Multiple interacting biological and social risk factors associated with poverty (eg, low maternal education, single parenthood, maternal substance use and abuse, low birth weight) have been associated with compromised functioning in later childhood.1 Biological and social risk factors can influence a child's prenatal and/or long-term health and development through a variety of paths. These factors could include exposure to environmental toxins, trauma, and poor nutrition; major parental factors such as parental stress and depression; lack of stimulation within the home environment and poor quality of out of home care; limited learning experiences including unsafe physical space in which to explore and play; and lack of toys or objects to manipulate and from which to learn.2 Ultimately, the risk factors associated with poverty exert both biological and social influences that potentially effect many areas of infant development including cognitive and motor functioning.1,2
Fortunately, not all infants being raised in economically disadvantaged homes will demonstrate developmental concerns. To determine the need for priority intervention services, physical therapists are asked often to evaluate infants born at term but considered at developmental risk due to issues related to low socioeconomic status.2 The accuracy of therapists' judgments is of critical importance because errors in either direction may lead to under- or overreferral for intervention services. As a result, families may experience potentially negative human and financial consequences.3
Despite a continuing clinical need, there is a paucity of empirically established guidelines or cost-effective infant neurodevelopmental assessments to help clinicians make these important referral decisions.4 In many states, clinicians are required to use standardized developmental instruments as one measure to determine infants' eligibility for special services.4 However, omnibus assessments may be too global to detect subtle sensorimotor and neurological problems, which may presage later developmental difficulties.4–7 In addition, test development of most extant infant assessments includes limited instrument evaluation with an array of specific groups of infants who are at risk.4 Furthermore, the heterogeneous nature of each specific at-risk group of infants requires that more in-depth instrument evaluation (eg, specificity, sensitivity, predictive values) with clinically relevant infant subsamples occur once the instrument is in general use.
One potentially useful and cost-effective assessment for this purpose is the Movement Assessment of Infants (MAI). The MAI is one of the few assessments of young infants beyond the newborn period that can be used to document neuromotor maturation in areas such as reflexes and quality of movement.8 The MAI is a 65-item criterion-based assessment originally developed to assess the status of four-month-old neonatal intensive care unit (NICU) graduates at risk of long-term neuromotor problems such as cerebral palsy. Most prior research studies using the MAI involved infants who were born preterm, required NICU care, and were at risk of later motor impairment such as cerebral palsy.9–11 Currently, professionals generally use the MAI with a broader range of infants for screening and assessment in clinical practice and research.12–14 However, studies evaluating the predictive validity of the MAI with term infants at social risk are rare. Those that exist often have methodological problems that limit interpretation and generalizability of the results, such as small sample size, which decreases statistical power, or the use of unmasked examiners, which increases the potential for measurement bias.15,16 Researchers in most prior follow-up MAI studies of infants born at term that are at social risk either have not examinined infants at least 24 months of age when developmental tests begin to be more predictive of longer term outcome or have evaluated motor outcomes only.13,14
The Bayley Scales of Infant Development (BSID) is an established, standardized test with determined normal values of infants' cognitive and motor functioning with well-described psychometric test properties.17 The BSID was developed to assess current developmental functioning, assist in the diagnosis of developmental delay, and identify the need for developmental intervention.17 BSID scoring of the Mental and Psychomotor Scales yields two age-normed standard scores: a Mental Developmental Index (MDI) and a Psychomotor Developmental Index (PDI). The MDI and PDI are each standardized to have a mean of 100 and a standard deviation of 16. An MDI or PDI falling one or more standard deviations below the population mean (≤84) is considered a marker of developmental concern because it is below the test standard of “within normal limits.”17
The goal of this analysis was to evaluate the predictive validity of the MAI in a relatively large sample of infants who were born at term and did not require NICU care but were considered at developmental risk due to issues associated with low socioeconomic status. We assumed that these infants' developmental risk status at four months as measured by the MAI would not predict severe developmental diagnoses such as cerebral palsy at two years of age but rather would be associated with more subtle developmental problems such as low BSID scores. Four specific questions were evaluated in this sample: 1) Are infants' MAI total risk scores (TRS) at four months of age correlated with their cognitive and motor developmental status at age two years, as assessed by the MDI and PDI of the BSID)? 2) Is a categorical variable of risk status on the MAI (eg, high risk = TRS >13) at four months of age associated with a categorical variable of risk status on the BSID at two years (ie, high risk = MDI or PDI ≤84)? Moreover, is a two-level or a three-level classification of risk on the MAI a more powerful predictor of two-year BSID risk status? 3) Does the presence or absence of specific sociodemographic variables (eg, maternal education <12 years) or biologic risk variables (eg, low birth weight <2500g) modify the relation between MAI risk status at four months and BSID MDI or PDI risk status at 24 months? 4) What specific MAI TRS produces the strongest prediction of developmental status at two years as measured by the sensitivity, specificity, positive predictive value, and negative predictive value using different MAI TRS cutoffs?
