankle movement. This item slope for children with CP was negative while its slope for children with TD and DD was positive, meaning that infants with CP are losing this skill while the other infants were either maintaining their skills or gaining this ability. This loss of skill made the discriminant values of children with CP lower, placing them further apart from the other infants. Changes in the ability to maintain head in midline, to bring the right hand to mouth, and to present left head righting after lateral tilt on vertical suspension were negatively correlated with discriminant function 2. Children with TD had a negative slope, although very small, in their ability to maintain the head in midline, whereas children with CP and DD were improving in this skill. Although at a different rate, all infants were improving in their ability to bring the right hand to mouth, and to demonstrate left head righting after lateral tilt on vertical suspension (Table 4).


The goal of this study was to assess the contribution of the TIMP items' rate of change, either individually or combined with others, in correctly classifying infants according to their developmental outcome. Our current findings confirm the results of previous analyses showing that total TIMP scores discriminate among children with different developmental outcomes.10,12,13 The present results are unique, however, because of the stepwise analysis, which was not only able to classify the infants into different groups with good predictive values based on the discriminant functions but also to identify a smaller and more significant number of independent items explaining these functions.

The number of items explaining the functions varied depending on the infants' age and on the specific analysis run, but in all cases, individual item slopes contributed to correct classification of the infants resulting in good predictive values for all three outcome groups. The predictive values were generally better when more items were forced into the discriminant analyses.

Surprisingly, with the exception of sensitivity for identifying children with DD, all of the predictive values were better at eight than at 13 weeks of age. This finding contradicts our previous results11–13 that have consistently shown performance at 13 weeks to have better overall predictive results. In previous studies analyses were done on total scores or on the item performance levels, but in this study items' rate of change was emphasized. Rate and sequence are two parameters frequently discussed in the motor development literature.7 Rate is usually referred to as the period of time an infant requires to progress from one motor skill to another. However, rates of change do not necessarily imply gains in skills but can also refer to loss of skills. This distinction between performance level and performance rate of change broadens the possibility of identifying sources of differences among children. At the specific ages studied, the rates of change of children with CP in some items were as high as the rates of change of children with TD, but their performance levels were lower. In other cases, the rates of change were similar between these groups but the directions of change were different, pointing to a different underlying developmental process. While one group was gaining skills, the other was losing them. It seems then that performance level and performance rate of change are separate aspects that should both be considered when looking for differences in the developmental profile of children with CP and with DD in comparison to children who are TD.

The rates of change among the different children appear to be more distinct at eight than at 13 weeks' CA. Comparison between Figures 1 and 2 reveals a similar pattern of discriminant functions at both ages but with the distinction among the groups dropping off at 13 weeks, especially between infants with DD and infants with TD. This might have resulted from a ceiling effect on many items at 13 weeks for children with TD so that children with DD appear to be catching up to those with TD. Although the TIMP is intended for children up to 16 weeks of CA and no ceiling effect for total scores has been found,8,10 some of the items that contributed to the discriminant functions at 13 weeks were easier, and ones in which children with TD generally acquire high performance at earlier ages. By 13 weeks children with TD are maintaining a stable rate of performance with decreasing variability in their rates of change on these items.

The results of this study were generally consistent with our previous results in terms of the specific items that presented differences among the outcome groups. At eight weeks, eight of the 10 item slopes that discriminated among the infants also presented differences among the infants either on a graphical representation of groups' mean scores over time or on a Rasch DIF analysis.13 At 13 weeks, all of the item slopes that discriminated among the infants also had differences on the item scores. Although further study with a larger sample size would be useful to confirm these results, a closer look at these items can inform us of some important developmental features that might be significant in the diagnostic process.

Individual Items' Contributions in Classifying Infants

In the stepwise analysis the items that contributed to the discriminant functions at both ages represented different difficulty levels covering the whole range of ability measured by the TIMP. At eight weeks, items of different difficulty levels were equally helpful in discriminating among the three outcome groups; at 13 weeks, however, the more difficult items were better at separating infants with TD and with DD while the easier items were better at discriminating infants with CP.

