**3. Implementing autism-specific screening**

Both the Social Security Act [40] and the Individuals with Disabilities Education Improvement Act [41] reaffirm the mandate for PCPs to play a central role in the identification of develop‐ mental disabilities. Between birth and 2nd birthday, children are recommended to participate in a sequence of 11 preventive pediatric healthcare visits (i.e., well-child visits). In the context of these visits, healthcare professionals have a unique opportunity to develop a long-term, trusting relationship with the children and their families. Parents expect that healthcare professionals take an interest in their child's development and behavior, competently identify strengths and weaknesses, and are able to help them access available community resources, if necessary [6].

#### **3.1. Limitations of developmental surveillance**

Developmental surveillance is defined as a "flexible, longitudinal, continuous, and cumulative process whereby knowledgeable healthcare professionals identify children who may have developmental problems" ([5] p. 407). Surveillance includes eliciting and attending to the parents' concerns, maintaining a developmental history, making accurate and informed observations of the child, identifying the presence of risk and protective factors, and docu‐ menting the process and findings. According to the AAP recommendations on developmental surveillance and screening [5], healthcare professionals are recommended to implement developmental screening during all well-child visits. At the same time, this policy statement also asserts that developmental surveillance by itself is insufficient for detecting developmen‐ tal concerns in a large number of children. Instead, the AAP recommends using a combination of developmental surveillance and formal screening tests. Research has shown that healthcare professionals who rely solely on clinical judgment are much less accurate in estimating developmental status than professionals who implement formal screening tests [42, 43]. That is, the clinical impressions of healthcare professionals tend to have excellent specificity (i.e., if children are identified, concerns tend to be valid). At the same time, their sensitivity in detection developmental delays is quite poor, leading to many un-identified children and revealing significant difficulties in detecting less obvious delays. For example, Hix-Small and colleagues [44] conducted a study where global developmental delays (at 12 and 24 months) were identified based on both, clinical impressions and a formal screening test (Ages and Stages Questionnaire, ASQ;[45, 46]). Results revealed that 48% of children with true develop‐ mental delays would have been missed if only clinical impressions were used. Similarly, in a study that aimed to identify toddlers with ASD, Robins [47] screened a total of 4,797 children, identifying 21 children who were later diagnosed with ASD. Of those 21 children, only 4 were previously red-flagged by the child's pediatrician.

### **3.2. Recommended and implemented screening practices**

surveys indicate 34% of parents reported some familiarity with the LSAE campaign. Significant changes in target parent behaviors were detected in that more parents knew of the behaviors likely to be associated with ASD, the best time to get help for ASD, and the developmentalmilestones their childshouldbe reaching[38].Similarly, substantiallymore healthcare professionals believe that they have the resources necessary to educate parents about monitoring child development, and fewer advocated a 'wait and see' approach,

Both the Social Security Act [40] and the Individuals with Disabilities Education Improvement Act [41] reaffirm the mandate for PCPs to play a central role in the identification of develop‐ mental disabilities. Between birth and 2nd birthday, children are recommended to participate in a sequence of 11 preventive pediatric healthcare visits (i.e., well-child visits). In the context of these visits, healthcare professionals have a unique opportunity to develop a long-term, trusting relationship with the children and their families. Parents expect that healthcare professionals take an interest in their child's development and behavior, competently identify strengths and weaknesses, and are able to help them access available community resources, if

