Descriptive Statistics

The mean (M) is a measure of central tendency and is the average score in a distribution. Because it can be affected by variations caused by extreme scores, the mean can be misleading in scores obtained from a highly variable sample. In Figure 7A-1, the mean score is 100.

The mode, also a measure of central tendency, is the most frequent or common score in a distribution.

The median is defined as the middle score that divides a distribution in half when all the scores have been arranged in order of increasing magnitude. It is the point above and below which 50% of the scores fall. This measure is not affected by extreme scores and therefore is useful in a highly variable sample. In the case of an even number of data points in a distribution, the median is considered to be halfway between two middle scores. Noteworthy is the fact that in the normal distribution depicted in Figure 7A-1, the mean, mode, and median are equal (all scores = 100), and the distribution is unimodal.

The range is a measure of dispersion that reflects the difference between the lowest and highest scores in a distribution (highest score − the lowest score +1). However, the range does not provide information about data found between two extreme values in the test distribution, and it can be misleading when the clinician is dealing with skewed data. In this situation, the interquartile range may be more useful: The distribution of scores is divided into four equal parts, and the difference between the score that marks the 75th percentile (third quartile) and the score that marks the 25th percentile (first quartile) is the interquartile range.¹⁶

The standard deviation (SD) is a measure of variability that indicates the extent to which scores deviate from the mean. The standard deviation is the average of individual deviations from the mean in a specified distribution of test scores. The greater the standard deviation, the more variability is found in test scores. In Figure 7A-1, SD = 15 (the typical standard deviation in normreferenced tests). In a normal distribution, the scores of 68% of the children taking a test will fall between +1 and −1 standard deviation (square root of the variance). In general, most intelligence and developmental tests that employ deviation quotients have a mean of 100 and a standard deviation of 15. Scaled scores, such as those found in the Wechsler tests, have a mean of 10 and a standard deviation of 3 (7 to 13 being the average range). If a child’s score falls less than 2 standard deviations below average on an intelligence test (i.e., IQ < 70), he or she may be considered to have a cognitive-adaptive disability (if adaptive behaviors are also impaired).

Skewness refers to test scores that are not normally distributed. If, for example, an IQ test is administered to an indigent population, the likelihood that more children will score below average is increased. This is a positively skewed distribution (the tail of the distribution approaches high or positive scores, i.e. the right portion of the x-axis). Here, the mode is a lower score than the median, which, in turn is lower than the mean. Probabilities based on a normal distribution will yield an underestimate of the scores at the lower end and an overestimate of the scores at the higher end. Conversely, if the test is administered to children of high socioeconomic status, the distribution might be negatively skewed, which means that most children will do well (the tail of the distribution trails toward lower scores or the left portion of the x-axis). In negatively skewed distributions, the value of the median < mean < mode scores at the lower end will be overestimated, and those at the upper end will be underestimated. Skewness has significant ramifications in interpretation of test scores. In fact, the meaning of a score in a distribution depends on the mean, standard deviation, and the shape of the distribution.

Kurtosis reflects the shape of the distribution in terms of height or flatness. A flat distribution, in which more scores are found at the ends of the distribution and fewer in the middle, is platykurtic, in comparison with the normal distribution. Conversely, if the peak is higher than the normal distribution, scores do not spread out and instead are compressed and cluster around the mean. This is called a leptokurtic distribution.

Transformations of Raw Scores

Sensitivity and Specificity

Frequently, interpretation of test results must take into account how well the instrument performs with set cutoff scores. Sensitivity is a measure of the proportion of children with a specific problem who are positively identified by a test, with a specific cutoff score. Children who have a disorder but are not identified by the test are considered to have false-negative scores. In developmental/behavioral pediatrics, the “gold standard” (criterion used to determine the presence of a given problem) often is not definitive but rather is a reference standard. Comparison with an imperfect “gold standard” may lead to erroneous conclusions that a screening test is inaccurate. As a result, sensitivity may be better conceptualized as copositivity. Desired sensitivity rates are 70% to 80%, and sensitivity is the true positive rate of a test.

Specificity is a measure of the proportion of children who actually are normal and who also are correctly determined by a given test to not have a problem. Children who are normal but who are incorrectly determined by a test cutoff score to be delayed or learning disabled are considered to have false-positive scores. Specificity is the true negative rate of a test. Again, in cases such as developmental screening, the presence of a reference (and not “gold”) standard makes the term conegativity more appropriate. A specificity rate of 70% to 80% is desirable. However, in the case of screening, it is better to have a higher sensitivity rate, perhaps at the cost of lowered specificity, so as to enhance identification of infants and children who might be at risk.

Cutoff scores can be adjusted to enhance sensitivity. By making criteria more inclusive, fewer children with true abnormalities will be missed; however, a more restrictive cutoff will also increase the probability of false-positive findings (overidentifying “normal” children as being abnormal). Conversely, if the cutoff score is made more exclusive to enhance specificity, the number of normal children inaccurately identified as abnormal is decreased, but some of those who are truly abnormal will be erroneously called normal (false-negative findings). Sensitivity and specificity are described in Figure 7A-2.

FIGURE 7A-2 Example highlighting sensitivity and specificity.

Positive predictive value refers to the proportion of children with a positive test result who actually are delayed or learning disabled. This reflects the probability of having a problem when the test result is positive. The lower the prevalence of a disorder, the lower is the positive predictive value. Sensitivity may be a better measure in low-prevalence problems. In developmental screening, positive predictive values often are in the range of 30% to 50%.

Negative predictive value refers to the proportion of children with a negative test result who indeed do not have developmental delays or learning problems. It is the probability of not having the disorder when the test result is negative. This value is influenced by the frequency or prevalence of a problem; in low-prevalence problems, specificity may be a better measure.