The rates of change in TIMP item performance varied according to the CA and the outcome groups. Only two items consistently contributed to the classification of infants at both ages: hand to mouth and fingering objects. Examining the infants' ability to bring hand to mouth across the groups we found that, at eight weeks' CA, children with DD and children with TD were bringing the hand to the mouth less than at an earlier age, while children with CP were attaining this skill. Losing this skill is not necessarily negative. The better head control in midline of children with TD and DD might have taken the mouth temporarily out of reach of the hand. On the other hand, the typical neck hyperextension of children with CP16 might have kept these infants' heads to the side, making it easier for them to place the hand in the mouth. At 13 weeks, children with TD and with CP improved in the ability to bring the hand to mouth using different movement strategies. In children with TD, improvement was related to better control of the head associated with gains in the antigravity movements of the arms so that they could bring the hand to the mouth while keeping the head in midline. The apparent improvement of children with CP might have been due to further development of neck hyperextension, keeping the head to the side and consequently at easier reach for the mouth.

At eight and 13 weeks, all infants were developing the ability to explore objects or the body surface with fingers. At eight weeks, children with DD demonstrated the slowest rate of change and children with CP were changing faster. At 13 weeks, this pattern was reversed. In children with TD gains were small and did not change from eight to 13 weeks, showing a steady developmental trend. These results appear useful in differentiating between children with DD and children with CP in that, at eight weeks children with CP might have used total movement patterns of raking when children with DD were slowly developing isolated voluntary finger control. At 13 weeks, children with DD are catching up to children with TD, while children with CP might be developing hypertonicity that limits finger mobility. Fingering objects requires alternating control of flexor and extensor muscles, reported to be difficult for children with CP.17 This might also explain why children with CP at eight weeks had few gains in their ability to perform individual finger movement while children with DD were the ones improving most. However, in these three items an inconsistency was seen among all three groups in which side of the body was changing faster or slower at each age, supporting reports in the literature that asymmetry might refer to some normal variability expected in children's patterns of movement3,18 and that asymmetry might result from myogenic causes (ie, maturation of different muscle fibers at different ages) rather than neurological causes,19making it a poor predictor of later motor outcome.1

Despite being the easiest item on the whole test,8,10 hip and knee flexion was one of the discriminating items at eight weeks' CA when children with CP and with DD tended to lose this skill. Children with DD may have decreased their hip and knee flexion against gravity because of muscle weakness3 and associated weight gain, while children with CP may have lost this skill as the result of increased hypertonic extension of the legs.2 In contrast, reaching, the most difficult TIMP item,8,10 might have been difficult for children with CP because of shoulder retraction,20 difficulty coordinating movements to a target,19 or difficulty performing antigravity movements,16,21 emphasizing the importance of this item for detecting differences among infants. Interestingly at eight weeks CA children with DD had the highest rate of change among the three groups, again possibly pointing to a catch up in the ability to reach for a person or a toy.

All the other item slopes contributing to the discrimination of infants with CP at eight weeks were related to head control in different positions (supine, prone, sitting, and standing), using different muscle groups (ie, extensors, flexors), or different sources of stimulation (with or without visual stimulation). In general, children who were TD or with DD were gaining better head control and children with CP either did not improve in these skills or improved very little in comparison with the other infants. Children with CP also lost the ability to control extension of the head when placed in sitting, possibly reflecting a lack of balance between flexor and extensor muscles22 combined with difficulty moving against gravity.16,21 At earlier ages these children may have used neck hyperextension to raise the head, but at older ages, with a bigger and heavier head, neck hyperextension may not have been sufficient to lift the head up against gravity.

