Construct validity of the test of infant motor performance.

The evolution of high-technology neonatal care units has resulted in the preservation of life for a growing number of infants at high risk for mortality, resulting in a large number of children who have high risk for developmental disabilities.[1] Approximately 25% to 30% of at-risk infants exhibit some form of neuromotor disturbance early in life.[2,3] As a result, physical therapists and occupational therapists have increasingly become incorporated as regular members of the neonatal special care team, providing assessment and treatment for babies at especially high risk for developmental dysfunction.[4] In addition, therapists provide education to those caring for these children.

Despite several decades of involvement by therapists in special care nurseries, little research exists to document the effectiveness of their services.[5] Studying the effects of these therapies for high-risk infants is hindered by the lack of functional motor outcome measures appropriate for this population. In addition, available tests do not provide an adequate range of items to allow for detailed quantitative assessment of motor development, and characterization of deviations from the norm, in infants during the first few months after birth. The purpose of this research was to assess the sensitivity to age-related change and to risk for poor developmental outcome of a new test, the Test of Infant Motor Performance (TIMP). The TIMP was developed for use by physical therapists and occupational therapists in assessing the components of functional motor performance in high-risk premature and very young term-born infants.

A Theoretical Approach to

Developmental Motor

Assessment

Movement is a way that the infant establishes control over the world, communicates needs, and explores the environment; movement is a basis for early learning as well as being important in its own right for developing and maintaining musculoskeletal integrity.[6] Infants use movement to organize their environment; to communicate and interact physically with objects or people; to change postures or to adjust positions; to respond to environmental demands; and for self-consolation, such as thumb-sucking. Active movement has been shown to be necessary for optimal perceptual development,[7] an area that is frequently problematic in prematurely born children.[8] The goal of physical therapy or occupational therapy(*) for young infants with neurologic dysfunction is to assist them in developing these functional uses of movement and to limit disability. No test has yet been developed, however, that reflects the movements that are basic to these functional competencies in newborns.

Because poor postural control can limit the rate at which function skills, such as manipulation and mobility, are acquired during development,[9-12] therapists often use assessment of postural control as a basis for determining the need for intervention to promote motor development. Shumway - Cook and Woollacott[12] define postural control as control of the body's position m space for stability (balance) and orientation. Stability is the ability to maintain the center of body mass within the base of support.[12] Orientation is the alignment of the body segments with respect to one another as appropriate for the task.[12]

Thelen and colleagues[13] place special emphasis on the physical and social contexts of actions in daily life as aspects of the organization of postural and movement responses. Interactions among environmental and task characteristics, physical and neurologic properties of the individual infant's body, and the inherent self-organazing properties of the human sensorimotor system are the principal components of their dynamical systems model of infant motor development.[13] The two related theoretical approaches, those of Shumway-Cook and Woollacott[12] and Thelen and colleagues,[13] have influenced the structure and development of the TIMP.

The systems model of motor development described by Shumway-Cook and Woollacott[12] assumes that the processes of postural control include (1) development of sensory systems and central neural processes that organize visual, vestibular, and somatosensory inputs for orientation of the body; (2) musculoskeletal change, such as increasing strength and changing body morphology; (3) development of neuromuscular synergies to maintain stability; (4) development of adaptive mechanisms to modify control processes for posture in response to changing environmental and task demands; and (5) development of anticipatory postural control processes to prepare for destabilizing internal and external forces. In their model, motor behaviors are divided into four functional categories: (1) sustaining posture; (2) regaining posture; (3) transitions between postures; and (4) integration of posture into movements such as locomotion, manipulation, and exploration. The TIMP assesses behaviors in the first three categories that we believe to be the basic skills leading to performance of activities at the fourth level. Head control and other functions are assessed in active and alert states, for example, by testing the ability of the child to maintain head stability in a variety of spatial orientations, right the head when the body is tilted, turn the head in various positions, and stabilize or orient the head in response to interesting visual or auditory events. A variety of positioning and stimulus materials (ie, toys, examiner's face) across items allows infants to demonstrate their abilities to perform in a number of environmental and task contexts.

Comparison of the Test at

Infant Motor Performance With

Available Infant Tests

Available tests primarily assess infant neurobehavioral responses, social interactions, and reflexes and muscle tone.[9,14-19] Each of these tests has one or more problems that limits its use in clinical practice, such as lack of summary or quantitative scores, excessive length, or questionable or inadequately demonstrated reliability or validity.[20] In the tests developed by Dubowitz and Dubowitz[16] and by Korner and colleagues,[17,18] the examiner stabilizes the child's head position when testing many items, thus limiting the infant's ability to express independent postural control capacities. Many of these tests, however, do contain small numbers of items that assess the ability to control the head and trunk in several spatial orientations and in response to interesting stimuli in the environment.