Frequency of a Disorder/Problem

Prevalence rate refers to the number of children in the population with a disorder, in relation to the total number of children in the population, measured at a given time. The incidence rate indicates the risk of developing a disorder: namely, new cases of a problem that develop over a period of time. The relationship between incidence and prevalence can best be illustrated by the following: prevalence rate = the incidence rate × the duration of the disorder. In essence, the predictive value of screening takes into account sensitivity and specificity of the screening procedure and the prevalence of the disorder.

Base rate is the naturally occurring rate of a given disorder. For example, the base rate of learning disabilities would be much higher in children referred to a learning and attention disorders clinic than in the general population. If a screening instrument were used to detect learning disabilities for this group, sensitivity and specificity values would differ from those found in the general pediatric population. For example, in the follow-up of low-birth-weight infants, the base rate for major handicaps (moderate to severe mental retardation; cerebral palsy; epilepsy; deafness or blindness) is 15%; therefore, in 85% of this population, the findings would be true negative. Low base rates increase the possibility of false-positive results. High base rates do not leave much room for improvement in terms of locating true-positive scores and result in an increase in false-negative findings. Tests can be most helpful in decision making when the base rate is in the vicinity of 0.50. Therefore, particularly in the case of screening, the relatively low base rates of developmental problems in very young children may increase the probability of false positive findings. However, in such situations, this scenario is more desirable than the converse: false negative findings.

Relative risk provides an alternative strategy for evaluating test accuracy.^17,18 This approach involves use of the likelihood ratio, which indicates the increased probability that the child will display a developmental problem, if the results of an earlier screening test were abnormal or suspect. This approach recognizes that not all children at early risk will later manifest a developmental problem, but there is a greater likelihood that they will. If a problem or disorder is rare, relative risk and odds ratios are nearly equal.

Test Characteristics

Age and Grade Equivalents

Age- and grade-equivalent scores are based on raw scores and portray the average age or grade placement of children who obtained a particular raw score. Although these metrics are useful in explaining results to parents and make conceptual sense, age and grade-equivalent scores are uneven units of measurement. For example, a six-month difference in performance at the age of 2 years is much more significant than a 6-month lag at age 8 years. Moreover, a 9-year-old with an age equivalent of 7 years is quite different from a 4-year-old functioning at a 7-year age equivalent, or an average 7-year-old. These measures lack precision, and in some test manuals, the same standard scores can produce somewhat different age/grade equivalents. Both metrics assume that growth is consistent throughout the school year and tend to exaggerate small differences in performance. These measures also vary from test to test. Furthermore, with achievement testing, it is necessary to know whether age or grade norms were used to obtain standard scores. For example, if age norms are used and the child had been retained in grade, he or she would be at a significant disadvantage because he or she would not have been exposed to the more advanced material. Conversely, if a child failed second grade and is being tested in early fall while repeating second grade, he or she may receive inflated scores if grade norms are used.

The IQ/DQ ratio (developmental quotient) is computed as mental age (obtained by the use of a test score) ÷ the child’s chronologic age and then multiplied by 100. Although developmental age refers to a level of functioning, DQ reflects the rate of development.¹⁹ IQ/DQ ratio scores are not comparable at different age levels because the standard deviation (variance) of the ratio does not remain constant. As a result, interpretation is difficult, and these scores generally are not used very much in contemporary standardized testing. Instead, the deviation IQ/DQ is employed. The deviation IQ is a method of estimation that allows comparability of scores across ages and is used with most major psychological and developmental test instruments. The deviation IQ/DQ is norm referenced and normally distributed, with the same standard deviation; typically, M = 100 and SD = 15. Therefore, a deviation IQ of 85 obtained at age 6 should have the same meaning as a score of 85 obtained at age 9.

The standard error of measurement (SEM) is an estimate of the error factor in a test that is the result of sampling or test characteristics, taking into account the mean, standard deviation, and size of the sample. The larger the standard error of measurement, the greater the uncertainty associated with a given child’s score. The SEM is produced by multiplying the standard deviation of the test by the square root of (1− the reliability coefficient of the test). In 95% of cases, the interval of approximately two times (1.96) the SEM above or below a child’s score would contain the “true” score: a 95% confidence interval. Stated differently, a 95% confidence interval indicates that if a test is given 100 times with different samples, scores will fall in this interval 95% of the time. In a 90% confidence interval, an interval of 1.64× the SEM above and below a child’s score would contain the “true” score. Such estimates are important in test-retest situations or in the case of a child who does not receive services because of missing a cutoff score by only a few points (e.g., a WISC-IV Full Scale IQ score of 72).

A final concern is the Flynn effect,²⁰ in which test norms increase approximately 0.3 to 0.5 points per year, which is equivalent to a 3- to 5-point increment per decade. This finding has ramifications in comparisons of scores obtained on earlier versions of tests to more contemporary scores (e.g., WISC-Revised to the WISC—Third Edition or WISC-IV; BSID to BSID-II; Stanford-Binet form LM to the 5^th edition). Caution is warranted when the practitioner attributes a decline in scores to a loss of cognitive ability, because in actuality this decline may be attributable to the fact that a newer test has mean scores that are considerably lower than those of an earlier version of the test (e.g., 5-8 points).²⁰ This issue would also have ramifications for children whose IQ score on an older version of a test is in the low 70s but decreases to below the cutoff for mild mental retardation on a newer version.

Although some practitioners may administer tests, all have occasion to respond to inquiries from parents about their child’s test performance or diagnosis derived from testing. The physician’s role includes explaining test results to parents, acknowledging parental concerns and advocating for the child, providing additional evaluation, or referring to other professionals.²¹

The Assuring Better Child Health and Development (ABCD) Program

Created by The Commonwealth Fund, the ABCD Program has identified policy strategies for state Medicaid agencies to strengthen the delivery and financing of early childhood services for low-income families. The emphasis is on assisting participating states in developing care models that promote healthy development, including the mental development of young children. Models include developmental screening, referral, service coordination, and educational materials and resources for families and clinical providers. The program has resulted in improvements in screening, surveillance, and assessment. Most notably, work in North Carolina facilitated a 75% increase in screening, increased enrollment rates in early intervention from 2.6% to 8% (in line with the Centers for Disease Control and Prevention’s prevalence projections), while simultaneously lowering referral age²⁶ (http://www.nashp.org; http://www.cdc.gov/ncbdd/child/interventions.htm).