Developmental surveillance is defined as a "flexible, longitudinal, continuous, and cumulative process whereby knowledgeable healthcare professionals identify children who may have developmental problems" ([5] p. 407). Surveillance includes eliciting and attending to the parents' concerns, maintaining a developmental history, making accurate and informed observations of the child, identifying the presence of risk and protective factors, and docu‐ menting the process and findings. According to the AAP recommendations on developmental surveillance and screening [5], healthcare professionals are recommended to implement developmental screening during all well-child visits. At the same time, this policy statement also asserts that developmental surveillance by itself is insufficient for detecting developmen‐ tal concerns in a large number of children. Instead, the AAP recommends using a combination of developmental surveillance and formal screening tests. Research has shown that healthcare professionals who rely solely on clinical judgment are much less accurate in estimating developmental status than professionals who implement formal screening tests [42, 43]. That is, the clinical impressions of healthcare professionals tend to have excellent specificity (i.e., if children are identified, concerns tend to be valid). At the same time, their sensitivity in detection developmental delays is quite poor, leading to many un-identified children and revealing significant difficulties in detecting less obvious delays. For example, Hix-Small and colleagues [44] conducted a study where global developmental delays (at 12 and 24 months) were identified based on both, clinical impressions and a formal screening test (Ages and Stages Questionnaire, ASQ;[45, 46]). Results revealed that 48% of children with true develop‐ mental delays would have been missed if only clinical impressions were used. Similarly, in a

indicating that the marketing campaign was effective [39].

**3. Implementing autism-specific screening**

8 Recent Advances in Autism Spectrum Disorders - Volume I

**3.1. Limitations of developmental surveillance**

necessary [6].

The use of general developmental screening instruments has been recommended by the AAP since 2001 [48]. Current recommendations for identifying children with developmental disabilities (not specifically ASD) suggest population-wide screening at 9, 18, and 30 months [5]. Empirical information on the extent to which these recommendations have been imple‐ mented is limited. Two national surveys of AAP members completed in 2002 and 2009 indicate that implementation has been slow [49, 50]. When asked about their screening practices, only 23.0% (2002) and 47.7% (2009) of physicians reported that they 'always' or 'almost always' administer a formal screening tool. Since many healthcare professionals may administer standard screening tools in a non-standard manner (e.g., by asking some but not other items) or only administer these tools to patients considered 'high risk', these numbers likely overes‐ timate true implementation [51]. Researchers who asked parents to report on the screening practices of their healthcare professionals found that less than 27% of parents of children between 10 to 35 months recalled completing a developmental screening questionnaire within the last 12 months [52]. Only two survey-based studies have evaluated the implementation of ASD-specific screening instruments. In a 2004 survey of 255 pediatricians licensed in Maryland and Delaware, dosReis and colleagues reported that only 8% of the respondents screened for ASD [53]. In a 2007 survey of 51 pediatricians licensed in Alabama and Mississippi, Gillis reported that 28% reported using ASD-specific screening instruments [54]. Importantly, only one pediatrician reported routine screening for ASD at 18- or 24-months, suggesting that many healthcare professionals administer ASD-specific screening tests only to children who are considered 'high risk'.

#### *3.2.1. Barriers to successful screening in the primary care setting*

Well-child visits are often the only routine, formalized, and longitudinal contact a child has with a healthcare professional and thus is an ideal place to implement population-wide screening. Given this widely acknowledged responsibility, it is striking that many practices do not implement developmental and ASD-specific screening measures as recommended by the AAP. In the following we will consider several key barriers that interfere with healthcare professionals' ability to implement effective screening practices.

**1.** Parental compliance with the preventive pediatric healthcare schedule

The Medical Expenditure Panel Survey provides nationally representative information on preventive care for children between 0 and 5 years of age [55]. Results indicate that the average compliance ratio is 71.3% (*SD* = 1.4%), indicating that on average, parents attend about 3 out of 4 AAP recommended well-child visits. In addition, findings reveal large variation in wellchild visit attendance based on the families' socioeconomic status, access to resources, and geography. That is, the compliance ratio was significantly below average if the child was without health insurance (*M* = 52.5%, *SD* = 3.8), lived below the poverty level (*M* = 62.8%, *SD* = 2.4), lived with a single parent (*M* = 65.5%, *SD* = 2.7), was Hispanic or Black (*M* = 64.5%, *SD* = 2.1, or *M* = 64.7%, *SD* = 2.9, respectively), or had parents without a high-school degree (*M* = 60.8%, *SD* = 3.1). Moreover, in terms of geography, the compliance ratio ranged between 90.9% (*SD* = 5.0) in New England and 51.3% (*SD* = 3.4) in the West South Central Census division. To account for this variability in parental compliance with the preventive pediatric care schedule, practices may be able to screen about 30% more children by flexibly administering screening measures during sick-visits, when necessary [56].