Subjects were 134 infants from a larger longitudinal prospective study (n = 250) who had complete data sets (both an MAI TRS at four months and BSID scores at two years). All infants were born at term but were considered to be at risk of developmental problems due to low socioeconomic status and associated risk factors (eg, low maternal education, single parenthood, Medicaid as source of health insurance, a proxy for low income). Prenatally, infants were exposed to a range of alcohol, marijuana, and/or tobacco, and 49% were additionally exposed to cocaine, as documented by biological assay in the newborn period (either positive maternal or infant urine screens or infant meconium assay) and/or positive maternal self-report. The remaining 51% had negative urine and/or meconium assays for cocaine and negative maternal self-reports for cocaine use during pregnancy. In the larger study, mother–infant dyads were recruited from October 1990 to March 1993, within two days of the infant's birth at Boston Medical Center (formerly Boston City Hospital). Mother–infant dyads were eligible for recruitment if they met the following criteria: 1) infant gestational age >36 weeks; 2) no requirement for level III (neonatal intensive) care; 3) no obvious major congenital malformations; 4) no diagnosis of fetal alcohol syndrome in the neonatal record; 5) no history of HIV seropositivity noted in the mother's or infant's medical record; 6) mother's ability to communicate fluently in English; 7) mother >18 years of age; and 8) no indication by urine and/or meconium assay screen and medical record of mothers' use during pregnancy of illegal opiates, methadone, amphetamines, phencyclidine, barbiturates, or hallucinogens. Trained interviewers screened maternity and nursery records seven days a week and then sought informed consent from eligible women who, with their infants, met the selection criteria. Details of recruitment and newborn characteristics have been reported previously for this cohort.18
Results of bivariate analyses (t tests or χ2 square analysis) revealed no significant differences on biologic or sociodemographic variables among the 134 infants in the present sample and those enrolled in the larger study who did not contribute data for analysis in the present study. Specifically, the two groups did not differ on infant birth weight (p = 0.45), gestational age (p = 0.25), maternal age (p = 0.91), maternal education (p = 0.25), parity (p = 0.14), ethnicity (p = 0.64), and Medicaid status (p = 0.49). Missing data and attrition in the larger sample can be attributed in part to the high-risk characteristics of the sample. For instance, as observed for families in other longitudinal studies with similar demographics, families in the present study as a group frequently 1) changed residences or were homeless, 2) temporarily or permanently moved out of state or country, 3) had telephones disconnected or changed repeatedly, and/or 4) resided in temporary drug treatment or domestic violence shelters. These circumstances may have caused some families to miss some time-dependent research appointments at four months and/or two years.
The MAI was administered to infants at four months of age by trained examiners. The MAI was developed to evaluate muscle tone, primitive reflexes, automatic reflexes, and volitional movement in infants who were biologically at risk, such as infants born prematurely and requiring NICU care. Standard MAI scoring was used, and individual test items were scored as either “risk” or “no risk” on the basis of published age-referenced criteria. The risk points were summed to create a TRS. Originally, the TRS cutoff values for a predominantly European-American sample of typically developing infants were: 0–7 = no risk; 8–13 = suspect; >13 = high risk.8 Subsequent studies of NICU graduates have recommended a cutoff MAI TRS of >10 (one standard deviation above the mean for healthy low birth weight infants) to detect later neurological abnormalities.9,11,12 Data are not available yet to identify the appropriate MAI TRS cutoff for infants born at term who have not required NICU care but are considered at developmental risk due to low socioeconomic status and associated risk factors. Therefore, in the present study, we evaluated which MAI cutoff score was the strongest predictor of two-year risk status on the BSID.
Examiners received initial training on the MAI at an on-site two-day training seminar led by a developer of the MAI. A physical therapist with established reliability using the MAI trained a team of examiners to reliability and served as the reference standard for ongoing reliability checks. Interrater reliability was evaluated after the testing of approximately every tenth child. Exact agreement on individual test items was calculated using the following formula [agreements/(agreements + disagreements) × 100] and averaged 85.8% (range = 77.1%–93.8%). This method was used so that scoring of specific items could be checked and discussed after each formal reliability testing and scoring session.
In the larger study, the MDI and PDI of the BSID were administered to infants at six, 12, and 24 months of age by an independent team of trained examiners who were masked to infant history and prior MAI scores. The hypotheses of the present MAI study were evaluated using only the two-year BSID outcomes, the oldest age at which the BSID was administered. The original BSID, rather than the second edition (BSID-II), was used because the BSID-II had not been published at the onset of the data collection for the larger study.
Scoring of the Mental and Psychomotor Scales of the BSID yields two age-normed standard scores (MDI and PDI), each standardized to have a mean of 100 and a standard deviation of 16. In addition to the continuous MDI and PDI scores, a dichotomous “risk” score was also derived. Infants were classified as being at risk if their MDI or PDI scores fell one or more standard deviations below the population mean (≤84). Infants with an MDI or PDI score >84 were classified as being “within normal limits.”17
Initial and ongoing BSID training and reliability for this study were led by a psychometrician who had trained and worked under a colleague of the author of the BSID. Formal interrater reliability was checked at the beginning, middle, and end of the study. Average interrater reliability (exact agreement) for individual MDI and PDI items was consistently above 90% (range = 81%–100%). As in the procedures used for evaluating MAI reliability, percentage of agreement at the individual item level rather than at the developmental index level was used so that scoring of specific items could be discussed after reliability documentation was completed.