Help Me Grow

A program of the Connecticut Children’s Trust Fund, Help Me Grow links children and families to community programs and services by using a comprehensive statewide network. Components of the program include the training of child health providers in effective developmental surveillance; the creation of a triage, referral, and case management system that facilitates access for children and families to services through Child Development Infoline; the development and maintenance of a computerized inventory of regional services that address developmental and behavioral needs of children and their families; and data gathering to systematically document capacity issues and gaps in services. The program has increased identification rates of at-risk children by child health providers and increased referral rates of such children to programs and services. For example, chart reviews conducted in participating practices noted an increase in documented developmental or behavioral concerns from 9% before training to 18% after training. Furthermore, training resulted in significant differences in referral rates for certain conditions. Behavioral conditions were involved in 4% of referrals from trained practices, in comparison with 1% from untrained practices. Four percent of referrals from trained practices were for parental support and guidance, in comparison with fewer than 1% from untrained practices²⁹ (http://www.infoline.org/Programs/helpmegrow.asp).

Promoting Resources in Developmental Education (PRIDE)

This program is a 3-year project funded by the Duke Endowment through a partnership of the Children’s Hospital, the Center for Developmental Services (a colocation of agencies serving children with developmental disorders), the regional office of the state’s early intervention system (BabyNet), the local school district’s Childfind and Parents-as-Teachers programs, and a parent-to-parent mentoring program for parents of children with special health care needs. The goal of PRIDE is earlier identification and intervention for children in Greenville County, South Carolina with developmental delays and improved support for their parents.

The program has targeted key players in the lives of infants and toddlers as follows: Parents sign up around the time of their child’s birth to receive milestone cards every 3 to 6 months during the first 3 years that describe the key developmental attainments, activities to promote development at that age, and red flags for potential developmental problems. Parents are instructed to discuss any concerns with their physician. Primary care physicians are provided with information and tools (the Parents Evaluation of Developmental Status questionnaire) to improve their system of developmental screening. A nurse practitioner employed by PRIDE as the “physician office liaison” works closely with practices, initially by setting up lunch meetings with physicians and staff that are also attended by the PRIDE developmental-behavioral pediatrician. With the agreement of the physicians, the liaison then assists the office staff in implementing the system and provides a “Resource Guide” with information on local developmental services and forms to facilitate referrals. Child care providers have the opportunity to attend educational sessions (for credit hours) in which they learn about child development, signs of developmental problems, and services that are available for these children. The training sessions are provided in collaboration with local programs that promote higher quality child care and early education (Success By 6 and First Steps), and the attendees receive “toolkits” with information on the topics discussed. Initial results of the program indicate success; 16 of 17 local pediatric practices (which previously had no standardized system of developmental screening) now utilize the Parents Evaluation of Developmental Status questionnaire. Over the first 18 months of the program, referrals to early intervention have increased almost 100% and referrals to the school’s Childfind program by 30%. Other service providers have seen increases in new referrals of up to 30%. The average age at referral to early intervention has also dropped slightly.

Not surprisingly, increasing rates of referral raised the likelihood of even longer waiting lists for tertiary-level developmental-behavioral pediatric evaluations. To address this challenge, the PRIDE staff sought funding from The Commonwealth Fund to study the feasibility and cost effectiveness of a model of “midlevel” developmental-behavioral pediatrics assessment (as a step between telephone triage/record review and comprehensive diagnostic evaluation) for children younger than 6 years.³⁰

First Signs

This national and international training effort is devoted to early detection of children with disabilities, with a particular focus on autism spectrum disorders. This detection is accomplished through a mix of print materials and broadcast press, direct mail, public service announcements, presentations (to medical and nonmedical professionals), a richly informative website (www.firstsigns.org), and detailed program evaluation. Although First Signs initiatives have been conducted in several states, including New Jersey, Alabama, Delaware, and Pennsylvania, the Minnesota campaign is highlighted here because of that state’s assistance in program evaluation. Minnesota is divided into discrete service regions. Centralized train-the-trainers forums were conducted to prepare 130 professionals as outreach trainers. These individuals were from all regions of the state, and most were early interventionists, family therapists, and other nonmedical service providers. They then provided more than 165 workshops to 686 medical providers, to whom they offered individualized training tailored for health care clinics, as well as training for more than 3000 early childhood specialists. First Signs Screening Kits (which include video, information about and in some cases copies of appropriate screening tools, wall charts and parent handouts on warning signs) were distributed to more than 900 practitioners and clinics. In addition, public service announcements were aired across the state in collaboration with the Autism Society of Minnesota. Within 12 months, there was a 75% increase in the number of young children identified in the 0- to 2-year age group and an overall increase of 23% in detection of autism spectrum disorders among all children aged 0 to 21 years in that same period. The state has now expanded the initiative to include childcare providers and is educating them about red flags and warning signs. In addition, physicians with the Minnesota Chapter of the American Academy of Pediatrics Committee for Children with Disabilities have begun incorporating First Signs information into physician training program at the University of Minnesota.²⁷