The TIMP is designed for prematurely born infants from 32 weeks' gestational age up to about 4 months after term-equivalent age or for full-term infants up to 4 months of age.[21] Some items in the test were selected from those in the previously mentioned infant tests.[15,16] These include items developed by Brazelton[15] involving orientation in response to visual and auditory stimuli and reactions to a cloth over the eyes; items from the test by Dubowitz and Dubowitz[16] involving control of the head in the upright seated position and when suspended in the prone position, and the ability to flex the arms from an extended position in a prone posture; and items developed by Amiel-Tison and Grenier[9] for eliciting evidence of developing postural control in sitting and side-lying positions. In each case, however, new scoring descriptors were developed to reflect the goals of the TIMP.

Movements reflecting the qualitative changes in coordination (oscillating movements and ballistic [ie, rapid, forceful] movements) recently identified by Hadders-Algra and Prechtl[22] and Cioni and Prechtl[23] are also scored when they occur spontaneously. The presence of ballistic movements (called "swipes" and "swats") at 7 to 10 weeks postterm is thought to reflect increasing amounts of reciprocal innervation in muscular coordination, which makes these rapid, forceful movements possible.[24] Because these qualitative developments do not appear in children with spastic cerebral palsy, their absence at an appropriate age may have diagnostic significance.[25] All other items in the TIMP are original in conception and scoring[26,27]; a complete list is presented in Figure 1.

Figure 1. List of items in the Test of Infant Motor Performance (Version 2.3). Missing items were deleted after psychometric analysis. (R/L=right/left.)

OBSERVED SCALE

1. Head in midline 2s 2/3. R/L head turn 180[degrees] 4. Hands together in midline 5/6. R/L mouths hand 7/8. R/L individual finger movements 9/10. R/L individual wrist movements 11/12. R/L financing objects 13. Pelvic lift 14. Bilateral hip and knee flexion 15/16. R/L individual ankle movements 17. Kicking 18/19. R/L head turn 180[degrees] in prone position 20. Head lift n prone position 23. Antigravity bilateral forearm movements 24. Antigravity bilateral arm movements 25. Ballistic movements 27. Reaching

ELICITED SCALE

1. Neck rotation with visual stimulation in semiupright position 2. Head control in seated position 3. Trunk extension in seated position 4. Head lifting from full neck flexion 5. Head lifting from full neck extension 6. Head control when lowered from sitting to supine position 7. Inhibition of trunk rotation with head turn 8/9. Midline head alignment without/with visual stimulation 10/11. R/L neck rotation with visual stimulation in supine position 12/13. Neck stretch/arm reactions to face covering 14. Antigravity hip\knee flexion n supine position 15. R/L rolling to prone position with leg adduction 16. R/L rolling to prone position with arm adduction 17. Neck flexion on pull-to-sit 18. Lateral head and trunk extension righting with elbow extension 19. Lateral hip abduction reaction 20. Neck and trunk extension in prone suspension 21. Crawling in prone position 22. Head lift in prone position 23. R/L head turning in prone position with auditory stimulation 24. Recovery of arm position after displacement in prone position 25. Standing 26. R/L head righting with tilt n upright suspension

The TIMP requires an average of 36 minutes to administer and to score, and it consists of two scales, one for rating the presence of spontaneous motor behaviors and the other for rating the infant's responses to being positioned and handled in a variety of spatial orientations and to interesting types of visual or auditory inputs. Items on both scales are scored only when infants are in quiet alert or active alert behavioral states, as defined by Brazelton.[15]

The Observed Scale on TIMP Version 2.3 (items denoted by O-) consists of 27 dichotomously scored behaviors reflecting the infant's spontaneous attempts to change positions or to orient the body in various ways, to selectively move individual body segments, and to perform the qualitative types of movements mentioned earlier.[22-24] Examples include the ability to center the head in fine with the midline of the body in the supine position, bring the hand to the mouth, produce ballistic movements of the arms or legs, and lift the head in the prone position (Fig. 2).

The second part of the test is the Elicited Scale (items denoted by E-), made up of 26 items (6 of which are scored separately for each side of the body), each rated on five- or six-point scales. Performance on these items reflects the infant's ability to solve movement "problems" posed to elicit evidence of developing postural control in a variety of spatial orientations. Examples include rolling to a prone position with head righting when the leg is rotated across the body (Fig. 3); inhibiting rolling when the head is turned to one side; turning the head to follow a visually interesting object or to search for a sound in a prone, supine, or supported sitting position; and aligning the legs, trunk, and head in a vertical orientation when placed in a standing position.