At 13 weeks, loss of the ability to isolate ankle movements, one of the test's easiest items, was one of the most predictive items for children with CP. This finding further validates the results of our previous study13; however, previously only qualitative differences were found in this item. In the present study rates of change in the ability to isolate ankle movement were statistically significant contributors to differentiating among the groups. The difference in these results was possibly due to the fact that in the previous analysis item performance level was used whereas in this study the variable of interest was the rate of change in this skill. Examining the original data of children with CP to evaluate the individual patterns of response in this item, it was found that two children showed a clear decrease in the ability to isolate ankle movement, while two others were inconsistently presenting this skill. It might be that if these infants with inconsistent performance were assessed at older ages they would have shown further decrease in this skill. The inconsistent presence of isolated ankle movements might explain the higher negative rate of change in this item for children with CP. It reveals fluctuation and oscillation in the developmental process of at-risk infants23,24 and it might reflect a progressive increase of muscle tonus25 with associated decrease in isolated movements.7,26

Defensive reaction arm response and standing were among the more difficult items contributing to the discrimination among the infants at 13 weeks. Defensive reaction discriminated mostly infants with DD, who despite being able to maintain the head in midline had less antigravity movement of the arms. The typical hypotonia reported for children with DD3 might have prevented them from bringing their arms towards the face to remove the cloth. Children with CP and children with TD had similar rates of change in these two items, with both groups improving their performance. However, inspection of their raw scores revealed that they were doing so at different ability levels. In the defensive reaction, children with CP had their heads rotated to the side until they were 13 weeks old when they started keeping the head in midline and directing the arms to midline but were still not able to coordinate arm movement to take the cloth off of the face. In children with TD, gains were at a higher functional level; these infants were now developing the skill to grasp the cloth off the face while maintaining the head in midline. In ability to stand children with TD were gaining the skill to bear weight for a longer period of time with feet flat and head upright, while children with CP were developing the ability to bear weight only at a lower level of performance in which foot and head position were allowed to vary. The underlying mechanism used to stand appears to be different for these infants; children with CP used leg and trunk hyperextension to maintain body posture while children with TD had better overall postural control. Children with DD were losing the ability to sustain body weight with trunk alignment in standing, which might have been due to strength limitations and increased body weight.

Two other items that contributed to discriminating among the infants at 13 weeks' CA were those related to head control either in supine position or in vertical suspension. While children with CP and with DD were improving the ability to maintain the head in midline, children with TD were losing this skill. This apparent loss of skills of children with TD might in fact reflect a more refined head control which allows them to move the head freely in and out of the midline while exploring the surrounding environment. Children with TD and with DD were significantly improving the head righting response to lateral tilting, while children with CP improved little in this skill. This item requires advanced motor skills involving head control when the body is displaced, a skill difficult for children with CP who are still improving in more basic head control such as maintaining the head in midline in static positions.

In summary, different items' rates and direction of change combined differently to maximize the differences among the outcome groups depending on the children's age. Children with DD were changing faster in some items while still developing slowly in others, resulting in a great variability in their motor performance. The faster developing skills in children with DD generally involved balance of flexor and extensor muscles or isolation of movement. Skills they were developing slowly or even losing required more muscle strength and antigravity control. At eight weeks, children with TD developed head control faster than children with DD or with CP, as measured by the more difficult items at this age, and improved less than children with DD or CP in easier items, indicating they had already mastered these skills. At 13 weeks, children with TD changed very little in some items (possible ceiling effect) while gaining competence in the more difficult TIMP items that either require more muscle strength, reflect a better balance of extensor and flexor muscles, or more coordinated movements to a target (turn head after visual stimulation and arms response in defensive reaction). It also appears that children with TD had sufficient muscle strength to allow them to perform movements against gravity. Children with CP usually demonstrated a lower rate of change than children without CP, but within their group they also had variable rates of change depending on age and the items. Children with CP were not improving on items that required head control, items that required isolated control of body segments, and items that required antigravity control. The items changing most were items that involved the control of leg movements. In these items they either lost antigravity skills or they started developing abnormal movement patterns in response to the environmental demand as for example using leg hyperextension to bear weight when placed in standing.

Clinical Implications

The findings of this study are helpful in informing clinicians about how rates of change in performance on individual TIMP test items at eight and 13 weeks' CA can be interpreted and may contribute to development of a diagnostic profile.13 Different items' rate of change at each age were found to be helpful in discriminating among infants with different outcomes showing that the rates of change in children's motor performance are at least as important as the ability level presented at a specific time.

