Primary ASD screening (CHAT-A). The CHAT-A is a parent-completed screening questionnaire used in this project. It comprises 23 fixed items with standardized instructions, which facilitates implementation in primary care settings. Its item framework is based on M-CHAT–related screening constructs and is used for parent-reported primary screening of children aged 18–24 months (3) (4).

Assessment of two social high-risk warning behaviors (Binomial Observation Test, BOT). The BOT includes two behavioral observations (1): response to name calling; and (2) response to verbal instructions (following commands). During routine well-child visits in a naturalistic setting, both the physician and the caregiver independently initiate name calling and simple verbal instructions, while observing whether the child shows responsive behaviors such as turning toward the caller, making eye contact, vocalizing, or carrying out the instruction. BOT is considered positive if any observation indicates insufficient or absent responsiveness.

Integrated determination of primary screening results. Primary screening results are determined by combining the two components. A child is classified as screen-positive at the primary level if either the parent questionnaire (CHAT-A) or any BOT component is positive, and the child then enters the subsequent referral pathway. Among these components, the physician-led behavioral observation is a critical element for identifying high-risk children at the primary screening stage and is therefore a key focus of subsequent screening supervision (Figure 1).

In July 2021 (one month after the centralized training), the Children’s Hospital of Fudan University served as the expert technical support institution and was responsible for training and supervisory guidance, whereas the district maternal and child health hospital acted as the district-level management and quality-control body, overseeing organization, implementation, and routine supervision. A joint supervisory team was formed and divided into three supervisory groups to conduct on-site quality-control supervision across community child health outpatient clinics throughout the district. Supervision indicated that implementation of the parent-completed questionnaire (CHAT-A) was relatively consistent at the primary level, owing to its fixed items and standardized instructions. In contrast, the main challenges were concentrated in the BOT component. As BOT relies on behavioral observation, operational quality was susceptible to multiple on-site factors, leading to variability in implementation. These interfering factors primarily included the screening environment and equipment configuration, table height and seating arrangement, preparation of materials, pre-assessment communication with parents, and operational precautions during the assessment process. Following supervision, the district maternal and child health hospital developed a “screening precautions and key points checklist” based on expert recommendations and disseminated it to the 13 community centers to support subsequent quality improvement.

Video recording. From August to October 2021, each community health service center organized video recording and presentation at the team level. After implementing improvements based on key points identified in primary screening, teams recorded videos of routine primary ASD screening procedures during daily practice. Each team reviewed and selected two to three primary screening videos that reflected its improvement focus for that month, which were then submitted for discussion at the subsequent month’s video-feedback meeting.

Video review and discussion. Video review was integrated into the monthly district-level child health routine meetings. Each community health service center took turns presenting the procedural changes implemented during that month’s video recording and discussing factors that could further improve screening performance. The remaining 12 community centers and the district maternal and child health hospital reviewed the submitted videos and conducted a structured appraisal of key points. Discussions focused on technical/professional elements, organizational factors, and the effectiveness of work processes. Teams also shared practical experience regarding screening procedures and communication with parents.

Two physicians from the project team at the district maternal and child health hospital served as evaluators. They reviewed the participant videos presented at each video-feedback meeting and assessed screening performance and operational execution, including: examiner position, equipment preparation, screening environment, management of parental interference, standardized verbal instructions, initiation timing, waiting time, number of operations, use of physical gestures during language items, and operational sequence. These 10 items were treated as key performance indicators and scored dichotomously (1 point if the criterion was met; 0 otherwise), yielding a total score ranging from 0 to 10 (5).

Specifically (1): Examiner position was defined as maintaining an appropriate distance and eye-level alignment with the child to ensure clear delivery of name-calling and instructions, while avoiding positioning that distracts the child or obstructs observation (2); Equipment preparation referred to having necessary materials, toys, and recording/registration tools prepared and placed for efficient use (3); Screening environment referred to a relatively quiet setting with distractions controlled to meet basic conditions for observing child responses (4); Parental interference management referred to communicating with parents in advance and minimizing prompts, answering on behalf of the child, or physical guidance during observation (5); Standardized verbal instructions referred to clear and concise name-calling and commands consistent with the project protocol, avoiding repeated explanations or leading cues (6); Initiation timing referred to initiating name-calling/commands when the child’s attention was relatively stable, rather than during crying or high-interference states (7); Waiting time referred to providing an adequate response window before proceeding to the next step, avoiding overly rapid progression (8); Number of operations referred to completing the prescribed number of name-calling/commands as specified, without omission or arbitrary increase/decrease (9); Physical gestures during language items referred to avoiding additional gestures or pointing that might cue the child when delivering verbal instructions; and (10) Operational sequence referred to completing the two behavioral observation steps in the order specified by the protocol.

The two evaluators scored independently. In cases of disagreement, the evaluators jointly reviewed the video and reached consensus; the final consensus score was used as the video score.

A questionnaire survey was administered among child health staff participating in the program. A total of 58 questionnaires were returned; 7 were deemed invalid due to missing key information or inconsistent responses. Ultimately, 51 valid questionnaires were included in the analysis. Among the respondents, 1 was male and 50 were female; 28 were aged ≤39 years and 23 were aged ≥40 years. Eight had a junior college degree and 43 had a bachelor’s degree. Twenty-one respondents were nurses and 30 were physicians. Nineteen held junior-level titles or below, and 32 held intermediate-level titles or above. Years of practice were ≤10 years for 19 respondents, 11–20 years for 17 respondents, and >20 years for 15 respondents. Duration of work in child health was ≤5 years for 16 respondents, 6–10 years for 19 respondents, and ≥11 years for 16 respondents (Table 1).