tion can be dramatically reduced by administering the follow-up interview (3:1); (3) even though the sensitivity of the M-CHAT has not been evaluated, comparisons to prevalence esti‐ mates of ASD suggest that the M-CHAT is likely to miss a considerable number of children [7]. Assuming current prevalence estimates (1:88), the sample reported by Roberts [47] likely in‐ cluded about 55 children with ASD. Even after considering participant attrition as a factor (i.e., families who did not complete the follow-up interview or diagnostic evaluations), the number of children with ASD actually identified by the researchers (*n* = 21) is considerably lower than

Promoting Early Identification of Autism in the Primary Care Setting: Bridging the Gap Between What We Know and

What We Do

11

http://dx.doi.org/10.5772/53715

The extent to which limited parent literacy interferes with the implementation of broad-band developmental and ASD-specific screening measures has not been investigated. However, da‐ ta presented by Davis and colleagues suggest that limited parental literacy has the potential to pose significant obstacles, particularly in the context of practices that primarily serve indigent or immigrant communities [60, 61]. Based on a convenience sample of 396 parents from one large medical center (i.e., Louisiana State University Medical Center, Shreveport, LA), 11% and 16% of parents showed a reading level below 4th and 7th grade, respectively. Limited parental literacy may be an important factor in explaining missing data problems, reported across many population-based screening studies. For example, Hix-Small and colleagues reported that only about 54% of the administered screening questionnaires were completed and re‐

The successful implementation of effective screening practices requires more than educational opportunities for individual staff members. Instead, what is needed is a context that supports organizational restructuring [51, 62]. That is, practices need to develop, evaluate, and refine office-wide implementation systems that divide responsibilities among staff members at multiple levels. For example, the screening instrument may be distributed by a member of the front desk staff, scored by a nurse, reviewed with the family by a provider, and possible referrals may be coordinated by a social worker. Developing such an office-wide implemen‐ tation system requires an 'internal champion' to lead the charge, a process for collecting data to monitor progress, and a seamless integration with the clinics' electronic medical record system. In order to be sustainable, the implementation system also needs to be sufficiently robust to be workable in the context of busy periods (e.g., the onset of the winter viral season)

Research suggests that traditional methods of education, including printed educational materials and didactic, lecture-based continuing medical education (CME) sessions, have little to no effect on the behavior of healthcare professionals [63, 64]. During recent years, several alternative approaches to the traditional CME format have been suggested. These approaches include specialized modules on ASD that are included as part of pediatric residency training programs (e.g., CDC Autism Case Training) [65, 66], and academic detailing where a focused

would be expected.

turned [44].

and staff turnover.

**4.** Limited parent literacy

**5.** Need for practice-wide system change

*3.2.2. Novel approaches to support autism-specific screening practices*

**2.** Competing priorities, time, and reimbursement constraints

During the last decade, the AAP [5, 6, 57] has published three separate statements on the early detection of developmental-behavioral problems. Recommended screenings include: (1) screening for maternal postpartum depression (in the first year), (2) broad-band developmen‐ tal screening (at 9, 18, and 24/30 months), (3) ASD-specific screening (at 18 and 24 months), (4) social-emotional screening (contingent upon abnormal developmental or ASD screeners), (5) kindergarten readiness screening (at 4 years), and (6) mental health/psychological function screening (age 5 years and thereafter). Implementing such a dense screening schedule in the context of short well-child visits filled with competing priorities (e.g., vaccinations, medical surveillance), and limited reimbursement options is often noted as a key challenge by health‐ care professionals [53].