The primary caregiver brought her infant to Boston Medical Center for MAI testing when the child was four months old (mean = 4.2 months, SD = 0.4) and again for BSID testing and other assessments when the child was 24 months old (mean = 24.8 months, SD = 1.5). Examiners at each visit were masked to infant history, including prenatal drug exposure and previous developmental assessment scores. To minimize possible scoring bias by knowledge of the earlier testing protocol, a different examiner conducted the assessments at each age whenever possible. Examiners scored all MAI and BSID examinations in the standardized manner. After each assessment, caregivers received store vouchers to thank them for their participation in the study.
Four sets of analyses19 were conducted to evaluate the study's four questions. Because of the nongaussian distribution of MAI TRS, we used nonparametric Spearman rank correlations in the first set of analyses to evaluate the relationship between infants' MAI TRS at four months and the BSID MDI and PDI scores at two years. Correlations are commonly used to quantify the degree of association between two measures in studies of neuropsychological research. In the second set of analyses, cross-tabulations were used to evaluate whether a two-level categorical measure of infants' risk status on the MAI (TRS >13 = high risk, TRS ≤13 = lower risk) or the traditional three-level categorical measure of risk status (TRS >13 = high risk, 8–13 = suspect, 0–7 = no risk) better discriminates infants' dichotomous risk status on the BSID at 24 months (BSID MDI or PDI ≤84 = high risk, MDI or PDI >84 = within normal limits). These analyses were performed because categorical variables are likely to be more useful in clinical practice than continuous variables, particularly when determining infants' eligibility for intervention services. We tested the differences in the distribution of BSID risk across TRS categories using χ2 tests. In the third set of analyses, we evaluated whether the presence or absence of specific social or biologic risk variables modified the relationship between MAI risk status at four months and BSID MDI and PDI risk status at two years. To carry out these analyses, we stratified the sample on an exhaustive, a priori list of biologic and social contextual variables that have been shown in the literature to influence directly or indirectly infants' developmental outcomes at these ages. These variables included the following: low birth weight (<2500 g), infant gender, African-American or African-Caribbean ethnicity, low maternal education (<12 years), and any prenatal exposure to cocaine, alcohol, marijuana, or to 10 or more cigarettes per day. We used a cutoff of 10 or more cigarettes per day because that level of cigarette smoking has been shown in prior research to differentiate light from heavy smokers and to affect infants' developmental outcomes.20 Because most of the infants qualified for a variety of possible community interventions due to social risk factors but may not have received the services, we also stratified the sample on the receipt of child-focused intervention prior to age two. This variable was obtained from caregiver report during the interview at each study visit. Receipt of intervention services was defined as participation in any type of child-focused early intervention up to age two years (eg, formal home- or clinic-based early intervention programs; home- or clinic-based professional visits from a nurse, educator, or physical, occupational, or speech therapists).
A fourth set of analyses was conducted to determine the optimal cutoff score on the MAI (ie, the cutoff score that best predicted BSID risk status in this sample). To accomplish this goal, we computed sensitivity, specificity, positive predictive value, and negative predictive value of different MAI TRF cutoff scores ranging from 9 to 15.19 Sensitivity is the probability that a screening test (or other marker) accurately detects individuals who are affected with the condition or disease of interest. In other words, if sensitivity of a screening test was 81%, then that test identified 81% of the infants who were actually at risk. We defined sensitivity of the MAI as the percentage of children classified as at risk at four months on the MAI among those classified as at risk on the BSID at two years.19,21 Specificity is the probability of a screening test (or other marker) to accurately detect individuals who are identified as not affected with the condition or disease of interest.19 In other words, if specificity of a screening test was 40%, then that test identified 40% who actually were not at risk. We defined specificity of the MAI as the percentage of children classified as not at risk at four months on the MAI among those within normal limits on the BSID at two years. Positive predictive value is the probability that an individual is not within normal limits among those who had a screening score that was defined as at risk. We defined positive predictive value as the percentage of infants who had BSID MDI or PDI scores of ≤84 among those with an at-risk MAI score at four months. We defined negative predictive value as the percentage of infants who had BSID MDI or PDI scores >84 among those who did not have an at-risk MAI score at four months. The C statistic, the area under the receiver operating curve, is a function of sensitivity and specificity and is defined and computed as a summary measure of the overall predictive performance of the MAI TRS to subsequent risk status on the BSID. Higher values of the C statistic (ie, those closer to 1.0) indicate more optimal prediction. We computed positive and negative predictive values assuming population prevalence for an at-risk BSID score at two years of 32%, the rate of BSID scores ≤84 in this at-risk sample. As previously indicated, a score of 84 is the demarcation of one standard deviation below the population mean, and scores equal to or below this number are recognized as a level of developmental concern identified in the BSID manual. Analytic results with two-tailed p values <0.05 were noted as statistically significant.
Descriptive statistics depicting the sociodemographic and biological characteristics of the 134 children in this sample are provided in Table 1. Of note, 77% of the study families were receiving Medicaid and the average level of maternal education was 11.4 years, attesting to the high social risk status of the majority of families in this sample. Forty-nine percent of the infants had documented in utero exposure to cocaine, and the majority of infants also had a range of documented in utero exposure to alcohol, marijuana, and/or cigarettes.
