Update on the EMI for Infants and Toddlers

Key Movement Skills

Five EMI key skills were identified and operationalized for measurement based on a conceptual review of the literature followed by validation with criterion measures of movement and motor ability (Greenwood & Carta, 2010; Greenwood et al., 2002). The EMI measures postural, locomotion, and object control skills. These skills are Transition in Position (postural movement); Grounded and Vertical (locomotion); and Throw/Roll and Catch/Trap (object control). These skills represent a continuum of movement skill acquisition expected for children 6 to 36 months of age. These key skills were further operationalized by their composite, a single indicator of rate of Total Movement (Greenwood & Carta, 2010).

Administration and Scoring

The EMI is administered during a 6 min, semi-structured, play-based session with a familiar adult. The majority of EMI assessments conducted in this study occurred in the child’s home (65%), a child care program (35%), or in other locations (0.6%, e.g., grandparents’ home, etc.). The EMI may be scored live during the administration or later by a video recording by a trained and certified coder (see below for description of certification). EMI score sheets are available from the IGDI website. Recording the occurrence of child’s movement key skills during the play session involves a simple tally of one per occurrence leading to a session total frequency count for key movement skills. These frequency data are entered into the password-protected website wherein all data for a particular program are securely managed. These data from each new session are added to each child’s cumulative record by date. Individual and group reports of children’s movement may be created in several forms as graphical displays and summary tables.

EMI Toys and Materials

The EMI consists of two equivalent toy sets (Forms A vs. B). Each set comprises a flexible, nylon pop-up, enclosure and a selection of blocks and balls (Greenwood & Carta, 2010). Each enclosure, for example, a bus or hut, is characterized by a door that children can walk or crawl through. Children may stand up in the enclosures, open and close the nylon doors, and look out of and in to the enclosure. Rubber blocks and ball in each set of various sizes, colors, and shapes can be rolled or thrown in and out of the enclosure, which encourages movement in and around the toy. All toys were selected because they evoked movement rather than sitting and exploration (Greenwood et al., 2002). The blocks and balls could be easily cleaned with antiseptic wipes and large enough to avoid becoming a choking hazard.

Testing Accommodations

The EMI administration allows for specific adaptations to ensure children are not disadvantaged due to diversity in language, vision, hearing, or physical disabilities (Americans With Disabilities Act, 1990; Walker & Buzhardt, 2010). The administration and scoring of the child’s skills requires adults who speak the child’s language(s) (including sign language or picture communication systems). Introductory experiences with the EMI format can be used with children with hearing or vision impairment using guided prompts and practice with the play partner. Toys may be placed within reach of the child. Children using postural or stability supports (braces, scooters), or assistive strategies may be included. For children who use alternative or compensatory movement replacement strategies to meet their needs, such as rolling across the floor rather than walking, interpretation of just the relevant EMI key skills scores is recommended. For example, young children with disabilities that impede or do not permit vertical locomotion, vertical locomotion should not be interpreted in lieu of focus on the other relevant key skills. Grounded locomotion may be a replacement for walking. If developing vertical locomotion skills is a targeted goal for a child (e.g., learning to use crutches, artificial limbs, etc.), then the vertical locomotion key skill becomes a valuable tool for measuring progress toward this goal.

EMI Training and Certification

Infant/toddler service providers (assessors) learned to administer and score the IGDIs through a combination of didactic training, online resources, and scoring practice. IGDI staff conducted an on-site workshop in which service providers learned to administer the EMI, score assessments, and interpret results for intervention decision making. Trainees were certified and ready to collect data after passing two benchmarks: coding of two EMI assessments in which their coding was within at least 85% agreement of a master coding, and a demonstration using the correct EMI administration steps (80% correct) during either a live or video-recorded assessment. If the trainee did not achieve these benchmarks at first attempt, they consulted their local certified EMI trainer for suggestions and corrective feedback on additional attempts until reaching the criterion.