**3.** Availability of high-quality screening instruments

Given that screening tools are being developed, revised, and evaluated on an ongoing basis, the AAP practice guidelines do not recommend any particular ASD-specific screening instru‐ ment [6]. Instead, the AAP guidelines include a review of several potential measures, leaving it up to the healthcare professional to select an instrument that fits the particular needs of their practice. To date, the ASD-specific screening measure that has been adopted most widely in primary care settings is the Modified Checklist for Autism in Toddlers (M-CHAT; [47, 58, 59]. The M-CHAT is a 23-item yes/no parent report screener for ASD. Screening positive (i.e., "fail‐ ing") is defined as failing any three items, or any two of six critical items. Failed items are re‐ viewed with a follow-up interview, typically administered by phone a few weeks after the screener is completed. Most of the research evaluating the efficacy of this measure has focused on the positive predictive value (PPV), defined as the number of true positive cases divided by number of cases that screened positive. Robins reported on the screening results of 4,797 tod‐ dlers (screened during their 15-, 18-, or 24-month well-child visits) [47]. The numbers of tod‐ dlers who failed the M-CHAT questionnaire, failed the M-CHAT follow-up interview, and were eventually diagnosed with ASD, were 466, 61, and 21, respectively. Thus, if the results of the follow-up interview were considered, the M-CHAT revealed a PPV of.57. However, with‐ out the follow-up interview, the PPV of the M-CHAT was as low as.06. These estimates are consistent with data reported by Kleinman and colleagues [59], who reported PPVs of.65 and. 11, depending on whether the follow-up interview was or was not considered. Based on these findings, at least three conclusions seem warranted: (1) without administering the follow-up interview, the M-CHAT is likely to over-identify children by a factor of 17:1; (2) over-identifica‐ tion can be dramatically reduced by administering the follow-up interview (3:1); (3) even though the sensitivity of the M-CHAT has not been evaluated, comparisons to prevalence esti‐ mates of ASD suggest that the M-CHAT is likely to miss a considerable number of children [7]. Assuming current prevalence estimates (1:88), the sample reported by Roberts [47] likely in‐ cluded about 55 children with ASD. Even after considering participant attrition as a factor (i.e., families who did not complete the follow-up interview or diagnostic evaluations), the number of children with ASD actually identified by the researchers (*n* = 21) is considerably lower than would be expected.

**4.** Limited parent literacy

without health insurance (*M* = 52.5%, *SD* = 3.8), lived below the poverty level (*M* = 62.8%, *SD* = 2.4), lived with a single parent (*M* = 65.5%, *SD* = 2.7), was Hispanic or Black (*M* = 64.5%, *SD* = 2.1, or *M* = 64.7%, *SD* = 2.9, respectively), or had parents without a high-school degree (*M* = 60.8%, *SD* = 3.1). Moreover, in terms of geography, the compliance ratio ranged between 90.9% (*SD* = 5.0) in New England and 51.3% (*SD* = 3.4) in the West South Central Census division. To account for this variability in parental compliance with the preventive pediatric care schedule, practices may be able to screen about 30% more children by flexibly administering screening

During the last decade, the AAP [5, 6, 57] has published three separate statements on the early detection of developmental-behavioral problems. Recommended screenings include: (1) screening for maternal postpartum depression (in the first year), (2) broad-band developmen‐ tal screening (at 9, 18, and 24/30 months), (3) ASD-specific screening (at 18 and 24 months), (4) social-emotional screening (contingent upon abnormal developmental or ASD screeners), (5) kindergarten readiness screening (at 4 years), and (6) mental health/psychological function screening (age 5 years and thereafter). Implementing such a dense screening schedule in the context of short well-child visits filled with competing priorities (e.g., vaccinations, medical surveillance), and limited reimbursement options is often noted as a key challenge by health‐