Blue Cross/Blue Shield of Tennessee

Blue Cross/Blue Shield of Tennessee requested that child health providers use standardized, validated screening at all EPSDT visits. To facilitate compliance, Blue Cross/Blue Shield of Tennessee piloted a program in 34 high-volume, Medicaid-managed care practices. Outreach nurses, called regional clinical network analysts, trained providers on site how to administer, score, interpret, and submit reimbursement for the Parents’ Evaluation of Developmental Status questionnaire (the standardized developmental-behavioral surveillance and screening instrument that elicits parents’ concerns about their children). After training, screening rates increased from 0% to 43.5% during the pilot phase. At the same time, the practices experienced a 16% increase in attendance at scheduled well-child visits, which suggests that focusing on parents’ concerns may increase their adherence to visit schedules. Blue Cross/Blue Shield of Tennessee, together with the Tennessee Chapter of the American Academy of Pediatrics, is now providing training across the state.²⁵ More information can be found through the Center for Health Care Strategies, “Best Clinical and Administrative Practices for Statewide” developmental and behavioral screening initiatives as established by the Center for Health Care Strategies [http://www.chcs.org/]

Healthy Steps for Young Children

This a national initiative improves traditional pediatric care with the assistance of an in-office child development specialist, whose duties include expanded discussions of preventive issues during well-child and home visits, staffing a telephone information line, disseminating patient education materials, and networking with community resources and parent support groups. Now in its 12th year, Healthy Steps followed its original cohort of 3737 intervention and comparison families from 15 pediatric practices in varied settings. In comparison with controls, Healthy Steps families received significantly more preventive and developmental services, were less likely to be dissatisfied with their pediatric primary care, and had improved parenting skills in many areas, including adherence to health visits, nutritional practices, developmental stimulation, appropriate disciplinary techniques, and correct sleeping position. In practices serving families with incomes below $20,000, use of telephone information lines increased from 37% before the intervention to 87% after; office visits with someone who teaches parents about child development increased from 39% to 88%; and home visits increased from 30% to 92%. Low-income families receiving Healthy Steps services were as likely as high-income parents to adhere to age-appropriate well-child visits at 1, 2, 4, 12, 18, and 24 months.^31,32 One program evaluation suggests that Healthy Steps offers a benefit comparable with that of Head Start at about one-tenth the cost,³³ although this claim is somewhat premature because Head Start data now extend to more than 35 years of follow-up research with a proven return rate of $17.00 for each $1.00 spent on early intervention, with savings realized through reductions in teen pregnancy, increases in high school graduation and employment rates, and decreased adjudication and violent crime.⁷ Nevertheless, Healthy Steps is extremely promising and inexpensive and includes a strong evaluation component that will answer questions about its long-term effect.

Structured and Semistructured Diagnostic Interviews

Assessment data obtained from unstructured clinical interviews tend to vary considerably and are largely interviewer dependent. As a result, unstructured interviews have particularly poor reliability and validity. When the primary assessment goal is to provide a diagnosis or a specific judgment with high interassessor reliability, as would be desired in research studies on specific psychiatric diagnoses, standardized, structured psychiatric interviews are often preferable. Structured interviews contain specific, predetermined questions with a format designed to elicit information efficiently and thoroughly. Key questions are followed by specified branch questions with restricted, closed (“yes”/“no”) or brief responses.

An example of a structured interview is the National Institute of Mental Health Diagnostic Interview for Children—IV.⁸ This instrument is a highly structured interview with nearly 3000 questions designed to assess more than 30, psychiatric disorders and symptoms listed in the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV)^8a in children and adolescents aged 9 to 17 years. Parent and child versions in English and Spanish are available, and lay interviewers can administer it for epidemiological research. The Diagnostic Interview for Children and Adolescents⁹ is another structured diagnostic interview for children ages 6 to 17. This instrument consists of nearly 1600 questions that address 28 DSM-IV diagnoses relevant to children. Interrater reliability estimates of individual diagnoses range from poor to good, and diagnoses are moderately correlated with clinicians’ diagnoses and self-rated measures.

Structured interviews result in higher interrater (or interobserver) reliability because there is little opportunity for the interviewer to influence the content of data collected. Although sometimes considered to be the “gold standard” for psychiatric diagnostic and epidemiological research, standardized interviews are not impervious to reporter bias. In addition, structured diagnostic interviews tend to rely on DSM-IV symptoms which may not be developmentally appropriate, particularly for very young children. Moreover, structured diagnostic interviews may take 1 to 3 hours to complete, which renders them impractical for most clinical settings, especially because they typically do not assess background and family factors that are necessary for developing and implementing an intervention plan.

Semistructured interviews combine aspects of traditional and behavioral interviewing techniques. Specific topic areas and questions are presented, but, in contrast to structured interviews, more detailed responses are encouraged. Semistructured formats also support use of empathic communication described previously (e.g., reflecting, paraphrasing). For example, the Semistructured Parent Interview³ contains sample questions organized around six topic areas: concerns about the child (open ended), behavioral or emotional problems (eliciting elaboration to begin a functional analysis of behavior), social functioning, school functioning, medical and developmental history, and family relations and home situations. Like other semistructured formats, the Semistructured Parent Interview encourages parent interviews built around a series of open-ended questions to introduce a topic, followed by more focused questions about specific areas of concern.

The Semistructured Clinical Interview for Children and Adolescents (SCICA)¹⁰ is an interview designed for children aged 6 to 16. It is part of the Achenbach System of Empirically Based Assessment (ASEBA)¹¹ and was designed to be used separately or in conjunction with other ASEBA instruments (e.g., Child Behavior Checklist [CBCL], Teacher Report Form). The SCICA contains a protocol of questions and procedures assessing children’s functioning across six broad areas: (1) activities, school, and job; (2) friends; (3) family relations; (4) fantasies; (5) self-perception and feelings; and (6) problems with parent/teacher. There are additional optional sections pertaining to achievement tests, screening for motor problem, and adolescent topics (e.g., somatic complaints, alcohol and drug abuse, trouble with the law). Interview information (observations and self-report) are scored on standardized rating forms and aggregated into quantitative syndrome scales and DSM-IV—oriented scales. Test-retest, interrater, and internal consistency evaluations indicate excellent to moderate estimates of reliability. Accumulating evidence for validity of the SCICA includes content validity, as well as criterion-related validity (ability to differentiate matched samples of referred and nonreferred children).