These results reinforce the need for repeated measures.27 Repeated examination enables the clinician to determine the meaning of the variability in an infant's performance, to understand how the infant learns motor skills or experiences regressions in skills, to identify critical variables that constrain or facilitate the infants' achievement of higher-level skills, to identify which skills are in transition or phase shifts, and to suggest what skills will emerge next.27

Early identification of infants who might develop a major handicap is important in deciding whether or not to refer a child for intervention. The very slow rate of change in some TIMP items associated with fast changes in other items suggests that initiating therapy at earlier ages for children at risk of developing CP might be beneficial. By instituting therapy we may influence the acquisition of motor abilities through the variations in the experiences and opportunities parents are able to provide.28 For example at 13 weeks, children with CP are rapidly developing the ability to stand; however, it appears that they are using atypical motor patterns. Intervention at this time could prevent the repetitive use of atypical movements that could result in abnormal motor memories and habits.29 Educating parents about the meaning of leg hyperextension might prevent them from encouraging their children to practice these abnormal standing patterns, thereby avoiding the consolidation of abnormal neurological paths. Rather, adequate postural control could be promoted through caregiving or handling techniques that parents could incorporate in their daily routine. Conversely, for infants who present with hypotonia or muscle weakness intervention might also facilitate strengthening and practice of skills.

Limitations and Suggestions for Further Research

For this study, we derived the coefficients of items' change only at eight and 13 weeks CA. It is recommended that rates of changes at other ages and their diagnostic usefulness at these ages also be determined. Also, some of the items that appear as the most discriminating (ie, hand to mouth; left head turn in supine and in prone) are among the items that have been deleted in the newest TIMP version (version 5), either because they were redundant with others or because they failed to conform to the Rasch psychometric model.9 However, it might be the case that these items are more important than they were thought to be and they deserve further investigation as part of a diagnostic profile.

One difficulty of discriminant analysis is choosing a set of compounding independent coefficients that will define function.15 The potential of rotating the discriminant functions to reveal a simpler structure than the one found with the principal component analysis deserves further research. Rotation of the discriminant function might lead to another representation of items' slopes that might work better for interpreting the items' arrangement.

Further exploration of the meaning of asymmetry in the TIMP items' slopes, comparing the slopes of items tested on both sides of the body per outcome group, should be conducted at both ages. The literature on asymmetry is controversial,20,25 with some people arguing that early asymmetries tend to persist in infants who present CP,26 and others arguing that asymmetry might refer to some normal variability expected in children's patterns of movement3,18 and thus is a poor predictor of later motor outcome.1 Despite this controversy asymmetries might provide some useful diagnostic information and could be signs alerting clinicians to the fact that the child is at risk of developmental problems.


In this study, children's diagnosis as TD, DD, or CP at one to one and a half years of age was used as the outcome criterion to evaluate each specific TIMP item's ability to separate the children into different groups. This study's results differ somewhat from our previous results13 in terms of the better age to identify differences among infants, but the items that appear to contribute to group discrimination were the same as the ones previously identified in the same dataset. These results suggest that the motor behaviors that could work as identification factors might be different at different ages, depending on whether the motor ability level itself or the speed with which children acquire different motor skills is being assessed. These findings also provide some insight about the underlying developmental process (ie, acquisition, loss of skills) of infants with different outcomes. These findings are in agreement with the more recent literature22,23 suggesting that neurological impairment might not be the same at different ages depending on the stage of brain and body development.