Purpose and Hypotheses

The purpose of this article is to present preliminary evidence of the construct validity of the TIMP. If the test is truly a measure of the postural and selective control underlying functional motor performance, it should possess certain characteristics. Performance measures (or total scores) should vary as a function of (1) age-related changes in motor performance; (2) the presence of risk factors for developmental disability, such as extreme prematurity or perinatal asphyxia; (3) medical diagnosis, such as chronic lung disease or cerebral palsy; (4) the presence of developmental deviance in any of the subsystems that contribute to the organization of functional movement (eg, visual or auditory impairment, altered reflex activity or defects in motor control); or (5) delivery of effective interventions to improve motor development.

The analyses presented in this article address the first two of these issues, as well as possible differences related to ethnicity of the infant. Our hypotheses were that ability on the TIMP would be positively correlated with postconceptional age and that ability on the TIMP would be negatively correlated with degree of perinatal medical complications related to increased mortality and developmental deviance. The literature on motor performance in children of different racial and ethnic backgrounds also suggests that Caucasian children might be expected to have lower scores than African-American children.[28-30]

Method

Subjects

The subjects in this study were 137 infants born prematurely or after a full-term pregnancy who were tested one to three times (with the exception of 1 child who was tested four times), yielding 174 measures of performance on the TIMP. Subject selection was made in accordance with a preestablished plan for stratifying infants by postconceptional age, medical risk, and race or ethnicity. Based on this plan, we attempted to recruit 3 infants of each racial or ethnic group at each of three levels of risk for developmental disability for a total of 27 infants in each of seven predefined postconceptional age ranges (Tab. 1). Most of the sample was cross-sectional (ie, tested once); however, 9 infants who were tested three or four times constituted a sample of convenience for longitudinal assessment. Children were tested in their homes or in one of two inpatient special care nurseries (the University of Illinois Hospital, Chicago, Ill, or Lutheran General Hospital, Park Ridge, Ill).

[TABULAR DATA OMITTED]

Age was calculated from the expected date of confinement for each infant's mother. For all subjects, the average gestational age at birth was 33 weeks (SD=5); 29 of the subjects were full term at birth (38-42 weeks), and the rest were born prematurely. Subjects were selected for stratification by postconceptional age into seven groups: three groups of prematurely born infants at postconceptional ages 32 weeks to 37 weeks; a term age group consisting of 10 full-term infants and 19 prematurely born infants who had reached term-equivalent age; and three groups of infants who were 2 to 13 weeks past term-equivalent age, defined as 40 weeks postconception (Tab. 1). The three oldest age groups consisted of 9, 11, and 11 full-term infants. The average gestational age at birth (last column of Tab. 1) of infants tested within each of the seven postconceptional age groups varied from 30 to 35 weeks.

Risk assignment (high, medium, low) was made based on scores on the Newborn form of the Problem-Oriented Perinatal Risk Assessment System (POPRAS),[31,32] derived from reviews of medical records. Scores of 2 to 60 were considered low risk, scores of 61 to 90 were considered medium risk, and scores over 90 were considered to be high risk.

Ethnicity/race was taken from each infant's medical record. Only children who were African-American, Latino, or non-Latino Caucasian were included in the sample. Thirty-five percent of the infants were non-Latino Caucasian, 36.5% were African-American, and 28.5% were Latino, distributed across all categories of age and risk.

Testing

Informed consent was obtained from the parents of each infant, and permission to test was obtained from the infant's physician or nurse before assessment began. Children who were observed repeatedly were never tested more than once while they were in the same age category. Chronologic age at the time of testing of the children below term-equivalent postconceptional age averaged 20 days (SD = 18, range = 3-68) for the 32- to 33-week-old infants, 26 days (SD=20, range=1-81) for the 34- to 35-week-olds, and 29 days (SD=27, range= 1-77) for the 36- to 37-week-olds.

Five therapists (three physical therapists, two occupational therapists) tested all infants in the study. Their intrarater and interrater reliability on 14 videotaped tests of infants from all three ethnic and racial groups and with a variety of ages and degrees of medical risk was analyzed with the FACETS Computer program.[33,34] The analysis allows the investigator to assess whether raters use items systematically across subjects and whether different raters use item ratings similarly. All testers met the predetermined criterion[35] of fewer than 5% misfitting[dagger] ratings.

[dagger] The omputer program is designed to identify as misfit any observed item rating that differs with high probability from he rating that was expected based on the psyschometric model of unidimensionality of item structure. Fit to the model requires that higher item ratings are achieved by persons with higher ability (greater total raw score) and that persons with given functional level have a greater probability of scoring higher on easier items than on more difficult items; the mean square fit statistic is used to identify ratings that deviate from the expectations.[36] Ratings can misfit because items are not well defined, because raters apply the item definitions inconsistently, or because infants perform in unexpected ways (eg, failing easy items relative to their overall ability). Misfitting item scorings were deleted from the reliability analysis when unusual child performance appeared to cause misfit to the model but examiners' ratings were in close agreement. A 5% occurrence of misfitting ratings could be expected by chance alone, so this criterion was used to establish an acceptable level for rater reliability, both within and across raters.