In summary, given the need for individual measures of growth in movement that practitioners can use, we sought to provide new information on the use of the EMI to inform the practice of movement assessment in early intervention, update the original benchmark trajectories informed by a large sample of children measured by program staff, evaluate the construct validity of the EMI relative to a conceptual framework of movement skills development, and rule out potential sources of measurement bias (moderators) relative to children’s sociodemographics, culture/home language, and disability characteristics (American Psychological Association, 2013; Greenwood & McConnell, 2011). Thus, the following research questions were addressed:

Research Question 1: Was use of the EMI scalable by practitioners in early childhood programs measured in terms of the volume of individual child data collected by practitioners in the number of programs, regions, states, children, and occasions?

Participants

Children

Infants/toddlers between 6 and 36 months of age at start participated (N = 628; see Table 1). The mean age of children at the start of the project was 18 months, with a range of 6 to 36 months (SD = 9.5 months). The number of boys and girls was nearly evenly split at 50%. About 85% of children were from disadvantaged backgrounds based on the low-income eligibility requirements of EHS and included 15% of children with IFSPs (see Table 1). In total, 35 (37%) of the 95 children with IFSPs had a movement concern identified. The majority home language reported was English (59%), followed by Spanish (36%) and other languages (5%).

OPEN IN VIEWER

Table 1.

Children’s Sociodemographics.

Variable	M	SD	Minimum	Maximum	N	%
Number of children					628
Age in months	18.0	9.5	6.0	36.0
Gender, % Male					317	50.5
Gender, % Female					311	49.5
Home language, % English					373	59.4
Home language, % Spanish					223	35.5
Home language, % Other					32	5.1
IFSP, % No					533	84.9
IFSP, % Yes					60	9.5
IFSP, % Yes (with movement concerns)					35	5.6
Total					628	100.0

Note. IFSP = Individual Family Service Plan.

Measurement Procedures

Statistical Analysis

The total EMI assessments completed were screened for outliers (N = 2,282). In total, 24 (1.0%) were identified falling more than +3.0 SD above mean and removed, leaving N = 2,258 assessments included in analyses. The mean number of observations collected per child was 3.6, ranging from 1 to 15 occasions. Simple descriptive statistics, mean (M) and standard deviation (SD) were used to address the scalability research question. To address the questions about children’s movement growth trajectories, multilevel growth curve modeling (MGCM) was used (Snijders & Bosker, 2012). MGCM was appropriate because it computes an equation (model) to the raw data over time that describes its prototypical growth curve trajectory across the 6- to 36-month age span. The linear form of the statistical model requires two parameters, an intercept and slope to define a trajectory in graphical space. A curvilinear form of the model adds an acceleration parameter that defines bend in the curve. Together, these parameters define the shape of the growth curve that best fits the raw data that is either accelerating or decelerating over time. An advantage of MGCM is that the intercept value, can be determined at any one age at the discretion of the researcher. We computed the intercept at 36 months of age because it is an endpoint for children exiting IDEA Part C (Infant/Toddler) services and a transition point for the onset of Part B (Preschool) services. This value is helpful in establishing one’s level of movement proficiency relative to normative benchmarks at this age, the impact of Part C, and/or the need for Part B services.

Levels in the EMI data addressed in the models were EMI assessments nested within children within programs using child age in months as the time variable of interest. A preliminary step in the analyses involved testing for the correct shape of the trajectory (i.e., linear vs. quadratic). We used the quadratic model, because the chi-square test of nested models indicated significant improvement in the fit of the linear model (χ = 123.07, p < .01).