Among the survey respondents, 7 reported serving as team leaders, 36 as team members, and 8 as part-time team members. Forty respondents reported spending 0–59 minutes per week on video recording, whereas 11 reported spending ≥1 hour. Regarding the average number of recordings per video, 17 reported recording once, 32 reported recording 2–4 times, and 2 reported recording ≥5 times. Thirty-six respondents rated the quality of their team’s videos as good and 15 as average. Forty-five respondents rated the quality of other teams’ videos as good and 6 as average. All respondents reported that video-feedback learning was helpful for their work. Forty-eight respondents considered video-feedback learning to be more efficient than traditional approaches; only 3 considered its efficiency comparable to or lower than that of traditional approaches. Forty-six respondents reported that, after video-feedback learning, they would be more likely to perform well in similar situations encountered in subsequent work (Table 2).

Regarding the perceived advantages of the video-feedback approach compared with traditional methods, 47 respondents reported that video feedback allows repeated viewing to identify and correct procedural details; 45 reported that videos can be viewed at any time, providing flexible learning schedules; and 45 reported that joint viewing facilitates collaborative learning and discussion, enabling participants to recognize strengths and weaknesses in their own and others’ practices. In addition, 17 respondents indicated that they felt more relaxed during video recording and that the recordings more closely reflected routine practice, and 26 respondents reported that video feedback helped provide educational guidance to parents regarding ASD. Conversely, 5 respondents stated that they preferred traditional learning methods (Table 3).

The interviews were conducted by two project physicians from the district maternal and child health institution, both of whom had experience in qualitative research methods. Five subthemes were identified: perceptions of the training, implementation of video feedback, perceived support for learning, safety concerns, and experiences with supervision using video feedback. Participants expressed generally consistent views regarding the video-feedback approach, and thematic saturation was reached. Overall, community child health staff reported good acceptability of video-feedback discussions. Participants commonly perceived video feedback as an efficient approach for consolidating screening skills and as important for the rapid implementation of primary-level screening in community settings. With respect to training perceptions, many participants indicated that video-feedback learning facilitated mastery of key screening points, such as eye contact and the child’s social-communication status.

Participants also indicated that video feedback not only enabled community screening teams to identify potential problems during video recording, but also supported more intuitive peer learning during cross-community video exchange by allowing teams to observe and emulate strengths in other teams’ practices and to identify and correct common issues. The video format was also perceived as conducive to repeated viewing and learning. In addition, some participants noted that each community was required to submit only one to two videos for presentation, which was not perceived as adding substantial workload; the process of video submission was also considered motivating for improving screening quality in order to produce higher-quality videos.

Regarding safety concerns related to video recording, when asked whether parents might refuse or be uncooperative due to concerns about exposure of personal information, participants reported that parental informed consent was obtained prior to video recording and that parental cooperation was sought accordingly. When discussing the acceptability of using video feedback for technical supervision in primary care, participants indicated that smartphone-based video recording was convenient. Recording in routine practice environments was perceived as promoting a more natural screening state and as better reflecting real-world workflow.

During primary care well-child visits, when a child is identified as high risk, frontline physicians are required to guide parents and facilitate adequate understanding of heterogeneity among high-risk children. Some parents may believe that the child is merely developing more slowly and will gradually “catch up,” which can delay further clinical evaluation and undermine the value of early screening and early intervention.

As frontline physicians’ behavioral observation skills improve—particularly with better mastery of initiation timing, waiting time, number of operations, use of physical gestures during language items, and operational sequence—more nuanced understanding may develop regarding behaviors such as reduced speech, reduced eye contact, reduced responsiveness, and reduced pointing. In addition, discrepancies between physician-led observation and parent-initiated prompts may help clinicians identify parental misunderstandings (15). As a result, clinicians may be better positioned to communicate core interpretations of high-risk behaviors in a timely manner, potentially increasing the likelihood that parents proceed to further diagnostic evaluation.

Post-diagnostic intervention for ASD is long-term and often requires substantial family time and financial investment (14). Social support is therefore essential. ASD intervention is comprehensive; in addition to behavioral intervention, nutrition, growth and development, and behavioral assessment are also important. If community physicians master behavioral observation skills and integrate them with routine child health monitoring competencies, and subsequently provide supportive health education, they may serve as a meaningful source of social support (16), thereby strengthening families’ capacity to sustain intervention efforts.

Research on AI-based technologies for ASD screening is expanding; however, nationwide coverage may require additional time. Because AI equipment is typically site-fixed, implementation in more remote areas may remain challenging in the near term. If frontline child health physicians possess behavioral observation skills, children may be observed and assessed across multiple settings and time points, enhancing flexibility of primary-level assessment.

In broadly implemented primary-level skills training, no single method can meet all educational needs. The present study indicates that video feedback—through recording routine practice and sharing feedback aligned with procedural steps—can provide a relatively accessible and convenient means of supporting team-based improvement in procedural quality among frontline staff (17).