Data Analysis

Test performance for all children on the TIMP was subjected to psychometric analysis for fit to a theoretical Rasch measurement model[37-39] using the BIGSTEPS computer program.[40] In this model, raw normal-level scores for success in passing items, or levels within items, are transformed into a linear scale of measurement.[41] The number of success (S) indicates the infant's level of function, and the number of levels the infant faded to pass (F) is the difficulty of the item. Taking the log(s) and the log(F) can yield B (the ability of individuals) and D (the difficulty of items), which, if plotted for the group, would be linear. The measurement model focuses on the difference between ability and difficulty; thus, B - D = log(S) - log(F) = log(S/F).[41] This statement is reformulated into a more general statement about probabilities (ie, the chances of success on items by infants), so that the model for dichotomously scored items is described as log[[P.sub.S]/[P.sub.F]] = B - D. The model has an additional term for the difficulty of the various steps in the item when several levels of ratings are possible within each item. Analysis of the raw ordinal-level scores by imposing this measurement model definition on the observed responses of tested individuals evaluates the extent to which the data fit the model. Those parts of the data that do not fit the model are identified with misfit statistics. If the data provide an overall fit to the measurement model, a scale has been formed that is assumed to be based on a unidimensional construct (all items measure the same thing) and to approximate an interval-level scale (ie, measurement units are the same size over the whole range of the scale), a property that generally does not hold true for the original raw scores on items. Data from a scale with these properties that be used to order individuals by ability and can be analyzed with parametric statistics.[41]

Items on the TIMP were calibrated according to difficulty level ad assessed for fit to the measurement model. Both item difficulties and person ability measures are expressed in logits, or logarithmically transformed probabilities of success given a particular level of ability (log-odds probability scaling), ranging from about - 5.0 to +4.0 for the TIMP and forming a hierarchical linear scale with equal intervals.[36] The internal consistency coefficient was .98, indicating that die items on the TIMP form a coherent scale. Clarity of the measure, as reflected by die item separation index, was 7.38 (root mean square error =0.19). The practical meaning of this measure is that the TIMP items can be separated into more than seven different average levels of difficulty across the 5-month age span of the children tested.

Item calibrations are shown in Figure 4, in which every dichotomous item in the test and each level within multi-level items is placed at its median difficulty calibration in logits. The easiest item (0-14) is observed spontaneous hip and knee flexion or being able to flex the hips and knees in the supine position so that the legs are of the supporting surface at least momentarily the most difficulty item (0-26) is spontaneously demonstration of an oscillating quality of arm or leg Movement.[22] offer difficult skills we reaching (0-27), and Amiel-Tison and Grenier,[9] items for evoking evidence of the development of lateral control of the head and body in space (E-18 and E-19). Items demonstrating similar levels of difficulty, such as items 0-2/3, E-10/11, and E-16 R/L, are primary items that rate the same activity but on different sides of the body. A few items of similar degree of difficulty involve use of different body parts (eg, 0-2/3 involve spontaneously turning the head a full 180[degrees], 0-13 involves pelvic control, and 0-23 involves antigravity forearm movements with elbows on the support surface). These abilities, therefore, seem to appear at about the same time during the course of development.

Floor or ceiling effects have not been noted (ie, item difficulties extend well below the lowest performance measures obtained for the subjects tested), and the highest or lowest ratings possible for a few items have not been attained by any subjects. We therefore suspect that the test could be sensitive to reflecting the ability of children who are younger and older than those in our sample. Before the TIMP could be used on older or younger children, however, further research will be needed. Items O-21 and O-22, spontaneous rolling to side-lying R/L, seriously misfit the Rasch model and have been eliminated from the test because they did not show consistent differences with increasing age. After consultation among the raters, several additional misfitting item definitions were subsequently revised slightly to eliminate ambiguities in descriptors we were able to identify as potentially responsible for misfit or to better reflect the developmental sequence observed in the Rasch analysis.

Descriptive statistics were calculated for each age group and risk group, and correlation and multiple regression analyses were used to assess the sensitivity of the test to reflect age-related changes in motor development, ethnicity and race, and the risk for developmental deviance.

Results

The calibration of child performance measures on the TIMP is given in logits in Table 2, which shows the means, standard deviations, and ranges for each age group and risk group in the stratified sample. The mean score progressively increased with age, reflecting the sensitivity of the test to development of motor skills. The Pearson Product-Moment Correlation Coefficient between postconceptional age in days and the performance measure in logits on the TIMP was .83. The Rasch analysis derived a person separation index of 6.02 (root mean square error=0.21), indicating that the children tested can be separated into at least six different levels of ability across the 5-month age range tested.