To address sensitivity to growth in the Total Movement trajectory (Research Question 2), we calculated benchmark growth trajectories representing the mean trajectory as well as the −1.5 and −1.0 and +1.5 SD trajectories over the age span. These benchmarks are helpful when deciding whether or not an individual child’s growth is falling within or outside of the range of expectation for their chronological age (Greenwood et al., 2010). In this growth model, we were interested in the parameters representing the means of intercept, slope and acceleration rate across all individuals, that is, γ₀₀, γ₁₀, and γ₂₀, respectively (Equation 1):

y i j = β 0 j + β 1 j Age i j + β 2 j Age i j 2 + e i j β 0 j = γ 00 + u 0 j β 1 j = γ 10 + u 1 j β 2 j = γ 20 + u 2 j

(1)

To address the Key Skills trajectories (Research Question 3), separate growth models were computed for each individual skill intended to reveal shape of each skill over time and to compare features of all in comparison to a theoretical continuum of growth in movement proficiency. For example, we expected to see grounded locomotion (crawl, a more basic, early EMI skill) peak and then plateau at approximately 12 months, in covariation with the onset and growth of vertical locomotion (a more advanced skill). We hypothesized growth in transitions in position to precede growth in ground locomotion, preceding growth in vertical locomotion, preceding growth in throws/rolls, and preceding growth in catches/traps.

To address Research Question 4, the moderating variables were added as cross-level interactions with the growth and intercept parameters in the multilevel growth models (see Equation 2):

y i j = β 0 j + β 1 j Age i j + β 2 j Age i j 2 + e i j β 0 j = γ 00 + γ 01 Moderator j + u 0 j β 1 j = γ 10 + γ 11 Moderator j + u 1 j β 2 j = γ 20 + γ 21 Moderator j + u 2 j

(2)

The key parameters of interest in this model were those associated with moderators, that is, γ₀₁, γ₁₁, and γ₂₁, which were referred to as cross-level interaction effects (see Table 4). These parameters represent the differences in intercepts, slopes, and acceleration rates, respectively, between two groups defined by the Moderator (i.e., gender, IFSP status, or home language status). Because it was possible to improve the precision of the IFSP question by separating the IFSP status of children into subgroups that were differentiated by whether or not movement domain concerns were expressed in the IFSP, an additional model was computed using: IFSP = none; IFSP = yes, but without movement concerns; and IFSP = yes, with movement concerns as fixed effects.

OPEN IN VIEWER

Table 4.

Growth Curve Modeling Results for the Early Movement Indicator.

	Model
	Gender				Language				IFSP status
	Estimate	SE	t	p	Estimate	SE	t	p	Estimate	SE	t	p
Intercept	11.297	0.835	13.524	<.01	11.018	0.719	15.323	<.01	12.312	0.678	18.172	<.01
Age (centered at 36 months)	−0.108	0.062	−1.744	.08	−0.153	0.058	−2.655	<.01	−0.127	0.048	−2.670	.010
Age2 (centered at 36 months)	−0.016	0.002	−8.232	<.01	−0.017	0.002	−9.658	<.01	−0.016	0.001	−11.203	<.01
Contrast value	0.029	0.721	0.040	.97	2.605	0.743	3.505	<.01	−1.403	0.989	−1.418	.16
Age × effect	−0.035	0.088	−0.402	.69	0.128	0.091	1.409	.16	0.032	0.126	0.255	.80
Age2 × effect	−0.001	0.003	−0.431	.67	0.004	0.003	1.276	.20	0.000	0.004	0.106	.92

Note. The contrast values were as follows: Gender = Female, where 0 = Female, and 1 = Male; Language = English, where 0 = English and 1 = Spanish; and IFSP= No IFSP where No IFSP = 0 and IFSP = 1. IFSP = Individual Family Service Plan.

Was Use of the EMI in Early Childhood Programs Scalable?