The Child and Adolescent Psychiatric Assessment¹² is another semistructured diagnostic interview for children and adolescents aged 9 to 17. One interesting feature of this instrument is the inclusion of sections assessing functional impairment in a number of areas (e.g., family, peers, school, and leisure activities), family factors, and life events.

Motivational Interviewing

Motivational interviewing is an empirically supported interviewing approach gaining considerable attention in medical and mental health settings. More than an assessment strategy, motivational interviewing is a brief, client-centered directive intervention designed to enhance intrinsic motivation for behavior change through the exploration and reduction of patient ambivalence.¹³ Based on a number of social and behavioral principles, including decisional balance, self-perception theory, and the transtheoretical model of change,¹⁴ motivational interviewing combines rogerian and strategic techniques into a directive and yet patient-centered and collaborative encounter. Assessment from a motivational interviewing perspective involves addressing the patient’s ambivalence about making a change in behavior, exploring the negative and positive aspects of this choice, and discussing the relationship between the proposed behavior change (e.g., compliance with mediations) and personal values (e.g., health). This information is elicited in an empathic, accepting, and nonjudgmental manner and is used by the patient to select goals and create a collaborative plan for change with the provider.

The effectiveness of motivational interviewing with children and young adolescents has not been established. However, there is emerging evidence of its utility with adolescents and young adults, particularly in the areas of risk behavior, program retention, and substance abuse.^15,16

Infancy and Early Childhood

Since the 1980s, there has been increased interest in the developmental evaluation of infants and young children.^17,18 This began with the 1986 Education of the Handicapped Act Amendments (Public Law 99-457) and continues with the Individuals with Disabilities Education Improvement Act of 2004 (Public Law 108-446), a revision of the Individuals with Disabilities Education Act (IDEA). These laws involve provision of early intervention services and early childhood education programs for children from birth through 5 years of age. Developmental evaluation is necessary to determine whether children qualify for such intervention services. Part C of the IDEA revision (Section 632) delineates five major areas of development: cognitive, communication, physical, social-emotional, and adaptive. However, definitions of delay vary, criteria being set on a state by state basis. These can included a 25% delay in functioning in comparison with same-aged peers, 1.5 to 2.0 standard deviations below average in one or more areas of development, or performance on a level that is a specific number of months below a given child’s chronological age. However, pressure to quantify development has caused professionals working with infants and young children to attribute a degree of preciseness to developmental screening and assessment that is neither realistic nor attainable. Additional problems include test administration by examiners who are not adequately trained and use of instruments that have varying degrees of psychometric rigor.¹⁹ Nonetheless, developmental evaluation is critical, because timely identification of children with developmental problems affords the opportunity for early intervention, which enhances skill acquisition or prevents additional deterioration.

Again, choice of the type of developmental assessment that is administered is driven by the purposes of the evaluation: for example, determination of eligibility for early intervention or early childhood education services, documentation of developmental change after provision of intervention, evaluation of children who are at risk for developmental problems because of established biomedical or environmental issues, documentation of recovery of function, or prediction of later outcome. Assessment of infants and young children is in many ways unique, because it occurs against a backdrop of qualitative and quantitative developmental, behavioral, and structural changes, the velocity of change being greater during infancy and early childhood than at any other time. The rapidly expanding behavioral repertoire of the infant and young child and the corresponding divergence of cognitive, motor, and neurological functions pose distinct evaluation challenges.^18,19

Another significant testing concern in this age range is test refusal.²⁰ Test refusal, where a child either declines to respond to any items, or eventually stops responding when items become increasingly difficult, occurs in 15% to 18% of preschoolers.^21–24 Occasional refusals occur in 41% of young children. In addition to the immediate ramifications problematic test-taking behaviors have on actual test scores, there is evidence that early high rates of refusals are associated with similar behaviors at later ages, and with lower intelligence, visual perceptual, neuropsychological, or behavioral scores in middle childhood.^22–25 Non-compliance has been reported to occur in verbal production tasks, gross motor activities, or toward the end of the testing session, and it occurs more in children born at biologic risk or those from lower socioeconomic households. Children who refuse any aspect of testing differ from those who refuse some items, or who are compliant and cooperative to a certain point and then refuse more difficult items. This situation prompted inclusion of the Test Observation Checklist (TOC) in the Stanford-Binet Scales for Early Childhood, 5th Edition (SB5).²⁶

A distinction is often made between developmental tests and intelligence tests,²⁷ and both are used in the age range under discussion. The assessment of intelligence originated from the need to determine which children would be able to learn in a classroom and which would be mentally deficient. In fact, this was the original purpose of the Binet test. Intelligence tests have become more psychometrically sophisticated but still assess different facets of primary cognitive abilities such as reasoning, knowledge, quantitative reasoning, visual-spatial processing, and working memory. In contrast, the purpose of early developmental measures such as the Bayley Scales of Infant Development (BSID)²⁸ or the Gesell Developmental Schedules²⁹ was to be diagnostic of developmental delays, providing a benchmark of developmental acquisitions (or lack thereof) in comparison to same-aged peers. Nonetheless, this distinction is often blurred, perhaps because there is no specific age at which a child shifts from “development” to “intelligence” (although the culmination of the infancy period is often indicated), nor is there a clear-cut transformation from a delay to a deficit. Developmental tests also tend to include motor and social-adaptive skills. Both tests of development and intelligence are driven by the theoretical model of the test developer and the constructs measured by the test. Those that assess the former are considered more dynamic or fluid; those that assess intelligence are more consistent and predictive. Herein, we discuss both developmental and intelligence tests that are used in children in this age level.

