Some studies have attempted to optimize the timing and behavioral repertoire for identification of the earliest signs of ASD. The first era of such studies was devoted to retrospective analyses of home videotapes of infants. For example, Adrien et al. compared home videotapes of children with and without ASD before and after their 1st birthday, using the Infant Behavioral Summarized Evaluation Scale. The children with ASD ignored other people, preferred being alone, and made no eye contact in a socialization context [7]. They showed a lack of vocal communication, lack of appropriate facial expressions, no social smile, a lack of gestures, and no or poor imitation of others, as compared to typically developing children. Using similar methodology, Osterling & Dawson identified 4 key behaviors that could differentiate children with and without ASD with 90% accuracy: low frequency of looking at others (including eye contact), low frequency of orienting to name call (at 8 months), absence of showing objects, and a lack of pointing. Repetitive motor actions could also differentiate ASD and intellectual disability [8]. The age of identification has been reduced, as many studies have been performed, as deficits in social smiling and eye contact at 6 months are significant indicators of the development of ASD [9].

Retrospective parental reports also provided similar insights into the early signs of ASD in the 1st year of life. In a retrospective parental interview using the First Year Inventory, Baranek et al. [10] suggested that markers differentiating ASD from developmental delay include markers involving social attention and communication, including orienting to name call, following pointing, social orientation, interest in their age, social smiling, facial expression, playing peek-a-boo, and demanding attention from the caregiver. Imitation, expressive communication, sensory processing, regulatory patterns, reactivity, and repetitive behaviors were found to be nonspecific signs, which can only differentiate ASD and developmental delay from typical development.

The first retrospective studies set the first milestones for early identification, emphasizing 1) the existence of early risk markers around 1 year of age and 2) the value of deficits in social communication and attention as early signs of ASD. However, these studies were limited by generalization, such as sampling biases and a lack of standardized social situations in home videos, incorrect recall, recall biases, distortion of events, biases related to parental alertness in recognizing behaviors, socioeconomic status, personality, intelligence, and parental mental health.

To compensate for the limitations of the retrospective studies, prospective infant follow-up studies have been performed in the past decade. Those studies focused on high-risk infants, i.e., the infants believed to have a higher chance of developing ASD than the general population. The first such subjects studied were neonates who had been cared for in neonatal intensive care units (NI-CUs) due to neonatal or obstetric causes. As the published prevalence of ASD among these high-risk infants is 1.3-3.2%, it may be debatable whether they actually have significantly higher risk of developing ASD than typical infants. However, birth complications, low birth weight, or prematurity have consistently been replicated as risk factors for ASD, and infants who have been hospitalized in NICUs are candidates for prospective surveillance of the effectiveness of follow-up studies.

The second, more feasible and reliable subjects for prospective studies are infants who are the younger siblings of children diagnosed with ASD. The sibling concordance rate is high, up to 10%, and there is an increased recurrence risk of ASD in siblings. A recent, large prospective follow-up study explored 664 younger siblings who have elder siblings diagnosed with ASD. The recurrence rate of ASD in the siblings was as high as 18.4%, and the risk was even higher if the sibling was male or from multiplex families including two or more older siblings with ASD [11]. Typically, babies were followed up in parallel with low-risk infants who did not have any older siblings with ASD, from 6 months of age up to 3 years, with specific attention to the first 18-24 months of development. Such a design is desirable, as it avoids the recall bias of the parents, provides multiple time points for grasping very early signs from the youngest age, and makes it possible to apply experimental measures as well as rating scales and observational instruments. Moreover, from a clinical perspective, this design allows early intervention if the subject shows significant red flag signs.

A few population-based prospective follow-up studies have also been performed. Such a research design is ideal for following up high risk behaviors in the general population, but not in those with high risk, who share genetic vulnerabilities and possible broad autism phenotypes, and may allow identification of red flag signs without the influence of genetic traits. However, such a design requires a very large cohort, time, and personnel, and more importantly, a consolidated system in the community for registering and following up subjects consistently and reliably.

One of the noteworthy attempts to identify risk signs and screen infants in the community is the development of the SACS [36], which was based on a previously established developmental surveillance system, including Maternal and Child Health (MCH) nurses and related practitioners in Australia. The MCH service is a free public service provided to all families with children younger than 6 years of age, with the focus on child and maternal health surveillance and screening. The authors provided a draft of the SACS highlighting potential early risk signs of ASD, in infancy and toddlerhood, selected from the literature, and trained MCH nurses to administer the instrument. The MCH nurses screened 20,770 children, a total of 216 subjects were referred for confirmation of diagnosis as "at-risk" infants, and the positive-predictive value of the SACS was found to be 81%, with a sensitivity of 69-84% and a specificity of 99.9% [37].

The SACS is one of the most efficient early screening instruments for ASD. It measures red flag signs of ASD at multiple points of development (8, 12, 18, and 24 months), compensating for the age limit of the Checklist of Autism for Toddlers (CHAT) and the Modified CHAT. It covers the risk signs of early ASD comprehensively, including social response, communication, reciprocity, imitation, pointing, joint attention, and developmentally appropriate interactive play. It is a combined rating scale and direct instrument, which can be administered by trained mental health professionals. The Korean translated version is also now available.