Overall, use of the EMI was scalable given the number and range of adopting early childhood programs in diverse geographical locations implemented by trained staff supported by the website and online data management system (see Table 2). In total, 15 community-based early childhood programs used the EMI and contributed child data through the efforts of 94 staff members who administered, collected, scored, and entered the EMI data. The total number of children with EMI data used in analyses was 628 with a total of 2,258 EMI completed assessments. The number of EMI assessments at each month of age (6–36 months) ranged from 41 to 86 with the overall rate of Total Movement growing from a mean of 2.8 (SD = 2.8) per min at 6 months to 14.0 (SD = 6.3) per min at 36 months (see Table 2). Children were not assessed on all occasions because they entered and left programs at different ages and for different reasons (e.g., moved in or away, aged out, etc.).

OPEN IN VIEWER

Table 2.

Number of Early Movement Indicator Assessments by Age and Total Movement M, SD.

	EMI
Months	Assessments	M	SD
6	70	2.8	2.8
7	51	2.5	2.0
8	41	4.2	3.1
9	62	5.0	3.5
10	75	6.2	4.9
11	58	7.5	4.9
12	56	7.4	4.1
13	72	8.6	4.5
14	77	9.2	4.4
15	72	9.8	5.1
16	76	10.3	5.5
17	79	10.8	4.6
18	63	12.3	5.9
19	68	11.7	6.0
20	73	11.4	5.4
21	86	12.3	5.4
22	57	11.4	5.3
23	91	12.0	4.8
24	64	11.9	4.5
25	72	12.8	6.7
26	81	12.7	4.8
27	67	12.5	5.0
28	60	12.8	6.2
29	83	13.0	4.7
30	72	12.8	5.9
31	55	13.2	6.8
32	75	13.8	5.5
33	74	14.5	6.9
34	68	14.3	5.7
35	54	12.6	5.0
36	57	14.0	6.3
37	42	13.4	4.5
38	40	14.4	4.5
39	17	14.4	9.8
40	26	14.3	5.7
41	9	10.3	3.3
42	15	11.7	5.3
Total	2,258
M	10.7
SD	5.6
Minimum	0
Maximum	33

Note. n = 628 children. Assessments after removal of 24 outliers. EMI = Early Movement Indicator.

Was the Total Movement Benchmark Trajectories Sensitive to Growth Over Time?

Similar to the original report (Greenwood et al., 2002), the new trajectories also were accelerating negatively; that is, increasing rapidly, slowing later, and reaching an inflection point (see Figure 1). The fitted mean growth model was 13.5 total movements per min (tm/min; mean intercept at 36 months of age), –0.118 (linear slope), and −0.016 (acceleration). All three values were significantly greater than zero. In this sample of N = 628 children, the mean values along the total movement trajectory were 2.7, 10.5, and 13.5 tm/min at the sixth, 18th, and 36th months of age, respectively. These same values on the original trajectory (Greenwood et al., 2002) were similar at 0.50, 7.9, and 13.8 tm/min at the sixth, 18th, and 36th months of age (N = 29 children).

OPEN IN VIEWER

Figure 1.

EMI M, and SD trajectories (intercept centered at 36 months).

Separating the overall Total Movement trajectory into SD bands for intervention decision making purposes, produced four Total Movement rate benchmark trajectories with mean intercepts of 20.2, 14.1, 8.6, and 6.5 tm/min for the +1.5 SD, M, –1.0 SD, and −1.5 SD trajectories, respectively. Comparisons between the shapes of these four benchmark trajectories key differences emerged. For children in the −1.0 SD trajectory below the mean, early acceleration was comparatively slower and trending downward sooner compared to the mean trajectory (see Figure 1). This same pattern was more extreme in the −1.5 SD trajectory. In contrast, for the children in the +1.5 SD trajectory above the mean, growth was higher and rapid at 6 months and but accelerated slowly through 36 months of age compared with the mean trajectory, where children continued to accelerate and did not reach an inflection point.

Did Patterns of Growth in EMI Key Skill Benchmark Trajectories Reflect a Continuum?