Developmental Assessment Instruments

Intelligence Assessment Instruments

Achievement Testing

Use of individually-administered achievement tests has increased dramatically since the introduction of Public Law 94-142 (Education of All Handicapped Children Act), and these tests continue to be a critical component in the evaluation of children with academic difficulties under the IDEA revision of 2004. The major reason is that achievement tests enable the delineation of aptitude-achievement discrepancies, a hotly debated requirement for establishment of a learning disability (versus response to treatment intervention). It is assumed that such tests identify children who need special instructional assistance; help recognize the nature of a child’s difficulties/deficiencies, thereby clarifying the nature of the learning problem; and assist in planning, instruction, and intervention. Unfortunately, achievement tests do not adequately meet these needs. In general, standard scores (with percentiles) are the most precise metric; age and grade-equivalent scores are least useful. With regard to the Wechsler tests, the Verbal IQ (or Verbal Comprehension Index) and FSIQ are most highly correlated with achievement, particularly reading; the Performance IQ (Perceptual Reasoning Index), with mathematics.⁵¹ Achievement tests differ in terms of content and type of response required (e.g., multiple choice vs. recall of information), and these differences sometimes cause one test to produce lower scores than another.

Neuropsychological Testing

There are three approaches to neuropsychological testing of children, and all involve the assessment of brain-behavior relationships. The first approach entails modification of traditional neuropsychological batteries such as the Halstead-Reitan Neuropsychological Battery or the Luria-Nebraska Neuropsychological Battery, to form corresponding children’s batteries.⁵⁹ The second approach involves interpretation of standard tests such as those measuring intelligence, with the use of a neuropsychological “mind-set.” In this case, results from standardized tests are tied into neuropsychological constructs and functions (e.g., the Kaufman Assessment Battery for Children—Second Edition). The third approach includes tests or rating scales designed to assess specific areas of neuropsychological function. Neuropsychological testing generally is more specific in terms of pinpointing strengths and deficits, and the results more precisely describe brain-behavior relationships. Neuropsychological testing may elucidate more subtle problems that contribute to cognitive, academic, or social difficulties; these problems may not be apparent from results of more routine measures used to detect learning disabilities. Noteworthy is the fact that standard intellectual assessment is typically part of a neuropsychological workup. Selected tests from this third approach are discussed as follows.

Behavioral Rating Scales

Behavior rating scales are an extremely useful and efficient method for obtaining data on child functioning. Most rating scales use a standard questionnaire, checklist, or Likert-response format for surveying areas of interest and usually are completed by caregivers without much assistance. Rating scales include brief screening measures that assess global, broad-based measures, and problem-specific scales.

Broad-based behavioral assessment instruments assess multiple dimensions of behavior in children. Most are empirically developed taxonomies that are symptom driven and do not necessarily correspond to specific diagnostic schemas. On rating scales, informants rate the child on a broad range of social competencies and problematic behaviors. Results produce empirically derived factor scores on broad dimensions (e.g., internalizing and externalizing problems) and specific symptom areas (e.g., depression or aggressiveness) based on age and gender norms. Parent, teacher, and self-report forms are available for crossinformant comparisons. Rating scales yield very useful information about a child’s functioning in comparison with children of the same age and gender, and generally are viewed as necessary components of most child assessments.

ACHENBACH SYSTEM OF EMPIRICALLY BASED ASSESSMENT/CHILD BEHAVIOR CHECKLIST^11,67–71

The CBCL was one of the first broad-based rating scales of behavior in children to be developed, and it continues to be the most widely used method for behavioral assessments in children. Achenbach began work on what would become the CBCL in the 1960s in an effort to differentiate child and adolescent psychopathology.⁶⁸ At that time, the DSM provided just two categories for childhood disorders: Adjustment Reaction of Childhood and Schizophrenic Reaction, Childhood Type. Achenbach and collaborators applied an empirically based approach to child psychopathology much like what was used in the development of the Minnesota Multiphasic Personality Inventory. This approach involved recording problems for large samples of children and adolescents, performing multivariate statistical analyses to identify syndromes of problems that co-occur, using reports to assess competencies and adaptive functioning, and constructing age and gender-specific profiles of scales on which to display individuals’ scores.¹¹ These taxonomic procedures revealed that most behavior problems in children could be broadly divided into “internalizing” and “externalizing” conditions. This pioneering work had enormous influence on clinical and research assessment practices and established the empirical foundation for contemporary conceptualizations of child psychopathology.

The CBCL was published first in 1983 as a measure of behavior problems in children aged 4 to 18 years. Currently, there are ASEBA materials for ages 1½ to older than 90 years. There are forms for preschoolers (1½ to 5 years, parent and teacher/daycare versions)⁶⁹ and school-aged children (parent, teacher versions for children aged 6 to 18 years and youth self-report for ages 11 to 18 years),⁶⁷ as well as for adults (18 to 59 years)⁷⁰ and older adults (60 to older than 90 years)⁷¹ (both with caregiver and self-report formats). For each problem listed, informants provide ratings on the following scale: 0 = “not true,” 1 = “somewhat or sometimes true,” and 2 = “very true or often true.” Hand-scored and computer-scored profiles are available, as are Spanish-language forms.

The Child Behavior Checklist for Ages 1½-5 (CBCL/1½-5) obtains parents’ ratings of 99 problem items along with descriptions of concerns and competencies. Scales are based on parent ratings of 1728 preschool children; norms are based on a national sample of 700 children. Raw scores can be translated into standard T-scores, yielding interpretative information on three summary scales (Internalizing, Externalizing, and Total Problems), as well as on clinical syndromes scales (Emotionally Reactive, Anxious/Depressed, Somatic Complaints, Withdrawn, Attention Problems & Aggressive Behavior, and Sleep Problems). A Language Development Survey is included to screen for language delays. DSM-oriented scales pertaining to affective problems, anxiety problems, pervasive developmental problems, attention-deficit/hyperactivity problems, and oppositional defiant problems are now available.