1. Harris SR. Movement analysis: An aid to early diagnosis of cerebral palsy. Phys Ther. 1991;71:215–221.
2. Harris SR. Early neuromotor predictors of cerebral palsy in low-birth-weight infants. Dev Med Child Neurol. 1987;29:508–519.
3. Lacey JL, Henderson-Smart DJ. Assessment of preterm infants in the intensive-care unit to predict cerebral palsy and motor outcome at 6 years. Dev Med Child Neurol. 1998;40:310–318.
4. Amiel-Tison C, Grenier A. Neurologic Assessment During the First Year of Life. New York: Oxford University Press; 1986.
5. Dubowitz L, Dubowitz V. The Neurological Assessment of the Preterm and Full-Term Newborn Infant. Clin Devel Med. Vol. 12. Philadelphia: JB Lippincott; 1981.
6. Chandler LS, Andrews MS, Swanson MW. Movement Assessment of Infants: A Manual. Rolling Bay, WA: Authors; 1980.
7. Piper MC, Darrah J. Motor Assessment of the Developing Infant. Philadelphia, PA: WB Saunders; 1994.
8. Campbell SK, Kolobe THA, Osten ET, et al. Construct validity of the Test of Infant Motor Performance. Phys Ther. 1995;75:585–596.
9. Wright B, Masters GN. Rating Scale Analysis: Rasch Measurement. Chicago: MESA Press; 1982.
10. 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;139:546–51.
11. Campbell SK, et al. 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.
12. Barbosa VM, Campbell SK, Sheftel D, et al. Longitudinal performance of infants with cerebral palsy on the Test of Infant Motor Performance and on the Alberta Infant Motor Scale. Phys Occup Ther Pediatr. 2003;23:7–29.
13. Barbosa VM, Campbell SK, Smith E, Berbaum M. Comparison of Test of Infant Motor Performance (TIMP) item responses among children with cerebral palsy, developmental delay, and typical development. Am J Occup Ther. 2005;59(4):446–456.
14. Kolobe THA, Bulanda M, Susman L. 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. Phys Ther. 2004;84:1144–1156.
15. Pedhazur EJ, ed. Multiple Regression in Behavioral Research. 3rd ed. Fort Worth, TX: Harcourt Brace; 1997.
16. Jelsma JP, Illif, Kelly L. Patterns of development exhibited by infants with cerebral palsy. Pediatr Phys Ther. 1999;11:2–11.
17. Samsom JF, Sie LTL, de Groot L. Muscle power development in preterm infants with periventricular leukomalacia in relation to outcome at 18 months. Dev Med Child Neurol. 2002;44:735–740.
18. Riccio GE, Information in movement variability about the qualitative dynamics of posture and orientation. In: Newell KM, Corcos DM eds. Variability and Motor Control. Champaign, IL: Human Kinetics Publisher; 1993;317–357.
19. Samsom JF, de Groot. The influence of postural control on motility and hand function in a group of “high risk” preterm infants at 1 year of age. Early Human Dev. 2000;60:101–113.
20. Person K, Stromberg B. Structured Observation of Motor Performance (SOMP-I) applied to preterm and full term infants who needed neonatal intensive care. A cross-sectional analysis of progress and quality of motor performance at ages 0–10 months. Early Human Dev. 1995;43:205–224.
21. de Groot L. Posture and motility in preterm infants. Dev Med Child Neurol. 2000;42:65–68.
22. de Vries AM, de 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.
23. Brandt I, et al. Transient abnormal neurologic signs (TANS) in a longitudinal study of very low birth weight preterm infants. Early Human Dev. 2000;59:107–126.
24. Darrah J, et al. Intra-individual stability of rate of gross motor development in full-term infants. Early Human Dev. 1998;52:169–179.
25. Fedrizzi E, et al. Developmental sequence of postural control in prone position in children with spastic dyplegia. Brain Dev. 2000;22:436–444.
26. Green EM, Mulcahy CM, Pourtney TE. An investigation into the development of early postural control. Dev Med Child Neurol. 1995;37:437–448.
27. 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.
28. Chiarello LA, Palisano RJ. Investigation of the effects of a model of physical therapy on mother-child interactions and the motor behaviors of children with motor delay. Phys Ther. 1998;78:180–94.
29. Thelen E. Dynamic Processes of Stability and Change in Perceptual-Motor Development. Chicago: D. S. Seminar; 2001.

cerebral palsy/diagnosis; developmental disabilities/diagnosis; infant; infant behavior/physiology; motor skill disorders/diagnosis; neuropsychological tests; predictive value of tests; psychomotor performance/physiology; psychomotor disorders/diagnosis; risk factors; time factors

© 2007 Lippincott Williams & Wilkins, Inc.