Means and Correlations of MAI TRS and BSID Scores
The mean MAI TRS was 10.7 (SD = 6.0). On the BSID, the mean MDI was 89.7 (SD = 15.5), and mean PDI was 102.5 (SD = 15.1). There was a statistically significant (albeit modest) inverse Spearman rank correlation between MAI TRS and BSID MDI (r = −0.23, 95% confidence interval: −0.38, −0.06; p = 0.007). The correlation between MAI TRS and BSID PDI was not significant (r = −0.07, 95% confidence interval: −0.23–0.10; p = 0.43).
Relationship Between Infant Risk Status at Four Months and Two Years
In Table 2, the relationship between infants' risk status on the MAI and their later risk status on the BSID at two years is shown. When a three-level MAI TRS classification variable (0–7 = no risk; 8–13 = suspect; >13 = high risk) was used (see Table 2), results of χ2 square analyses indicated that infants' MAI risk status was significantly associated with their risk status on the BSID, but only for MDI (p = 0.01). BSID risk status on the PDI was not associated with MAI risk status (p = 0.52). As can be seen in Table 2, the strongest relationship between MAI TRS and BSID MDI was observed in the group with TRS >13. Fifty percent of infants with TRS >13 had an MDI ≤84, whereas only 25% of infants with MAI TRS of 0–7 and 23% of infants with a TRS of 8–13 had an MDI ≤84. Clinically stated, infants with an MAI TRS >13 had approximately three times the risk of achieving an MDI ≤84 on the BSID at two years (p = 0.003) compared with infants with an MAI TRS of ≤13 (odds ratio = 3.2; 95% confidence interval 1.5–6.9).
The two-level MAI TRS classification variable results shown in Table 3 support the above findings in that 50% of the 42 children who had MAI TRS >13 had at-risk BSID MDI scores, whereas only 24% of the 90 children who had MAI TRS ≤13 had at-risk BSID MDI scores. Only 7% of the 42 children who had MAI TRS >13 at four months had at-risk PDI scores at two years, as compared with 10% of the children who had MAI TRS ≤13 (p = 0.75).
Because BSID PDI was not associated with MAI TRS in any of the above analyses, it was not evaluated further as a dependent variable in the subsequent analyses described below.
To evaluate whether biological and social risk factors modify the relationship between MAI risk status and BSID MDI risk status, we stratified the sample on presence or absence of each of the biological and social risk variables described above and then ran cross-tabulations between MAI risk and MDI risk status. None of the results for the stratification variables were statistically significant. However, the results of χ2 analyses indicated that receipt of child-focused intervention services in the first two years of life (unlike the other stratification variables) moderated the relationship between MAI TRS risk status and BSID MDI risk status. Among the 108 children who did not receive early intervention, 45% of children with TRS >13 had MDI scores that were at risk compared with 22% of those who had TRS ≤13, an effect quantifiable by an odds ratio of 3.0. In contrast, among the 26 children who received early intervention, 67% of children with TRS >13 had MDI scores that were at risk versus 23% of those who had TRS scores ≤13, equal to an odds ratio of 6.5.
Sensitivity and Specificity Analyses
To determine the best MAI TRS cutoff score to predict BSID risk status, we computed C statistics, sensitivities, specificities, positive predictive values, and negative predictive values for a series of different two-category MAI TRS risk classifications (>9 through 15 in one-point TRS increments). In Table 4, the cutoffs of TRS >12 and TRS >13 are associated with the relatively largest C statistic (0.629). Both cutoffs have identical sensitivity (0.49) and specificity (0.77) values. However, TRS >13 also has the greater positive (0.50) and negative (0.76) predictive values of the two cutoffs, indicating that TRS >13 is the lowest cutoff tested that best discriminates BSID MDI risk in this sample.
This longitudinal study is one of the first to evaluate the validity of the MAI as a predictor of infants' cognitive and motor developmental functioning at two years of age in an understudied sample of infants born at term from urban, low-income families. Despite being born at term and relatively healthy at birth, the infants in this sample were presumed to be at risk of developmental delays due to the presence of social and/or biological risk factors associated with poverty. The descriptive statistics of the sample's demographic characteristics in Table 1 including low maternal education, high use of public health insurance (a proxy for low income), as well as the relatively poor developmental scores at four months and two years confirm the sample's social risk status and are representative of the children and their families who receive medical services at the urban hospital where the present sample was drawn. Studies of developmental risk in this population are important because young children have one of the highest poverty rates of any age group in the United States, and poverty experienced during early childhood is thought to be more detrimental to future development than poverty during later life.2
The infants born at term in this sample differ from those included in many previous MAI studies due to the larger study's inclusion criteria. For instance, prematurely born infants, infants requiring ventilation and/or extended NICU care, infants of young teen mothers, and mothers who at the time of giving birth were HIV positive were excluded. Each of these conditions is known to be associated with compromised infant outcomes. Thus, the relatively healthy infants born at term in this sample were not expected to be at risk of severe neurological disorders such as cerebral palsy. Indeed, by two years of age, none of the infants in this sample were diagnosed with cerebral palsy.