Given that screening tools are being developed, revised, and evaluated on an ongoing basis, the AAP practice guidelines do not recommend any particular ASD-specific screening instru‐ ment [6]. Instead, the AAP guidelines include a review of several potential measures, leaving it up to the healthcare professional to select an instrument that fits the particular needs of their practice. To date, the ASD-specific screening measure that has been adopted most widely in primary care settings is the Modified Checklist for Autism in Toddlers (M-CHAT; [47, 58, 59]. The M-CHAT is a 23-item yes/no parent report screener for ASD. Screening positive (i.e., "fail‐ ing") is defined as failing any three items, or any two of six critical items. Failed items are re‐ viewed with a follow-up interview, typically administered by phone a few weeks after the screener is completed. Most of the research evaluating the efficacy of this measure has focused on the positive predictive value (PPV), defined as the number of true positive cases divided by number of cases that screened positive. Robins reported on the screening results of 4,797 tod‐ dlers (screened during their 15-, 18-, or 24-month well-child visits) [47]. The numbers of tod‐ dlers who failed the M-CHAT questionnaire, failed the M-CHAT follow-up interview, and were eventually diagnosed with ASD, were 466, 61, and 21, respectively. Thus, if the results of the follow-up interview were considered, the M-CHAT revealed a PPV of.57. However, with‐ out the follow-up interview, the PPV of the M-CHAT was as low as.06. These estimates are consistent with data reported by Kleinman and colleagues [59], who reported PPVs of.65 and. 11, depending on whether the follow-up interview was or was not considered. Based on these findings, at least three conclusions seem warranted: (1) without administering the follow-up interview, the M-CHAT is likely to over-identify children by a factor of 17:1; (2) over-identifica‐

measures during sick-visits, when necessary [56].

10 Recent Advances in Autism Spectrum Disorders - Volume I

**3.** Availability of high-quality screening instruments

care professionals [53].

**2.** Competing priorities, time, and reimbursement constraints

The extent to which limited parent literacy interferes with the implementation of broad-band developmental and ASD-specific screening measures has not been investigated. However, da‐ ta presented by Davis and colleagues suggest that limited parental literacy has the potential to pose significant obstacles, particularly in the context of practices that primarily serve indigent or immigrant communities [60, 61]. Based on a convenience sample of 396 parents from one large medical center (i.e., Louisiana State University Medical Center, Shreveport, LA), 11% and 16% of parents showed a reading level below 4th and 7th grade, respectively. Limited parental literacy may be an important factor in explaining missing data problems, reported across many population-based screening studies. For example, Hix-Small and colleagues reported that only about 54% of the administered screening questionnaires were completed and re‐ turned [44].

**5.** Need for practice-wide system change

The successful implementation of effective screening practices requires more than educational opportunities for individual staff members. Instead, what is needed is a context that supports organizational restructuring [51, 62]. That is, practices need to develop, evaluate, and refine office-wide implementation systems that divide responsibilities among staff members at multiple levels. For example, the screening instrument may be distributed by a member of the front desk staff, scored by a nurse, reviewed with the family by a provider, and possible referrals may be coordinated by a social worker. Developing such an office-wide implemen‐ tation system requires an 'internal champion' to lead the charge, a process for collecting data to monitor progress, and a seamless integration with the clinics' electronic medical record system. In order to be sustainable, the implementation system also needs to be sufficiently robust to be workable in the context of busy periods (e.g., the onset of the winter viral season) and staff turnover.

#### *3.2.2. Novel approaches to support autism-specific screening practices*

Research suggests that traditional methods of education, including printed educational materials and didactic, lecture-based continuing medical education (CME) sessions, have little to no effect on the behavior of healthcare professionals [63, 64]. During recent years, several alternative approaches to the traditional CME format have been suggested. These approaches include specialized modules on ASD that are included as part of pediatric residency training programs (e.g., CDC Autism Case Training) [65, 66], and academic detailing where a focused training session on early identification of ASD is presented to the entire office staff of individual practices [67]. In the following section, we will describe a third approach that has been used widely to improve the delivery of high-quality healthcare, although it has yet to be imple‐ mented with a focus on improving early detection of children with ASD.

knowledge with practical voices from peers. In between the Learning Sessions, teams engage in Action Periods during which they implement change in a cyclical fashion: a) teams develop a PLAN to implement change, b) they DO the work to implement the chance, c) they STUDY their progress by measuring clinical behaviors, and d) they ACT upon the results by refining

Promoting Early Identification of Autism in the Primary Care Setting: Bridging the Gap Between What We Know and