Initial demonstration of the sensitivity of the TIMP to age-related change in individual children is shown in Figures 5 and 6 for nine children who were assessed on three to four different occasions at least 12 days apart (with one exception when two tests were separated by only 6 days). Figure 5 shows the performance of five infants at low or moderate risk for poor motor outcome; Figure 6 shows the data for four infants at high risk for poor motor outcome. Based on the data points from all infants assessed repeatedly, the average change in scores over a 2-week period would be expected to be about 0.48 logit, greater than twice the average error value (0.21). Three children demonstrated essentially no change in scores - decreases of 0.04, 0.04, and 0.12 - over periods of 12, 14, and 24 days, respectively. Two infants were at high risk and one infant was at low risk for poor motor outcome. Each was a young premature infant who subsequently showed gains between the second and third tests of 0.93, 1.09, and 1.35 logits, respectively. The numbers of days between these later tests ranged from 16 to 47 days.

Multiple regression analysis demonstrated the joint and unique contributions to explaining variance in TIMP performance measures of postconceptional age, medical risk, and 0/1 dichotomous variables for ethnicity/race. The combination of these independent variables in predicting TIMP logit measures resulted in a multiple R of .85 (P<.00001). Seventy-two percent of the variance in TIMP scores was explained by the combination of these three variables. Postconceptional age was a significant variable (beta=.80; P<.00001). Medical risk, as reflected in the infant's total Newborn POPRAS score, was also significant at P<.0001 (beta= - 3.92). The negative coefficient indicates that ability decreased with increasing degree of medical risk; the simple correlation between TIMP and POPRAS scores was -.29. Differences in performance based on race/ethnicity did not reach the .05 level of probability. Figure 7 shows the relationship among postconceptional age, POPRAS risk score, and TIMP performance measures in a three-dimensional plot.

DISCUSSION

The results of these analyses demonstrate that the TIMP meets two important criteria necessary for potential usefulness as a tool for assessing infants at risk for poor motor outcome. Scores in systematically with increasing postconceptional age, and children with greater numbers of medical complications do less well than same-age peers. In infants assessed over time, the average change in TIMP scores over 2 weeks was more than twice the error of measurement, and rates of change were approximately linear. These findings suggest that it will be possible to develop norms for performance of premature and young full-term infants that will be useful in identifying those children whose motor performance is poorer than expected.

Of other tests for newborns that have been published, only the test developed Korner and colleagues[17,18] was designed for the purpose of measuring age-related change in behavior. Their test has been shown to be sensitive in preterm infants to weekly changes in alertness and orientation to sensory stimulation, irritability, and muscle extensibly (degree to which a muscle can be passively extended or elongated.) Despite including a dimension called "motor development," however, the test of Korner and collegues is not useful for therapists interested in the development of postural control because many of the items require that the child's head be passively positioned in a midline orientation.

The test developed by Dubowitz and Dubowitz[16] similarly requires fixation of the infant's head in midline when most items are tested. Testing the infant with a passively positioned head may explain the insensitivity of the test in Darrah and colleagues' research on the effects of waterbed use to promote more flexed postures in premature infants,[42] which we would have expected to result in improved ability to independently center the head. Because head control is an important aspect of postural development in the early months of life and is frequently impaired in children with cerebral palsy, the infant's ability to independently control head position in a variety of spatial orientations and in response to a variety of sensory and social stimuli is a major construct assessed in the TIMP.

Further evidence of the construct validity of the TIMP is found in the hierarchy of average item difficulty identified by the Rasch analysis. For example, oscillating (0-26) and ballistic (0-25) movements are ranked as very difficult (ie, passed only by infants with overall high scores), commensurate with Hadders-Algra and Prechtl's[22] and Cioni and Prechtl's[23] identification of the initial appearance of these aspects of movement at about 7 to 10 weeks of age postterm. Other difficult items are those involving lateral righting of the head and trunk (E-18, E-19, E-26). Diagonal and rotational components of movement require advanced levels of skill relative to sagittal-plane movements of flexion and extension, as clearly demonstrated by the developmental sequence of behaviors on the Alberta Infant Motor Scale.[43]