Nevertheless, the infants exhibited a range of developmental functioning on the MAI at four months of age and on the BSID at two years of age. Moreover, the average level of their performance on these assessments was relatively poor, compared with published age norms. The mean MAI TRS was 10.7 (range = 1–26), approximately one standard deviation above (in the less optimal direction) what has been reported for groups of infants who are typically developing and from economically secure families.8,11 Although the mean BSID MDI (mean = 89.7) and PDI (mean = 102.3) reported in this study were within normal limits, the average MDI was 12.6 points (approximately three fourths of a standard deviation) lower than the PDI. The careful reliability process in this study using an outside evaluator as the reference standard and the use of masked examiners decreases possibilities of evaluator bias as a reason for the decreased scores. These less optimal MAI and BSID scores provide further support for the social risk status of the sample as a whole.
The discussion of the results of this study is organized by the study's four questions. In addition to addressing the validity of the MAI in predicting infants' developmental functioning at two years of age within this full-term but socially and biologically vulnerable sample, the clinical implications and limitations of the study are considered.
The first aim of the study was to evaluate whether infants' MAI TRS at four months of age was correlated with their cognitive and motor developmental status at age two years, as assessed by BSID MDI and PDI. Our finding of a statistically significant though modest correlation of four-month MAI TRS to 24-month MDI is consistent with similar magnitudes of findings in other studies evaluating other subtle longitudinal relationships such as newborn neurobehavior with two-year BSID21 or specific infant findings with school-age cognition.22 Given that significant differences are often difficult to identify in a relatively homogeneous sample, the degree of association of MAI TRS to BSID MDI findings at two years of age in these data is noteworthy.
Additionally, the association between MAI TRS and infants' later cognitive but not motor outcomes on the BSID mirrors results from previous longitudinal studies. For instance, Ellison22 reported several studies of children who had transient or subtle motor findings in infancy and lower cognitive measures at preschool or school age compared with those children without the early minor neurological findings. In addition, Wilden et al23 reported that in young infants of very low birth weight, more abnormal neurological scores predicted greater deceleration of later cognitive development. Also, a growing body of recent neuroanatomical and neurophysiological evidence indicates that motor and cognitive function fundamentally may be interrelated via several neurological structures, most importantly the neocerebellum and the dorsolateral prefrontal cortex.24 Although the cerebellum often had been associated with motor functions and the prefrontal cortex with cognition, recent functional neuroimaging studies have shown activation in both structures during cognitive tasks such as those involving language or memory tasks as well as novel motor tasks and motor tasks involving timing.24
In regard to instrumentation, the MAI may predict to the BSID because the MAI is a clinical and research tool that documents subtle early neuromotor differences between infants originating from a variety of nonspecific developmental influences. For example, while the MAI documents muscle tone, primitive reflexes, automatic reactions, and volitional movement, it also documents right-left qualitative asymmetries in these areas. In the case of an observed asymmetry, the less optimal of the two scores is recorded for the item and therefore influences the MAI TRS. Given the limited observable repertoires of four-month-old infants, the early subtle or seemingly transient neuromotor concerns such as those documented within the context of the MAI may indicate an immature or diffusely impaired central nervous system. That immaturity in turn may be associated with later and specific cognitive, language, and behavioral abilities.5–7,25 Also, the MAI may reflect early infant differences in visual perception, attention, self-regulation that are not tested specifically by the MAI or the BSID but could influence the infants' abilities on both assessments. Without more complex evaluation at either of the two ages, possible four-month difficulties could appear to be within the motor output because of the relatively limited variety of output of an infant compared with a preschool or school-aged child. Potential perceptual and attentional issues in this sample could be more obvious in school-age outcomes. Evaluation of these areas in this cohort is planned.
BSID instrumentation also must be considered. To put the correlation between MAI TRS and two-year BSID MDI into additional perspective in our study, we did a post hoc correlation of the six-month MDI (closest age to the four-month MAI) to the two-year MDI. Notably, we found the correlation between six-month and two-year BSID MDI (an 18-month time span) was only 0.16, in contrast to the higher correlation (0.23) between MAI TRS and BSID MDI (a 20-month time span). Therefore, it appears that the four-month MAI is a better predictor of two-year MDI than the six-month MDI. This post hoc finding most likely is due to the neurological emphasis of the MAI described above and the developmental task emphasis of the BSID assessment at each age.17 Future studies should investigate the relationship of the MAI to later infant outcomes compared with neurological or other neurologically based assessments. The second version of the BSID (BSID-II) is addressed later.
The second aim of the study was to evaluate whether a categorical variable of risk status on the MAI (eg, high risk = TRS >13) at four months of age was associated with a categorical variable of risk status on the BSID at two years (ie, high risk = MDI or PDI ≤84). We also evaluated whether a two- or three-level classification of risk on the MAI was a more powerful predictor of two-year BSID risk status.
Results showed that infants with a MAI TRS >13 had three times the odds26 of having a low MDI (≤84) than those with MAI TRS of 0–7 or 8–12. These results support the clinical use of a two-level categorization of MAI TRS (low to medium and high) with infants who are at risk born at term rather than the three-level interpretation (low, medium, and high) of TRS in the MAI manual for infants who were NICU graduates and of predominantly European descent. Knowing that an infant born at term from an impoverished background with an MAI TRS >13 has a threefold increase of developmental risk at two years of age clearly is important information in clinical planning and decision making, particularly related to intervention services.