The Child Behavior Checklist for Ages 6-18 (CBCL/6-18) similarly obtains reports from parents, close relatives, and/or guardians regarding school-aged children’s competencies and behavioral/emotional problems. The competency scale includes 20 items about a child’s activities, social relations, and school performance. Specific behavioral and emotional problems are described in 118 items that are rated along the 0-to-2 scale described previously, along with two open-ended items for reporting additional problems. A scoring profile provides raw scores, T-scores, and percentiles for three competence scales (Activities, Social, and School); Total Competence; eight crossinformant (clinical scale) syndromes; and Internalizing, Externalizing, and Total Problems (broad scales). The eight clinical scales scored from the CBCL/6-18 Teacher Report Form and Youth Self-Report are Aggressive Behavior; Anxious/Depressed; Attention Problems; Rule-Breaking Behavior; Social Problems; Somatic Complaints; Thought Problems; and Withdrawn/Depressed. Now available are also six DSM-oriented scales associated with affective problems, anxiety problems, somatic problems, attention-deficit/hyperactivity problems, oppositional defiant problems, and conduct problems. The school-age scales are based on new factor analyses of parents’ ratings of nearly 5000 clinically referred children, and norms are based on results from a nationally representative sample of 1753 children aged 6 to 18 years¹¹ (Box 7C-3).

BOX 7C-3 CASE 2: BEHAVIORAL AND EMOTIONAL ASSESSMENT DISCUSSION

The behavior problem profiles obtained on the CBCL/6-18 and the Youth Self-Report for Rachel are shown in the following two illustrations. On the CBCL problem scales (completed by her mother), Rachel’s Total Problems, Internalizing, and Externalizing scores and syndrome scales were all in the normal ranges for girls aged 12 to 18. Similarly, a teacher completed a Teacher Report Form, and results were all within the normal range. However, on the Youth Self-Report problem scales, Rachel reported more problems than are typically reported by teenage girls, particularly withdrawn behavior, somatic complaints, problems of anxiety or depression, problems in social relationships, thought problems, attention problems, and problems of an aggressive nature. Rachel’s responses on the Minnesota Multiphasic Personality Inventory—Adolescent indicated that she was experiencing high levels of general distress. Elevations on clinical scales 2,3,7,8,0 suggested that she may have felt anxious, lonely, and pessimistic much of the time and may have felt isolated from others and inferior. In other words, Rachel reported having high levels of internalizing symptoms, as well as difficulties managing social relationships and aggression. Crossinformant comparisons indicate that adults in Rachel’s life were not aware of the level of her internal distress. Discrepancies between Rachel’s self-report of symptoms and the ratings by her mother became a springboard for validating Rachel’s need for mental health attention and led to better communication within the family.

ASEBA materials are backed by extensive research in their development and have been used in more than 6000 studies pertaining to a broad range of behavioral health topics. There is strong support for its use with multidimensional child assessments in pediatric settings, (e.g., Mash and Hunsley²; Riekert et al,⁷² Stancin and Palermo⁷³), although criticisms have been raised about the validity of the CBCL for populations of chronically ill children.⁷⁴

Projective Techniques

Projective assessment techniques encourage a respondent to “project” issues, concerns, and perceptions onto ambiguous stimuli such as an inkblot or a picture. The basic premise is that when the child is faced with an ambiguous stimulus or one requiring perceptual organization, underlying psychological issues affecting the child will influence interpretation of these stimuli. The most commonly used projective techniques with children include use of child human figure or family drawings, storytelling responses to pictures or photographs, and reactions to Rorschach inkblots. Once the mainstay of personality assessment, projective assessment techniques have fallen out of favor in the era of evidence-based assessment techniques. However, some techniques continue to have clinical utility and validity with specific assessment purposes. They can provide clues that subsequently can be pursued with interviews and other techniques. For example, family drawings can be a helpful source of qualitative information about a child’s view of family relations, especially with younger children with more limited verbal expressions. Responses to incomplete sentences, story cards, and “3 wishes” (“if you could have 3 wishes, what would they be?”) can reveal insights into a child’s internal representations of relationships. In addition, the Rorschach has been shown to be a valid method for examining perceptual accuracy in youth with possible thought disorders when used with validated scoring systems such as John E. Exner’s system for scoring the Rorschach test.⁷⁹

Assessing Peer Relationships

Peer perspectives contain unique and important information about children but are usually missing in multi-informant clinical assessments. Peers play critical social roles in children’s lives and have access to information that adults may not have and that children may be reluctant to self-report. For example, social acceptance within a peer group is an important aspect of a child’s functional status, but it can be difficult to assess accurately by interview or parent report. Sociometric assessments that use peer nomination methods have been developed as a systematic way of gathering information about the extent to which a child is accepted or rejected within a peer group.⁸⁰ Strategies may involve asking children by interview or on paper to nominate three classmates with whom they most like to play (positive nominations/peer acceptance) and three classmates with whom they would least like to play (negative nominations/peer rejection). An alternative method is for children to rate how much they like to play with each classmate, for example, on a scale from 1 (“I don’t like to”) to 5 (“I like to a lot”). Using various statistical classification schemes, children can be considered to be popular, accepted, rejected, neglected, or controversial.