To be useful for the purpose of identifying children who might benefit from physical therapy or occupational therapy in the period of early infancy, the TIMP should demonstrate (1) acceptable rater reliability among therapists who were not involved in test development and (2) test score stability across short time periods. Other aspects of construct validity also remain to be assessed, such as responsiveness to change produced by effective interventions and sensitivity to important deviations from the norm, including the presence of signs of motor performance deficits or delayed development. For example, we are currently planning a research project to assess the discriminative validity of the TIMP in identifying differences in developmental growth curves of defined groups of infants. These groups will include infants with documented brain insults, chronic lung disease, and extreme prematurity at birth without brain or severe lung impairments, as well as both premature and full-term infants at low risk for poor motor outcome. Developmental outcome at 1 year will be assessed with the Alberta Infant Motor Scale.[43] This will be done to assess the predictive validity of the TIMP using a test with a similar theoretical basis emphasizing postural control during functional activities. Because the regression coefficients for ethnicity/race were close to significance (P=.10, non-Latino Caucasian children's ability was lower than African-American children's ability; African-American and Latino children performed similarly), differences in performance related to this variable also warrant further attention with larger samples before the test is normed for diagnostic use.

Future research might also assess the TIMP from the perspective of dynamical systems theory.[44] For example, if this theory holds, children with visual or auditory impairment should demonstrate poorer performance on those head control items in which stimuli are used to which they are unable to respond when compared with performance on head control items that do not require use of their impaired subsystem. If task characteristics are determinants of motor responses, altering the toys used to elicit responses might also reveal variations in age-related performance.

Dynamical systems theory also suggests that children might use different developmental paths to achieving similar motor milestones.[45] The overall item sequencing and individual item scaling on the TIMP therefore, should be evaluated to determine whether the sequence of item difficulty indeed forms a hierarchy of postural control skills that all children need to achieve to reach the functional motor outcomes of sitting alone and beginning to grasp objects, which are expected of children at 4 to 6 months of age, or whether other paths to achievement of these skills are possible. Longitudinal research would be necessary to investigate this question and related questions that could be posed based on Shumway-Cook and Woollacott's model.[12]

Although item scaling is ordered hierarchically, the Rasch model does not assume strict adherence to an order on the part of an infant when deriving a logit ability measure.[39,41] Performance on items that is strictly hierarchical generates a misfit statistic because such regularity is considered suspect. The statistical model is instead a probability model that generates an expectation of obtaining certain performances, given the difficulty calibration of each item and the child's overall ability as estimated from the total raw score. Because of this characteristic, one advantage of the Rasch model for clinical use is that small numbers of item ratings can be missing from a child's test without compromising the ability to derive an overall ability measure based on the item performances available. The presence of multiple items (or steps within items) with similar difficulty calibrations also means that not all items may need to be used in any given test administration.

One aspect of our research plan involves attempting to develop individually tailored testing. With this approach, the therapist would assess the child on a small number of items that cover a wide age range (eg, items E-3, E-13, E-14, and E-17 in Fig. 4) to derive an initial estimate of the child's ability. Then only items with difficulty calibrations sligthly above and below the child's estimated ability would be used for further testing, and items expected to be too hard or too easy for the child could be omitted. Although infants have tolerated the time currently needed for testing with no adverse effects, tailored testing would be useful to reduce the time required of the therapist and the demand on the infant.

Two other major issues of construct validity currently being addressed in our research on the TIM are (1) die relationship between postnatal age and TIMP performance and (2) the ecological validity of TIMP items. A group of infants born at a variety of gestational ages is being assessed at 36 weeks postconceptional age to evaluate whether earlier exposure to the extrauterine environment affects TIMP performance and, if so, which items vary by postnatal age. Ecological validity is being studied by comparison of the items involving use of handling during test administration (E-scale) with actual demands for movement provided by parents' handling of their own infants in daily routines such as bathing, dressing, and playful interaction. Eventually, we could also study whether parents change the way they interact with their infants after observing administration of the TIMP. We have found that parents enjoy watching the test and often report seeing evidence of skills they did not realize their infant possessed. Ultimately, our plans also include a national normative study to establish standards for children of different ages and risk (and ethnicity and race, if necessary) and investigation of whether the TIMP has value in developmental diagnosis by defining the limits of its specificity and sensitivity for identification of developmental deviance.

Conclusion

Our research to date has produced a standardized version of a unique, new test of infant motor performance that is based on a clinical perspective focused on the postural control and spontaneous movements characterizing the development of normal motor behavior. The test was found to form an internally consistent scale that is highly correlated with age-related changes in motor performance and that is sensitive to differences in developmental risk as a result of perinatal medical complications. The full spectrum of research activities described in this article will, however, be necessary to fully establish the construct validity of the TIMP for its intended purposes.

Acknowledgments

Appreciation is expressed to Mary Murney, PT, Joyce Laskey, RN, David Sheftel, MD, Lucky Jain, MD, Dharmapuri Vidyasagar, MD, and the nursing staff of the special care nurseries at Lutheran General Hospital and University of Illinois Hospital for assistance in subject recruitment. We also thank the parents of our subjects for their willingness to allow their babies to be tested and videotaped.