The third goal of the study was to evaluate whether the presence or absence of specific sociodemographic risk factors within this low-income cohort (eg, maternal education <12 years) or biological risk variables (eg, low birth weight <2500 g) moderates the relationship between MAI risk status at four months and BSID MDI or PDI risk status at two years. Contrary to our expectations, the results of these analyses were not statistically significant for the risk variables evaluated. This was presumably due to the relatively homogeneous, high-risk status of the sample as a whole. With the exception of early intervention, stratification of this sample by specific biological and social risk factors showed nearly identical patterns of relationships between MAI TRS and BSID. It may be that cumulative, interacting risk factors may be of greater importance than the single risk factors in this low-income sample. Alternatively, it is possible that the same analyses in future studies carried out in larger or less homogeneous samples might provide additional information in this area.
Although the purpose of the present study was not intended to evaluate specifically the impact of unique risk factors, including prenatal cocaine exposure on developmental outcome, cocaine is a highly controversial risk factor that requires some elaboration.27 In a well-controlled, prospective study of 28 infants who were exposed to cocaine and 22 infants who were unexposed with demographics similar to those of the infants in the present study, Fetters and Tronick13 reported significant MAI TRS differences between infants who were cocaine exposed and unexposed at seven but not at four months on the MAI or at 15 months on the Peabody Developmental Motor Scales (PDMS). Unfortunately, that study did not report cognitive assessments. Other studies finding significant MAI differences by cocaine exposure either had subjects who were predominantly European American rather than African American/African Caribbean28 or were retrospective studies.29
Interestingly, results of the stratification analyses identified early child-focused intervention services as a significant moderator of the relationship between MAI risk status and BSID MDI. Unlike the pattern of the other stratified variables, the MAI TRS had a notably better positive predictive value (67%) among children who received early intervention than among those who did not (45%) and with approximately the same negative predictive value (22% and 23%).
In interpreting these results, it is important to note that the larger study from which the present sample was drawn was not designed to provide intervention services for study subjects nor was the evaluation of the effects of intervention on child outcome the study's objective. The researchers did document the children's receipt of intervention services as broadly defined but did not evaluate the length, intensity, or quality of services provided. Nonetheless, caregiver report of the child's receipt of early intervention even broadly defined may be a marker for higher or more obvious developmental risk within an already at-risk sample. To be eligible for enrollment in intervention programs during the first two years of life, infants must have been identified by outside service providers (unaware of the infant's MAI scores in this study) as being at developmental risk. If this assumption is correct, then poor MAI scores (TRS >13) at four months of age may be a reasonable predictor of poor BSID MDI (>84) at two years within this subsample of infants.
The fourth aim of the study was to evaluate what specific cutoff score of the MAI TRS produces the strongest prediction of developmental status at two years as measured by sensitivity, specificity, positive predictive value, and negative predictive value analyses. The results indicate that in a sample such as in the present study, an MAI TRS of 13 may be the most optimal cutoff score. However, the MAI is more useful in identifying infants who are more (rather than less) likely to have BSID MDI scores that fall within normal limits at two years of age. Therefore, the MAI, using a TRS cutoff of 13, may be one valid neurodevelopmental measure to help clinicians and other professionals rule out the need for more intensive developmental services in infants from backgrounds similar to those in the present study. Ideally, the screening instrument should help professionals identify those infants who do rather than do not need intensive developmental services. Furthermore, ideal sensitivity and specificity of a screening instrument should be >0.80 and the positive predictive value of >0.70.30 However, those ideal values are more likely to be attained statistically in heterogeneous samples such as that representative of a national population than in relatively homogeneous samples such as the present sample that is representative of a particular population subsample of clinical interest and importance. Also, given the lack of available neurodevelopmental screening instruments with known and ideal predictive values for very young infants born at term at developmental risk due to social factors, it is clinically important to know the characteristics including strengths and weaknesses of extant assessment instruments that currently are being used to evaluate infants in this understudied population.
The results of this study do not appear to be influenced by differential attrition or differences between the present sample and the larger group on demographic, maternal, or infant characteristics, based on results of analyses comparing the subjects in the present sample with those in the larger recruited sample.
However, the specific findings from this study most likely were influenced by the recruitment criteria and the population base of the hospital, which resulted in a sample of infants from predominantly African-American/Caribbean ethnicities who were born at term and clinically healthy at birth but still developmentally vulnerable due to the presence of risk factors associated with poverty. For similar reasons, the generalizability of the results of this study is limited to infants born at term from similar low-income, urban backgrounds and mostly African-American/Caribbean ethnicities. Our findings may not apply to infants with higher levels of biological risk, such as prematurely born infants or those requiring extended NICU stays or to infants from more advantaged socioeconomic backgrounds.