Peer nomination assessment instruments have been used to measure specific domains of child functioning besides peer acceptance. Techniques often involve presenting children in a classroom with a list of behavioral descriptions and asking them to select which of their peers best match each descriptor. Peer nomination approaches with acceptable reliability and validity have been developed to obtain peer ratings for a number of specific behavioral or emotional problem domains in children, such as ADHD symptoms, aggression and withdrawal, and depression.⁸¹

The Peer-report Measure of Internalizing and Externalizing Behavior⁸¹ was developed to assess a broad range of peer-reported externalizing and internalizing child psychopathology. As with other peer-nomination inventories, students are provided with classroom roster sheets that contain listings of all of the children in the classroom. Then, they are asked to select up to three classmates (either gender) who best fit the description read to them (e.g., “worry about things a lot” or “get mad and lose their temper”). Preliminary reports suggest that this measure demonstrates adequate reliability and validity as a broad measure of psychopathology.

Peer nomination procedures may be useful for psychosocial screening in the classroom, evaluating the effectiveness of mental health interventions on social behaviors in school settings, and conducting research on a range of child social behaviors. However, for clinical purposes, it may difficult (and impractical) to obtain peer ratings of an individual child. For this reason, parent, teacher, and rating scales of behavior in children can be used to as a more practical alternative for multi-informant assessments of peer relationships and functioning. For example, the ASEBA scales (e.g., the CBCL Teacher Report Form) include positive peer relationship items on the competence scales and a social problems scale highlighting peer difficulties on the problem scales.

Testing for Specific Problems

Family Assessment

Evaluations in developmental and behavioral pediatrics often include a family assessment in order to understand the interpersonal dynamics of the family system.¹¹¹ Using an unstructured interview format, a clinician may inquire about family structure, roles, and functioning and explore each family member’s perception of a presenting issue or problem. This assessment approach is often useful in family therapy sessions. Structured interviews may be employed to ensure that specific areas or topics are covered. Genograms are graphic representations of families that begin with a family tree and may include additional details about family structure, cohesiveness or conflicts, timelines of events, and family patterns (e.g., domestic violence, substance abuse, divorce, suicides, health conditions, presence of behavioral disorder). Formal, validated observational approaches to family assessment typically involved trained observers who coded ratings during live or videotaped observations of family interactions and are mostly confined to research settings.

There are many family self-report questionnaires targeting different aspects of functioning that may be useful in family assessments, especially in research settings.¹¹² Although questionnaires have psychometric appeal, they carry biases of the individual completing them, which is counter to the spirit of family assessment. Moreover, questionnaires may have limited utility when specific treatment recommendations are developed in clinical settings for a particular family’s set of concerns.¹¹¹ A popular example of a parent report family questionnaire with research and clinical applications is the Parenting Stress Index.¹¹³ This index consists of 120 items about child characteristics, parent personality, and situational variables, and it yields a Total Stress Score, as well as scale scores for child and parent characteristics. It has been translated and validated for use with a variety of international populations and has been shown to be useful in a clinical contexts.

Functional Outcomes

Measures of global functioning are typically ratings of a clinician’s judgment about a child or adolescent’s overall functioning in day-to-day activities at school, at home, and in the community.¹¹⁴ Measures of global functioning are useful for identifying need for treatment, as well as for monitoring treatment effects and predicting treatment outcome. The importance of global functioning is reflected in the placement of the Global Assessment of Functioning—which stipulates that impairment in one of more areas of functioning is necessary in order to meet criteria for a diagnosis—as Axis V on the DSM-IV. The Global Assessment of Functioning is a scale of a mental health continuum from 1 to 100 with 10 anchor descriptions; higher scores reflect better functioning. For example, a score between 31 and 40 would be given for a child with major functional impairment in several areas (frequently beats up younger children, is unruly at home, and is failing in school); a score between 61 and 70 is given to a child with mild symptoms (mild depressed mood) or some difficulties in functioning (disruptive in school) but who generally functions fairly well and who has good social relationships. Shaffer and colleagues modified the anchors of the Global Assessment of Functioning to pertain better to youth, creating the Children’s Global Assessment Scale (CGAS).¹¹⁵ This instrument yields one score and has been used in a large number of psychiatric outcome studies, especially medication-related research.¹¹¹

A widely used measure of functioning is the Child and Adolescent Functional Assessment Scale.¹¹⁶ This measure is a clinician-rated instrument consisting of behavioral descriptions (e.g., is expelled from school, bullies peers) grouped into levels of impairment for each of five domains: role performance (school/work, home, community), behavior toward others, moods/self-harm, substance use, and thinking. The Child and Adolescent Functional Assessment Scale has been shown to have considerable criterion-related and predictive validity and is widely used to evaluate outcome in clinical settings and in clinical research.¹¹¹

Adaptive functioning measures such as the Vineland Adaptive Behavior Scales¹¹⁷ are used to assess personal and social skills needed for everyday living and are especially useful for identifying children with mental retardation, developmental delays, and pervasive developmental disorders. The Vineland scales include survey interview and parent/caregiver rating forms that yield domain and adaptive behavior composite standard scores (M = 100, SD = 15), percentile ranks, adaptive levels, and age-equivalent scores for individuals from birth to age 90 years. Domains assessed include Communication, Daily Living Skills, Socialization, Motor Skills, and an optional Maladaptive Behavior Index.

Health-related quality-of-life (HRQOL) measures have been developed to evaluate functional outcomes in clinical and health services research. HRQOL measures differ from more traditional measures of health status and physical functioning by also assessing broader psychosocial dimensions such as emotional, behavioral, and social functioning. The Pediatric Quality of Life Inventory (PedsQL 4.0)¹¹⁸ is an example of an HRQOL measure that has been developed and validated for use in pediatric settings. The PedsQL 4.0 Generic Core Scales assess physical, emotional, social, and school functioning with child self-report (ages 5 to 18) and parallel parent proxy-report formats (for children aged 2 to 18 years). Physical Health and Psychosocial Health summary scores are transformed to a scale of 0 to 100 in which higher scores reflect better health-related quality of life. The PedsQL 4.0 had excellent internal consistency reliability in a large pediatric sample, distinguished healthy children from those with chronic health conditions, and was related to other indicators of health status.¹¹⁸