SK Campbell, PhD, PT, FAPTA, is Professor and Director of Graduate Studies, Department of Physical Therapy, University of Illinois at Chicago, M/C 898, 1919 W Taylor St, Chicago, IL 60612 (USA) ([e-mail]). Address all correspondence to Dr Campbell.

THA Kolobe, PhD, PT, is Lecturer, Department of physical Therapy, University of Illinois at Chicago.

ET Osten,OT, is in private practice in Skokie, Ill. She is also Clinical Instructor, Department of Occupational Therapy, University of Illinois at Chicago.

M Lenke, OT, is Clinical Supervisor Pediatrics, Department of Occupational Therapy, Lutheran General Hospital, Park Ridge, IL 60068.

GL Girolami, PT, is Director, Pathways Center for Children, Glenview, IL 60025.

This study was approved by the University of Illinois at Chicago (H-89-422) and Lutheran General Hospital (#888) institutional human subjects review boards.

This project was supported in part by a grant from the Foundation for Physical Therapy Inc. During the course of this project, Dr Campbell and Dr Campbell and Dr Kolobe were partially supported by Grant MCJ IL 179590, Maternal and Child Health Bureau, US Public Health Service.

This work was previously presented at the 1993 Annual Meeting of the American Academy for Cerebral Palsy and Developmental Medicine and received the Patricia Miller Award for research in physical therapy; as a result, an expanded version of the paper was presented at the 1994 Annual Meeting of the American Academy for Cerebral Palsy and Development Medicine.