Nevertheless, our results expand and extend findings from prior studies using the MAI. As noted in the introduction, many previous longitudinal studies using the MAI and BSID have evaluated infants who were specifically at risk of motoric delays or deficits secondary to prematurity or other perinatal complications. Moreover, these studies have typically evaluated only motor outcomes. Thus, those studies have limited generalizability to infant samples such as ours.11–14 One MAI study of 77 infants with demographic characteristics similar to those of infants included in our study reported that there were no clinically significant relationships between the Peabody Developmental Gross Motor Scale or Frostig Eye-Motor Coordination Subtest at four years of age and the MAI at four months of age.29 Cognitive outcomes, however, were not assessed.
Another potential limitation of this study was that the BSID rather than the BSID-II was used to evaluate children's developmental outcomes. It is uncertain what the actual effect on the results of this study would be if the BSID-II rather than the BSID had been used for the two-year outcome. According to the BSID-II manual, scores on the newer test version would be an average of 12 points lower on the MDI and 6 points lower on the PDI than if the older BSID version had been used.17 However, other test modifications in the newer version also may change the relationship between the two test versions with this at-risk sample making a simple adjustment of scores unfeasible. The BSID standard administration ensured a relatively stable basal level for each individual infant tested by directing the examiner to identify the basal level as the point of the first failed item after ten sequentially passed items. In contrast, the standard BSID-II ensures a relatively stable basal level across infants of a particular age by directing the examiner to test the child on a specific set of items for the child's age, with minor objective modifications. The assumption of the BSID-II is that all items prior to the first item of the recommended item set are passed even though no items were administered prior to the basal level. This test assumption is maintained even if the first administered items actually are failed and there is no evidence that prior items had been passed. Solid development prior to the first administered BSID-II item is less likely for children who are at risk of poorer developmental outcomes and who may have scattered developmental abilities or inconsistencies within the developmental domains assessed by the BSID-II's mental and motor scales.31 If this were the case, the BSID-II scores might be inflated in comparison with BSID scores. In contrast, the BSID-II was developed to address the concerns that the original normative sample was out of date, thereby causing inflation of developmental indexes of infants tested with the original BSID.32
In addition to these two conflicting issues related to the BSID outcome measurement, there are other identified and unidentified factors that may have influenced our results. Further studies are needed to evaluate the validity of the MAI for predicting children's developmental outcomes using the BSID-II in similar samples of infants born at term but at social and biological risk of development.
The results of this study contribute to a growing literature evaluating the usefulness of the MAI as an early neuromotor measure of developmental risk. The MAI is one instrument used in clinic settings to determine the need for closer developmental follow-up including the need for referral to early intervention. Optimally, developmental measures should help clinicians identify infants who should, rather than should not, receive priority for early intervention services. Unfortunately, there is a paucity of neurodevelopmental measures with high predictive validity for infants who are born at term and did not require NICU care yet still are biologically vulnerable and environmentally at risk due to risk factors associated with poverty. The identification of criteria for use with extant developmental measures would be an important step toward improved, appropriate, and inclusive evaluation for this relatively large segment of the United States population, who are infants at risk developmentally.2,33 The results of this study provide clinicians and researchers with important new information about how the MAI, an extant assessment instrument, might appropriately be used and interpreted with a specific but large population of infants at risk who are underserved.
In summary, the results of this study suggest that the MAI is an instrument with some predictive validity and should be considered useful when implemented in a battery of measures evaluating infants born at term from low socioeconomic backgrounds who are at diverse biological and social risk of overall delayed development. Infant neuromotor development as measured by the MAI TRS at four months of age correlates in our sample with infants' cognitive functioning at two years of age, as measured by the MDI of the BSID. Results of sensitivity and specificity analyses suggest that an MAI TRS cutoff of 13 may identify infants similar to those in the study sample who are at risk of development but not at risk of cerebral palsy.10 Clinically, in children who were born at term and are at biological and social risk, the MAI may be a better measure for ruling out, rather than ruling in, the need for priority intervention services.
There is evidence to suggest that all infants at social and biological risk should, if adequate resources are available, receive compensatory intervention to optimize later development.2,3,34 However, in a time of limited intervention resources, it is useful to have an empirically determined basis for prioritizing intervention. Further clarification of the MAI as a neuromotor assessment for infants at risk will come with future studies using the MAI and other neurodevelopmental instruments with more diverse infant samples that are followed to later ages. These studies need to identify the relationship of the MAI and other neurodevelopmental assessments with other later infant, preschool, and school-age measures and documentation, including the BSID-II, and other cognitive, motor, and child development indicators.
We are grateful to the families who participated in this longitudinal study. We also thank Drs. Marie Anzalone and D. Rebecca Brown, Nancy Flaherty, Grace Brilliant Gustafson, Alica High, Judith James, and others for their valuable assistance with data collection or subject tracking and coordination.
1. Sameroff AJ, Chandler MJ. Reproductive risk and the continuum of caretaking casualty. In: Horowitz FD, Herrington EM, Scarr-Safapatick S, et al, eds. Review of Child Development Research, Vol 4
. Chicago: Society for Research in Child Development
2. Aber JL, Jones S, Cohen J. The impact of poverty on the mental health and development of very young children. In: Zeanah CH, eds. Handbook of Infant Mental Health
. 2nd ed. New York: Guilford Press; 2000:113–128.