What We Do

13

http://dx.doi.org/10.5772/53715

When multiple risk factors are present (e.g., the child has a family history of ASD, the child's parents or other family members are concerned about behaviors related to ASD, the child's pediatrician notes concerns about ASD, or the child fails an ASD-specific screener), the AAP clinical practice guidelines recommend that the child's PCP simultaneously (1) provides parental education on ASD, (2) refers the child for a comprehensive ASD evaluation, (3) refers the child for an audiology evaluation, (4) refers the child to Early Intervention or Early Childhood Education Services, and (5) schedules a targeted follow-up visit within one month [6]. Accomplishing all five actions simultaneously is considered crucial to avoid costly delays.

Little is currently known about how parents experience autism-specific screening, parent education, and referrals provided by the child's PCP in accordance with the AAP clinical practice guidelines. By definition, these interactions between PCP and parent precede a formal ASD diagnosis. Thus, PCPs typically communicate a certain level of uncertainty about the child's ASD diagnosis, emphasizing that the child is considered to be at 'high risk for ASD', and focusing on the need for a comprehensive evaluation to 'rule out ASD'. Even at this tentative level, communicating concerns about ASD requires initiating or engaging parents in a difficult conversation. For some parents, this conversation confirms suspicions that lay dormant for various amounts of time. For other parents, it comes out of nowhere and confronts them with concerns they had not previously contemplated. In either case, parents might not yet be emotionally ready to consider the possibility that their child has autism. Parents often feel responsible for their child's behavior or delays, feel guilty about not noticing or acting upon these concerns more promptly, and fear the stigma all too often associated with devel‐

When prompted to reflect upon the time when their child was first diagnosed, many parents of older individuals with ASD express significant levels of discontentment, in some cases resentment and anger [72, 73]. In part, these strong emotions stem from the fact that most families experience significant delays between the time when they first note concerns [73], the time when they choose to share these concerns with their child's physician [9, 74], the time when their child is first evaluated [11], and the time when the ASD diagnosis is first given [7, 11]. On average, parents report seeing four to five doctors before an ASD diagnosis is made, which occurs on average at age of four or five years [7, 11, 72, 73]. Parental perceptions of delay in diagnosis due to 'watchful waiting' are associated with (1) lower satisfaction with the diagnostic process, (2) lower satisfaction with the help physicians offer after the diagnosis, and

(3) lower parental confidence in the physician's ability to help [72, 73, 75].

their approach (Plan-Do-Study-Act cycles of learning).

opmental disabilities.

**4. Communicating concerns about autism to parents**

#### The Breakthrough Series Collaborative Model

The Breakthrough Series Collaborative Model (BSCM) has been developed by the Institute of Healthcare Improvement (IHI, www.ihi.org) [68], and used successfully to improve the delivery of preventive services by pediatric practices [69], follow-up to newborn hearing screening [70], and child mental health service use [71]. In this model, several healthcare provider teams partner with external experts to overcome specific barriers that impede the delivery of high-quality care within their organization. Figure 2 presents the key elements of a Breakthrough Series model.

**Figure 2.** Breakthrough Series Model

The breakthrough series starts with the selection of a specific topic that is considered ripe for improvement. Even though data on the efficacy of learning collaboratives to increase early identification of ASD has not been published to date, we suggest that this would be a very appropriate topic, (1) because of the high prevalence rates of ASD, and (2) because the existing knowledge in this area is sound but not widely used. Once the topic is selected, a faculty team is assembled that combines expertise in the subject area as well as an improvement advisor who coaches teams on improvement methods. Organizations elect to join the collaborative through an application process, appointing multi-disciplinary teams within the organization (a champion's committee). The multi-disciplinary teams from all organizations are then brought together for Learning Sessions that combine the exchange of formal academic knowledge with practical voices from peers. In between the Learning Sessions, teams engage in Action Periods during which they implement change in a cyclical fashion: a) teams develop a PLAN to implement change, b) they DO the work to implement the chance, c) they STUDY their progress by measuring clinical behaviors, and d) they ACT upon the results by refining their approach (Plan-Do-Study-Act cycles of learning).