Individually, each of the Key Skill trajectories depicted differences at age of onset, increasing and some waning in rate of occurrence over time (see Table 3 and Figure 2). At 6 months of age, Transitional Movement and Grounded Locomotion were in use, with Transitional Movement at a much higher rate. By 8 months, Vertical Locomotion and Throw-Roll movements were beginning to grow. Growth curved for Transitional Movement, Vertical Locomotion, and Throw/Catch, accelerating negatively over time. Growth in Grounded Locomotion after a fast start was slow and linear over time. By 36 months, the rates of all five skills had increased substantially in frequency, with Transitional Movement, Vertical Locomotion, Throw-Roll, Grounded Locomotion, and Catch trap, in order of highest to lowest. The slowest growing key skills were Catch/Trap and Grounded Locomotion (Figure 2; Table 3).

OPEN IN VIEWER

Table 3.

EMI Key Skills Growth Models.

Model	Intercept	Age (centered at 36 months)	Age2 (centered at 36 months)
Transitional movement
Estimate	5.293	−0.024	−0.005
SE	0.194	0.025	0.001
t	27.337	−0.937	−6.600
p	<.01	.349	<.01
Grounded locomotion
Estimate	1.372	0.017	—
SE	0.140	0.005	—
t	9.818	3.249	—
p	<.01	.001	—
Vertical locomotion
Estimate	3.444	−0.72	−0.007
SE	0.379	0.021	0.001
t	6.456	−4.948	−11.679
p	<.01	<.01	<.01
Throwing/rolling
Estimate	2.130	−0.022	−0.003
SE	0.284	0.016	0.000
t	7.497	−1.395	−7.416
p	<.01	.163	<.01
Catching/trapping
Estimate	0.481	0.013	—
SE	0.119	0.005	—
t	4.024	2.647	—
p	<.01	<.01	—

Note. EMI = Early Movement Indicator.

OPEN IN VIEWER

Figure 2.

Early movement key skill elements fitted growth trajectories for all children.

Taken together, the ages at onset and shapes of growth reflected a complex picture of a continuum in that multiple skills were nearly always in the repertoire with each changing differentially adding to overall proficiency in movement. This was in contrast to a simple stair stepping continuum—wherein one skill started and ended and another skill began. Instead, Vertical Locomotion soon exceeded Grounded Locomotion over the age range, as expected and did not drop from the repertoire. In another case, Transitional Movement emerged early, grew over time, but was never exceeded by Vertical Locomotion in children this age.

Was Growth in Total Movement Moderated by Children’s Characteristics?

Results were comparisons made between the values of each moderating variable. Value 1 of the three moderators were Females, English home language, and no IFSP status and their interactions, respectively. The contrasting value of the moderators were males, IFSP status, and Spanish–Other home language. For the purpose of interpreting growth effects, the first three growth parameters (intercept, age, age²) of each moderator were for the value 1 comparisons only (see Table 4). Thus, the last three growth parameters of each moderator in Table 4 indicated the between values differences, or the influence of the moderator on explaining growth in the mean Intercept (36 months), the linear slope (Age × effect), and acceleration (Age² × effect).

Results indicated no significant moderation of children’ growth in Total Movement by Gender or IFSP status (Yes vs. No) or their interaction at α = .05. However, when the IFSP analysis included information on an expressed movement concern, the difference was significant (Estimate = −3.78, SD = 0.68, t = −5.54, p < .01). Children with IFSP movement concerns were lower in Total Movement rate at 36 months of age (M = 9.4 tm/min) versus those with an IFSP but without a movement concern (M = 13.8 tm/min). Children’s home language was a significant moderator. Children whose home language was Spanish or Other than English attained a significantly higher rate of Total Movement at 36 months of age: M = 15.4 per min compared to M = 12.4 for English only (Estimate = 2.95, SE = 0.74, t = 3.988, p <.001).