3. Barnett S. Economics of early childhood intervention. In: Shonkoff JP, Meisels SJ, eds. Handbook of Early Childhood Intervention
. 2nd ed. Cambridge, UK: Cambridge University Press; 2000:589–610.
4. Meisels SJ, Atkins-Burnett. The elements of early childhood assessment. In: Shonkoff JP, Meisels SJ, eds. Handbook of Early Childhood Intervention
, 2nd ed. Cambridge, UK: Cambridge University Press; 2000:231–257.
5. Nelson KB, Ellenberg JH. Children who outgrew cerebral palsy. Pediatrics
6. Saigal S, Rosenbaum P, Szatmari P, et al. Learning disabilities and school problems in a regional cohort of extremely low birth weight children: a comparison with term controls. J Dev Behav Pediatr
7. Siegel L. Infant
tests as predictors of cognitive and language development two years. Child Dev
8. Chandler LS, Andrews MS, Swanson MW. Movement Assessment of Infants: A Manual
. Rolling Bay, WA: Chandler, Andrews, and Swanson; 1980.
9. Paban M, Piper MC. Early predictors of one year neurodevelopmental outcomes for at risk infants. Phys Occup Ther Pediatr
10. Harris SR. Early diagnosis of spastic diplegia, spastic hemiplegia, and quadriplegia. Am J Dis Child
11. Swanson MW, Bennett F, Shy KK, et al. Identification of neurodevelopmental abnormality at four and eight months by the Movement Assessment of Infants. Dev Med Child Neurol
12. Piper MC, Pinell LE, Darrah J, et al. Early developmental screening: sensitivity and specificity of chronological and adjusted scores. J Dev Behav Pediatr
13. Fetters L, Tronick EZ. Neuromotor development of cocaine-exposed and control
infants from birth through 15 months: poor and poorer performances. Pediatrics
14. Arendt R, Singer L, Angelopoulos J, et al. Sensorimotor development in cocaine-exposed infants. Infant Behav Dev
15. Rose-Jacobs R, Cabral H, Posner M, et al. Do “we just know?” Masked assessors' ability to accurately identify children with prenatal cocaine exposure. J Dev Behav Pediatr
16. Eyler FD, Behnke M, Stewart N, Wobie K. Examiner masking in research on the effects of prenatal cocaine exposure. J Drug Issues
17. Bayley N. Manual for the Bayley Scales of Infant Development
. New York: Psychological Corporation; 1969, 1993.
18. Tronick EZ, Frank DA, Cabral H, et al. Late dose-response effects of prenatal cocaine exposure on newborns neurobehavioral performance. Pediatrics
19. Rosner B. Fundamentals of Biostatistics
. Belmont, CA: Duxbury Press; 1995.
20. Fried PA. Prenatal exposure to tobacco and marijuana: effects during pregnancy, infancy, and early childhood. Clin Obstet Gynecol
21. Singer LT, Arendt R, Minnes S: Relationship between newborn neurobehavioral outcomes and 6, 12, and 24 month Bayley Scales of Infant
Development. Presented at the International Conference of Infant
Studies, Toronto, Canada, April 18–21, 2002.
22. Ellison PH. Relationship of motor and cognitive function in infancy, preschool, and early school years. J Clin Child Psychol
23. Wildin SR, Smith K, Anderson A, et al. Prediction of developmental patterns through 40 months for 6- and 12-month neurologic examinations in very low birth weight infants. J Dev Behav Pediatr
24. Diamond A. Close interrelation of motor development and cognitive development and of the cerebellum and prefrontal cortex. Child Dev
25. Murray EA, Cermak SA, O'Brien VO. Relationship between form and space perception, constructional abilities, and clumsiness in children. Am J Occup Ther
26. Fleiss JL. Statistical Methods for Rates and Proportions
. New York: John Wiley & Sons; 1981.
27. Frank DA, Rose-Jacobs R, Beeghly M, et al. Level of prenatal cocaine exposure and scores on the Bayley Scales of Infant
Development: modifying effects of caregiver, early intervention, and birth weight. Pediatrics
28. Swanson MW, Streissguth AP, Sampson PD, et al. Prenatal cocaine and neuromotor outcome at four months: effect of duration of exposure. J Dev Behav Pediatr
29. Deitz JC, Crowe TK, Harris SR. Relationship between infant
neuromotor assessment and preschool motor measures. Phys Ther
30. Glascoe FP, Bryne KE, Ashford LG, et al. Accuracy of the Denver-II in developmental screening. Pediatrics
31. Gauthier SM, Bauer CR, Messinger DS, et al. The Bayley Scales of Infant
Development. II: Where to start? J Dev Behav Pediatr
32. Campbell SK, Siegel E, Parr CA, et al. Evidence for the need to renorm the Bayley Scales of Infant
Development based on performance of a population-based sample of twelve-month-old infants. Top Early Child Special Educ
33. Cicchetti D, Wagner S. Alternative assessment strategies for the evaluation of infants and toddlers: An organizational perspective. In: Meisels SJ, Shonkoff JP, eds. Handbook of Early Childhood Intervention
. New York: Cambridge University Press; 1990:246–277.
34. Hofkosh D, Pringle JL, Wald HP, et al. Early interactions between drug-involved mothers and infants. Arch Pediatr Adolesc Med