References

[1] Leviton A, Paneth N. White matter damage in preterm newborns. Early Hum Dev. 1990; 24:1-22. [2] Coolman RB, Bennett RC, 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. 1985;6:327-333. [3] Piper MC, Mazer B, Silver KM, et al. Resolution of neurological symptoms in high-risk infants during the first two years of life. Dev Med Child Neurol. 1988;30:26-35. [4] Campbell SK. Organizational and educational considerations in creating an environment to promote optimal development of high-risk neonates. Physical and Occupational Therapy in Pediatrics. 1986;6(3/4):191-204. [5] Kahn-D'Angelo L. The special care nursery. In: Campbell SK, Vander Linden DL, Palisano RJ, eds. Physical Therapy for Children. Philadelphia, Pa: WB Saunders Co; 1994:787-821. [6] Campbell SK. The child's development of functional movement. In: Campbell SK, Vander Linden DW, Palisano RJ, eds. Physical Therapy for Children. Philadelphia, Pa: WB Saunders Co; 1994:3-37. [7] Bertenthal BI, Campos JJ. New directions in the study of early experience. Child Dev. 1987;58:560-567. [8] Saigal S, Szatmari P, Rosenbaum P, et al. Cognitive abilities and school performance of extremely low birthweight children and matched control children at age 8 years: a regional study. J Pediatr. 1991;118:751-760. [9] Amiel-Tison C, Grenier A. Neurological Assessment During the First Year of Life. New York, NY: Oxford University Press Inc; 1986. [10] Shumway-Cook A, Woollacott M. The growth of stability: postural control from a developmental perspective. Journal of Motor Behavior. 1985;17:131-147. [11] Shumway-Cook A, Woollacott M. Dynamics of postural control in the child with Down syndrome. Phys Ther. 1985;65:1315-1322. [12] Shumway-Cook A, Woollacott M. Theoretical issues in assessing postural control. In: Wilhelm IJ, ed. Physical Therapy Assessment in Early Infancy. New York, NY: Churchill Livingstone Inc; 1993:161-171. [13] Thelen E, Kelso JAS, Fogel A. Self-organizing systems and infant motor development. Developmental Review. 1987;7:39-65. [14] Als H, Lester BM, Tronick EC, Brazelton TB. Towards a research instrument for the assessment of preterm infants behavior (APIB). In: Fitzgerald HE, Lester BM, Yogman MW, eds. Theory and Research in Behavioral Pediatrics, Volume 1. New York, NY: Plenum Publishing Corp; 1982:1123-1132. [15] Brazelton TB. Neonatal Behavioral Assessment Scale. 2nd ed. Philadelphia, Pa: JB Lippincott Co; 1984. [16] Dubowitz L, Dubowitz V. The Neurological Assessment of the Preterm and Full-Term Newborn Infant. Philadelphia, Pa: JB Lippincott Co; 1981. [17] Korner AF, Kramer HC, Reade EP, et al. A methodological approach to developing an assessment procedure for testing the neurobehavioral maturity of preterm infants. Child Dev. 1997;58:1478-1487. [18] Korner AF, Brown BW, Dimiceli S, et al. Stable individual differences in developmentally changing preterm infants: a replicated study. Child Dev. 1989;60:502-513. [19] Prechtl HFR, Beintema D. The Neurological Examination of the Full-Term Newborn Infant. London, England: William Heinemann Medical Books Ltd; 1964. [20] Cloptin N, Martin AS. A criticism of interrater reliability procedures for the Brazelton Neonatal Behavioral Assessment Scale. Physical and Occupational Therapy in Pediatrics. 1985;4(4):55-65. [21] Campbell SK, Osten ET, Kolobe THA, Fisher AG. Development of the Test of Infant Motor Performance. Physical Medicine and Rebabilitation Clinics of North America. 1993; 4:541-550: [22] Hadders-Algra M, Prechtl HFR. Developmental course of general movements in early infancy, I: descriptive analysis of change in form. Early Hum Dev. 1992;28:201-213. [23] Cioni G, Prechtl HFR. Preterm and early postterm motor behavior in low-risk premature infants. Early Hum Dev. 1990;23:159-191. [24] Hadders-Algra M, Van Eykern LA, Klip-Van den Nieuwendijk AWJ, Prechtl HFR. Developmental course of general movements in early infancy, II: EMG correlates. Early Hum Dev. 1992;28:231-251. [25] Ferrari F, Cioni G, Prechtl HFR. Qualitative changes of general movements in preterm infants with brain lesions. Early Hum Dev. 1990;23:193-231. [26] Girolami GL. Evaluating the Effectiveness of a Neurodevelopmental Treatment Program to Improve Motor Control of High-Risk Preterm Infants. Chapel Hill, NC: Department of Physical Therapy, University of North Carolina at Chapel Hill; 1987. Master's thesis. [27] Girolami GL, Campbell SK. The efficacy of a neuro-developmental treatment program for improving motor control in preterm infants. Pediatric Physical Therapy. 1994;6:175-184. [28] Cintas HM. Cross-cultural variation in infant motor development. Physical and Occupational Therapy in Pediatrics. 1988;8(4):1-20. [29] Lee AM. Child-rearing practices and motor performance of black and white children. Res Q Exerc Sport. 1980;51:494-500. [30] Plimpton CE, Regimbal C. Differences in motor proficiency according to gender and race. Percept Mot Skills. 1992;74:399-402. [31] 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. [32] Ross MG, Hobel CJ, Bragonier JR, et al. A simplified risk-scoring system for prematurity. Am J Perinatol. 1986;3:339-344. [33] Linacre JM. FACETS: Computer Program for Many-Faceted Rasch Measurement. Chicago, Ill: MESA Press; 1988. [34] Linacre JM. Multi-Faceted Rasch Measurement. Chicago, Ill: MESA Press; 1989. [35] Osten ET. Examination of the Rater Reliability the Test of Infant Motor Performance. Chicago, Ill: Department of Occupational Therapy, University of Illinois at Chicago; 1993. Master's thesis. [36] Kilgore KM, Fisher WP Jr, Silverstein B, et al. Application of Rasch analysis to the Patient Evaluation and Conference System. In: Granger CV, Gresham GE, eds. New Developments in Functional Assessment. Philadelphia, Pa: WB Saunders Co; 1993:493-515. [37] Andrich D. Rasch Models for Measurement. Newbury Park, Calif. Sage Publications Inc; 1988. [38] Hambleton RK, Swaminathan H, Rogers HJ. Fundamentals of Item Response Theory. Newbury Park, Calif: SAGE Publications Inc; 1991. [39] Wright BD, Masters GN. Rating Scale Analysis: Rasch Measurement. Chicago, Ill: MESA Press; 1982. [40] Wright BD, Linacre JM. A User's Guide to BIGSTEPS, Version 2.2. Chicago, Ill: MESA Press; 1992. [41] Wright BD, Linacre JM, Heinemann AW. Measuring functional status in rehabilitation. In: Granger CV, Gresham GE, eds. New Developments in Functional Assessment. Philadelphia, Pa: WB Saunders Co; 1993:475-491. [42] Darrah J, Piper M, Byrne P, Watt MJ. The use of waterbeds for very low-birthweight infants: effects on neuromotor development. Dev Med Child Neurol. 1994;36:989-999. [43] Piper MC, Darrah J. Motor Assessment of the Developing Infant. Philadelphia, Pa: WB Saunders Co; 1994. [44] Heriza C. Motor development: traditional and contemporary theories. In: Lister MJ, ed. Contemporary Management of Motor Control Problems: Proceedings of the II Step Conference. Alexandria, Va: Foundation for Physical Therapy Inc; 1991:99-126. [45] Thelen E, Corbetta D, Kamm K, et al. The transition to reaching: mapping intention and intrinsic dynamics. Child Dev. 1993;64:1058-1098.