Limitations

While the new estimates of EMI Total Movement and Key Skills were improved in size, authenticity, and diversity of contributing programs, they are not yet nationally representative. Programs contributing data to this report were self-selected, interested adopters from EHS and Part C IDEA (2004) programs and not randomly selected. It was a strength that the programs (a) were geographically diverse, (b) varied in the number of children served, and (c) differed in duration of using the EMI. Inclusion of the data for the original validation (Greenwood et al., 2002) accounted for only 9% of the total number of assessments. We concluded these data were appropriate to boost sample size and also because the small numbers argued against making a statistical comparison to the original.

It was not feasible that two people were present for live agreement checking, in this enterprise, full program scenario, particularly when the EMI was most often administered in the home. Re-scorings of videos in annual calibration agreement checks has been much more feasible. However, the impact on reliability and drift in reliability/scoring accuracy during the interim has yet to be addressed. We were pleased to see that a simple comparison between the original versus new sample findings indicated that Total Movement scores were highly similar at 6, 18, and 36 months of age points, additional evidence that scoring had not drifted. Also, shapes of the Total Movement between original and new sample growth curves were similar, together suggesting a degree of score accuracy/agreement.

Implications for Future Research

A need in future research is an effort to develop more representative EMI growth benchmarks. Traditionally, the development of universal measures for screening and program monitoring has been to grow incrementally, beginning with local program data collection, moving to accumulated wide-scale program data collected via centralized data infrastructures supported by websites (e.g., Kaminski & Good, 1996). This is in contrast to collections guided by a nationally representative sampling plan as is typically used in standardized test norming. A part of this work should be a comparison of live versus video scoring.

The finding of accelerated EMI growth by children experiencing Spanish and Other home languages was intriguing and our limited explanation of this difference, suggested that research is needed. Such differences are consistent with racial/ethnic and environmental factors (cultural, parenting practices, socioeconomic) that do explain differences in language, cognitive, and academic development. Such research may also explore associations between children’s movement development and codevelopment in other domains. The present findings and the advent of the EMI and other IGDIs for infants and toddlers, such as the ECI, provide a stimulus and a method for doing so. IGDI measures combined with latent MGCM techniques research on dual and multiple developmental processes feasible.

Another need is to demonstrate that growth in the EMI is facilitated by interventions and use of the EMI for data-driven decision making, as has been the case with the ECI and use of language promoting strategies in controlled trials (Buzhardt et al., in press). Such studies could profitably be conducted in collaboration with OT/PT and Early Childhood Special Education (ESCE) professionals as part of implementing and monitoring individual children’s progress given goals and targets within a child’s IFSP. Experimental studies of groups of children receiving alternative movement interventions are needed to add to our knowledge of what works as measured by the EMI.

The EMI’s key skills appeared to capture growth well up until 30 months of age, at which time acceleration slowed and a plateau reached. A similar reduction in older children with DS has been reported (Palisano et al., 2001). Research is needed to better understand whether or not this ceiling is a true development outcome or a weakness in the EMI protocol. For example, the EMI’s Catching/Trapping movement skills begins early but remains slow growing over time to 36 months. Additional research is needed to improve the EMI’s sensitivity to growth.

Implications for Practice

The EMI represents a practical, new support for early educators and interventionists interested in monitoring the development of infant and toddlers’ emerging movement skills. It also represents an advance in evidence-based practice, given that local data collected in the course of typical duties with the children are being used for decision making, as opposed to relying only on data collected by researchers or outside assessors. The EMI can be used as part of a multi-tier support system (MTSS) framework to identify children who need more support and monitor progress given interventions as part of the IFSP.

Use of the web-based EMI by both local and distant early childhood programs indicated that new options exist for fulfilling programs’ accountability needs for tools and strategies that promote movement outcomes. The improved EMI benchmarks can be used for individual decision making that may include: referral for additional assessment by the PT/OT, collaboration between the PT/OT with parents and early educators in implementing an intervention, or testing new strategies designed to promote the development of movement skills. In conclusion, given the lack of practitioner/program-level movement measures, the need for such measures, and Internet access and scalability, the EMI appears to be a viable alternative.